Juan David Bastidas-Rodrguez

Modeling and diagnosis of photovoltaic arrays

Doctoral Thesis

Supervised by
Dr. Edinson Franco, Dr. Giovanni Spagnuolo and Dr. Carlos Andres
Ramos-Paja

Escuela de Ingeniera Electrica y Electronica - Universidad del Valle,

Dipartimento di Ingegneria dellinformazione, Ingegneria Elettrica e
Matematica Applicata - Universit`a degli Studi di Salerno and
Departamento de Energa Electrica y Automatica - Universidad
Nacional de Colombia

UNIVERSIT DEGLI STUDI DI SALERNO

Cali
2014
2
 To my beloved wife ... Margarita,
my mother ... Gilma,
my father ... Jaime,
my sister ... Lina
and all my family.
A mi amada esposa ... Margarita,
mi mam ... Gilma,
mi pap ... Jaime,
mi hermana ... Lina
y toda mi familia.

3
4
Acknowledgments
I thank God for being the force, light and guidance to continue moving forward in difficult times
and for putting in my way all the people and opportunities that have allowed me to reach this
accomplishment. Without Him nothing is possible.

I thank my beloved wife because she is not only my engine, my strength and my joy but also my
advisor and support to make the right decisions to always move forward. I also thank her for all the
sacrifices she made to reach this achievement together. I thank my Mom -Gilma-, my Dad -Jaime-
and my sister -Lina- because they were who gave me the best lessons in my life with their love,
example and dedication. In addition, the support and encouragement of my Mom, from the sky,
my Dad and my sister gave me the strength to overcome the difficult times. I thank my Aunts and
Uncle for always being available for anything I needed, but above all for all their love, unity and
unconditional support always provided. I also thank my Grandparents and my Cousins for all their
unconditional love and support and for all the encouragement and strength that have accompanied
me in these years.

I also thank my tutors, professors Edinson Franco, Carlos Ramos and Giovanni Spagnuolo for their
guidance, support and knowledge sharing in professional and personal area, without them this thesis
had not been possible. I thank to Professor Carlos Ramos (Department of Electrical and Automatic
Energy - Universidad Nacional de Colombia) for all his guidance and support during the research
stay on his University. I am also grateful to Professors Giovanni Spagnuolo and Giovanni Petrone
(Laboratorio di Circuiti Elettronici di Potenza e Fonti Rinnovabili - Universit`a degli Studi di Salerno)
for all their help, guidance and support during the internship conducted in their laboratory, for car-
ing both academics and personal aspects and for making me feel part of their group.

I also thank my friends at the Universidad del Valle: Sergey Edna and Ramiro, for their friend-
ship and support throughout this process and for helping me in the assembly of some laboratory
equipment. I also want to thank to professor Nicola Femia and my friends at the Universit`a degli
Studi di Salerno: Patrizio, Davide, Massimiliano, Giulia and Mattia for their friendship, guidance
and support during the internship and for making me feel part of their group.

Thanks to all of you.

5
6
Agradecimientos
Doy gracias a Dios por ser la fuerza, la luz y la gua para seguir siempre adelante en los momentos
difciles y por poner en mi camino a todas las personas y oportunidades que me han permitido
alcanzar este logro. Sin El nada es posible.

Agradezco a mi amada esposa, Margarita, porque ella no solo es mi motor, mi fuerza y mi alegra,
sino tambien mi consejera y apoyo para tomar las decisiones correctas para poder seguir siempre
hacia adelante. Le agradezco tambien por todos sus sacrificios para poder alcanzar este logro juntos.
Agradezco a mi Mama - Gilma -, a mi Papa - Jaime - y a mi Hermana - Lina - porque fueron
ellos quienes me han dado las mejores ense
nanzas de la vida con su amor, ejemplo y dedicacion.
Adem
as, el apoyo y aliento de mi Mama, desde el cielo, de mi Papa y mi Hermana me han dado
la fuerza para superar los momentos difciles. Agradezco a mi Tas y a mi To por estar siempre
disponibles para cualquier cosa que necesitara, pero sobre todo por todo su amor, uni
on y por el
apoyo incondicional bridado siempre. Agradezco tambien a mis Abuelos y a mis Primos por todo
su amor y apoyo incondicional y por todas sus palabras de aliento y fuerza que me han acompa
nado
en estos a
nos.

Tambien agradezco a mi tutores, los profesores Edinson Franco, Carlos Ramos y Giovanni Spagnuolo
por su gua, por su apoyo y sus ense
nanzas en el area profesional y personal, sin ellos esta tesis no
hubiera sido posible. Agradezco al profesor Carlos Ramos (Departamento de Energa Electrica y
Automatica - Universidad Nacional de Colombia) por todo su guia y apoyo durante la estancia de
investigacion realizada en universidad. Igualmente, agradezco a los profesores Giovanni Spagnuolo
y Giovanni Petrone (Laboratorio di Circuiti Elettronici di Potenza e Fonti Rinnovabili - Universit`a
degli Studi di Salerno) por toda su ayuda, gua y apoyo durante la pasanta realizada en su labora-
torio, por preocuparse tanto de la parte academica como personal y por hacerme sentir parte de su
grupo.

Quiero agradecer tambien a mis amigos de la Universidad del Valle: Sergey, Edna y Ramiro, por
su amistad y apoyo a lo largo de este proceso y por su ayuda en el montaje de algunos equipos de
laboratorio. Igulamente, quiero agradecer al profesor Nicola Femia y a mis amigos de la Universit`a
degli Studi di Salerno, Patrizio, Davide, Massimiliano, Giulia y Mattia por su amistad, guia y apoyo
durante la pasanta y porque me hicieron sentir parte de su grupo.

Gracias a todos.

7
8
 Modeling and diagnosis of photovoltaic arrays
by
Juan David Bastidas-Rodrguez

Escuela de Ingeniera Electrica y Electr

onica - Universidad del Valle, Dipartimento di Ingegneria
dellinformazione, Ingegneria Elettrica e Matematica Applicata - Universit`a degli Studi di Salerno
and Departamento de Energa Electrica y Automatica - Universidad Nacional de Colombia
Cali, 2014

Abstract
In most of the photovoltaic (PV) system the modules that form the PV generator do not operate
under the same conditions (mismatching conditions). Such a phenomenon is one of the main causes of
the efficiency reduction of the PV systems. In this thesis, the mismatching problem is faced starting
with an overview of the Maximum Power Point Tracking (MPPT) techniques aimed at extracting the
maximum power available in the PV modules of a given array that works in mismatched conditions.
In such an overview, the general architectures of the MPPT techniques is presented along with their
main advantages and disadvantages and hardware requirements in order to provide to the reader
useful information to select an MPPT technique for a particular application.
The overview of MPPT techniques puts into evidence the necessity of developing accurate and
fast models to represent PV arrays with different electrical configurations operating under uniform
and mismatching conditions since they are the base for accurate and realistic simulation and analysis
of PV arrays that is mandatory in MPPT techniques design. Therefore, in this thesis models for
PV arrays in the three widely adopted configurations, i.e. Series-Parallel (SP), Total Cross-Tied
(TCT) and Bridge-Linked (BL), are also proposed. The structures of such models facilitate their
implementation in standard programming languages (e.g. C or Matlab) in order to let them be used
in the development of new MPPT techniques, energy production estimation, power distribution
network analysis and other applications.
The second Chapter also puts into evidence an increased interest in the dynamical electrical
reconfiguration of the PV arrays to mitigate the mismatching effects, which consist in changing the
electrical connection of a PV panel in the array through a matrix of switches. However, the methods
found in literature do not use the arrays models in the decision of the configuration that provide
the highest output power. Hence, a method to determine the SP array configuration that provides
the highest Global Maximum Power Point (GMPP), based on the array model, is also proposed and
validated using simulations and experimental data.
The accurate PV generators models developed in this thesis are also very useful for putting into
evidence the electrical stresses a PV panel might be subjected to during its lifetime. Moreover,
the mismatching conditions and outdoor operation may produce degradation in the module of a PV
array, which may increase significantly the return of investment time of a PV system. Therefore, two
model-based indicators to detect and quantify the degradation of a PV module are also proposed
in this thesis. Such indicators use the a model to represent a PV module without degradation
and combine such information with the experimental measurements to estimate variation in two
parameters. The estimation of those variations allows to quantify the module degradation regardless
the irradiance and temperature conditions. Moreover, the usefulness of the indicators is tested with
simulation and experimental data.
Finally, the thesis conclusions are presented along with the possible future work in this field.

Thesis Supervisors: Dr. Edinson Franco (Universidad del Valle), Dr. Giovanni Spagnuolo (Univer-
sit`a degli Studi di Salerno) and Dr. Carlos Andres Ramos-Paja (Universidad Nacional de Colombia)

9
10
Contents

Preface 21

1 Introduction 23
1.1 PV cells, modules, panels and arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.2 Single diode model of a PV module . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.3 PV arrays operating under uniform and mismatching conditions . . . . . . . . . . . 27
1.4 Maximum Power Point Tracking (MPPT) techniques . . . . . . . . . . . . . . . . . . 29
1.5 PV panels used for simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.6 Experimental platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.7 Scope of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

2 MPPT Architectures for PV Systems in Mismatching Conditions 35

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.2 The static approach for mitigating the mismatching problems . . . . . . . . . . . . . 38
2.3 Classification of MPPT architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.3.1 Centralized MPPT architectures (CMPPT) . . . . . . . . . . . . . . . . . . . 41
2.3.2 Distributed MPPT architectures (DMPPT) . . . . . . . . . . . . . . . . . . . 42
2.3.3 Reconfiguration-MPPT architectures (RMPPT) . . . . . . . . . . . . . . . . 45
2.4 Literature survey of CMPPT-based solutions . . . . . . . . . . . . . . . . . . . . . . 46
2.4.1 Sampling P-V curves + hill climbing . . . . . . . . . . . . . . . . . . . . . . . 47
2.4.2 Computational intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.4.3 Other CMPPT techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.5 Literature survey of DMPPT-based solutions . . . . . . . . . . . . . . . . . . . . . . 50
2.5.1 DMPPT-Us connected in SP configuration . . . . . . . . . . . . . . . . . . . 52
2.5.2 DMPPT-Us connected in parallel with a string of modules/panels . . . . . . 54
2.5.3 DMPPT-Us connected in parallel . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.5.4 DMPPT-Us where the PV sub-array is a string or small array . . . . . . . . . 56
2.6 Literature survey of RMPPT-based solutions . . . . . . . . . . . . . . . . . . . . . . 58

11
 2.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

3 Modeling of Series-Parallel PV arrays under mismatching conditions 67

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.2 Basic equations of the SP array elements . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.3 Calculation of the inflection voltages . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.4 String current calculation using inflection voltages . . . . . . . . . . . . . . . . . . . 73
3.5 PV field simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.6 Performance evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.6.1 Simulation of a small PV field . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.6.2 Simulation of medium size PV fields . . . . . . . . . . . . . . . . . . . . . . . 77
3.6.3 Simulation of large PV fields . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.6.4 Prediction of the energy production . . . . . . . . . . . . . . . . . . . . . . . 81
3.6.5 Dynamic simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.6.6 Validation through experimental measurements . . . . . . . . . . . . . . . . . 83
3.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

4 Modeling of Total Cross-Tied and Bridge-Linked PV arrays under mismatching

conditions 89
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.2 Modeling of TCT arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.2.1 Experimental comparison of TCT and SP configurations . . . . . . . . . . . . 93
4.2.2 TCT configuration analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.2.3 Inflection points in TCT configurations . . . . . . . . . . . . . . . . . . . . . 97
4.2.4 Calculation of the TCT array current . . . . . . . . . . . . . . . . . . . . . . 99
4.2.5 MPP calculation in TCT arrays . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.2.6 Experimental validation of the TCT model . . . . . . . . . . . . . . . . . . . 102
4.2.7 Energy evaluation example in TCT configuration . . . . . . . . . . . . . . . . 105
4.3 Modeling of BL arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.3.1 BL configuration analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.3.2 Simulation results of the BL model . . . . . . . . . . . . . . . . . . . . . . . . 109
4.3.3 Simulation of a small and medium BL array . . . . . . . . . . . . . . . . . . . 110
4.3.4 Energy estimation in a medium BL array . . . . . . . . . . . . . . . . . . . . 111
4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

5 Reconguration analysis of photovoltaic arrays based on parameters estimation 115

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

12
 5.2 Calculation of PV array current by using inflection points . . . . . . . . . . . . . . . 118
5.2.1 Inflection points in mismatching conditions . . . . . . . . . . . . . . . . . . . 119
5.2.2 String and array current calculation . . . . . . . . . . . . . . . . . . . . . . . 122
5.3 Detection of inflection and open-circuit points from PV string data . . . . . . . . . . 123
5.4 Estimation of PV modules parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.4.1 Model parameters and data sets . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.4.2 Impact of the model parameters error . . . . . . . . . . . . . . . . . . . . . . 129
5.4.3 Parameters estimation using classical estimation algorithms . . . . . . . . . . 130
5.4.4 Parameters estimation using approximated analytical expressions . . . . . . . 131
5.4.5 Application example: PV string parameters estimation . . . . . . . . . . . . . 132
5.5 PV array reconfiguration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
5.5.1 Reconfiguration system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.5.2 Experimental measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

6 Diagnostic of PV modules 143

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.2 Degradation effects in the module I-V curves . . . . . . . . . . . . . . . . . . . . . . 146
6.3 Traditional PV module performance indicators . . . . . . . . . . . . . . . . . . . . . 147
6.3.1 NFF behavior for different degradation conditions . . . . . . . . . . . . . . . 148
6.4 Proposed PV module performance indicators . . . . . . . . . . . . . . . . . . . . . . 150
6.4.1 N Rs and N Rh behavior for different degradation conditions . . . . . . . . 151
6.5 Effects of the sensors uncertainties on traditional and proposed indicators . . . . . . 152
6.6 Experimental results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

7 Conclusions and future works 159

A Estimation of single-diode model parameter using a Genetic Algorithm 163

A.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
A.2 Proposed method to calculate the parameters . . . . . . . . . . . . . . . . . . . . . . 164
A.3 Simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
A.4 Parameters variations with different irradiance and temperature conditions . . . . . 170
A.5 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172

B Publications 173

Bibliography 175

13
14
List of Figures

1-1 PV Cells, models, panels, strings, and arrays . . . . . . . . . . . . . . . . . . . . . . 25

1-2 Single-diode model of a PV cell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1-3 Example of I-V and P-V curves for three different irradiance conditions of the ERDM
85SM/5 panel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1-4 Example of I-V and P-V curves for three different temperature conditions of the
ERDM 85SM/5 panel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1-5 PV field under mismatching conditions. The clouds dimensions identify the reduction
in the received irradiance. Big cloud: 0.5 kW/m2 . Small cloud: 0.75 kW/m2 . No
cloud: 1.0 kW/m2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1-6 Modules and array power vs. voltage curves. . . . . . . . . . . . . . . . . . . . . . . 29
1-7 Experimental platforms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

2-1 PV field configurations. SP: black lines. TCT: black lines + red lines + blue lines,
BL: black lines + red (blue) lines. The clouds dimensions identify the percentage
of the shade: Big cloud means PV module shaded at 50%. Small cloud means PV
module shaded at 25%. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2-2 Modules and array P-V curves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2-3 General structure of CMPPT architecture. . . . . . . . . . . . . . . . . . . . . . . . . 41
2-4 Characteristic curves for a PV array of 2 2 operating in uniform conditions with 1
kW/m2 (continuous line) and 0.7 kW/m2 (dashed line). . . . . . . . . . . . . . . . . 42
2-5 General structure of DMPPT architectures. . . . . . . . . . . . . . . . . . . . . . . . 43
2-6 General structure of the RMPPT architecture. . . . . . . . . . . . . . . . . . . . . . 45
2-7 Further DMPPT architectures. The arrows identify the energy flows. . . . . . . . . . 53

3-1 PV Field with m strings of n modules parallel connected. . . . . . . . . . . . . . . . 68

3-2 Single-diode model of a PV module with bypass diode. . . . . . . . . . . . . . . . . . 68
3-3 Piecewise linear approximation of bypass and blocking diodes. . . . . . . . . . . . . . 72
3-4 Inflection voltages of one string with four PV modules. . . . . . . . . . . . . . . . . . 73

15
3-5 Flow chart of the proposed model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3-6 Characteristic curves of the first scenario simulations. . . . . . . . . . . . . . . . . . 78
3-7 Characteristic curves of the second scenario simulations (PV field 15x15). . . . . . . 79
3-8 Characteristic curves of the third scenario simulations (PV field 50x50). . . . . . . . 80
3-9 Simulations for typical winter day, South of Italy. . . . . . . . . . . . . . . . . . . . . 81
3-10 Simulations for one week of summer, center of Colombia. . . . . . . . . . . . . . . . . 82
3-11 Battery charger scheme in PSIM considering the proposed model. . . . . . . . . . . . 84
3-12 Power vs. Voltage characteristic curves of the 11x1 PV array. . . . . . . . . . . . . . 84
3-13 Battery charger dynamic simulation in PSIM. . . . . . . . . . . . . . . . . . . . . . . 85
3-14 Experimental characteristics for Panel 1. . . . . . . . . . . . . . . . . . . . . . . . . . 85
3-15 Experimental characteristics for string of Panel 1 and Panel 2. . . . . . . . . . . . . 86

4-1 PV arrays of 3 columns and 3 rows in SP (connections in thin continuous lines), TCT
(SP connections plus white dashed lines) and BL (SP connections plus thick black
lines) configurations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4-2 I-V and P-V curves for a 3x3 PV array with a fixed mismatching pattern in SP, TCT
and BL configurations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4-3 PV arrays with two strings and two rows. . . . . . . . . . . . . . . . . . . . . . . . . 94
4-4 Connection scheme of experimental test bench. . . . . . . . . . . . . . . . . . . . . . 94
4-5 PV arrays characteristics in mismatching conditions. . . . . . . . . . . . . . . . . . . 96
4-6 Equivalent circuit of a row in a TCT array. . . . . . . . . . . . . . . . . . . . . . . . 97
4-7 Model and circuital simulation of a 4 2 TCT array. . . . . . . . . . . . . . . . . . . 100
4-8 MPP calculation for a 4 2 TCT array. . . . . . . . . . . . . . . . . . . . . . . . . . 102
4-9 Relative size of the non-linear equation systems to find the MPP. . . . . . . . . . . . 103
4-10 MPP calculation for a 4 2 TCT array. . . . . . . . . . . . . . . . . . . . . . . . . . 104
4-11 Simulations for typical winter day, South of Italy, for TCT arrays. . . . . . . . . . . 106
4-12 Ideal single-diode model with bypass diode. . . . . . . . . . . . . . . . . . . . . . . . 107
4-13 Flow chart to write non-linear equation system and Jacobian matrix for BL arrays. . 109
4-14 Characteristic curves of a small and medium BL arrays. . . . . . . . . . . . . . . . . 110
4-15 GMPP for a 20x3 BL array in two configurations. . . . . . . . . . . . . . . . . . . . . 112

5-1 PV Field with m strings of n modules parallel connected. . . . . . . . . . . . . . . . 118

5-2 String of two modules under uniform and mismatching conditions. . . . . . . . . . . 120
5-3 Characteristic curves of 2x1 PV array. Blue lines: module 1,1. Dashed-red lines:
module 2,1. Purple lines: 2x1 string. . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5-4 Inflection points effect on PV strings. . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5-5 Inflection points detection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

16
5-6 Flow chart of the parameters estimation method. . . . . . . . . . . . . . . . . . . . . 127
5-7 Points used for parameter estimation. . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5-8 Effect of parameters errors on the PV module current. . . . . . . . . . . . . . . . . . 129
5-9 Parameters estimation procedure from data. . . . . . . . . . . . . . . . . . . . . . . . 131
5-10 PV string measurements and estimated model simulation. . . . . . . . . . . . . . . . 133
5-11 Estimation of the new strings power characteristics. . . . . . . . . . . . . . . . . . . 135
5-12 PV modules characteristics and estimated models. . . . . . . . . . . . . . . . . . . . 135
5-13 Switches matrix scheme for the 3 2 PV array reconfiguration. . . . . . . . . . . . . 137
5-14 Experimental and estimated strings powear curves. . . . . . . . . . . . . . . . . . . . 138
5-15 Experimental and estimated PV modules characteristics. . . . . . . . . . . . . . . . . 139
5-16 Experimental and estimated power curves. . . . . . . . . . . . . . . . . . . . . . . . . 140

6-1 I-V curves distortion with variations in Rs and Rh. . . . . . . . . . . . . . . . . . . . 147

6-2 Single-diode model with degradation represented by additional series and parallel
resistances. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
6-3 NFF for variations in Rs and Rs and different irradiances and temperatures. Con-
tinuous line: minimum temperature, dashed lines: maximum temperature. . . . . . . 149
6-4 N Rs and N Rh for variations in Rs from 0 to RsST C 4 and different irradiances
and temperatures. Continuous line: minimum temperature, dashed lines: maximum
temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
6-5 N Rs and N Rh for variations in Rh from to RhST C /80 and different irradiances
and temperatures. Continuous line: minimum temperature, dashed lines: maximum
temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
6-6 Monte Carlo analysis of NFF for variations in Rs and Rh as well as different irradi-
ances and temperatures. Continuous line: maximum values, dashed lines: minimum
values. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6-7 Monte Carlo analysis of N Rs for increments in Rs and different irradiances and
temperatures. Continuous line: maximum values, dashed lines: minimum values. . . 154
6-8 Monte Carlo analysis of N Rh for reductions in Rh and different irradiances and
temperatures. Continuous line: maximum values, dashed lines: minimum values. . . 155
6-9 Example of reference and experimental I-V curves for the different additional Rs
values. Dashed lines: experimental I-V curves. Continuous lines: reference I-V curves. 155
6-10 Experimental results (dots) of NFF for different increments in Rs and operating
conditions. Dashed lines represent the trend of the experimental points. . . . . . . . 156
6-11 Experimental results (dots) of N Rs for different increments in Rs and operating
conditions. Dashed lines represent the trend of the experimental points. . . . . . . . 157

17
A-1 Five key points of an I-V curve proposed in Kings Model. . . . . . . . . . . . . . . . 165
A-2 I-V curves obtained form calculated parameters for different irradiances. . . . . . . . 169
A-3 I-V curves obtained form calculated parameters for different temperatures. . . . . . . 169
A-4 Parameters values for different irradiance and temperature conditions. . . . . . . . . 171

18
List of Tables

1.1 Electrical characteristics of the PV panels used in simulations . . . . . . . . . . . . . 31

2.1 CMPPT techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

2.2 Main advantages and disadvantages of DMPPT sub-classes . . . . . . . . . . . . . . 51
2.3 DMPPT techniques part 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.4 DMPPT techniques part 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.5 Central units of DMPPT techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.6 RMPPT techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.7 Main advantage and disadvantages of DMPPT sub-classes . . . . . . . . . . . . . . . 66

3.1 Parameters for the Fast, Complex, and Tradeoff modes . . . . . . . . . . . . . . . . . 77

3.2 Irradiance profiles for simulation of 2x1 and 3x1 PV fields [% of IscST C ] . . . . . . . 77
3.3 Relative mean errors for simulations of 2x1 and 3x1 PV fields . . . . . . . . . . . . . 78
3.4 Relative mean errors and simulation time of the 15x15 PV field . . . . . . . . . . . . 79
3.5 Relative mean errors and simulation time of the 50x50 PV field . . . . . . . . . . . . 80
3.6 Relative mean errors for simulations of 2x1 and 3x1 PV fields . . . . . . . . . . . . . 82
3.7 Model parameters for the PV panels . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

4.1 Mismatching profile of the 3x3 PV array. . . . . . . . . . . . . . . . . . . . . . . . . . 92

4.2 Mismatching profile to simulate the TCT array. . . . . . . . . . . . . . . . . . . . . . 100
4.3 Mismatching profile for experimental validation of the model. . . . . . . . . . . . . . 104
4.4 Mismatching profile of the 3x3 BL array. . . . . . . . . . . . . . . . . . . . . . . . . . 111

5.1 PV string parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

5.2 Possible configurations of a 3 2 PV array . . . . . . . . . . . . . . . . . . . . . . . 136
5.3 Switches matrix states for the 3 2 PV array configurations. . . . . . . . . . . . . . 137
5.4 Experimental and estimated GMPP . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

6.1 Main PV module degradation mechanisms . . . . . . . . . . . . . . . . . . . . . . . . 146

6.2 Main characteristics of selected sensors . . . . . . . . . . . . . . . . . . . . . . . . . . 153

19
A.1 Genetic Algorithm settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
A.2 BP585 (2002) parameters for different irradiances and fixed temperature . . . . . . . 168
A.3 BP585 (2002) parameters for fixed irradiance and different temperatures . . . . . . . 169

20
Preface
This thesis reports the results of my Ph.D. study perfomed in cooperation between the Escuela
de Ingeniera Electrica y Electr
onica of the Universidad del Valle, the Departamento de Energa
Electrica y Autom
atica of the Universidad Nacional de Colombia and the Dipartimento di Ingegne-
ria dellinformazione, Ingegneria Elettrica e Matematica Applicata of the Universit`a degli Studi di
Salerno. Such study was developed from August 2009 to June 2014 and supported by the Universi-
dad del Valle under the program Asistente de Docencia (two and a half years) and the Departamento
Administrativo de Ciencia Tecnologa e Innovaci
on of Colombia (COLCIENCIAS) with the schol-
arship APCC-ND-66-197 (two and a half years). I developed my studies as a Ph.D. student of the
Research Group of Industrial Control (GICI) of Universidad del Valle and the Power Electronics
and Renewable Sources Laboratory of the Universit`a degli Studi di Salerno.

The thesis contains results that have been published, accepted or will be submitted for publication
as papers in international journals and conference proceedings (see appendix B), but the thesis
format makes it easier for the reader to gain a better understanding of the overall work and the
improvements to the state of the art.

The photovoltaic modules are connected in different configurations to reach the voltage and cur-
rent levels required by a given load or the input constraints of PV inverters. The resulting set
of photovoltaic modules can be denominated as photovoltaic array, field or generator. The terms
photovoltaic array or photovoltaic field are used indistinctly along the document, while the term
photovoltaic generator can be used also to refer a photovoltaic module or panel; hence, its meaning
depends on the context in which it is used. The use of one of the discussed terms depends on which
of them matches better with the writing style of a particular paragraph.
The thesis follows common publishing guidelines given by international journals. The bibliographical
citations have been numbered in order of appearance, as well as the equations, figures and tables.
Finally, the simulations of the photovoltaic modules, panels and arrays, as well as power electronics
circuits, have been carried out by using Matlab/Simulink/C code and PSIM software.

21
22
Chapter 1

Introduction

Development of power systems employing renewable energy sources has exhibited an acceleration
in the last years as consequence of the global challenges to mitigate the climate change, reduce the
CO2 emissions and contribute to sustainable development. The photovoltaic (PV) energy is one of
the most promising renewable energies, since the International Energy Agency (IEA) predicts that
in 2030, 5 % of the global electric energy consumption will be from PV and it will grow up to 11 %
in 2050 [1]. The IEA also anticipates that the cost of each installed kW will reduce by more than
50 % for 2020 with respect to the costs in 2009 [1].
In general, the PV systems have the following advantages: modularity, zero emissions, the energy
source source is free and easy to find, they do not have mobile parts, they can coexist with other
energy sources and their lifetime is longer than 20 years. All these advantages have led to an increase
in the PV installations all over the world, especially in distributed generation systems connected
to the grid [2]. However, the PV generators also have some disadvantages: relatively low efficiency
(between 5 and 20 % depending on the technology), their behavior is highly non-linear, the energy
production depends directly on the environmental conditions (irradiance and temperature) and they
may have a reduced lifetime due to early degradation produced by the outdoor operation [3, 4].
This Chapter presents some basic concepts related to PV systems that will be used in the
following Chapters of the thesis. Moreover, the PV panels used for the simulations along the thesis
are introduced and the experimental platforms are described. Finally, the scope of the thesis is
presented in order to give to the reader a general idea of the content and organization of the thesis.

1.1 PV cells, modules, panels and arrays

PV generators are groups of semiconductor cells (usually made of silicon) series and/or parallel con-
nected that directly transform the sun light into electricity by means of the photovoltaic effect. Each
cell is composed by, at least, two thin semiconductor layers, one n-type, exposed to solar irradiance,

23
and other one p-type. When a photon hits the n-layer and it is absorbed by a valence electron, the
photon energy is delivered to the electron; if the photon energy is higher than the semiconductor
band gap, the electron jumps to the conduction band where it moves freely in the presence of an
electric potential. On the contrary, if the photon energy is smaller than the semiconductor band
gap, the energy received by the electron is turned into kinetic energy, thus increasing the device
temperature [5].
To obtain output voltage values for different applications it is necessary to connect several PV
cells in series, since each one generates a voltage near to the conduction voltage of one diode (0.6 V
for silicon, approximately). That group of cells forms a basic PV unit called PV module, which is
commonly protected by a diode connected in antiparallel (Fig. 1-1).
A PV panel is composed by one or several PV modules connected in series (Fig. 1-1) and a set
of panels series connected form a string. The strings can be connected in parallel for constructing
arrays (or fields) to reach the current and voltage levels required for feeding a load. The bypass
diodes in each module provide an alternative path for the string current to avoid its reduction when
a module is receiving an small amount of irradiance or if it is damaged. The bypass diodes are
important because protect the cells in the module and avoid the reduction of the current into a
string and, therefore, a reduction in the power.
The configuration of the array described in the previous paragraph is known as Series-Parallel
(SP); however, the panels can be connected in different configurations, like Total Cross-Tied (TCT),
Bridge-Linked (BL), among others, that may provide some advantages depending on the applications.
The models of the widely adopted arrays configurations will be presented with more detail in the
following chapters.

1.2 Single diode model of a PV module

A PV cell can be represented by the circuit shown in Fig. 1-2, which is denominated single-diode
model and it is widely adopted in literature [6, 7, 8]. Although there are other electrical models
to represent more accurately a PV cell (e.g. double diode model [9, 10]) this thesis considers the
single-diode model since its accuracy is enough for the scope of the thesis.
In Fig. 1-2 the current source (Iphcell ) represents the electron flow produced by photon collisions
in the semiconductor material and the non-linear behavior of the p-n junction is modeled through
a diode in parallel. The losses caused by the leakage current are represented by a parallel resistor
(Rhcell ) and the resistance of the associated junctions and the metal-semiconductor contacts are
modeled with a series resistor (Rscell ). By using the Kirchhoff laws, (1.1) is obtained, where Icell
and Vcell are the current and the voltage of the cell, respectively, Isatcell is the inverse saturation
current of the diode, is the ideality factor of the p-n junction and V tcell is the thermal voltage of

24
 Figure 1-1: PV Cells, models, panels, strings, and arrays

the cell. The thermal voltage is defined as V tcell = k T /q, where k is the Boltzmann constant, q is
the electron charge, and T is the cell temperature in Kelvin degrees.

   
Vcell + Icell Rscell Vcell + Icell Rscell
Icell = Iphcell Isatcell exp 1 (1.1)
V tcell Rhcell

A PV module is usually composed by a number of series connected cells (N s). Considering that
all the cells in the module are operating under the same irradiance and temperature conditions, it
is possible to extend the single-diode model for a PV module. The equivalent expression of (1.1) for
a PV module is presented in (1.2), where the module parameters (Iph, Isat, V t, Rs and Rh) and
variables (I and V ) can be calculated as shown in (1.3).

   
V + I Rs V + I Rs
I = Iph Isat exp 1 (1.2)
Vt Rh

V = N s Vcell , I = Icell , Iph = Iphcell , Isat = Isatcell ,

V t = N s V tcell , Rs = N s Rscell , Rh = N s Rhcell (1.3)

25
 Rscell Icell

I Dcell I Rhcell

Iphcell
VDcell Vcell
Rhcell

Figure 1-2: Single-diode model of a PV cell.

From the implicit function (1.2) it is possible to obtain an explicit expression for the module
current as function of the module voltage (1.4) by using the Lambert-W function [11]. Similarly,
the module voltage can be expressed as an explicit function of current as shown in (1.5). In (1.4)
and (1.5) the Lambert-W function is represented by W0 .

In some cases Rs and Rh can be neglected, i.e. Rs = 0 and Rh = , which lead to a simplified
expressions for the voltage and current of the module [12, 13]. In such cases the parameters of the
model need to be calculated in order to fit the model with a given reference current vs. voltage
curve, as will be shown in Chapter 5.

Vt Iph + Isat V /Rh

I = W0 () + (1.4)
Rs (1 + Rs/Rh)
 
Isat Rs V + Rs (Isat + Iph)
= exp
(1 + Rs/Rh) V t (1 + Rs/Rh) V t

V = V t W0 () I (Rs + Rh) + Rh (Iph + Isat) (1.5)

 
Isat Rh Rh
= exp (Iph + Isat I)
Vt Vt

Fig. 1-3 and Fig. 1-4 show an example of the current vs. voltage (I-V) and power vs. voltage (P-
V) curves of an ERDM 85SM/5 [14] PV panel. Those curves illustrate the non-linear behavior of a
PV module for three different irradiance (denoted by G) and three different temperature (denoted by
T ) conditions, respectively. It can be observed that the short-circuit current (Isc) strongly depends
on the irradiance (Fig. 1-3(a)), while the open-circuit voltage (V oc) is inversely proportional to the
temperature (Fig. 1-4(a)). Moreover, there is only one operating point where the module delivers
its maximum power for a given irradiance and temperature condition. Such a point is defined as
Maximum Power Point (MPP) and it is marked with a circle in Fig. 1-3(b) and Fig. 1-4(b) for each
different operating condition.

26
 6
2
G = 1 kW/m
80 2
5 G = 0.8 kW/m
G = 0.6 kW/m2
4 60
Current (A)

Power (W)
3
40
2
2
G = 1 kW/m 20
1 2
G = 0.8 kW/m
G = 0.6 kW/m2
0 0
0 5 10 15 20 0 5 10 15 20
Voltage (V) Voltage (V)
(a) I-V curves. (b) P-V curves.

Figure 1-3: Example of I-V and P-V curves for three different irradiance conditions of the ERDM
85SM/5 panel.

6 100
o
T = 50 C
5 o
80 T = 25 C
o
T=0 C
4
Current (A)

Power (W)

60
3

o
40
2 T = 50 C
T = 25 oC
20
1 T=0 C
o

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Voltage (V) Voltage (V)
(a) I-V curves. (b) P-V curves.

Figure 1-4: Example of I-V and P-V curves for three different temperature conditions of the ERDM
85SM/5 panel.

1.3 PV arrays operating under uniform and mismatching

conditions

When a PV array (independently on its configuration) is operating under uniform conditions, namely
all the parameters of the PV modules assume exactly the same and all the modules operate under
the same irradiance and temperature conditions, there is a particular voltage value (V mpp) at
which the power extracted from the PV field assumes the maximum value (MPP). Unfortunately,
a real PV system does not operate under ideal conditions for two principal reasons: first, the
modules parameters are not the same due to manufacturing tolerances; second, the irradiance and
temperature conditions of each module may be different [15, 16, 17] due to shadows of surrounding
objects, dust, trees leaves, clouds, and other elements presented in rural and urban areas.
When one module into a string receives low irradiance level (in comparison with the other ones),
the current produced by such a module is lower than the one produced by the other modules, so

27
that the current of the whole string is limited by the module with the lowest current. The typical
solution to that problem is to connect a diode in antiparallel to each module. Such a diode processes
the current difference and prevents the power dissipation on the PV module. Nevertheless, this
solution generates multiple maximum power points, but only one of them corresponds to the global
maximum (GMPP) in the power versus voltage curve of the PV string. This phenomenon is known
as mismatching and it degrades the overall efficiency of the PV system significantly, even if a small
difference of irradiance and/or in the PV modules parameters occurs [18].
To illustrate such a concept, an SP arrays of nine ERDM 85SM/5 PV panels (three strings of
three modules each) is considered (see Fig. 1-5). Fig. 1-6(a) presents the P-V curves of the array.
In this example, it is assumed that all the modules have the same temperature (25 o C) and that the
irradiace over each module is uniform. Moreover, the irradiance of unshaded modules is 1 kW/m2 ,
while the irradiance of the shaded modules can be 0.75 kW/m2 (25 % of shade opacity) or 0.5
kW/m2 (50 % of shade opacity). The irradiance distribution over the PV array is illustrated in Fig.
1-5, where big clouds, small clouds and no cloud represent irradiances of 0.5 kW/m2 , 0.75 kW/m2
and 1 kW/m2 , respectively. The P-V curves of a module for the different irradiance conditions is
presented in Fig. 1-6(b).

Bypass Diode

BD BD
Shadowing
BD
Shadowing

BD BD BD
PV Array:
Shadowing Shadowing
SP

BD BD BD

Shadowing
Shadowing

Figure 1-5: PV field under mismatching conditions. The clouds dimensions identify the reduction in
the received irradiance. Big cloud: 0.5 kW/m2 . Small cloud: 0.75 kW/m2 . No cloud: 1.0 kW/m2

Under uniform conditions, all the modules have the same P-V curve marked as 1 kW/m2 in Fig.
1-6(b). The equivalent P-V curve of the array corresponds to the sum of the P-V curves of each
module (see Fig. 1-6(a)) and MPP is 776.1 W . While in mismatching conditions the P-V curve
is different since it exhibits three maximums, two local (Local MPP - LM P P ) at 18 V and 36 V ,
approximately, and one global (Global MPP in SP configuration - GM P PSP ) at around 57 V of 408

28
W . Such GMPP does not correspond to the sum of the MPPs of all the modules in the array (black
line in Fig. 1-6(a)) and it is significantly lower than the MPP under uniform conditions.

Sum of the MPP of all the Uniform

modules in mismatching conditions conditions
800
SP mismatch. 100
2
SP uniform 1 kW/m
600
Array power (W)

80

Module power (W)

2
0.75 kW/m
LMPP 2
400 60 0.5 kW/m

40
200
GMPP 20

0 0
0 10 20 30 40 50 60 70 0 5 10 15 20
Array voltage (V) Module voltage (V)

(a) Power vs. voltage curve of the 33 array in SP and (b) Power vs. voltage curve of the modules with high
TCT configuration under uniform (SP: blue, TCT: red) (blue), mid (green) and low (red) irradiance.
and mismatching (SP: violet, TCT: green) conditions.

Figure 1-6: Modules and array power vs. voltage curves.

This example considered only a SP array; however, the effects of the mismatching in the P-V
curves depend on the array configuration. If there are connections between the strings of the array,
as in TCT connection, the current of the modules can take alternative paths through the other
modules and bypass diodes producing different P-V curves. Such, curves may generate a different
GMPP and different numbers of LMPPs.

1.4 Maximum Power Point Tracking (MPPT) techniques

To extract the maximum energy from a PV array it is necessary to implement a control strategy
to identify the PV operation point characterized by the maximum power (MPP); such a control
strategy is named Maximum Power Point Tracking (MPPT). The traditional MPPT techniques most
widely used in commercial products are the Perturb and Observe (P&O) [19] and the Incremental
Conductance (IC) [20]. Those techniques guarantee satisfactory performances [21] provided that the
values of their parameters are suitably designed.
Nevertheless, those algorithms are effective only in arrays operating in uniform conditions. Even
if in these conditions the MPP is unique, its position depends strongly on the irradiance and tem-
perature variation, so that the traditional MPPT algorithms are optimized for having fast tracking
response during the environmental variations and a high efficiency under stationary conditions.
In urban contexts and other applications, it is almost impossible to find a large space without
shadows or, even worse, the shadow could appear due to a new building realized after the installation
of the PV system. Therefore, many real PV systems can be rarely assumed in uniform conditions,
even by part of the day. That is why, PV arrays operating in mismatching conditions require

29
different control strategies to extract the maximum power available in the array and avoid energy
losses, which are translated into an increase in the return of investment time of the system.

In literature is possible to find a large variety of MPPT techniques for PV arrays operating in
mismatching conditions. Although such techniques propose different algorithms, hardware require-
ments and implementation costs, they can be divided into the following three classes according to the
hardware structure of the system: Centralized MPPTs (CMPPT), Distributed MPPTs (DMPPT)
and Reconfiguration MPPTs (RMPPT).

In CMPPT techniques only one power stage is used to track the maximum power from the array.
The power converter is controlled by an algorithm aimed to find the highest MPP in the array.
DMPPT techniques divide the array into sub-arrays (e.g. panels, strings or small arrays) to simplify
the tracking of the MPP in each one of them. Each sub-array is connected to a power stage that
tracks the maximum power in the corresponding sub-array. On the other hand, reconfiguration
techniques (RMPPT) use a matrix of switches to modify, periodically, the electrical interconnection
of the panels in the array; this to mitigate the mismatching effects and increase the maximum
power available in the PV array (e.g. [22]). The CMPPT, DMPPT and RMPPT techniques will be
discussed in detail in Chapter 2.

1.5 PV panels used for simulations

Three PV panels were used along the thesis as reference to perform the simulations. Those panels
were selected in order to have mono-crystalline and poly-crystalline panels as reference and because
they were available to perform experimental measurements.

One of the panels is the ERDM 85SM/5 [14] produced by ERDM Solar. This is a mono-crystalline
PV panel that contains one module of 36 series connected cells. The second panel is the BP 585
[23] manufactured by BP Solar. Such a panel has two modules (i.e. two bypass diodes) of 18
mono-crystalline series connected cells each. The third panel used is the KC 120-1 [24] produced
by Kyocera and is composed by 36 poly-crystalline cells connected in series and organized in two
modules. The main electrical characteristics of the three panels in Standard Test Conditions (STC)
are presented in Table 1.1, where the temperature constants of the short-circuit current and the
open-circuit voltage for the KC 120-1 have been taken from System Advisor Model (SAM) database
provided by National Renewable Energy Laboratory (NREL) of the United States since they are
not available in the datasheet.

30
 Table 1.1: Electrical characteristics of the PV panels used in simulations
Parameter (STC) ERDM 85SM/5 BP 585 KC 120-1
Open Circuit Voltage (V ocST C ) 21.78 V 22.1 V 21.5 V
Short Circuit Current (IscST C ) 5.13 A 5.00 A 7.45 A
MPP Voltage (V mpp) 17.95 V 18.0 V 16.9 V
MPP Current (Impp) 4.8 A 4.72 A 7.10 A
Voc Temperature Constant (V oc ) -0.34 %/o K -0.080 V/o C -0.0842 V/o K
o
Isc Temperature Constant (Isc ) 0.020 %/ K 0.065 %/o C 0.04 %/o K
Number of Cells (N s) 36 36 36
Number of bypass diodes 1 2 2

1.6 Experimental platforms

Two experimental platforms were used to obtain the experimental results presented along this thesis.
Such platforms are presented in Fig. 1-7, in which the experimental system shown in Fig. 1-7(a)
was used at the Universidad del Valle and the test bench in Fig. 1-7(b) was used at the Universit`a
degli Studi di Salerno.

The experimental system 1 (Fig. 1-7(a)) is composed by four ERDM 85SM/5 panels, a connection
table and a BK-8514 electronic load. Such a load is connected by USB to a PC equipped with
Intel(R) Core(TM)2 Duo of 3.0 GHz and 4 GB of RAM. The four panels are connected at the input
of the connection table, which allows to interconnect the four PV panels in different configurations
depending on the specific experiment. The output of the connection table is connected to the input
of the electronic load. Such a load can be controlled from the PC to modify the voltage or current
of the array and it is equipped with voltage and current sensors in its input; those values can be
requested from the PC and used by the software that is controlling the system.

In some experiments it is necessary to simulate mismatching conditions over one or more panels.
In those cases, a plastic net with holes of different diameters is used to simulate a reduction in the
irradiance over a specific panel. Such nets are marked as artificial shades in Fig. 1-7(a).

The experimental system 2 was used at the Universit`a degli Studi di Salerno (Fig. 1-7(b)) is
composed by: one KC120-1 panel, one boost converter, a battery bank, one laser temperature sensor,
an interface board and a PC. The DC/DC converter has the PV panel connected to its input and
the battery bank to its output and it is controlled by a XC6SLX45 Xilinx Spartan-6 FPGA. The
DC/DC converter also has voltage and current sensors in its input, such information is acquired by
the interface board and sent to the computed to save the data.

The test bench of Fig. 1-7(b) also includes a box with different resistance values (0.022 , 0.168
, 0.517 , 1.013 , 2.028 and 3.000 ) that is connected between the positive terminal of the
panel and the input of the DC/DC converter to simulate increments in the series resistance of the
panel.

31
 DC Load Connection Artificial
Table Shade

(a) Experimental platform 1 (Universidad del Valle).

Battery Battery
bank bank switch KC 120

DC power FPGA Power

supply Converter

(b) Experimental platform 2 (Universit`

a degli Studi di Salerno).

Figure 1-7: Experimental platforms.

1.7 Scope of this thesis

Any PV system is equipped with a control device, which allows to extract as much energy as possible
from the PV modules in the array; this, to improve the overall profitability of the system and to
reduce the return of the investment time.
Mismatching conditions are among the main problems that affect the overall performance of the
PV systems since it reduces considerably the maximum power available at the output of the PV
array. Moreover, the mismatching conditions also produce stress in the cells and bypass diodes of
the mismatched modules, which may led to their degradation and/or damage in the medium and
long term.
In the first part of this thesis the mismatching problem is studied as well as different MPPT
techniques aimed at extracting the maximum power from PV arrays operating under mismatching
conditions. The results of such study are presented in Chapter 2 as an overview of the general

32
architectures used by different PV systems to mitigate the mismatching effects.

Most of such techniques considers a PV array or sub-array (i.e. one string or and small array)
connected to a power converter in order to track the corresponding GMPP. Such PV arrays can
have different electrical configurations (e.g. SP, TCT, BL, etc.) and different number of modules
as an static approach to mitigate the mismatching effects. Hence, it is important to have reliable
models of PV arrays (with different sizes and electrical configurations) operating under uniform and
mismatching conditions not only for the test and development of new MPPT techniques, but also
for a number of different applications based on simulation procedures: the design PV systems, the
estimation of the energy production of a PV array, the analysis of dynamic electrical reconfiguration
of the PV fields, the evaluation of the effects of a real PV system in power distribution grids [25],
among others.

Suitable models for the different PV arrays configurations should offer a tradeoff between accu-
racy and calculation burden, the possibility to easily change the dimensions of the PV arrays, and
the flexibility to be implemented in different programming languages (like C or Matlab). This to
have a reliable model that can be used in different simulation environments for the different appli-
cations mentioned in the last paragraph. However, in the overview process, it was identified a lack
PV array models offering those characteristics.

Such a problem is faced in this thesis and the results are presented in Chapter 3 and Chapter 4.
In Chapter 3 a procedure to model SP PV arrays operating under uniform and mismatching condi-
tions is proposed. Besides, some practical examples are introduced to illustrate the usefulness of the
models in the energy prediction of a PV array and the evaluation of MPPT techniques performance
under different mismatching conditions using circuit simulation software. On the other hand, in
Chapter 4 models for TCT and Bridge-Linked (BL) configurations are proposed. This Chapter also
includes simulation and experimental results to illustrate the usefulness of the models.

Once the models of the PV arrays of different configurations were proposed, it was decided to
apply one of those models in the analysis of the dynamical electrical reconfiguration of a SP array
and the results are presented in Chapter 5. Such an application were selected since the overview
presented in Chapter 2 shows an increased interest in the RMPPT techniques in the last years and
a lack of reconfiguration analysis based on the model of the entire array.

In Chapter 5 a method to determine the PV SP array configuration that provides the highest
GMPP is introduced. Such a procedure was designed to only require measurements of voltage and
current of each string, which avoids to perform experiments in each module. The parameters of each
module in the string are obtained from the analysis of the I-V characteristics of each string. Using
the estimated parameters, all feasible PV array configurations are evaluated to determine the array
configuration that provides the highest output power.

33
 Along the lifetime of a PV arrays the PV modules, cells and bypass diodes may present degrada-
tion and/or damages [3, 4] due to the mismatching conditions, the outdoor operation, the grounding
problems, quality problems, among others. Such degradation modifies the parameters the single-
diode model of the affected PV modules; therefore, the identification and quantification of degrada-
tion are very important for the monitoring and maintenance of PV fields as well as the update of
the models parameters in some applications like the energy production estimation. Moreover, such
a degradation affects considerably the economical evaluation of a PV systems, since the degraded
panels may not fulfill the guarantee offered by the panel manufacturer, which reduces the energy
produced the PV system.
Taking into account the importance of the degradation detection and quantification in the PV
modules, as well as a lack of model-based procedures for that purpose, the last part of the thesis
(Chapter 6) proposes two model based indicators to detect and quantify the degradation of a PV
module. Such indicators use the single-diode model to represent a PV module without degradation
and combine such information with the experimental measurements (MPP voltage and current,
short-circuit current and temperature) to estimate the module degradation.
Finally, the general conclusions along with some possible future work, in the field of modeling
and diagnostic of PV systems, are presented in Chapter 7.

34
Chapter 2

MPPT Architectures for PV

Systems in Mismatching
Conditions

The mismatching problem can be faced from different aspects of a PV system because it is present
along all the system life. One important approach is to mitigate the mismatching by using MPPT
techniques aimed to extract the maximum power from a PV array operating under mismatching
conditions. That is why in the last years there is an increased interest in the research community
of proposing new MPPT control strategies, algorithms and hardware architectures. Considering the
importance of the MPPT techniques as an approach to solve the mismatching problems, this thesis
started with the study of those techniques to establish the state of the art, which is introduced this
Chapter.

The general architectures used by sixty-one different MPPT techniques is presented, by discussing
their main advantages and disadvantages, this to provide to the reader a comprehensive view of both
the control strategies and the architectures for extracting the maximum power from a mismatched
PV field. Moreover, the widely adopted techniques for each hardware structure are presented in a
structured and compact way, thus providing to the reader some guidelines regarding to the techniques
operation principle and hardware requirements.

The content of this Chapter has been accepted in the IET Power Electronics Journal in a paper
entitled Maximum Power Point Tracking Architectures for Photovoltaic Systems in Mismatching
Conditions: A Review [26].

35
2.1 Introduction
The manufacturing tolerances (e.g. in [14] the datasheet reports 10 %), the shadows produced by
surrounding objects (trees, antennas, buildings, etc.), dirt, and/or passing clouds are some of the
most important causes of the performance degradation of a PV installation [27]. Such mismatching
conditions are cumulative and very penalizing, especially in building integrated PV arrays. There-
fore, it is important to take measurements to mitigate those effects in order to avoid: excessive
power and energy losses in a and the enlarge of the return of investment time of the system.
In general, the effects of mismatching on the power production of the PV generator can be
reduced by performing one or some of the following three actions:

1. Modifying the MPPT algorithm by controlling the converter(s) used in the system architecture
(e.g. [28]).

2. Modifying the power system architecture, i.e. using multiple converters instead of a single one
(e.g. [18]), modifying the interconnection among such converters (e.g. [13]) and/or changing
the converters type itself (Buck, Boost, Cuk, etc.) (e.g. [29]).

3. Changing the electrical connection of the panels into the PV field, i.e. exchanging dynamically
the position of the panel in the array (e.g. [22]) or the type of connection from Series-Parallel
to Total Cross-Tied, Bridge-Linked or other.

Theoretically, the optimal results can be obtained by performing the three actions simultaneously;
however, such an approach would result in a PV system that requires large amount of hardware and
complex control algorithms. Hence, different MPPT strategies implement one of the three actions,
other ones try to find trades offs between them, this resulting in a large number of solutions with
different characteristics, advantages and disadvantages.
Considering the large variety of MPPT techniques, it is not possible to compare them in terms
of energy production only. In fact, it depends on the specific mismatching conditions, the hardware
requirements, the implementation and maintenance complexity, the implementation costs, among
others; therefore, a general rule for selecting the best strategy does not exist.
For such a reason, this Chapter is mainly aimed to classify and organize different approaches for
performing the MPPT in PV systems operating in mismatching conditions to support the selection
of the MPPT solution for a particular application. The MPPT techniques are divided into the
following three classes according to the hardware structure of the system:

1. Centralized MPPTs (CMPPT): use one power stage to track the maximum power available in
the PV array.

2. Distributed MPPTs (DMPPT): divide the PV array into sub-arrays and track the maximum
power available in each sub-array.

36
 3. Reconfiguration MPPTs (RMPPT): modify the connections between the modules or panels of
the PV array to increase the maximum power available.

This Chapter also highlights the main characteristics of the selected techniques by describing
their operation principles and hardware requirements. In this way, an overview on the state-of-the-
art in this field is given and some issues that could be the subject of further research activities are
pointed out.

In literature some comparisons among MPPT techniques for PV arrays in uniform conditions
(e.g. [30, 20, 31, 32]), as well as general comparisons between centralized and distributed MPPT
strategies ([33, 34]) and overviews of the MPPT techniques for PV arrays in mismatching condition
have been found ([35, 36]). In some cases a qualitative description of the reviewed algorithms and of
the circuit topologies is given separately, but a comprehensive discussion on the interactions between
the architectures and the algorithms for the MPPT in mismatching conditions has not been given
yet. Moreover, some of the papers mentioned above do not discuss the hardware requirements and
they cover only part of the techniques presented in a wide literature.

Instead, the aim of this Chapter is to provide a comprehensive survey of different approaches
proposed in literature for extracting the maximum power from a mismatched PV field by classifying
some widely adopted solutions according to their structure. Moreover, the operation principle and
the main hardware requirements of sixty-one MPPT solutions are presented without proposing
comparisons among them. In fact, as is evidenced in the Chapter, there is not a unique MPPT
strategy that always provides the best results for any operating condition of the PV generator.
Hence, the designer has to select the most appropriate solution on the basis of the PV installation
objective, cost, energy production and complexity of the practical implementation.

The Chapter is organized as follows. Section 2.2 presents a brief description of the mismatching
problems and some static solutions acting on the PV configuration. Section 2.3 introduces a basic
classification of the power architectures and MPPT strategies applied in PV arrays operating in
mismatching condition. In particular, they have been divided in three classes identified with the
acronyms CMPPT, DMPPT and RMPPT and the characteristics of those approaches have been
highlighted. Sections 2.4, 2.5 and 2.6 are devoted to describe and comment the classified CMPPT,
DMPPT and RMPPT techniques, respectively. Each section includes a table summarizing the main
architectural characteristics as well as the hardware resources required by each reviewed method.
Conclusions close the Chapter.

37
2.2 The static approach for mitigating the mismatching prob-
lems

There are different methods to interconnect the PV modules to form an array, as mentioned in
Chapter 1. Some of the widely adopted PV array configurations are SP, TCT and BL. The SP
configuration is illustrated in Fig. 2-1 with the connection represented by black lines. The TCT
and BL configuration have additional interconnections between the strings of an SP array. The
additional connections of a TCT are represented in red and blue dashed-lines in Fig. 2-1, while the
additional connections used by BL arrays are represented by the blue lines only.

In a PV system, independently of the array configuration, it is quite frequent that the PV pan-
els in the string have mismatched parameters [37] as well as different irradiance and temperature
conditions, producing P-V curves with irregular shapes. However, the effects of the mismatching
conditions in the P-V characteristics depend on the array configuration because different intercon-
nections of the modules produce different distributions of the array current in the PV field.

To illustrate such a concept, two arrays of nine ERDM 85SM/5 PV modules are considered. One
of them connected in SP configuration and the other one connected in TCT configuration according
to Fig. 2-1. Fig. 2-2(a) presents equivalent P-V curves of both arrays. In this example, the following
assumptions holds: the ambient irradiance and temperature are 1 kW/m2 and 25 o C, respectively,
the irradiance over each module is uniform and each PV module can work in one of three different
shadowing conditions: 50 % (i.e. 0.5 kW/m2 ), 25 % (i.e. 0.75 kW/m2 ) and 0 % (i.e. 1 kW/m2 ), it
represented by a big, small or no cloud in Fig. 2-1. The P-V curves of a module in each shadowing
conditions are presented in Fig. 2-2(b).

Under uniform conditions, all the modules have the same P-V curve marked as 0 % shadowed
(i.e. 1 kW/m2 ) in Fig. 2-2(b). The equivalent P-V curve of the array in both, SP and TCT
configurations, corresponds to the sum of the P-V curves of each module (see Fig. 2-2(a)).

In mismatching conditions the P-V curves of SP and TCT arrays are different since the SP
configuration exhibits three maximums, two local (Local MPP in SP configuration - LM P PSP ) at
18 V and 36 V , approximately, and one global (Global MPP in SP configuration - GM P PSP ) at
around 57 V ; while TCT configuration has only two maximums, one local (Local MPP in TCT
configuration - LM P PT CT ) around 36 V and one global (Global MPP in TCT configuration -
GM P PT CT ) around 56 V . For the mismatching conditions selected in this example, it is clear
the benefit of the TCT over SP configuration, since the maximum available power in TCT array is
higher (P GM P PT CT > P GM P PSP ) and the number of local maximums is lower. In general the
power improvement obtained with TCT configuration (T CT ) can be expressed as (2.1), which in
this case is 30.53 %

38
 BD BD BD
Shadowing

Shadowing

PV Array:
BD BD BD SP
Shadowing Shadowing
TCT
BL

BD BD BD

Shadowing
Shadowing

Figure 2-1: PV field configurations. SP: black lines. TCT: black lines + red lines + blue lines, BL:
black lines + red (blue) lines. The clouds dimensions identify the percentage of the shade: Big cloud
means PV module shaded at 50%. Small cloud means PV module shaded at 25%.

SP mismatch. SMPP mismatching Uniform

conditions conditions
TCT mismatch.
800
SP uniform
100
TCT uniform Irradiance: 1 kW/m2
600 0 % shaded
Array power (W)

80
Module power (W)

25 % shaded
LMPP
SP 50 % shaded
400 60

40
200 GMPP
TCT
LMPPTCT
GMPPSP 20
0
0 10 20 30 40 50 60 70 0
0 5 10 15 20
Array voltage (V) Module voltage (V)

(a) P-V curve of the 33 array in SP and TCT configura- (b) P-V curve of the unshaded (blue), mid shaded (green)
tion under uniform (SP: blue, TCT: red) and mismatching and large shaded (ref) modules.
(SP: violet, TCT: green) conditions.

Figure 2-2: Modules and array P-V curves.


P GM P PT CT P GM P PSP 
T CT = (2.1)
P GM P PSP 
%

533.5W 408.7W 
=  = 30.53 %
408.7W %

However, it is worth noting that only one pattern of the solar irradiance has been considered,
thus the improvement introduced by the TCT connection does not ensure that this configuration
is the best one in other mismatching conditions. Moreover, TCT requires addition cabling (with
respect to SP) to its implementation, which represents an increase in the installation cost; hence,

39
the designer of the PV system should evaluate the specific application to determine the viability
of this array configuration. In any case, TCT and BL solutions do not remove the multiple power
peaks; therefore, they must be operated with a suitable MPPT algorithm able to track the GMPP.

Traditional MPPT techniques are able to track only one MPP in a P-V curve [20]. Therefore,
if such techniques are used in a PV array in mismatching conditions, they might be trapped in an
LMPP, this causing a drop in the energy extracted from the PV array. For example, if a traditional
MPPT is trapped in the lower LM P PSP of the Fig. 2-2(a), only the 61.12 % of the power available
at the GM P PSP would be extracted.

Moreover, for a PV array with N modules in a given mismatched condition, the maximum power
extracted from the PV array is lower than (or at most equal to) the sum of the maximum powers
(PMPP) that can be delivered by each module (named SMPP power) as illustrated in Fig. 2-2(a).

N
X
SM P P = P M P Pi (2.2)
i=1
SM P P P GM P PT CT

SM P P P GM P PSP

Such conditions are generated by the electrical connections used in any PV array configuration,
which introduce constraints that can activate some of the by-pass diodes. Thus, the corresponding
PV modules are short-circuited reducing the maximum power at the PV array terminals. This means
that the maximum power available in any mismatched array (e.g. P GM P PSP and P GM P PT CT )
does not correspond to the sum of the maximum power available in the modules (SM P P ), even if
the MPPT technique applied to the PV array is able to track the GMPP. Therefore, in a PV array
operating in mismatching conditions, part of the energy transformed form light to electricity will be
lost.

2.3 Classification of MPPT architectures

The following classification takes into account the solutions able to perform the MPPT in the PV
array operating in mismatched conditions. The techniques are grouped on the basis of the hardware
architecture used to extract the maximum power from the PV array. According to the number of
the power converters used in the solutions and to the way they are connected with the PV modules,
the techniques can be classified into the three groups described in the following sub-sections.

40
 PV Field

Output Sensors
Input Sensors
Power
Load
Stage

iPV , vPV io , v o

Environmental Manipulated
Sensors variable

Tamb , G
Control
Algorithm

Figure 2-3: General structure of CMPPT architecture.

2.3.1 Centralized MPPT architectures (CMPPT)

In this type of architecture the PV array, typically exhibiting a SP or TCT configuration, is connected
to one power converter, which performs the tracking of the GMPP by executing an algorithm that
modifies the voltage or current of the PV array. The CMPPT architectures are commonly based on
the scheme shown in Fig. 2-3, where current and voltage sensors (placed at the input and/or at the
output of the power converter) as well as environmental sensors, are used to calculate the PV power.
This information is elaborated by the MPPT algorithm and used to modify the operation point of
the power stage. Depending on the specific power stage implementation, the MPPT technique can
act on the duty cycle of the converter, or on the reference of a controller devoted to regulate the PV
array voltage or current.

MPPT techniques operating on the voltage reference give some advantages because they reduce
both, the settling time of each step of the MPPT algorithm, and the PV voltage oscillations induced
by the inverters in the grid-connected PV systems [38]. On the other side, an MPPT method
operating on the current reference, although it improves the dynamic response, it is more sensible
to the irradiance changes and it may produce large power drops when fast irradiance changes occurs
[39]. Fig. 2-4 illustrates a 30 % reduction in the irradiance over a 2 2 SP array with ERDM
85SM/5 panels. In such a case, if the system is operating in Impp1 and the irradiance changes
rapidly between two consecutive action instants of the perturbative MPPT, in the time between the
irradiance change and the next action instant of the MPPT algorithm, the PV current loop has the
initial reference value (Impp1 ). Because of the fact that Isc2 is lower than Impp1 , the current control
loop will saturate the duty-cycle of the converter by trying to follow the reference. Such a saturation
drives the system to operate, temporarily, to a current value close to Isc2 , which produces an output
power close to zero (power drop close to 100 %). Such a drawback can be avoided by combining the

41
current loop with an external voltage loop controller as in [40] and [41].

60

Voltage (V)
40
Isc2 Isc1
20

0
0 2 4 6 Impp2 8 Impp1 10

400

Power (W)
200

0 Impp1 10
0 2 4 6 Impp2 8
Current (A)

Figure 2-4: Characteristic curves for a PV array of 2 2 operating in uniform conditions with 1
kW/m2 (continuous line) and 0.7 kW/m2 (dashed line).

In general, CMPPT techniques have three main advantages:

They use a single converter, thus the switching losses and the cost of implementation are lower
compared with the other MPPT architectures.

The harvested energy can be higher than the one ensured by other architectures when there
are shallow mismatching conditions [33, 34].

Few sensors are used since the PV array is considered as a single PV generator.

The main drawback of this architecture concerns the fact that it is not able to track the SMPP;
moreover, to assure the convergence towards the GMPP, the MPPT algorithm might be much more
complex than well consolidated MPPT techniques like P&O and IC.

2.3.2 Distributed MPPT architectures (DMPPT)

In this type of architectures, the PV array is divided into sub-sections (string, panel, module, etc.)
and each sub-section is connected to a power converter, which is devoted to perform the MPPT on
it. By assuming that each sub-section operates in uniform conditions, then a simple MPPT tracking
algorithm (e.g P&O or IC) can be used on each power converter.
Fig. 2-5(a) shows the basic unit of a DMPPT architecture (DMPPT-U), which is similar to the
CMPPT, except for the fact that it works on smaller sub-sections and not on the whole PV array.
Moreover, such units can be considered as totally independent or they can interact with other units.
Fig. 2-5(b) shows the typical connection of multiple DMPPT-Us. They are usually connected in
series and parallel according to the voltage and power constraints. Such a configuration reproduces

42
the SP connection of the PV modules (as in Fig. 2-1), but connecting the converters outputs of each
DMPPT-U.

PV
sub-section D-MPPT Unit

Output Sensors
Input Sensors
series parallel connection
Dc/Dc
of D-MPPT units

Manipulated
TPV , iPV , vPV variable io , v o

Local
Control

from central unit to central unit

(a) Basic DMPPT unit.

1 1,m

D-MPPT D-MPPT
Unit
Unit 1,1
1 Unit
Unit1,m
1

Bus Sensors

DC Load/
k,1 k,m Inverter
Bus

D-MPPT D-MPPT
Unit
Unit k,1
k Unit
Unitk,m
k

data from local units data sent to local

Environmental units
Sensors

amb
(sensors values,
converters status) Control
Unit

Centralized control

(b) Basic DMPPT architecture.

Figure 2-5: General structure of DMPPT architectures.

The DMPPT architectures can, theoretically, track the SMPP since each sub-section should
operate at its MPP, but some limitations arise in this configuration. The first limitation is related
to the voltage constraint of the DMPPT unit, because the output voltage of each unit is:

43
 Pj
vo,j = vbus PN (2.3)
i=1 Pi

where vbus is the voltage of the DC-bus (see Fig. 2-5(b)), Pj is the power delivered by the j-th
PN
DMPPT unit, and i=1 Pi is the total power delivered by all the N DMPPT units connected in
series.

Such a voltage constraint is specially critical in applications where the bus voltage is regulated
to a fixed value (like in grid-connected systems). In such a case (2.3) implies that, in presence of
deep mismatched conditions, the output voltages of the DMPPT-Us are not well distributed and
the converter output voltage rating can be violated. Thus, to protect the sub-sections delivering
the higher power, they are forced to work far from the MPP, so that the SMPP cannot be tracked
[38, 42].

For example, if a string of two DMPPT-Us, where each DMPPT-U is composed by one BP 585
panel and one Boost converter, is connected to a regulated DC bus of 90 V , the output voltages of
the DMPPT-Us (vo1 and vo2 ) under uniform conditions are the same (45 V for an irradiance of 1
kW/m2 ), as shown in Ramos-Paja et al. [43]. But, if a shadowing of 50 % appears on the panel
connected to the second DMPPT-U, then vo1 and vo2 are different (vo1 = 60 V and vo2 = 30 V ).
In such a case, the output capacitor and the MOSFET of the Boost converter of the first DMPP-U
may be damaged if they have been selected taking into account the operation in regular conditions
(e.g. 50 V ) [43].

Instead, if each DMPPT-U is directly connected to the DC bus voltage, the unbalancing due to
the series connection is avoided. This configuration requires that the output voltage range of each
DMPPT-U matches with the voltage levels of the bus, but such a voltage depends on the application.
For grid connected applications the bus voltage is high, thus the converter of each DMPPT-U must
be designed for having a high boosting factor.

High boosting factor DC/DC converters are also frequently used in PV micro-inverter solutions,
where each panel is straightforwardly connected to the grid by means of a cascade of a DC/DC
converter and a DC/AC stage. In this case, the DMPPT architecture is easily scalable since each
PV module is a plug-in system that can be integrated in whichever electrical power plant. However,
those systems are complex, not only for the high-voltage gain needed, but also because each DMPPT-
U needs to manage the grid connection (synchronization, protections, etc.), which leads to an increase
of the implementation costs.

Another limitation of the DMPPT architecture concerns the switching losses introduced by the
power stage of each DMPPT-U, which not always operates at its own maximum efficiency since the
operating points vary in a wide range of voltage and currents along the day. This aspect is especially
important in those cases when the mismatch occurs for a limited time period (e.g a shadow that

44
covers part of the PV array only for a few hours). In fact, if the converters are permanently connected
even if they are not necessary, the energy they dissipate during this period may counterbalance the
increased energy harvested (over the CMPPT architectures) when the DMPPT architecture operates
in the mismatched condition. This problem can be solved by disconnecting the DMPPT units (pass-
through configuration) when they are not needed [44, 45].
MPPT techniques that combine some benefit of the CMPPT and DMPPT concepts are also
available. Such solutions have been classified as CU-DMPPT when a central unit (CU) works as a
supervisor system. Here, each PV sub-section is equipped with a switching converter but the MPPT
algorithm runs on the CU. In that way, the information coming from each module can be used for a
global optimization of the power produced. Moreover, such a solution is suitable for implementing
monitoring and diagnostic functionality at PV sub-section level.

2.3.3 Reconfiguration-MPPT architectures (RMPPT)

In this architecture each PV module can be interconnected dynamically with the other ones by means
of multiple switches connected to the positive and negative terminals. Fig. 2-6 shows a conceptual
scheme of such an approach. The connections among the modules are performed by means of a
matrix relay box controlled by a PV-Reconfiguration Algorithm (PV-RA), which selects the best
electrical configuration for the PV modules. It is worth to note that, as explained in previous
sections, a specific PV array connection which is able to eliminate the effects of any mismatching
conditions does not exist, thus the solution must be identified in real time by exploring the space of
all the possible solutions.

PV module
PV Sensors

PV Switch
Output Sensors
Input Sensors
...

Matrix
Relay Dc/Ac Load
PV Sensors

ibus , vbus Manipulated

variable io , v o
Climate Tpv , iPV , vPV
Sensors PV
MPPT - Control
Re-conguration
Algorithm
Algorithm
Tamb , G

Figure 2-6: General structure of the RMPPT architecture.

The PV-RA periodically measures the voltage and current of each PV module, as well as the
environmental conditions; and on the basis of the results of such measurements the algorithm de-

45
termines the new configuration. To set the new configuration the PV array is disconnected from
the load by the PV switch (see Fig. 2-6) to guarantee the safety of the system by avoiding that
the matrix relay commutates when the current is still flowing inside it. The PV array configuration
obtained by the PV-RA can exhibit multiple LMPPs, thus the RMPPT requires a power converter
at the output of the matrix relay. The operating principle of such a converter is the same of the
CMPPT architecture and it can be equipped with an MPPT technique for tracking the GMPP.
The main advantage of such an architecture concerns the fact that the dynamical reconfigura-
tion can reduce the distance between the GMPP and the SMPP by producing higher power than
CMPPT architectures. An additional benefit is the increased system safety because, for example,
this architecture allows to disconnect the PV panel if a failure like an electric arc is detected, thus
avoiding that such a failure may produce a permanent damage in the panel or even trigger a fire
[46, 47]. Moreover, the installation is simplified because the best PV array configuration is selected
automatically and, as in the DMPPT architectures, monitoring and diagnostic functionalities can
be performed at the module level.
To obtain a reasonable cost matrix relay it should be constructed with elecro-mechanical switches,
which increases the number of elements of the system and introduce mobile parts in the hardware.
Such additional devices increase the probability failures and the maintenance costs of the system.
Other drawbacks of the RMPPT approach are the increase of the cabling connection and the
fact that the PV array must be periodically disconnected to perform the re-configuration. Moreover,
the re-configuration produces energy losses for at least two reasons. The first one is that during the
disconnection of the PV array it does not produce energy and the duration of such phase depends
significantly on the PV-RA capability for selecting the new connection. The second one is because
the relay performing the reconfiguration requires a power supply.

2.4 Literature survey of CMPPT-based solutions

The aim of this Chapter is to present a classification of the widely adopted solutions, proposed in
the scientific literature, for mitigating the mismatched effects. In this way, the reader can have a
personal opinion on which solution should be used in a specific case. The approaches have been
organized on the basis of the hardware resources required to implement each CMPPT technique.
The presentation of the different solutions follows the order given in Table 2.1, whose columns
were defined to represent the main parts of the CMPPT architecture presented in Fig. 2-3. The
techniques were divided into the following three sub-classes specified in the first column of Table
2.1:

A Techniques in which the GMPP is tracked by sampling the P-V curves and by using hill climbing
algorithms.

46
B The GMPP tracking is performed by using computational intelligence approaches.

C The GMPP is tracked with approaches not falling into the categories listed above.

Columns 2-4 show the variables sensed for evaluating the MPP, and specify the power stage
topology and the load, respectively. Columns 5-7 give more details concerning the MPPT algorithm:
in particular, the variable on which the algorithm acts, the tracking method and the hardware used
for implementing the algorithm have been also reported. Column 8 specifies if in the literature the
parametric design procedure is explained or not. The names shown in those columns are directly
related to the variables and blocks introduced in Fig. 2-3.
Table 2.1 shows that about 44 % of the CMPPT techniques belongs to sub-class A and 38 %
belongs to sub-class B; therefore, it can be deduced that there is not a clear dominance of a specific
sub-class. Moreover, all the CMPPT techniques use the measurement of the array voltage and more
than the 90 % of the techniques use the array current. Only the technique presented in [38] uses
irradiance sensors to feed an Artificial Neural Network that estimates the location of the GMPP.
To track an LMPP or the GMPP, 50 % of the CMPPT techniques manipulates the PV array
voltage and 40 % manipulates the duty cycle of the converter, while the converters more used are
the Boost (44 %) and the Buck (31 %). The percentages presented before show that there is not a
clear trend in the selection of the manipulated variable and of the power converter used in CMPPT
techniques.
Although there are different CMPPT methods to track the GMPP of the PV array, most of them
can be implemented in digital signal processors (30 %) or microcontrollers (25 %): the selection of
one or another depends on the complexity of the algorithm.

2.4.1 Sampling P-V curves + hill climbing

Basically, such techniques perform a partial or a total scan of the P-V curve to find some or all
LMPPs and set the operation point at the GMPP.
The MPPT algorithms proposed by Lei et al. [48] and Alajmi et al. [49] track all LMPPs in
the P-V curve and save the values of the power and the manipulated variable of each LMPP. Then,
the algorithms set the operation point of the array in the GMPP. The difference between those two
MPPT algorithms is that Alajmi et al. [49] propose to scan the entire P-V curve by manipulating
the duty cycle of the DC/DC converter to track the GMPP with a fuzzy control system; while Lei et
al. [48] propose to only scan parts of the P-V curve because the algorithm sets the starting voltage
reference close to the next LMPP to reduce the LMPP tracking time of the Extremum Seeking
Control (ESC) technique.
The techniques presented in [28, 50, 51, 52] and [53] propose methods to find the GMPP of the
array without scanning all the LMPPs in the P-V curve. They propose to take few samples of the

47
 Table 2.1: CMPPT techniques
CMPPT Algorithm
Sub- Measu- Power Load Manipu- Tracking Imple- Design Ref.
Class rements stage lated method mented pro-
Variable in cess
iP V , v P V N.S. N.S. vP V ESC N.S. No [48]
iP V , v P V Boost Battery D Fuzzy uC Yes [49]
iP V , v P V Boost R vP V P&O uC Yes [28]
A iP V , v P V 3P Inverter + vP V P&O DSP Yes [50]
DC/AC grid
iP V , v P V N.S. DC load iP V P&O PC Yes [51]
iP V , v P V Buck Battery D P&O uC Yes [52]
iP V , v P V Boost Inverter + vP V IC DSP No [53]
grid
iP V , v P V Buck R vP V Chaos RT-LAB No [54]
iP V , v P V Boost R vP V Fibonacci uC Yes [55]
iP V , v P V Boost DC load D Fibonacci DSP Yes [56]
B
iP V , v P V Boost N.S. D PSO DSC Yes [57]
iP V , v P V Buck- R D PSO DSP [58], No [58,
Boost N.S. [59] 59]
iP V , Buck N.S. vP V ANN + PC No [60]
vP V , G1 Fuzzy
vP V Boost Battery+R vP V kVoc DSP Yes [61]
C iP V , Buck R D Optimum N.S. Yes [62]
v P V , io , control
vo
iP V , v P V Buck N.S. D P&O N.S. Yes [63]
vP V : PV array voltage, iP V : PV array current, vo : converter output voltage, io : converter
output current, G: irradiance, D: converter duty cycle, N.S.: Not Specified, 3P: three phase, R:
resistive, uC: Micro-controller, DSP: Digital Signal Processor, DSC: Digital Signal Controller,
PC: Personal Computer.
1
It requires one irradiance sensor for each group of 4 modules

P-V curve in order to determine the location of the GMPP or a point close to it. Afterward, they
use a P&O [28, 50, 51, 52] or IC [53] technique to keep the array operating point close to the GMPP.
Unlike [50, 51, 52], Patel and Agarwal [28] and Ji et al. [53] propose a method to detect a significant
change in the irradiance to run the GMPP tracking algorithm. This procedure avoids unnecessary
samples of the P-V curve while the conditions are within a given rage. In addition, Koutroulis and
Blaabjerg [52] propose a new method to generate the PWM signal for the DC/DC converter by
comparing a reference value with the inductor current. With such a method, a control strategy to
manage the DC/DC converter as a constant input-power load, controlled by the reference value, is
proposed. On the other hand, an equation to calculate a voltage close to the GMPP is proposed in
[53]. The equation is based on the open-circuit voltage and on the short-circuit current of the array:
unfortunately, it may fail for arrays with multiple LMPPs because the algorithm is designed for a
P-V curve with three LMPPs only.

48
2.4.2 Computational intelligence

Different kinds of global optimization techniques for tracking the GMPP of a PV array by making
multiple samples of the P-V curves have been used.
The authors in [54, 55] and [56] use methods that measure the power in a wide range of the
P-V curve: according to the results of such an analysis, the search range of the GMPP is tightened.
Zhou et al. [54], Ramaprabha et al. [55], and Ahmed and Miyatake [56] use different procedures
to calculate the voltages where the system has to measure the power. Zhou et al. [54] uses a
double carrier Chaos algorithm, while Ramaprabha et al. [55] and Ahmed and Miyatake [56] use the
Fibonacci series. Both, Zhou et al. [54] and Ahmed and Miyatake [56], propose different methods
to detect a change in the irradiance conditions in order to avoid unnecessary tracking of the GMPP.
The MPPT algorithms proposed in [57, 58] and [59] are based on the Particle Swarm Optimization
(PSO) method, which is a meta-heuristic global optimization technique. The system has multiple
agents sampling the P-V curve and, depending on the power obtained from the different agents, the
algorithm determines the next position of the agents themselves. The three papers propose methods
to detect significant changes in the irradiance conditions to start the GMPP procedure. Liu et al.
[57] propose a different method to adjust the MPPT depending on the size of the PV array, and
it also proposes an equation to calculate the values of the parameters of the equations used by the
PSO. Meanwhile Ishaque et al. [59] propose equations to find the minimum and maximum duty
cycle imposed to the DC/DC.
In [60] the authors propose an Artificial Neural Network (ANN) for predicting the GMPP voltage
(Vgmpp ) from the irradiance and temperature measurements of the PV array. The ANN calculates
the Vgmpp and delivers it as a reference to a Fuzzy controller for ensuring the PV array reaches the
desired value.

2.4.3 Other CMPPT techniques

There are other CMPPT techniques that use different concepts to track the GMPP or to reduce the
effects of the mismatching conditions over the PV array.
In [61] the authors compare the energy extracted by one array with more rows than columns
and one array with more columns than rows, both connected to a DC/DC converter operating at
a fixed voltage, that is the nominal value of the MPP voltage (Vmpp ). Such a method is defined as
fractional open-circuit voltage (kVoc) as shown in [20]. The experimental results presented in [61]
show that the energy extracted with the parallel connection is the highest in all the experiments.
A feedback controller for tracking the MPP in a PV system operating in uniform conditions is
proposed in [62], it assuming the PV array is a time-variant voltage source connected to a Buck
converter. The authors show that the controller works for a small array in uniform and shallow

49
mismatching conditions: the control design is developed by considering the PV generator in uniform
conditions, so that the algorithm may fail for medium and large arrays operating in mismatching
conditions.
Karatepe and Hiyama [63] use the TCT configuration to mitigate the mismatching effect over
the array and the GMPP is tracked with a DC/DC converter controlled by a traditional P&O. The
authors compare the performance of the proposed system with the same array connected in SP
configuration to show the advantage of TCT under a set of mismatching conditions.

2.5 Literature survey of DMPPT-based solutions

The main objective of this section is to collect the basic information about the hardware resources
and the operating principle of the DMPPT techniques, and to group the techniques based on such
information. Taking into account which part of the PV array is connected to each DMPPT-U and
the way in which the DMPPT-Us are connected each other, the techniques are classified into four
sub-classes:

A Techniques that interconnect the DMPPT-Us following the SP configuration (Fig. 2-5). Some
commercial products based on this solution are now available in the market [64, 65].

B Techniques that connect the power stage of the DMPPT-U in parallel with the modules or panels
connected in a string. The two more common structures of those techniques are presented in Fig.
2-7.

C Techniques that interconnect the DMPPT-Us in parallel to a bus. Only one commercial product
based on this solution [66] has been found by the author.

D Techniques where the PV sub-section connected to each DMPPT-U is a string or small array.

The main advantages and disadvantages of the four sub-classes are presented in Table 2.2. Those
ones should be considered in the selection of a structure for a given application, since, depending
on the application, an advantage may become into a disadvantage. For example, in a stand-alone
application, where the diagnostic procedures are important for maintenance programming, the re-
duced number of sensors of sub-classes B and D turns into a disadvantage because the information
about each module or panel becomes unavailable; so that, the diagnostic quality worsens.
The information about the different DMPPT-Us is summarized in Tables 2.3 and 2.4. The
columns in such Tables are similar to those ones given in Table 2.1 and they have been defined to
include the main information presented in Fig. 2-5(a) and Fig. 2-7. The variables and abbreviations
are defined as follows: vP V and iP V are the DMPPT-U input voltage and current, respectively;
vo and io are the DMPPT-U output voltage and current, respectively; vbus and ibus are the DC

50
 Table 2.2: Main advantages and disadvantages of DMPPT sub-classes
DMPPT General advantages
Sub-class
A It is possible to extract SMPP
with certain limitations.
Simpler implementa-
tion of modules monitor-
ing/diagnostic/security proce-
dures.
B Low switching losses under low
mismatching conditions because
only part of the power passes
through the converters.
Reduced number of sensors.
Reduced number of processors.
General disadvantages
Significant switching losses under low mis-
matching conditions because all the power
passes through the power converters.
High number of sensors.
One processor is required in each DMPPT-
U.
High output voltage in some DMPPT-Us un-
der deep mismatching conditions.
It is not possible to extract SMPP.
It is not clear which is the best method to
control the DMPPT-Us.

C It is possible to extract SMPP. Significant switching losses under low mis-

Simple addition and removal
of individual DMPPT-Us.
Simpler implementa-
tion of modules monitor-
ing/diagnostic/security proce-
dures.
D Reduced number of sensors.
Reduced number of converters.
Reduced number of processors.
matching conditions because all the power
passes through the power converters.
High number of sensors.
Complex converters due to high boosting
factors are required.
One processor is required in each DMPPT-
U.
It is not possible to extract SMPP.
Significant switching losses under low mis-
matching conditions because all the power
passes through the power converters.
MPPT techniques may be trapped in an
LMPP because each converter is connected to
a group of modules/panels.

bus voltage and current, respectively; D is the duty cycle of the DMPPT-U, TP V and G are the
temperature and irradiance of the PV sub-section, respectively; Mod., Pan. and Str. represent a PV
module, panel and string, respectively; the word Array represent a small part of the total array, uC
means Micro-computer, uP means Micro-processor, FPGA is the acronym of Field Programmable
Gate Array and R represent a resistive load.

The differences between Table 2.1 and Tables 2.3 and 2.4 are that the last ones include one
additional column for specifying if the DMPPT technique uses a central unit or not (column 9).
Instead, Tables 2.3 and 2.4 do not include the information about the design procedure of the tracking
methods since, in general, those methods are simple.

The main information of the central units (see Fig. 2-5(b)) is presented in Table 2.5, where
the column 1 shows the measurements performed by the central units; columns 2 and 3 present the
information sent by the DMPPT-Us and the communication links used to transmit such information,

51
respectively; column 4 presents the objectives of the central units and column 5 shows if the design
procedure has been provided. Finally, column 6 gives the reference to each technique.
In DMPPT techniques the structure presented in Fig. 2-5 (sub-class A) is dominant, since the
41 % of the analyzed DMPPT techniques belongs to such a sub-class, while the 23 % belongs to
sub-class B and the 23 % to class C. Most of the DMPPT techniques perform measurements of
the voltage (64 %) and current (64 %) in each PV sub-section, while only [82] requires irradiance
measurements; this because most of the algorithms in each DMPPT-U are relatively simple and
require few information.
In the analyzed documents, most of the DMPPT-Us are implemented with DC/DC converters
based on Buck and Boost topologies. Almost 40 % of the studied DMPPT techniques use Boost or
modified Boost converters, while around 13 % uses converters that can work as a Buck, Boost or
bridge converters. Other 10 % of the DMPPT-Us are more flexible and require different conversion
topologies. Moreover, some custom converters are proposed to obtain high-voltage gains with the
purpose of connecting a module or panel directly to the grid.
The tracking algorithms running in each DMPPT-U are usually simple. One third of the DMPPT
techniques use the P&O or its variants; however, in a significant part of the analyzed papers (20
%) the MPPT algorithm adopted is not specified. In general, the MPP tracking algorithms in each
DMPPT-U manipulate the PV sub-section voltage (36 %) or the duty cycle of the converter (28
%); nonetheless, there are some algorithms that do not specify the manipulated variable (15 %) or
operate over another variable of the converter (e.g. the input current: 10 %, output voltage: 7 %
and output current: 2 %). The hardware used to implement the tracking algorithms is reported in
a small number of the cases (28 %); nonetheless, micro-processing devices (uC, DSP, etc.) are the
most common. Besides, large part of the techniques is intended to grid connected applications (51
%) or the load is not specified (30 %).
Table 2.5 shows that less than half of the DMPPT techniques (43 %) implement Central Units.
Such Central Units usually measure the voltage (47 %) and current (53 %) of the bus that collects
the power of the entire PV system. Other techniques collect data from the DMPPT-Us (47 %);
however, a communication channel is not required or its adoption is not specified in most of the
cases. The information measured on the bus and collected from the DMPPT-Us is used for different
purposes: for improving the MPPT algorithm or implementing a centralized MPPT algorithm for
all the DMPPT-Us (47 %), maintenance and diagnostic procedures (30 %), and to avoid damages
in the DMPPT-Us by performing protection routines (23 %).

2.5.1 DMPPT-Us connected in SP configuration

Such techniques connect N DMPPT-Us to form a string and M strings in parallel depending on
the voltage and power requirements, respectively. Generally, the part of the PV array connected to

52
 P V1 P V1

RECC

from DC/DC
central unit
from
central
unit
RECC Load/ Load/
from
Shadowing Inverter Shadowing Inverter
central
unit

from
P Vk central unit DC/DC

P Vk

RECC

(a) Returned Energy Current Converters (RECC) (b) Power shuffle structure.
structure.

Figure 2-7: Further DMPPT architectures. The arrows identify the energy flows.

each DMPPT-U is a module or a panel, but different sub-arrays can be even connected.
The DMPPT techniques proposed in [67, 44, 68, 29, 69, 70] and [71] use DC/DC converters with
step-up or step-down operations; therefore, a simple MPPT algorithm (traditional or modified P&O)
in each DMPPT-U is able to set the voltage of the module regardless of the current flowing in the
string. Nevertheless, the operation range where all DMPPT-Us can track their MPPs is inversely
proportional to the PMPP (Power in MPP) difference between the less and the most irradiated
module/panel as explained in [29].
In [67] the authors consider the DC/DC converter is connected to a PV panel made of three
modules, and they propose a method to track the GMPP of each panel by performing a voltage scan
from the open-circuit voltage. Linares et al. [44], Erickson et al. [68] and Tsao et al. [29] propose
to connect the module directly to the string when the MPP current in the module is equal, or close,
to the current of the string of DMPPT-Us; this, to avoid switching losses. While in such conditions
the current flows through two switches in [44] and [68], the power stage presented in [29] uses an
additional switch to perform a direct connection of the module to the string.
The operating principle of the techniques presented in [69] and [70] are similar to the ones
presented above. In particular Perreault et al. [69] propose a high-frequency/low-power DC/DC
converter and a modified P&O method to perform the MPP tracking in each PV cell of a module.

53
An interesting fact presented in [70] is an analog implementation of the MPPT algorithm in each
DMPPT-U to reduce implementation costs.
In [18, 72, 73, 74, 75, 76, 77] and [80] the authors use step-up converters in the DMPPT-Us
controlled by a traditional MPPT algorithm (Roman et al. [72, 73] use IC, Femia et al. [18] and
Kadri et al. [74] use P&O, and Bratcu et al. [75] use ESC). Such converters may restrict the
MPP tracking when the module current in the MPP is smaller than the current of the string of
DMPPT-Us, since in the step-up converters the output current is smaller or equal to the input one.
The authors in [18] present the small-signal model of a string of DMPPT-Us and demonstrate the
stability of the system; moreover they show that a high voltage appears at the output of the highly
irradiated DMPPT-Us associated to highly irradiated modules when one or more modules operate in
deep mismatching conditions (according to (2.3)). They also propose to limit the duty cycles of the
DC/DC converters to avoid such high voltages. Another method to avoid such high output voltages
is presented in [75] and [43], which can change the operation mode of each DMPPT-U (between
MPPT or output voltage regulation) depending on the mismatching conditions. In [76] the high
output voltages are avoided by using a battery in each DMPPT-U to compensate the power drops
during the mismatching conditions.
A simple method to control two DMPPT-Us (operating in the same conditions) at the same
time is presented in [77]. Such a method considers two DMPPT-Us connected in parallel and only
measures the output current of each DMPPT-U since the output voltages of the two units are the
same. From those two measurements an analog control is able to converge to a point close to the
MPP in each module. To extend the system to higher voltage and powers, various groups of two
DMPPT-Us are connected following the SP configuration.
The system presented in [78] uses a Buck converter at the output of each string of DMPPT-Us
to produce a bus voltage that follows the reference of the absolute value of the grid voltage; in other
words the voltage of that bus is the voltage of the grid rectified. The second stage of the proposed
system is an H-bridge converter that transforms the voltage of the bus into an AC voltage to connect
the system to the grid.
Another interesting concept is proposed in [79, 80, 72, 73] and [45], where the authors implement
a communication link in each DMPPT-U to send, or exchange, information with a centralized
controller. Although such a system is useful for maintenance, diagnostic and/or improvements in
the MPPT methods in each DMPPT-U, it also increases the hardware and memory requirements of
the system.

2.5.2 DMPPT-Us connected in parallel with a string of modules/panels

In these techniques the DMPPT-U does not follows the structure presented in Fig. 2-5(a) and
Fig. 2-5(b). Instead, the PV modules/panels are connected in series to form a string, and the

54
power stages are connected in parallel to those modules/panels. The most common structures are
presented in Fig. 2-7, where the basic objective is to generate an alternative path to the current of
each module/panel to compensate or mitigate the effects of the mismatching conditions.

In Shimizu et al. [13], at each node of a string of modules is connected a converter following the
Power Shuffle structure presented in Fig. 2-7. Each DC/DC converter is composed by one switch,
one capacitor and one inductance. Such a string of N DMPPT-Us is named Generation Control
Circuit (GCC), and it permits to drive the difference between the current of the string and the
current produced by each PV module; such functionality allows to eliminate the multiple maximums
from the P-V curve. A Power Shuffle structure is also used by other DMPPT solutions like in
[82], where the authors use an additional battery or super-capacitor to improve the energy sharing
between the DMPPT-Us in the mismatching conditions.

The techniques in [83, 84] and [85] use the same concept of the GCC but each DMPPT system
uses a different converter in each DMPPT-U. Buthker et al. [83] follow the Power Shuffle structure
with a redundancy in the converters to facilitate the power transfer of a given module with the
previous and the next ones in the string. Meanwhile Stratakos and Ikiannikov [84] use bidirectional
converters in each DMPPT-U and a common capacitor to carry the energy from one to another
module. Instead, Villa et al. [85] use a custom converter composed by a set of switches, diodes,
capacitors (in parallel to each module) and a single inductor to compensate the current differences
between the shaded and unshaded modules.

The DMPPT technique presented in [86] proposes the Returned Energy Current Converters
(RECC) presented in Fig. 2-7(a). That structure is aimed to take part of the current of the PV
string to compensate the current reduction that occurs in the modules under mismatching conditions.
The objective of such a compensation is to eliminate the multiple LMPPs in the P-V curve of the
string to use a simple MPPT technique implemented in a centralized controller.

Although the technique proposed by Kadri et al. [74] was already presented in Section 2.5.1, it is
also included in this section because the authors propose additional converters connected in parallel
at the outputs of the DMPPT-Us, to balance their voltages in deep mismatching conditions. Such
additional converters solve the large output voltage limitation presented in Section 2.3.2; however,
it requires additional hardware that increases the implementation costs.

Some techniques follow the structure presented in Fig. 2-5(a) and Fig. 2-5(b) but just implement
one string of the system. Although each module has one converter the control algorithm may be
centralized or distributed depending on the technique. Two examples are presented in [87] and
[88]. In [87] each DMPPT-U has an H-Bridge converter whose outputs are connected in series with
a single inductor to form a multilevel inverter, where each module is controlled separately by a
traditional P&O and a PI voltage controller. The DMPPT-Us in [88] are controlled by a single
algorithm (Multivariable P&O). Such a method modifies the duty-cycles of the DMPPT-Us, one by

55
one, and monitors the output power of the string of DMPPT-Us to find the best operation point of
each DC/DC converter to optimize the whole system.

2.5.3 DMPPT-Us connected in parallel

This group of MPPT techniques are composed by DMPPT-Us connected in parallel. As in the group
presented in Section 2.5.1, each sub-array (module, panel, string, etc.) interacts with a converter
but the control algorithm may be centralized or distributed depending on the technique.
Some of those techniques are aimed to connect one module/panel directly to the grid or to the
DC bus of a centralized inverter. To achieve that goal it is necessary high-gain converters.
The solutions presented in [90, 91] and [92] propose high-gain converters without using trans-
formers. The converter proposed in [90] is denominated current-fed semi-quadratic Buck-Boost and
the voltage of the module can be controlled by a simple MPPT technique. Abdel-Rahim et al.
[91] present a modified Boost converter, whose duty cycle is controlled by an IC technique. Such
a converter is connected to a Full-Bridge inverter to fed the grid. In [92] each DMPPT-U has a
Boost converter with a battery at the output. The algorithms that controls the converters has to
operate in different modes (e.g. MPPT, battery charge, etc.) to guarantee the State-of-Charge of
each battery. The outputs of all DMPPT-Us are connected in parallel and fed an additional Boost
converter to reach the voltage level required by the inverter DC input.
The converters used in [94, 93, 95, 96] and [97] transform the DC current of the module into
AC and they use a transformer to obtain high voltage gains; besides, each DMPPT-U has its own
simple MPPT algorithm. Erickson et al. [94, 93] and Yao et al. [95] transform the DC voltage
of the module into a high frequency AC voltage to reduce the size of the transformer required by
each converter. One drawback of such techniques concerns the requirement of rectifying the high
frequency AC voltage to fed an inverter; therefore, the energy flows through three power stages
(DC/AC-transformer-AC/DC) before reaching the inverter. The DMPPT-Us presented in [96] and
[97] use modified flyback inverters connected directly to the module; hence the low frequency voltage
and current increase the size of the converters.
Different approaches are presented in [98] and [99], where a Multi-input Boost and Multi-input
Buck converters, respectively, are proposed. Each module is connected through an independent
branch, which is composed by a switch and an energy storage component (inductance in [98] and
capacitor in [99]). In [98] each switch is controlled by a P&O and all the modules share the same
output capacitor; while the modules in [99] share the inductor and the output capacitor.

2.5.4 DMPPT-Us where the PV sub-array is a string or small array

Some techniques propose to divide the PV array into multiple sub-arrays (i.e. one or more strings)
in order to reduce the number of converters required and the voltage conversion ratio of the power

56
converters thereof. The reduction of the number of converters, in some cases, is used to centralize
the control of the different converters, while the increment in the voltage range of the PV sub-array
connected to each DMPPT-U allows to use simpler converters whose outputs can be connected
in parallel. Therefore, these solutions can be seen as a tradeoff between CMPPT and DMPPT
architectures.

In some of these techniques the basic principle is similar to the CMPPT solutions presented
in Section 2.4.1, because the algorithms take samples of the P-V curve to find the GMPP in each
DMPPT-U. One example is presented in [100], where the authors divide the PV array into strings.
Each string is connected to a Boost converter to reach the voltage level of a DC bus that feeds an
inverter. To find the GMPP in each string, Noguchi et al. [100] measure the short-circuit current
and perform periodic scans of the P-V curve by using additional hardware at the input of each
DC/DC converter.

Other solutions use the same principle presented in Section 2.4.2 to track the GMPP in each
DMPPT-U. The technique introduced in [101] divides the PV array into strings, each string having
its own DC/DC converter. Between two consecutive strings of N modules there are N 1 irradiance
sensors, which are used by an ANN to determine the voltage of the GMPP or an LMPP close to
the GMPP in each string. Such voltages are set as the reference to be followed by each DC/DC
converter of the array.

The technique proposed in [102] divides the PV array into small arrays (sub-arrays) and connects
one DC/DC converter to each sub-array. The outputs of the converters are connected in parallel to
form a DC bus. All the converters are controlled by a centralized PSO technique, which searches
the GMPP by measuring the voltage and current at the DC bus.

There also are some techniques aimed at compensating the effects of the mismatching in the PV
array by including a DC/DC converter as part of the PV array. In Karatepe et al. [103, 104] and
Hargis [105], a DC/DC converter is connected in series with each string in order to compensate the
voltage drops caused by the modules whose bypass diodes are active and to avoid displacements in
the voltage operation of the modules without mismatching. Each DC/DC converter is fed by the
output of the string which it is connected, and it is controlled separately since the algorithm needs to
determine the voltage reference of each DC/DC converter depending on the number of mismatched
modules in each string. The output of the array is connected to another DC/DC converter with a
simple MPPT technique. Karatepe et al. [103] and [104] propose a simple method to determine
the number of mismatched modules in the array. Meanwhile, Hargis [105] presents different isolated
converters that can be used to implement this solution.

57
2.6 Literature survey of RMPPT-based solutions
This section introduces some widely adopted techniques that use a RMPPT architecture. The
main features concerning the RMPPT techniques are summarized in Table 2.6, including the main
hardware feature of each technique by considering the general structure presented in Fig. 2-6. The
first column shows the part of the PV array that can be reconfigured, the second column presents
the measurements needed by the reconfiguration algorithm, the third column illustrates the array
configuration adopted. The fourth, fifth and sixth columns present the method used to find the
best configuration, if the related document presents the design procedure, and in which conditions
the reconfiguration algorithm is executed, respectively. Finally the last three columns show the
information concerning the load, the GMPP tracking method and the reference, respectively.
It is worth noting that the RMPPT techniques mostly use the current and voltage measurements
of each reconfigurable element of the PV field. Moreover, four of the six studied techniques establish
a reduced number of possible configurations, which depend on the hardware and software used in
each PV system. In general, an MPP tracking method should link the PV-RA with the MPPT
algorithm to track the GMPP after the reconfiguration process; nevertheless, in two approaches the
P&O MPPT algorithm is adopted, while in the other papers the algorithm used or the existence of
a link between the reconfiguration process and the MPPT algorithms is not detailed.
A group of these techniques assumes that each module of the string can be connected in any row
(in TCT configurations) or in any column (in SP configurations) of the array through a matrix of
switches. The reconfiguration algorithm needs to determine which of all the possible configurations
provides the largest GMPP at the output of the array.
The RMPPT presented in [22] uses a reconfigurable TCT array. The objective is to reduce the
difference between the average irradiance of the rows as much as possible. To do that, the algorithm
measures the voltage and current of each module to estimate the irradiance. Afterwards it evaluates,
off-line, all the possible configurations to find the one that provides the minimum difference between
the rows with the highest and the lowest irradiance averages.
Another approach is presented in [106], where a reconfigurable SP array is proposed. The re-
configuration algorithm performs one scan of the array voltage and it measures the current in each
column of the array to obtain the Current vs. Voltage (I-V) curves of each string. On the basis of
such information, the model parameters of each module in the string are calculated and, by using
the model presented in [12], all the feasible configurations are evaluated to determine the one that
provides the highest GMPP.
The two techniques mentioned above require a high number of switches to connect all the modules
of the array in any row (in TCT) or column (in SP); that is why a reconfiguration solution available
in the market [107] divides the array in groups of 2 strings, and it allows to connect each panel in
one of the 2 strings or disconnect it. To find the best configuration, the system acquires the I-V

58
curve of each panel but the methodology used is not available. The information of the I-V curves is
also used for diagnostic and maintenance purposes.
Another group of RMPPT techniques, like those ones proposed by Nguyen and Lehman [108] and
by Shamseldein et al. [109], divide the PV array into fixed and reconfigurable parts. The modules
in the fixed part are connected in TCT configuration, while the adaptive part can be connected
to any row of the TCT array to compensate the effects of the mismatched modules. Nguyen and
Lehman [108] propose two methods to determine the connection of the adaptive modules. The first
one connects, one by one, the adaptive module with the highest voltage to the row of the fixed
array with the lowest voltage; such an action is repeated until all the reconfigurable modules have
been connected. The second method estimates the PV current in the fixed part and evaluates,
off-line, the best distribution of the adaptive modules to reduce the mismatching effects. The
reconfiguration algorithm presented in [109] requires the estimation of the irradiance on all the
modules in both fixed and adaptive parts, this to determine the best connection of the modules. In
general, these techniques reduce the number of possible configurations and switches required in the
array; nonetheless the capacity of mismatching compensation is also reduced.
Finally, the technique proposed in [110] uses a reconfiguration system with a low number of
switches and a simple algorithm. Basically, the method classifies the modules as shaded or unshaded.
The shaded modules are bypassed by switches and the unshaded modules are connected in series to
form strings of N modules connected in series. If the number of modules in one string is smaller than
N , then such a string is connected to a step-up converter to reach the voltage level of the complete
strings. All the strings are connected in parallel to feed a grid connected inverter.

2.7 Conclusion

A review of the state-of-the-art in the field of MPPT architectures and techniques for PV arrays,
operating in mismatching conditions, has been presented. The MPPT techniques were classified
into three architectures according to the hardware requirements and to the interconnection among
the elements of the PV array and the power converters. The general description of each architec-
ture, with its advantages and disadvantages, and the main hardware structures were also presented.
Moreover, the operation principle and basic hardware requirements of some widely adopted MPPT
techniques (61 techniques in total) were discussed. The details concerning the different approaches
and techniques proposed in this Chapter have been presented in a synthetic and structured way
to provide to the reader some basic elements that can be used as guidelines to determine the best
solution fitting with the application peculiarities and constraints.
The overview shows different ways to address the mismatching problem in PV arrays, where the
trend to use the DMPPT architecture is clearly identified, since it represents more than the 60 %

59
of the reviewed solutions. However, it is not possible to ensure which architecture and/or approach
provides the maximum benefit for the final user in all the applications, this because the criteria
to select the best MPPT solution for a specific application must be defined by the designer of the
PV system. Such criteria may vary from one application to another, since each application has its
own specific features like, among others, the shadowing profiles (fixed, variables, predictable, not
predictable, etc.), the irradiance profiles, the differences in the parameters of the PV modules of
the array, the maximum and minimum voltage constraints, and the interaction with other energy
sources and energy storage systems.

Other important aspect that should be considered in the selection of the MPPT solution is the
final objective of the PV installation, because this aspect significantly affects the decision process.
For instance, if the application is aimed at injecting energy into the grid, the analysis should provide
the architecture and the MPPT technique that reduce the time for the return of the investment.
Instead, if the PV system is used to supply a remote load, the analysis should consider the architec-
ture and the particular MPPT technique ensuring the minimum number of failures. In this kind of
applications the monitoring and diagnostic procedures may be of high importance.

The general advantages and disadvantages of the three architectures classes presented in this
Chapter are summarized in Table 2.7, this to guide the reader in the selection of the most suitable
architecture for a given application. It is worth noting that the advantages and disadvantages
in Table 2.7 are general and should be considered only as a guideline because, depending on the
application, an advantage may turn into a disadvantage. For example, a DMPPT system is able to
extract the SMPP since each module/panel of the array is connected to a power converter; however,
the use of converters for each module/panel becomes into a disadvantage in a system that operates
with low mismatching conditions since the switching losses in all the power converters degrade the
system efficiency in comparison with a CMPPT solution.

The last remarks put in evidence the lack of a clear methodology to determine what is the
best MPPT architecture and technique for a given installation. Therefore, it is necessary to de-
velop methodologies, guidelines and/or selection criteria, to help the designers of the PV systems
to choose the best MPPT architecture and technique once the features of the specific application
have been defined. One possible approach for middle and large PV plants could be to divide the
PV array according to the complexity of the mismatching conditions and use the most appropriate
architecture for each part according to the requirements and constraints of the specific application;
in any case, the process and the final performance of the system will depend on the particular design
adopted.

The overview presented in this Chapter reveals that the model-based MPPT techniques represents
a minority, since only few RMPPT techniques use the single-diode model of the PV modules to

60
determine the configuration that increases the output power of the array. Moreover, in most of
the documents the model used to represent PV array or sub-array is not provided or the circuital
representation of the array is used. This shows that there is a necessity of reliable models for PV
arrays (with different electrical configurations and dimensions) providing a tradeoff between accuracy
an calculation speed that can be used not only to develop model based MPPT techniques but also
to simulate and test new MPPT techniques and/or architectures, to estimate the energy production
of a PV system, to evaluate the effects of the PV systems in a power distribution system, among
others.
To can use the array models in all the mentioned applications, they also should be implementable
in different programming languages to can perform multiple simulation analysis that may require a
variety of simulation environments. Chapter 3 and Chapter 4 present models for the three widely
used configurations (SP, TCT and BL) and their applications in the I-V and P-V curves predictions
as well as dynamical simulations and energy production estimation.

61
 Table 2.3: DMPPT techniques part 1
DMPPT Unit (DMPPT-U)
Sub- PV Measu- Power Mani- Tracking Imple- Load Cen- Ref.
Class input rements stage pulated method mented tral
Vari- in unit
able
Pan. iP V , v o Buck/Boost/ vo Modified uC N.S. No [67]
Bridge P&O
Mod. iP V , Buck/Boost/ vP V P&O uC Inverter No [44,
vP V Bridge + grid 68]
Mod., iP V , Buck/Boost/ N.S. N.S. N.S. N.S. No [29]
pan. vP V Bridge
Cell iP V , Multiple D Modified uC, Inverter No [69]
A vP V DC/DC P&O analog + grid
Mod. io , v o Multiple D P&O Analog Battery No [70]
DC/DC + in-
verter
+ grid
Mod., iP V , Buck/Boost/ D Modified N.S. Inverter No [71]
pan., vP V Bridge P&O + grid
str.
Mod. iP V , Boost vP V IC uC Inverter Yes [72,
vP V + grid 73]
Mod. iP V , Boost vP V P&O N.S. Inverter No [18]
vP V + grid
Mod. iP V , Boost vP V P&O N.S. Inverter Yes [74]
vP V + grid
Mod. iP V , Boost vP V ESC N.S. Inverter Yes [75]
vP V , vo + grid
Mod. N.S. Modified vP V N.S. N.S. R No [76]
Boost
2 io Multiple vP V P bal- Analog Inverter No [77]
Mods. DC/DC ance + grid
Mod., N.S. Multiple N.S. N.S. N.S. Inverter Yes [78]
pan. DC/DC + grid
Mod., N.S. Multiple N.S. N.S. uP Inverter Yes [79]
pan. DC/DC + grid
Mod., iP V , Boost N.S. N.S. uP Inverter Yes [80]
pan. vP V , + grid
T P V , io ,
vo
Mod., iP V , Buck/Boost/ D Modified N.S. Inverter Yes [45]
pan. v P V , io , Bridge P&O + grid
vo
Mod. No Modified D No uComputerInverter Yes [13,
Buck-Boost + grid 81]
B
Mod., iP V , Modified vP V Modified N.S. R Yes [82]
pan. vP V Buck-Boost IC
Mod., vP V Custom N.S. V bal- N.S. N.S. No [83]
pan. DC/DC ance
Mod., iP V , Bi- vP V I bal- N.S. Inverter Yes [84]
pan. vP V directional ance + grid
DC/DC

62
 Table 2.4: DMPPT techniques part 2
DMPPT Unit (DMPPT-U)
Sub- PV Measu- Power Mani- Tracking Imple- Load Cen- Ref.
Class input rements stage pulated method mented tral
Vari- in unit
able
Mod. iP V Custom iP V I bal- N.S. DC Yes [85]
DC/DC ance load
B Mod. ibus , Multiple iP V I bal- N.S. N.S. Yes [86]
vbus DC/DC ance
Mod. iP V , Boost vP V P&O N.S. Inverter Yes [74]
vP V + grid
Pan., iP V , H-bridge vP V P&O DSP Inverter Yes [87]
str. vP V + grid
Mod., No Boost D No No Inverter Yes [88,
pan., + grid 89]
str.
Pan. iP V , Modified vP V N.S. N.S. Inverter No [90]
vP V Buck-Boost + grid
Pan. iP V , Modified D IC uC N.S. No [91]
vP V Boost
C Pan. iP V , Boost D N.S. uC DC No [92]
vP V , load
T P V , io ,
vo 1
Mod. iP V , Isolated vP V N.S. N.S. Inverter No [93,
vP V DC/DC + grid 94]
Mod., iP V , Isolated iP V P&O DSP N.S. No [95]
pan. vP V DC/DC
Mod., vP V Flyback in- D P&O N.S. N.S. No [96]
pan. verter
Pan. N.S. Modified N.S. N.S. N.S. N.S. No [97]
Flyback
Mod. iP V , Multi-input D P&O N.S. R No [98]
vP V Boost
Mod. vP V Multi-input io No FPGA Battery Yes [99]
Buck
Str. iP V , Boost iP V kIsc N.S. N.S. No [100]
vP V
D Str., G2 N.S. vP V ANN N.S. N.S. No [101]
array
Str., No Boost D PSO DSP DC Yes [102]
array load
Str. iP V , Buck vo 3 V bal- N.S. N.S. No [103,
vP V ance 104]
Str. iP V , Multiple vo 3 V bal- N.S. N.S. Yes [105]
vbus DC/DC ance
1
This technique also measures the ambient temperature and the current and voltage of the
battery in each DMPPT-U.
2
It requires one irradiance sensor for each group of 4 modules.
3
These applications modify the output voltage of the one converter for each string.

63
 Table 2.5: Central units of DMPPT techniques
Measure- Data from Commu- Objective Design Ref.
ments DMPPT nication proce-
unit link dure
No iP V , v P V , io , PLC- Maintenance, failure de- No [72, 73]
vo FSK1 tection
No No No Balance vo Yes [74]
vbus vo N.S. Avoid high vo in DMPPT- Yes [75]
Us
vbus No No Control custom DC/AC No [78]
No MPP N.S. Track GMPP in each No [79]
DMPP-U
No iP V , vP V , N.S. Improve MPP in each No [80]
TP V DMPPT-U, maintenance
ibus , vbus PP V N.S. CMPPT, diagnostic CMPPT: [45]
(CMPPT), Yes, di-
N.S. (diag- agnostic:
nostic) No
ibus , vbus No No Balance vP V Yes [13, 81]
No iP V , v P V No MPPT in each DMPPT No [82]
U.
ibus , vbus N.S. N.S. MPPT, diagnostic, main- No [84]
tenance
ibus , vbus iP V No MPPT in each str. of No [85]
DMPPT-U
ibus , vbus No No Balance iP V Yes [86]
ibus , vbus No No Regulate vbus Yes [87]
ibus , vbus No No MPPT in each DMPPT-U Yes [88, 89]
ibus No No MPPT in each DMPPT-U Yes [99]
No N.S. N.S. Maintenance, failure de- No [105]
tection
ibus , vbus No No GMPPT No [102]
1
Power Line Communication (PLC) with Frequency-Shift Keying (FSK)

64
 Table 2.6: RMPPT techniques
Reconfiguration algorithm
Reconfi- Measu- Possible Configu- Design Recon- Load Track- Ref.
gurable rements config- ration pro- figuration ing
unit ura- search ce- runs method
tions method dure
Mod. iP V , vP V Total Evaluate Yes N.S. Inverter P&O [22]
TCT all config- + grid
urations
1
Mod. istr , vbus Total Evaluate Yes N.S. DC N.S. [106]
SP all config- load
urations
Pan. iP V , vP V Reduced Evaluate No N.S. Inverter N.S. [107]
SP some + grid
configura-
tions
Cell iP V , vP V 2 Reduced I balance Yes Mis- R P&O [108]
TCT matching
detected
Mod., iP V , v P V Reduced I balance Yes N.S. N.S. N.S. [109]
pan., or G3 TCT
str.,
array
Mod. iP V , Reduced Disconnect Yes N.S. R N.S. [110]
vP V , SP shaded
TP V modules
vP V : reconfigurable unit voltage, iP V : reconfigurable unit current, vo : DMPPT-U output
voltage, TP V : reconfigurable unit temperature, G: reconfigurable unit irradiance, Mod.: PV
module, Pan.: PV panel, Str.: PV string, Array: small part of the total array, R: resistive.
1
istr refers to the measurement of the current in each string.
2
One method of this technique uses both sensors, the other one only uses voltage sensors.
3
The irradiance is estimated from image processing.

65
 Table 2.7: Main advantage and disadvantages of DMPPT sub-classes
Archi- General advantages General disadvantages
tecture
CMPPT Low cabling requirements.
Low number of sensors.
Only one power converter is required.
Low switching losses.
Only one processor is required.
Low number of components in all the
system, which reduces the probability
of failures.
DMPPT Low cabling requirements.
The power of the PV array is dis-
tributed over multiple converters.
If one or more converter/module fail
only part of the available power is lost.
High performance for complex mis-
matching pasterns.
Simple MPPT algorithms.
Simple processors/hardware to imple-
ment the MPPT algorithms.
Simple implementation of monitor-
ing/diagnostic/security procedures.
RMPPT Only one power converter is required.
If one or more module fail only part
of the available power is lost.
Increment of the maximum available
power in the PV array through the re-
configuration.
Mitigation of the mismatching effects
in the PV array.
Simple MPPT algorithms.
Simple implementation of monitor-
ing/diagnostic/security procedures.
66
All power handled by one power con-
verter.
All power is lost if the power converter
fails.
Complex MPPT algorithms.
Middle and high performance pro-
cessors are required to implement the
MPPT algorithms.
If one module fails a significantly part
of available power is lost.
Low performance for complex mis-
matching pasterns.
High number of sensors.
High number of power converters.
High switching losses in low mis-
matching conditions.
High number of components in all the
system, which increases the probability
of failures.
Communication links may be re-
quired.
Additional algorithms to manage the
DMPPT-Us may be required.
High cabling requirements.
High number of sensors depending on
the reconfiguration technique.
Power consumption of the matrix re-
lay.
All power is lost if the power converter
fails.
Complex algorithms to find the best
configuration.
High performance processors are re-
quired to implement the reconfiguration
algorithm.
A reasonable cost matrix relay have
electro-mechanical parts, which in-
crease the probability of failures.
The PV array need to be disconnected
from the load to perform the recon-
figuration, which introduces additional
power losses.
Chapter 3

Modeling of Series-Parallel PV
arrays under mismatching
conditions

The Series-Parallel is the configuration that is mostly used to interconnect PV modules in small,
medium and large PV arrays. Such a configuration simplify the electrical connections among the
panels and facilitates the power capacity of an array by adding one or more strings in parallel.

In literature there are some models for SP arrays, some of them offer high accuracy with high
computational burden [16], while other models propose methods to reduce the computational burden
with a reduced accuracy [12] or performing some interpolation approximations [15]. However, a
suitable model should provide a compromise between accuracy and computational cost and the
possibility to be implemented in different programming languages.

This Chapter presents a model of PV fields in SP configuration operating under mismatching

conditions, which provides a tradeoff between accuracy and calculation time in comparison with
other solutions proposed in literature. The electrical model of each PV module is the one presented
in Chapter 1, while the bypass and blocking diodes are represented with a piecewise linear model.
The model profits from the possibility to express each PV module voltage as an explicit function of
the current by using the Lambert-W function. The balancing of the string voltage with those ones
of the active and inactive modules, as well as of the blocking diode, allows to construct only one
non-linear equation depending on the string current. Consequently, the solution of only one equation
allows to calculate the string current for a given string voltage, and the current of the whole field
is calculated by adding the currents of all parallel strings. The advantages of the proposed model
are illustrated by comparing it with two approaches published in literature. The model accuracy

67
is evaluated with small, medium and large PV fields, and its usefulness for performing energetic
analysis and dynamical simulations is also illustrated. Experimental validation of the model on a
small scale PV array are also provided.
The information presented in this Chapter was published in Electric Power System Research
journal in a paper entitled A model of photovoltaic fields in mismatching conditions featuring an
improved calculation speed [17].

3.1 Introduction
The general structure of an SP array is presented in Fig. 3-1. In such a configuration multiple
modules are connected in series to form strings, and multiple strings are connected in parallel to
form the array. The number of modules in each string depends on the voltage levels required by
the application, while the number of strings is selected according to the power requirements of the
system.

Pv module 1,1 Vbld1 Pv module 1,m Vbldm

Ipv
V1,1 V1,m

I1 Im
Pv module n,1 Pv module n,m Vpv

Vn,1 Vn,m

Figure 3-1: PV Field with m strings of n modules parallel connected.

Each string has one diode in the top (blocking diode) to avoid backflow currents from the other
strings and each module has one diode connected in antiparallel (bypass diode) to provide an alter-
native path to current of the string when it is higher than the short-circuit current of the module.
The electrical model of a module with a bypass diode is shown in Fig. 3-2.
I

Rs
Iph ID IRh Rh V Vbd

Figure 3-2: Single-diode model of a PV module with bypass diode.

When the string current is lower than the short-circuit current (Isc) of a PV module, such a
module is active (its bypass diode is inactive) and delivers power to the system. On the other hand,

68
when the string current is higher than the Isc of a module, the bypass diode is activated. In such
a case, the module does not provide power to the system and the difference between Isc and string
current passes through the bypass diode. Moreover, the output voltage of the module is polarized
negatively due to the positive voltage of the bypass diode; hence, in those conditions the module is
consuming power.

A fast and accurate model of PV arrays (SP, TCT, BL, etc.) in mismatching conditions is needed
to evaluate the performance of the MPPT strategies, to perform an economic analysis of the PV
installations, to estimate the amount of energy fed by a PV system, to evaluate the impact of the
PV system on the reliability of a transmission and distribution system [25], to design further control
strategies, among others.

A large number of models for simulating PV systems have been presented in literature, e.g.
[16, 12, 111, 112, 6, 9, 113, 114]. Some of them do not consider the mismatching phenomenon [112, 6],
some others are very complex [16, 9] or use a linearization process to reduce the computational
burden [113, 114]. The one proposed in [12] considers the ideal single-diode model (i.e. neglecting
Rs and Rh) for each PV module and approximates the bypass diode by an ideal switch. This choice
simplifies the modeling of large n m PV fields, where n is the number of PV modules in a string
and m is the number of strings in parallel. Such a solution requires a small computational cost and
simulation time because it uses n explicit non-linear equations as maximum per each string. The
other side of the coin is that the series and parallel resistances are neglected, so that the model
accuracy in presence of a complex shading pattern is not guaranteed.

In [16] both series and parallel resistances in the single-diode model as well as the detailed model
of the bypass diode are considered. The use of the Lambert-W function [11] allows to express the
PV current as a function of the voltage explicitly. The high accuracy of such a model [16] in presence
of mismatched conditions is counterbalanced by a simulation time that is longer in comparison with
the one required by the approach given in [12] due to the implicit relations and the solution of n + 1
non-linear equations per each string.

In [111] the PV field is divided in order to calculate the strings voltage for a given current value.
Then, string currents at the same voltage are added to calculate the PV field voltage vs. current
characteristic. This model requires to pre-calculate the complete characteristics of all the strings
for calculating the current of the PV field at a given voltage. Because of the fact that not all the
strings voltages coincide, some errors are introduced, and the model parameters of all modules are
considered equal.

This Chapter presents a PV field modeling approach that includes the series and parallel re-
sistances of the single-diode PV model and uses a piecewise linear approximation for the bypass
and blocking diodes. The proposed solution is optimized for PV strings, where a single equation
is constructed by equating the string voltage, which is assumed to be given, with the sum of the

69
voltages of: active PV modules, active bypass diodes (or inactive PV modules), and the blocking
diode. Such a procedure only requires to solve one non-linear equation, per each string, to calculate
the string current instead of solving a system of n + 1 non-linear equations with n + 1 unknowns as in
the more classical PV modeling approaches [16] and [12]. Moreover, when a bypass diode is active,
the associated PV module is neglected thus reducing the number of non-linear terms of the string
voltage equation and furthermore simplifying the current calculation. Therefore, the proposed model
provides a tradeoff between the high accuracy (but with long calculation times) of some models [16]
and the short calculation time (but with lower accuracy) ensured by other ones [12]. Finally, the
overall PV array current is calculated by adding the currents of each string.

3.2 Basic equations of the SP array elements

When a PV module in a string is active, the equations presented in Section 1.2 are valid. From
now on, the sub-index i represents the position of the module in the string and the sub-index j
represents the number of the string (according to Fig. 3-1).
To simplify the reading and explanation of this Chapter, the explicit function of the PV module
voltage presented in Section 1.2 is re-written in this Section as shown in (3.1). In such an equation
V ami,j (Ij ) is the output voltage of the active module in the position i of the string j as function
of the string current (Ij ), Iphi,j is the photovoltaic current, Rsi,j and Rhi,j are the series and
parallel resistances, Isati,j is the inverse saturation current of the diode and Bi,j is defined as
Bi,j = N s k i,j Ti,j q 1 . Ti,j is the temperature of the semiconductor, i,j is the ideality factor,
N s is the number of series connected cells in the module, k is the Boltzmann constant and q is the
electron charge.

V ami,j (Ij ) = Bi,j W0 () Ij (Rsi,j + Rhi,j ) + Rhi,j (Iphi,j + Isati,j ) (3.1)

 
Isati,j Rhi,j Rhi,j
= exp (Iphi,j + Isati,j Ij ) (3.2)
Bi,j Bi,j

The values of Iphi,j , Isati,j , i,j , Rsi,j and Rhi,j can be calculated from the datasheet informa-
tion of a PV panel [6, 115]; however, Iphi,j varies with irradiance and temperature (3.3) and Isati,j
depends on the temperature (3.4). Some authors [6, 116] assume i,j , Rsi,j and Rhi,j as constants;
nevertheless, there are some works, as [117], which suggest that those parameters are also affected
by the irradiance and temperature. Appendix A introduces a method to estimate the parameters of
the single diode model for different G and T conditions by using a genetic algorithm and illustrate
the effects of environmental conditions on the parameters.
Usually, the irradiance measurement in a particular module is not practical because the irradiance

70
sensors are expensive. Besides, the value provided by the sensor may not represent the effective
irradiance received by the PV module due to soiling in the module surface, degradation in the
transparent cover of the cells, the irradiance in the location of the sensor may be different to the
one in the location of the module, among others. That is why the approximation given in (3.5) is
commonly used to obtain an estimation of Iph by means of a direct measure of Isc [7].

Gi,j
Iphi,j = (IphST C + Isc (Ti,j TST C )) (3.3)
GST C
IscST C + Isc (Ti,j TST C )
Isati,j =   (3.4)
V ocST C +V oc (Ti,j TST C )
exp Bi,j 1
Iphi,j Isci,j (3.5)

In equations (3.3) and (3.4), IscST C and IphST C are the short-circuit current and photovoltaic
current in Standard Test Conditions (STC), respectively, V ocST C is the open-circuit voltage in STC,
Isc and V oc are the temperature coefficients of the short-circuit current and open-circuit voltage,
respectively, Ti,j is the module temperature, TST C is the module temperature under STC (both in
Kelvin degrees), Gi,j is the irradiance and GST C is the STC irradiance.

For practical applications, Ti,j can be estimated using (3.6) from the ambient temperature
Tamb in Kelvin degrees, the Nominal Operating Cell Temperature (N OCT ) in Kelvin degrees, the
transmittance of the cells cover ( ), the fraction of the radiation incident on the surface of the
cells that is absorbed (), the efficiency of the module (c ), and GN OCT is the NOCT irradiance
(GN OCT = 800 W/m2 ). N OCT and c can be extracted from datasheets, while 0.9 as
described in [118].

Gi,j  c 
Ti,j = (N OCT Tamb ) 1 + Tamb (3.6)
GN OCT

When the i-th bypass diode of the j-th string is active, the associated PV module is deactivated.
Therefore, the voltage at the terminals of such a module (V imi,j ) is the inverse voltage of the bypass
diode (3.7), where V thbdij and Rbdij are the threshold voltage and the resistance of the piecewise
linear model of the bypass diode. The current that crosses the bypass diode is the difference between
the string current and the short-circuit current of the module (Isci,j ). Fig. 3-3 shows circuital
representation of the approximation used to represent the bypass and blocking diodes.

The voltage of the blocking diode of the j-th string (V bldj ) is calculated by means of (3.8),
where Rbldj and V thbldj are the series resistance and the threshold voltage of the blocking diode,
respectively.

71
 V imi,j (Ij ) = ((Ij Isci,j ) Rbdi,j + V thbdi,j ) (3.7)

V bldj (Ij ) = Ij Rbldj + V thbldj (3.8)

Vdiode
Idiode

Idiode

Ideal
diode Vthdiode Rdiode

Figure 3-3: Piecewise linear approximation of bypass and blocking diodes.

3.3 Calculation of the inflection voltages

In this Section, a PV field consisting of n modules connected in series and m strings in parallel
(Fig. 3-1) is considered. Each string has a blocking diode and each module has a bypass diode. The
bypass diode of the i-th module in the j-th string turns on when the string current (Ij ) is higher than
the short circuit current of such a module (Isci,j ), thus the voltage at which this condition occurs
is named inflection voltage [12]. Moreover, the voltage and current of the PV strings are calculated
individually, thus the analysis of the j-th string performed in this section holds for each string of the
PV field.

The calculation of the inflection voltages is done according to [12], but including the series
and parallel resistances and using the piece-wise linear approximation for the bypass and blocking
diodes. The n modules in the PV string are firstly sorted for decreasing values of their PV currents
(Iphk1,j > Iphk,j ). The inflection voltage (V ok,j , k [2, n]) is the minimum voltage of the j-th
string that activates the k-th PV module (or deactivates the k-th bypass diode). Such a condition
is fulfilled when the current of the string is equal to the short-circuit current of the k-th module
(Ij = Isck,j ). When the string is operating at an inflection voltage, the associated bypass diode
is active: the modules having a current higher than Isck,j (from P V1 to P Vk1 ) are active, and
the remaining PV modules (from P Vk to P Vn ) are inactive. Hence, for calculating the inflection
voltages it is necessary to determine the string voltage by adding the voltage of all the elements
in the string with Ij = Isck,j as in (3.9). In (3.9) the first sum (from left to right) represents the
voltage contribution of the active PV modules, while the second sum and the last term are the
contributions of the inactive PV modules and of the blocking diode, respectively.

72
 k1
X n
X
V ok,j = V amm,j (Isck,j ) + V imp,j (Isck,j ) V bldj (Isck,j ) (3.9)
m=1 p=k

From the previous analysis, a string with n modules can exhibit no more than n 1 inflection
voltages. For example, Fig. 3-4 presents the current vs. voltage (I-V) curve of a PV string made of
four ERDM 85SM/5 panels, each one of them receiving a different irradiation, where three inflection
points were calculated (3.9).

Isc
1,1
5

Isc2,1
4
Vo2,1=Vam1,1Vim2,1
Current (A)

Isc3,1
Vim3,1Vim4,1Vbld1
3

Vo3,1=Vam1,1+Vam2,1 Isc4,1
2 Vim3,1Vim4,1Vbld1

1 Vo4,1=Vam1,1+Vam2,1
+Vam3,1Vim4,1Vbld1

0
0 10 20 30 40 50 60 70 80 90
Voltage (V)

Figure 3-4: Inflection voltages of one string with four PV modules.

3.4 String current calculation using inflection voltages

The proposed model gives the string voltage (V strj ) as the sum of the voltages of the PV modules
(active and inactive) and the blocking diode. Expressions (3.1), (3.7) and (3.8) give the voltages of
the different elementas of the array as functions of the string current. Consequently, equation (3.10)
gives the value of the j-th string voltage as a function of its current (Ij ), where ac is the number of
active PV modules and n is the number of PV modules in the string.

ac
X n
X
V strj = V ami,j (Ij ) + V imp,j (Ij ) V bldj (Ij ) (3.10)
i=1 p=ac+1

The string voltage value is supposed to be assigned because it is typically fixed by the power
converter controlling the PV array, so that Ij is the only unknown variable in (3.10). Ij is determined
by means of the Newton-Raphson method. By using the properties of the Lambert-W function [11],

73
the first derivative of (3.1) is given by (3.11), where , Rsi,j , and Rhi,j were defined in Section
3.2. The derivatives of the inactive PV modules and blocking diode voltages are Rbdi,j and Rbldj ,
respectively.

dV ami,j Rhi,j
(Ij ) = Rsi,j (3.11)
dIj (W0 () + 1)

Then, the derivative of V strj with respect to Ij is:

ac n
dV strj X dV ami,j X
= (Ij ) Rbdp,j Rbldj (3.12)
dIj i=1
dIj p=ac+1

The Newton-Raphson (NR) method requires the iterative calculation of (3.13), which allows to
obtain the string current with the desired accuracy.

V strj
Ij(p+1) = Ij(p) (3.13)
dV strj /dIj

An important aspect of the model concerns the number of non-linear terms in (3.10) and (3.12),
which changes depending on the number of active PV modules. For example, in a string with
five modules and one blocking diode, where three PV modules are active and two are inactive, the
right side of (3.10) has three non-linear terms corresponding to the active modules and three linear
terms due to the two inactive modules and to the blocking diode. The number of active modules
is calculated from the string voltage and the inflection voltages: if V strj is higher than V ok,j , then
k modules are active, otherwise if V strj is lower than V ok,j but higher than V ok1,j , then k 1
modules are active. In addition, the inflection voltages also constrain the range of the Ij possible
solution: if the string voltage is greater than V ok1 and lower than V ok , then the solution of Ij is
between Isck1,j and Isck,j , which reduces even more the calculation time by providing an improved
guess solution to the NR method.

Another important aspect of the proposed model is that it does not require to calculate the
inverse of the Jacobian matrix, which is in turn needed for solving the system of non-linear implicit
equations given in [16, 12].

Finally, taking into account that in (3.2) is always positive because Isati,j , Bi,j and Rhi,j
are positive, and Iphi,j > Ij ; the term W0 () > 0 and the derivative of the module voltage (3.11)
is always negative. Therefore, the derivative of the string voltage (3.12) is always negative and
the string voltage in (3.10) is a monotonically decreasing function of the string current. Such a
condition guarantees the existence of a single solution for (3.10) since for each string voltage exists a
unique string current. In addition, the voltage derivative (3.12) is always different than zero, which
guarantees that (3.13) can be always processed.

74
3.5 PV field simulation

As described in Section 3.2, the proposed model considers a PV field consisting of m parallel strings,
each one made of n modules connected in series (Fig. 3-1). Therefore, in order to simulate the
whole PV field, the analysis shown in Section 3.4 must be repeated for each string and the currents
obtained are added to calculate the total PV field current.
Fig. 3-5 shows a flow chart that describes the proposed algorithm, which has been implemented in
both Matlab and C languages to analyze the performance of the proposed solution in the prediction of
the PV field electrical characteristics under different mismatching conditions. The C implementation
was compiled in a dynamic link library that is useful to test the model in PSIM environment.

   
   

   

     

  
   
  

          

        



      

     
  
          


   
    

    !      

  
      

    

         8
    "       

$ % #
  "    

 
 
        



 
 
 # $   
     

 

%
 

&   $ %
#
  "
  

' ( ( ) * +   !  , 


,   
     
      

$ % #

  "    

$ % #
'   

 

  &  
!- 


  
-  
 -     ,  
8
!

Figure 3-5: Flow chart of the proposed model.

3.6 Performance evaluation

The proposed model is numerically validated by comparing its performances with those ones of two
reference modeling approaches presented in literature. An experimental validation is also provided.
The first reference model [12] approximates the bypass diode by an ideal switch and neglects the

75
parallel and series resistances in the PV model. The second reference model [16] uses the the Shockley
equation for the bypass and blocking diodes as well as both the series and parallel resistances in the
single-diode model of the PV modules. The former model [12], in the sequel referred to as Fast
model, calculates the string current by solving multiple non-linear explicit equations. The latter
model [16], in the sequel referred to as Complex model, calculates the string current by solving a
system of non-linear implicit equations. The model presented in this Chapter, in the sequel referred
to as Tradeoff model, calculates the string current by solving a single explicit equation.
The model performance was evaluated in six scenarios processed in a PC equipped with Intel(R)
Core(TM)2 Duo of 3.0 GHz and 4 GB of RAM. The first scenario is aimed at comparing the
accuracy of the models in the simulation of a small PV string, considering also the single diode
circuital simulation implemented in both PSIM (by using the Renewable Energies Toolbox) and
Simulink (by using the SimElectronics toolbox). The second scenario considers a larger PV field of
15 strings with 15 PV modules each (15x15 PV field) in order to compare the performance of the
models with the circuital simulation performed in Simulink. The third scenario is useful to compare
the accuracy of the models and of the Simulink circuital implementation in the simulation of a large
PV field of 50x50 modules. The fourth scenario considers the energetic evaluation of a large PV field
(of 50x50 modules) for two different intervals of time by means of the Fast and Tradeoff models,
since such solutions provide reasonable simulation times. The fifth scenario illustrates the practical
application of the Tradeoff model in the simulation of a photovoltaic battery charger controlled by
a MPPT algorithm. The last scenario tests the proposed model accuracy for the reproduction of
experimental data measured from PV panels in different irradiance conditions. The simulations
and experiments consider ERDM 85SM/5 PV panels, which have one bypass diode and its main
electrical characteristics were presented in Section 1.6 (see Table 1.1).

3.6.1 Simulation of a small PV field

PV fields made of one string of two (2 3) and three (3 1) PV modules were implemented in
PSIM and Simulink in order to test the accuracy of the Fast, Complex and Tradeoff models, which
have been implemented in Matlab language. In the PSIM and Simulink circuits, the complete single
diode model with the bypass diode has been considered.
The models parameters adopted in the simulations are shown in Table 3.1, where Isatbd and
Isatbld are the inverse saturation currents of the bypass and blocking diodes, respectively, for the
Complex model. In this case, the parameters of the blocking and bypass diodes are assumed equal.
Three cases were simulated for the 2x1 PV field (2x1a, 2x1b, 2x1c), while two cases were simulated
for the 3x1 PV field (3x1a, 3x1b). For all cases, Iph currents are given in Table 3.2 as a percentage
of the IscST C current in Table 1.1.
The I-V and P-V curves of the 2x1 and 3x1 PV fields for the different irradiance profiles are

76
 Table 3.1: Parameters for the Fast, Complex, and Tradeoff modes
Fast model Complex model Tradeoff model
Param. Value Param. Value Param. Value
Isat 9.9410e-9 A Isat 1.5415e-8 A Isat 1.5415e-8 A
B 0.9264 V B 1.1088 V B 1.1088 V
- - Rs 0.0045 Rs 0.0045
- - Rh 109.405 Rh 109.405
- - Isatbd 1e-6 A Rbd 0.003
- - Bbd 0.015 V Vthbd 0.2166 V
- - Isatbld 1e-6 A Rbld 0.003
- - Bbld 0.015 V Vthbld 0.2166 V

Table 3.2: Irradiance profiles for simulation of 2x1 and 3x1 PV fields [% of IscST C ]
PV current 2x1a 2x1b 2x1c 3x1a 3x1b
Iph1 50 94 94 94 90
Iph2 20 50 60 60 80
Iph3 - - - 20 70

presented in Fig. 3-6. The values of the characteristic curve were calculated with voltage steps of 0.1
V; therefore the curves representing the 2x1 and 3x1 PV fields have 419 and 655 points, respectively.
The agreement between the results obtained by means of the circuital implementations (PSIM and
Simulink) and the Complex and Tradeoff models is evident. Instead, the curves calculated by means
of the Fast model exhibit voltage shifts, which causes errors in the inflection voltages prediction and
overestimations of the maximum power.
Table 3.3 presents the relative mean errors in the prediction of the current (Ei ), the power (EP )
and the maximum power (EM P ) of the different modeling approaches with respect to the Simulink
circuital simulation, as well as the simulation time (tsim ) of the three models implemented in Matlab
(Complex, Tradeoff and Fast models). The results show that both circuital simulations (PSIM and
Simulink) are equivalent. The Fast model requires the lowest calculation time but exhibits the largest
error. In contrast, the Complex model ensures the lower error but requires the larger calculation
time. The Tradeoff model ensures an error that is close to the one given by the Complex model,
which is much lower than the one given by the Fast model. At the same time the simulation time of
the tradeoff model is lower than one required by the Complex model (between 3 and 4 times lower).
Such results confirm that the proposed model runs significantly faster than the detailed model and
gives an accuracy similar to the one obtained by means of circuital simulations.

3.6.2 Simulation of medium size PV fields

This scenario considers a PV field of 15 strings with 15 PV modules each, where all the modules
have the same parameters given in Table 3.1. The irradiance of the PV modules, represented by
Iph currents, were defined as follows: in the first string, all PV modules have the same irradiance

77
 2x1c
4

Current (A)
2x1a

2
2x1b
PSIM
0 SIMULINK
0 Fast10 20 30 40
Tradeoff
100 Complex

Power (W)
2x1c
50 2x1b

2x1a
0
0 10 20 30 40
Voltage (V)

(a) 2x1 PV field

3x1b
Current (A)

2 3x1a

PSIM
0 SIMULINK
0 10 20 30 40 50 60
Fast
200 Tradeoff 3x1b
Complex
Power (W)

100
3x1a

0
0 10 20 30 40 50 60
Voltage (V)

(b) 3x1 PV field

Figure 3-6: Characteristic curves of the first scenario simulations.

Table 3.3: Relative mean errors for simulations of 2x1 and 3x1 PV fields
Model 2x1a 2x1b 2x1c 3x1a 3x1b
Ei [%] 3.620 2.517 2.636 2.198 2.515
Fast EP [%] 6.484 4.813 4.362 3.817 3.141
EM P [%] 10.757 7.488 6.462 7.131 5.363
tsim [s] 0.250 0.022 0.046 0.033 0.042
Ei [%] 1.696 1.833 1.951 0.793 1.152
Tradeoff EP [%] 3.288 3.727 3.504 1.647 1.842
EM P [%] 0.160 0.010 0.045 0.181 0.076
tsim [s] 5.651 7.021 6.977 12.068 15.540
Ei [%] 1.641 1.809 1.929 1.012 1.428
Complex EP [%] 3.199 3.681 3.470 2.010 2.240
EM P [%] 0.197 0.8e-3 0.027 0.074 0.030
tsim [s] 22.028 22.974 22.672 49.368 45.650
Ei [%] 0.500 0.547 0.571 1.487 2.033
Circuit in PSIM EP [%] 0.661 0.782 0.739 3.039 3.267
EM P [%] 0.170 8.2e-3 0.023 0.069 0.028

78
given by Iphk,1 = IscST C with k={1..15}. In the second string, only one PV module has a different
photovoltaic current equal to the 5 % of IscST C . In the third string, 13 PV modules have photovoltaic
currents equal to IscST C , one equal to 5 % of IscST C and another one equal to 11.79 % of IscST C .
The number of PV modules with Iph different from IscST C increases in one, uniformly distributed
from 100 % to 5 % of IscST C . In the last string, all the photovoltaic currents are different. In this
case, the circuital simulation is performed in Simulink only, this because large PV fields simulation
in PSIM gives convergence problems.
Fig. 3-7 shows the current and power curves of the whole PV field. The current and power values
were calculated with voltage steps of 1 V (thus resulting in 328 points) in order to obtain the curves
of the PV field. The curves of each string have to be calculated with the same voltage step (1 V);
therefore, each curve of Fig. 3-7 requires the calculation of 4920 points (15 238).
Again, the satisfactory agreement between the Complex and Tradeoff models, with the circuital
simulation, is evident. Instead, the Fast model exhibits errors in both the current and power estima-
tions. The relative errors of the three models with respect to the circuital simulation are summarized
in Table 3.4, where Ei represents the current prediction error, EP the power prediction error, EM P
the error in the prediction of the maximum power, and tsim the simulation time. Such results
demonstrate, again, that the proposed model provides a satisfactory accuracy in the current and
power predictions, requiring a simulation time much shorter than the Complex model.

80

60
Current (A)

40 SIMULINK
Fast
20 Tradeoff
Complex
0
0 50 100 150 200 250 300

10
Power (kW)

0
0 50 100 150 200 250 300
Voltage (V)

Figure 3-7: Characteristic curves of the second scenario simulations (PV field 15x15).

Table 3.4: Relative mean errors and simulation time of the 15x15 PV field
Fast Tradeoff Complex
Ei [%] 2.3210 6.7921e-2 6.4387e-2
EP [%] 3.3841 1.3633e-1 1.2404e-1
EM P [%] 5.5763 9.7720e-3 1.3232e-2
tsim [min:s] 00:0.31 11:30.56 28:54.20

79
3.6.3 Simulation of large PV fields

A large PV field made of 50 strings, each one consisting of 50 PV modules, is adopted for illustrating
the accuracy of the Tradeoff model as well as the short simulation time it needs. As in the previous
scenario, the irradiance profile goes from all PV modules with the same irradiance (first string) to all
PV modules with different irradiance equally distributed from 100 % to 5 % of the STC irradiance
(last string). The I-V and P-V curves, obtained using the models and the circuital approach, are
presented in Fig. 3-8. The current and power values were calculated for voltage steps of 1 V (1087
points); nevertheless, for obtaining each curve of the Fig. 3-8 it was necessary to calculate 54350
points (50 1087) collecting the values of the 50 strings.

Again, there is a satisfactory agreement of both the Complex and the Tradeoff models with
the circuital simulation, while the Fast model exhibits voltage and power shifts that degrade the
predictions. Table 3.5 presents the relative mean errors and simulation times for all the modeling
approaches, which demonstrate that the proposed model provides an accuracy very close to the
Complex model but with a significantly shorter simulation time.

200
Current (A)

SIMULINK
100 Fast
Tradeoff
Complex
0
0 200 400 600 800 1000

100
Power (kW)

50

0
0 200 400 600 800 1000
Voltage (V)

Figure 3-8: Characteristic curves of the third scenario simulations (PV field 50x50).

Table 3.5: Relative mean errors and simulation time of the 50x50 PV field
Fast Tradeoff Complex
Ei [%] 2.0342 6.6152e-2 6.3484e-2
EP [%] 2.9230 1.3928e-1 1.2755e-1
EM P [%] 4.4003 4.4438e-2 1.5959e-2
tsim [h:min:s] 00:00:12.57 08:40:52.00 46:59:40.00

80
3.6.4 Prediction of the energy production

The proposed model (Tradeoff) and the Fast model were implemented in C language in order to
evaluate the potential energy produced in a given period of time, e.g. days, months or years. The
Complex model is not considered because of the long calculation time it would require, with an
accuracy that is similar to that one ensured by the Tradeoff model. It is worth to note that for
evaluating the potential energy produced by a PV field, the calculation of the P-V curve is needed
and the global maximum power for each value of the irradiance profile must be taken.
In this case, the same PV field (50x50) and the same mismatching conditions used in the previous
simulations are considered for four irradiance profiles with different time scales. For such a PV field,
54350 points (see section 3.6.3) have to be calculated for each irradiance value.
The first profile, presented in Fig. 3-9(a), refers to experimental data taken in the South of Italy
during a typical winter day sampled every 30 s (1224 irradiance points). The maximum power of
the PV field for each irradiance condition of such a profile is presented in the Fig. 3-9(b). The Fast
model overestimates the potential power profile. The total daily energy predicted is 238.024 kWh
with the Fast model and 211.449 kWh with the Tradeoff model: thus the Fast model predics a 12.57
% more energy. Such an error may produce a wrong reliability analysis of the system and/or the
undersizing of the PV array. The simulation times of the Fast and Tradeoff models were 26 min
12.41 s and 43 min 27.32 s, respectively, which indicates that in the C implementation the Tradeoff
model only requires 1.66 times the calculation time of the Fast model.

0.5 50
Fast
0.4 40
Irradiance (kW/m )

Tradeoff
2

Power (kW)

0.3 30

0.2 20

0.1 10

0 0
08:10 10:10 12:10 14:10 16:10 08:10 10:10 12:10 14:10 16:10
Time (hn:mm) Time (hh:mm)

(a) Irradiance profile. (b) Predicted power.

Figure 3-9: Simulations for typical winter day, South of Italy.

The second, third and four irradiance profiles were taken in the center of Colombia during the
summer for periods of 15 days, one month, and two months, respectively, with samples taken every
hour. As an example, in Fig. 3-10(a) the irradiance profile of the first seven days is shown and in
Fig. 3-10(b) the maximum power prediction of both Fast and Tradeoff models is presented. The
energy predictions and simulation times of the Fast and Tradeoff models for the three irradiance
profiles are given in Table 3.6. In the three periods the energy over estimation of the Fast model is
12.3 % in average, such an error leading to a wrong economic analysis or PV plant design. Regarding

81
the calculation time, the Tradeoff model only requires, in average, 1.72 times the calculation time
of the Fast model. Therefore, the proposed model is adequate for energetic evaluations in different
time periods ensuring small errors.

Table 3.6: Relative mean errors for simulations of 2x1 and 3x1 PV fields
Model 15 days One month Two months
Fast Energy [MW-h] 4.1324 8.4917 20.376
tsim [min:s] 04:13.85 08:20.61 16:58.82
Tradeoff Energy [MW-h] 3.6597 7.5361 18.314
tsim [min:s] 07:01.03 14:11.00 30:35.91

0.8
Irradiance (kW/m2)

0.6

0.4

0.2

0
1:01 2:16 4:08 6:00 7:16
Time (day:h)

(a) Irradiance profile.

80 Fast
Tradeoff
60
Power (kW)

40

20

0
1:01 2:16 4:08 6:00 7:16
Time (day:h)

(b) Predicted power.

Figure 3-10: Simulations for one week of summer, center of Colombia.

3.6.5 Dynamic simulations

The proposed model was implemented in C++ in order to have a dynamic link library (DLL) to
be used in PSIM. Because of the simplicity of the operations needed to implement the proposed
model, only one specialized function (GSL - GNU Scientific Library [119]) was used to calculate the
Lambert-W function.
The model parameters are specified in a text file loaded by the DLL block. The PV modules
parameters Isat, B, Rs and Rh are given in matrices with dimensions n m, where n represents
the number of modules in the strings and m the number of strings. The bypass and blocking diodes
parameters, V thbd and Rbd, are given in matrices with dimensions (n + 1) m, where the first n
rows correspond to the bypass diodes and the rows n + 1 correspond to the blocking diodes. In

82
addition, the mismatching pattern for the photovoltaic current of the PV modules is defined with
a n m matrix, with numbers between 0 and 1, while the nominal irradiance is an input of the
block. Moreover, the solution tolerance for the numerical calculation is defined in the configuration
window of the DLL block.
In order to validate the proposed model in dynamic conditions, a practical application concerning
a battery charger controlled by a MPPT algorithm was implemented in PSIM. Fig. 3-11 shows the
PSIM scheme, where the block at the left of the circuit is the DLL block implementing the proposed
model, which interacts with a boost converter to charge a battery. The dc/dc converter is controlled
by a P&O algorithm implemented by means of another DLL block for tracking the Maximum Power
Point (MPP) [19].
The PV array consists of one string of 11 ERDM 85SM/5 PV panels connected in series, where
three of them have a shadow of 50 %, hence there are two MPPs. The simulation considers two
different conditions: an irradiance of 0.5 kW/m2 with two MPPs at 316 W and 230 W, and an
irradiance of 1 kW/m2 with two MPPs at 685 W and 510 W. The power curves for both irradiance
conditions are given in Fig. 3-12.
Two simulations were carried out in order to illustrate the usefulness of the proposed approach
in testing MPPT techniques for mismatching conditions. In both simulations the calculation step
time was 1 s, the total simulation time was 0.6 s, and the irradiance starts at 0.5 kW/m2 and
changes to 1 kW/m2 at 0.3 s. In the first simulation, presented as continuous lines in Fig. 3-13,
the initial duty cycle was D = 0.6, which sets the PV voltage at the left of the global MPP (see
continuous vertical line in Fig. 3-12), and the system successfully tracks such an MPP (319 W). At
t = 0.3 s the irradiance step from 0.5 kW/m2 to 1 kW/m2 occurs, and the P&O algorithm tracks
the new global MPP (689 W). In the second simulation, presented as dashed lines in Fig. 3-13, the
initial duty cycle was D = 0.4, where the system tracks the LMPPs: 234 W for 0.5 kW/m2 and 515
W for 1 kW/m2 (see dashed vertical line in Fig. 3-12).
The simulation results demonstrate the model usefulness for testing MPPT algorithms. As
expected, in such conditions the P&O algorithm is not reliable for tracking the global MPP.

3.6.6 Validation through experimental measurements

In order to validate the accuracy of the proposed model, two experiments are carried out by using
the experimental platform 1 presented in Section 1.6 (see Fig. 1-7(a)). In addition, a Matlab script
was used to perform the voltage sweeps on the PV string and to the aim of recording the PV current
and voltage values through the DC load sensors. Moreover, since the experiments were performed
in a single string, the blocking diode is neglected.
The first experiment considers a single ERDM 85SM/5 PV panel, named Panel 1; its current
and power curves were scanned from 0.5 to 19.5 V with steps of 0.5 V (39 points) in intervals of

83
 DC/DC converter

PV array model

MPPT
control

Figure 3-11: Battery charger scheme in PSIM considering the proposed model.

G=1 kW/m2
600
G=0.5 kW/m2
Power (W)

400

200
D=0.6 D=0.4
0
0 50 100 150 200 250
Voltage (V)

Figure 3-12: Power vs. Voltage characteristic curves of the 11x1 PV array.

30 ms. The current and power values predicted by the model were calculated for the experimental
voltages, where the model parameters are presented in Table 3.7. Fig. 3-14 shows both experimental
and predicted characteristic curves. Such results illustrate that the relative mean errors for the PV
current and power were 0.262 % and 0.255 %, respectively, thus agreeing with the errors obtained
by the simulations in the Section 3.6.1.

The second experiment considers an additional PV panel, named Panel 2, whose parameters
are also given in Table 3.7. Fig. 3-15 shows experimental and model predicted data obtained with
voltage sweeps from 0.5 V to 38 V with steps of 0.5 V (76 points) and time intervals of 30 ms.
Such a figure shows the satisfactory agreement between the experimental and predicted curves,
including the accuracy in the inflection voltage calculation. Moreover, the relative mean errors of
this experiment, 0.418 % for the current and 0.419 % for the power, validate the accuracy of the

84
 Power (W)
500
Initial D = 0.6
Initial D = 0.4
0
0 0.1 0.2 0.3 0.4 0.5 0.6

Voltage (V)
200

100 Initial D = 0.6

Initial D = 0.4
0
Duty Cycle 0 0.1 0.2 0.3 0.4 0.5 0.6

0.5
Initial D = 0.6
Initial D = 0.4
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Time (s)

Figure 3-13: Battery charger dynamic simulation in PSIM.

Table 3.7: Model parameters for the PV panels

Model Parameter Panel 1 Panel 2
Iph (A) 4.39 2.71
A (A) 0.719 0.494
B 1.241 1.206
Rh () 2.362e3 0.261e3
Rs () 0.532 0.572
Rbd () 0.603 0.603
Vthbd (V) 0.988 0.988

4
Current (A)

2
Tradeoff
Experiment
0
0 5 10 15 20
Power (W)

40

20 Tradeoff
Experiment
0
0 5 10 15 20
Voltage (V)

Figure 3-14: Experimental characteristics for Panel 1.

proposed model in mismatched conditions and illustrate the agreement with the simulation data of
Section 3.6.1.

Although the previous experiments consider small PV arrays, the same validation can be easily
extended to larger PV plants. In any case, the accuracy of the proposed model was validated with

85
15 15 and 50 50 PV arrays in sections 3.6.2 and 3.6.3, respectively, contrasting the model results
with the circuital implementations.

Current (A)
2
Tradeoff
Experiment
0
0 10 20 30 40
80

Power (W) 60
40
Tradeoff
20
Experiment
0
0 10 20 30 40
Voltage (V)

Figure 3-15: Experimental characteristics for string of Panel 1 and Panel 2.

3.7 Conclusions

The model of PV fields operating in mismatched conditions, presented in this Chapter, shows a sat-
isfactory compromise between accuracy and calculation speed. This makes it suitable for simulating
small, mid, and large PV fields as well as to perform energetic evaluations in different time periods
(days, weeks, months, etc.) and dynamical simulation in circuital simulation software like PSIM.
The consideration of the parallel and series resistors in each PV module and the piecewise linear
approximation of the bypass and blocking diodes guarantee the accuracy of the proposed model.
Moreover the bypass and blocking diode modeling simplification reduces the computational burden.
This allows to express the voltage of each PV module as an explicit function of the current, which
allow to write one non-linear equation per string as a function of the string current for a given string
voltage. This equation can be easily solved by using the classical Newton-Raphson method. Moreover
the calculation of the inflection voltages allow to reduce the non-linear terms in the equation that
describes each string and restricts the current range of the solution.
The proposed model has been compared with a Complex model [16], a Fast model [12] and a
circuital implementation in Simulink for small, mid and large PV fields with relative mean errors
very close to the Complex model (smaller than 2%), which demonstrate the accuracy of the model
with respect to the circuital implementation. Two energetic evaluation for two different time periods
were performed with the Fast and proposed models, where the energy calculated by the Fast model
was 12.68 % higher than the one obtained by the proposed model. Such conditions show that
the proposed model should be used for PV plants design. The proposed model has been also
implemented in PSIM for demonstrating its application in the MPPT testing and design, and finally

86
it was experimentally tested with current relative mean errors smaller than 0.5 %. The performance
evaluation presented in this Chapter demonstrates the flexibility of the model for different PV fields
analysis and further applications.
An important point that need to be considered is the calculation of the model parameters of
each module. Although Iph and Isat can be estimated with the equations presented in Section 3.2,
the values of , Rs and Rh may vary with the irradiance and temperature conditions [117] as well
as the degradation of the module due to the outdoor operation [3]. Hence, if the behavior of a PV
array need to be reproduced accurately along the operation time, it is necessary to have a method
that updates the parameters values; otherwise, the accuracy of the model can not be guaranteed.

Although the SP is the more usual PV array structure, there are other configurations (like TCT
or BL) that can be used to construct a PV field, which include additional connections between
the strings, with respect to SP, to produce alternative paths for the currents of the panels; this to
mitigate the effects of the mismatching conditions for some shadowing patterns. However, it can not
be guaranteed that an specific configuration will provide the highest GMPP for any mismatching
condition (as shown in Chapter 2); therefore, suitable models for configurations different than SP
are required to compare their performance. This could help the designers in the selection of the
configuration for a specific application according to its mismatching patterns.
The following Chapter presents modeling procedures for TCT and BL arrays that can be im-
plemented in programming language like C or Matlab, in order to can use them in for different
simulation analysis of PV systems.

87
88
Chapter 4

Modeling of Total Cross-Tied and

Bridge-Linked PV arrays under
mismatching conditions

Array configurations that implements connections between the strings of the PV array have been
proposed in literature to reduce the adverse effects produce by the mismatching conditions. Par-
ticularly, TCT and BL configurations have been used as alternatives for the the traditional SP
arrays.

The additional connections implemented by TCT and BL arrays provide alternative paths for the
currents of the panels that, in some cases, can compensate the current reduction in a panel produced
by a mismatching condition. Such a compensation may avoid the appearance of additional MPPs
in the P-V curve or generate a P-V curve with a single MPP depending on the specific mismatching
pattern, which increases the maximum power at the output of the array and simplifies the GMPP
tracking.

Those compensation or mitigation of the mismatching effects of a configuration like TCT, BL or

other is a characteristic that need to be considered in the designing and planning of a PV system to
improve the performance of the systems. For example, if the shades over the PV field are produced
by fixed surrounding objects, the shadowing patterns can be predicted and, using suitable models,
different configurations can be tested to determine the best option for a particular application or a
model based RMPPT can determine the best interconnection of the panels in the array.

This Chapter presents two mathematical models for TCT and BL PV arrays operating under
mismatching conditions, where the PV modules are represented by the ideal single-diode model, this
is, neglecting the series and parallel resistances of the model presented in Section 1.2.

89
 In the model for TCT arrays each row is represented as an equivalent non-linear PV circuit with
a bypass diode (approximated as an ideal switch), which allows to represent the TCT array as one
string of equivalent PV circuits. The inflection voltages (array voltages that turn off the bypass
diodes) of the string are calculated in order to solve only the non-linear equation system related
to the active equivalent PV circuits for calculating the array current for a given voltage. Such a
procedure generates a significant reduction on the computational burden required to calculate the
array current and power in comparison with classical approaches. In addition, the model is validated
by means of simulations and experimental data, obtaining satisfactory results. The accuracy and
improved calculation speed of the proposed model are useful to perform energetic evaluations of
photovoltaic arrays for viability analysis, and make the model suitable to work altogether with
online reconfiguration techniques to maximize the power production.
For BL arrays, a method to pose the system of n m non-linear equations required to obtain
the voltages of each module of the array is proposed. Such a method applies the Kirchhoff current
law to each internal node of the array and the Kirchhoff voltage law to: each loop created in the
upper part of each connection of two consecutive strings and each string with respect to the output
voltage of the array. Moreover, the model provide a procedure to obtain the Jacobian matrix to
reduce the solution time of the equation system. The circuital model of two arrays (small: 3x3 and
medium: 20x3) were implemented in Simulink to validate the proposed model. The accuracy and
improved calculation speed of the proposed model allow its use to perform energetic evaluations of
bridge-linked arrays or its comparison with other typical array configurations, which can be useful
in the designing of photovoltaic plants.
The information and results presented in this Chapter have been published in two papers pub-
lished in the journals Dyna [120] and Tecno Lgicas [121], as well as in two conference papers
[122, 123].

4.1 Introduction

In the TCT configuration the PV modules are connected in parallel forming rows and such rows are
connected in series to form the array. Fig. 4-1 shows an example of TCT configuration where the
connections are represented by both black thin lines (the same for SP configuration) and the dashed
white lines.
The BL arrays are a tradeoff between SP and TCT (see Fig. 4-1) since they implement approx-
imately half of the additional connections between two consecutive strings with respect to the TCT
configuration. The additional connections start after the second module between the first and the
second string and are implemented every two modules. Between the second and third string the
additional connections start after the first module and are implemented, again, every two modules.

90
The additional connections of the further strings are connected following the same pattern [124].

Ipv

+ + +
V1,1 V1,2 V1,3
I1,1 - I1,2 -
I1,3
- +
Vpv
+ + + -
V2,1 V2,2 V2,3
I2,1 I2,2 I2,3
- - -

+ + +
V3,1 V3,2 V3,3
I3,1 - I3,2 - I3,3 -

Figure 4-1: PV arrays of 3 columns and 3 rows in SP (connections in thin continuous lines), TCT (SP
connections plus white dashed lines) and BL (SP connections plus thick black lines) configurations.

The connections between the strings used by TCT and BL configurations provide alternative
paths to the currents of modules working under mismatching conditions; hence, a PV array with a
given mismatching pattern can have different I-V and P-V curves depending on the configuration in
which the modules are connected. An example of this phenomena is illustrated in Fig. 4-2 for an
array of 3x3 modules. In such an example it is assumed that all the module parameters are the same
and the mismatching profile is described in Table 4.1, whose values refer to the level of irradiance
that reaches the module with respect to the STC irradiance (1000 W/m2 ). In this way the value of
0.534 with n = 2 and m = 3 means that the second module in the third string has an irradiance of
534 W/m2 .
For the mismatching pattern used in the example it can be observed that the the SP configu-
ration provides the lower Global Maximum Power Point (GMPP) among the three configurations.
The additional connections of the BL array mitigate, in part, the effects of the mismatching con-
ditions because the GMPP increases and it only has two maximums (one Local Maximum Power
Point (LMPP) around 56 V and one GMPP at 36 V) instead of the three maximums of the SP
configuration. In this example, the TCT configuration show the best behavior because it provides
the highest GMPP.
It is important to note that there is not a configuration that ensures the best performance for all
the possible mismatching patterns because the characteristic curves vary with the array configuration
and with the mismatching distribution on the array; therefore, depending on the specific mismatching
pattern, SP or BL configurations can provide a better performance.
The TCT configuration has gain popularity in PV installations since it mitigates the effects of

91
 Current (A)
10

0
0 20 40 60
Voltage (V)
TCT
400 BL
Power (W) SP
200

0
0 20 40 60
Voltage (V)

Figure 4-2: I-V and P-V curves for a 3x3 PV array with a fixed mismatching pattern in SP, TCT
and BL configurations.

Table 4.1: Mismatching profile of the 3x3 PV array.

m=1 m=2 m=2
n = 1 0.283 0.551 0.652
n = 2 0.843 0.176 0.534
n = 3 0.371 0.788 0.415

the mismatching conditions as reported in [124, 125]. But, despite its increasing utilization in PV
installations, it has been found little information concerning the modeling of TCT arrays: [124]
proposes the equations for solving the equivalent circuit of a rectangular TCT configuration to
validate a piece-wise linearized model. However, for a TCT configuration with n rows of m PV
modules (connected in parallel) it is necessary to solve a system of n m + n non-linear equations as
proposed in [124], which results in a high computational burden and long calculation times.
Due to the lack of mathematical models for TCT configurations, Shams El-Dein et al. [125] use
circuital simulations of the PV system, implemented in Simulink from Matlab, to detect the TCT
configuration that produces higher power. Such a solution is intended for a TCT array with a fixed
number of PV modules and a particular mismatching profile. Basically, the solution reported in [125]
constructs a cost function that calculates the energy of the array in a period of time by considering
the voltages of each row and the current of the array, which is optimized by means of the Branch
and Bound (BB) algorithm. Then, the optimal TCT configuration (OTCT) is contrasted with the
SP and original TCT configurations by using circuital simulations again. Since a circuital simulator
is required, the solution of Shams El-Dein et al. is not suitable for online reconfiguration, moreover
it requires a large amount of calculations to solve the implicit Kirchhoff circuit laws of the PV array.
Regarding BL arrays, the authors, so far, have found few information about modeling and simula-
tion of BL arrays in mismatching conditions. Only in [124] the authors pose the system of equations

92
to be solved for an specific BL array (9x4) to validate a piece-wise linearized model proposed in
such a paper. Nevertheless, in [124] there is not any generalized procedure to model a BL array.
Moreover, in [126] the authors present some general indications to be considered in the definition of
the non-linear equation system when there is a connection between two columns of a PV array in
SP configuration; nonetheless, any general procedure to model BL arrays is presented.
Taking into account the reduced information about mathematical models for TCT and BL con-
figurations, it is evident the necessity of such models to: define the best PV modules arrangement in
PV installations, select between SP, TCT and BL configurations, among others. Moreover, finding
online the correct array reconfiguration is the new frontier in small PV systems, where active recon-
figuration devices are used to maximize the PV power. In such an application the TCT configuration
is widely adopted [22, 108], hence a fast and accurate model-based MPP calculation procedure is
required.
This Chapter proposes a mathematical model to calculate the current and power of TCT and
BL arrays, this aimed at reducing the calculation time and removing the requirement of circuital
simulators in energy evaluation analysis. In such a way, the proposed model is suitable to be
implemented in standard programing languages such as Matlab script or C, which also simplifies its
implementation in embedded controllers for reconfiguration procedures.

4.2 Modeling of TCT arrays

4.2.1 Experimental comparison of TCT and SP configurations

Fig. 4-3 shows both SP and TCT configurations of a PV array composed by four PV modules
connected in two strings (columns) and two rows (2x2). From circuit of Fig. 4-3(a) is noted that
in the SP configuration, the PV modules in the same string (column) have the same current but
different voltage. Instead, from circuit of Figure 4-3(b) is noted that in the TCT configuration, the
PV modules in the same row have the same voltage but different current.
Moreover, in SP configuration the same mismatching level on PV modules of the same row have
different effect on the array current and power. Therefore, as discussed in literature [22], exchanging
row-arranged PV modules could generates different maximum power points. In example, if the four
modules in Fig. 4-3(a) have different irradiance conditions, exchanging the position of modules 3
and 4, e.g. move module 3 to the second string and move module 4 to the first one, would change
the operating conditions of each string, which modify the power curve of the array. In contrast, the
same exchange of modules in the TCT configuration of Figure 4-3(b) has no impact on the array
current or power. Such a condition is due to the current of the row is the same regardless the PV
modules position whereas any module is removed or added.
To compare the SP and TCT electrical behavior, the experimental system 1 presented in Section

93
 (a) Series-parallel configuration.

(b) Total cross-tied configuration.

Figure 4-3: PV arrays with two strings and two rows.

Figure 4-4: Connection scheme of experimental test bench.

1.6 was used. The four ERMD85 PV modules were interconnected by means of a connection table,
and its PV voltage was imposed by an electronic DC load controlled by means of a computer with
Matlab as shown in Fig. 4-4. Such a computer also registered the current/voltage data of the PV

94
array.
From the experiments, the I-V and P-V curves for the SP and TCT configurations of Fig. 4-3
were obtained under deep mismatching conditions: the I-V curves of each PV module was registered
and used to plot the electrical characteristics of the PV arrays. The obtained data are shown in Fig.
4-5, where both TCT and SP configurations generate the same power profile if the same mismatching
profile is given (TCT and SP configuration 1). When the PV array is partially shaded by a fixed
obstacle as a post or a tree, which is typical in urban environments, the change of the sun position
also changes the irradiance profile, it causing that the shaded modules becomes non-shaded and
vice versa. Cloud movement may cause the same effect. This condition can be modeled by the
exchange of modules 3 and 4 irradiances, which in TCT does not cause any effect on the PV power,
while in SP it generates different current and power curves (SP configuration 2) with a lower global
MPP. Therefore, TCT configuration is an interesting option if the mismatching profile is not fixed
or previously known, which is a common condition.

4.2.2 TCT configuration analysis

Using an explicit relation between the current and voltage of a PV module by neglecting the parallel
and series resistances of the single-diode model presented in Section 1.2, equation (4.1) is obtained,
where I and V are the current and voltage of the PV module, respectively. The parameters A, B, and
Isc can be evaluated from the datasheet information for a given irradiance (G) and temperature (T )
by using (4.2)-(4.5), where IscST C , V ocST C are the short-circuit current and open-circuit voltage in
Standard Test Conditions (STC), respectively; TST C , and GST C are temperature and irradiance of
the PV module in STC, respectively. BST C is the value of parameter B in STC, Impp and V mpp
are the current and voltage of the PV module in the MPP for a given irradiance and temperature
conditions, respectively. Finally, Isc and V oc are the current and voltage temperature coefficients.
Note that in this case the parameters Isat and the thermal voltage are not defined in the same
way as in Section 3.2, this because when the resistances of the single-diode model are neglected
(single-diode model) the factors that multiply the exponential function (A) and the voltage inside
the exponential function (B) need to be adjusted to fit the model with some reference values of the
I-V.

I = Isc A exp (B V ) (4.1)

G
Isc = IscST C (1 + Isc (T TST C )) (4.2)
GST C
BST C
B = (4.3)
1 + V oc (T TST C )
ln (1 (Impp/IscST C ))
BST C = (4.4)
V mpp V ocST C

95
 (a) Current vs. voltage curve.

(b) Power vs. voltage curve.

Figure 4-5: PV arrays characteristics in mismatching conditions.

A = IscST C exp (BST C V ocST C ) (4.5)

From Fig. 4-3(b), as anticipated above, it is evident that in TCT configurations the current of
a row of PV modules does not depend on the modules position in the row. Using the explicit PV
model (4.1), the current of a PV row with m parallel modules and voltage Vr is:

Ir = Ieq (Vr ) (4.6)

i=m
X i=m
X
Ieq = Isci , (Vr ) = Ai exp (Bi Vr ) (4.7)
i=1 i=1

where Ai and Bi are the modules parameters, while Ieq and (Vr ) represent the equivalent electrical
circuit of the row as depicted in Fig. 4-6. From the derivative of the row current given in (4.8),

96
which is always negative since modules parameters are always positive, it is noted that each row
voltage Vr produces a unique row current Ir . In addition, (4.8) ensures that Ir is monotonically
decreasing.

i=m
Ir X
= Ai Bi exp (Bi Vr ) < 0 (4.8)
Vr i=1

Figure 4-6: Equivalent circuit of a row in a TCT array.

Another important factor from a practical TCT array is revealed in Fig. 4-3(b): commercial
PV modules have bypass diodes, inserted by the manufacturer, aimed at protecting the modules
from negative currents that cause hot spots that degrade the array [127, 128]. Bypass diodes are
placed in parallel with the PV module, and such diodes force zero voltage to the modules if the
current imposed to the module is higher than its short-circuit current, which mainly depends on the
irradiance.
Since the activation of a bypass diode forces null power production of the associated row (zero
voltage in one module), it also causes an inflection point [12] on both current and power curves.
Such a condition is illustrated in Fig. 4-5(b), where near 15.19 V the TCT current and power curves
exhibit an instantaneous derivative variation.

4.2.3 Inflection points in TCT configurations

A TCT configuration with n series and m parallel modules can be represented using (4.6) by n
series equivalent circuits, each one of them with a bypass diode. Therefore, the PV array current
Ia is defined by the higher row current, while the rows with lower currents will have the associated
bypass diodes active, which in turn imposes zero voltage to such rows; therefore, the currents of all
bypassed modules are short-circuit ones (currents at zero voltage).
The previous condition can be used to reduce the array current/power calculation time, since to
evaluate such variables for a given array voltage it is required to calculate the current of each PV
module. Therefore, the rows voltages must be calculated, and by means of (4.6), the array current
can be obtained. This means that n row voltage variables are unknown, and a non-linear n-equation
system, composed by multiple instances of (4.6), must be solved. But if the imposed array voltage

97
precedes an inflection voltage, at least one bypass diode is active; hence, at least one row voltage is
zero and the non-linear system to be solved has one equation less. In general, if the imposed array
voltage precedes J inflection voltages, the non-linear system to be solved has J equations less, which
strongly reduces array current/power calculation time.
Since the PV rows are in series, as in Fig. 4-3(b), the position of a particular row in the array has
no impact on the current and power curves. Therefore, without loss of generality, the description
of the inflection points calculation considers the PV rows placed in descendant order of Ieq , hence
Ieq,j Ieq,k with j < k and j, k [1, n]. Since current of j-row is higher than current of k-row,
in a string with j-row and k-row connected in series, the inflection point occurs when, at least, one
bypass diode of the row k becomes active, which also causes:

Ir,j = Ir,k , Vr,k = 0 (4.9)

Replacing (4.6) into (4.9), the inflection voltage is given by the voltage of the j-row (V oj,k ) as
shown in (4.10), where indexes {i, j} and {i, k} refer to the i-th modules of the j-th and k-th rows,
respectively. Moreover, (4.10) can be used to calculate the inflection voltage between PV rows with
different number of modules mj and mk . For a rectangular TCT array, i.e. n m, mj = mk . It is
noted that (4.10) is a nonlinear equation that could be solved by traditional numerical methods for
obtaining the inflection voltage between j-th and k-th row (V oj,k ).

mj mj mk mk
X X X X
Ai,j exp (Bi,j V oj,k ) = Isci,j Isci,k + Ai,k (4.10)
i=1 i=1 i=1 i=1

However, in a string with more than two rows, V oj,k represents the contribution of the j-row
to the minimum array voltage that turn off the bypass diode of the k-row (V ok ); therefore V ok is
calculated as the sum of the inflection voltage contributions of the rows with Ieq greater than Ieq,k
(modules from 1 to k 1). In general the contribution of the m-row to the inflection voltage of the
k-row (with m < k) is obtained from the solution of V om,k from (4.11), and the value of V ok is
calculated with (4.12).

mk
X
Ieq,m m (V om,k ) = Ieq,k Ai,k (4.11)
i=1
k1
X
V ok = V om,k (4.12)
m=1

From (4.12) is evident that for calculating an inflection point, (4.11) must be solved k 1 times.

98
Hence, with n PV rows there are maximum nr = n 1 inflection points, and to find all the inflection
points (worst case) (nr +1)nr /2 non-linear equations must be solved. But since the inflection points
do not change with the imposed array voltage for a given irradiance and temperature conditions,
the solution of those equations is required only one time to reconstruct the PV array current/power
curves or to find the array MPP.

4.2.4 Calculation of the TCT array current

As anticipated above, the array current Ia for an imposed array voltage Va is calculated by means
of (4.6), where the PV rows voltages Vr,i must be found. Such voltages are obtained by taking into
account that the current of the rows is the same, which define the following non-linear equation
system:

Ia = Ir,1 = Ir,2 = Ir,i = Ir,nac (4.13)

nac
X
Vr,i = Va (4.14)
i=1

Such a system has nac + 1 non-linear equations, where nac depends on the value of Va with
respect to the inflection voltages and represent the number of non-bypassed rows. Therefore, the
number of calculations required to obtain the array current/power curves is significantly reduced in
comparison with classical techniques, where the calculation of each point of the curves requires the
solution of a n m + n non-linear equation system as in the TCT array representation reported in
[124].
The non-linear equation system of (4.13)-(4.14) can be solved by means of classical approaches
like the Newton-Raphson method, or by means of modern approaches like the fsolve() function of
Matlab. But in both cases the searching domain of the solution is constrained by the inflection
points: the array current is always constrained by the currents of the inflection points that surround
the array voltage. Therefore, if the array voltage Va is between the inflection points k and k + 1,
i.e. V ok < Va < V ok+1 where V ok and V ok+1 are the voltage inflection points, the array current
Ia fulfills Io,k > Ia > Io,k+1 where Io,k and Io,k+1 are the current inflection points, which are
calculated from (4.6). Such a characteristic reduces the calculation time since the zone where the
solution occurs is known. Moreover, since the current derivative is negative (4.8), the solution Ia can
be found by iteratively searching the solution starting from the inflection points, but more efficient
algorithms like Newton-Raphson method or fsolve() function can converge to the solution faster.
The proposed mathematical model for TCT arrays is useful in both uniform and mismatching
conditions, but, due to the presence of multi-peaks and inflection points, the mismatching conditions
provide a higher challenge. Therefore, the proposed model was validated using a TCT array with

99
n = 4 and m = 2 and the deep mismatching profile described in Table 4.2, which values refer to
the level of irradiance that reaches the module: 1.0 refers to an unshaded module while 0.2 refers to
a module with 20 % of the maximum irradiance. The simulation considers a maximum irradiance
Gmax = 1000 W/m2 , hence, G1,1 = 1000 W/m2 for n = 1 and m = 1, while G4,2 = 200 W/m2 for
n = 4 and m = 2. Besides, in this example it is assumed that all the modules operate under the
same temperature (25 o C).
Each module in this example corresponds to one module of the BP 585 panel, whose principal
electrical parameters are: IscST C = 5.0 A, V ocST C = 11.05 V , Impp = 4.72 A, V mpp = 9 V ,
Isc = 80 mV /C, and V oc = 0.065 %/C. Moreover, the parameters A and B of the modules are
8.941 A and 1.406 1/V , respectively, which can be calculated by using (4.4) and (4.5).

Table 4.2: Mismatching profile to simulate the TCT array.

m=1 m=2
n=1 1.0 0.9
n=2 0.6 0.8
n=3 0.4 0.3
n=4 0.2 0.2

10

8
PV array current [A]

6 Inflection Points

2 PSIM circuital simulation

Mathematical model
0
0 5 10 15 20 25 30 35 40
PV array voltage [V]

(a) Current vs. voltage curve.

120
PV array power [W]

100

80

60
Inflection Points
40
PSIM circuital simulation
20
Mathematical model
0
0 5 10 15 20 25 30 35 40
PV array voltage [V]

(b) Power vs. voltage curve.

Figure 4-7: Model and circuital simulation of a 4 2 TCT array.

Fig. 4-7 presents the simulation of such a TCT array by means of the proposed model and using
a circuital scheme, similar to the one given in Fig. 4-3(b), implemented in the power electronics
simulator PSIM. The simulation results of Fig. 4-7 put in evidence the correctness of the inflection

100
points calculations and the accurate reconstruction of the current and power curves. Therefore, the
proposed model allows simulating the power production of TCT arrays without using a circuital
simulator, which allows a fast implementation of PV arrays of different sizes by only modifying
the matrices with the parameters of the models. Such size changes of the PV arrays in circuits
simulators, like PSIM, would imply the addition, removal, and connection of multiple blocks, which
is unpractical for large PV fields; moreover, some problems in the circuits simulation may appear
when the circuital representation and/or mismatching pattern complexity increase. Other advantage
of the proposed model is that the simulations can be performed iteratively in standard programing
languages, such as C/C++ or Matlab script, to predict the potential energy production of a TCT
array in periods of hours, days, months, years, etc. Besides, the model implemented in C++/Matlab
can be used in PSIM/Simulink by using a General DLL block or a Matlab function, this to use the
proposed model together with power electronics circuits and MPPT techniques as shown in Section
3.6 for SP arrays.

4.2.5 MPP calculation in TCT arrays

The calculation of the maximum power of a PV array, and the associated optimal operating point
MPP, is essential to estimate the array power production. Such a procedure is required to evaluate
the economic and technological viability of PV power plants [129, 130]; and also it could significantly
improve the operation efficiency of PV reconfiguration systems by providing information of the
best configuration and its optimal operating point [22, 108]. Similarly, a fast and accurate MPP
calculation could improve MPPT algorithms by reducing the convergence time to the MPP [63].
Several papers have been published concerning the estimation of power production in PV arrays
[12, 131], mainly devoted to SP configurations, where the most common practice is to perform a
sweep to the power curve to detect the GMPP [12] in a similar way as the calculation of Fig. 4-
7(b). But such long simulations waste processing time calculating points at both the left and the
right of the LMPPs to find, among them, the GMPP. Taking into account that after a LMPP there
always exists an inflection point, as shown in [48], the LMPPs can be found by calculating the points
only at one side of them, e.g. left or right, instead of both of them. Moreover, since the MPP of
commercial PV modules occurs near its open circuit voltage V oc (minimum voltage at which the
module current is zero), there is a high probability that the MPP of a row of PV modules occurs
near the row open-circuit voltage V ocr . Therefore, this Chapter proposes to search the LMPPs
staring from the inflection points and the array open-circuit voltage (V oca ).
The first step to find the GMPP of a TCT array is to calculate the inflection points, then from
the first V ok the array voltage is reduced meanwhile the array power is increased; when the array
power is not further increased the sweep stops and the first LMPP is detected. The same procedure
is performed for all the inflection points (V ok , k [1, nr ]) and for V oca , where a maximum of n

101
LMPPs may exist; the GMPP is the LMPP that provides higher power. V oca is calculated by
solving (4.6) with Ir = 0 for each row in the array to obtain n V ocr values, which are added to
Pn
obtain V oca = i=1 V ocr .

10

GMPP
8

PV array current [A]

6
LMPP
4
Tracking
2 trajectory

Inflection Points
0
0 5 10 15 20 25 30 35 40
PV array voltage [V]

(a) Current vs. voltage curve.

140
GMPP
120 LMPP
PV array power [W]

100

80

60 Tracking
trajectory
40
Inflection Points
20

0
0 5 10 15 20 25 30 35 40
PV array voltage [V]

(b) Power vs. voltage curve.

Figure 4-8: MPP calculation for a 4 2 TCT array.

The GMPP calculation, using the proposed procedure, for the mismatched TCT array simulated
in the previous section is illustrated in Fig. 4-8. The figure shows the trajectory (red dots) to detect
the four LMPP starting from the inflection voltages and array open-circuit voltage, obtaining the
same results from Fig. 4-7 (complete voltage sweep) but calculating only the 22.56 % of the points
used to construct such a figure. Therefore, the proposed MPP calculation procedure significantly
improves the estimation of the potential energy that could be produced by a TCT array in comparison
with the classical voltage sweep practice.

4.2.6 Experimental validation of the TCT model

The size of the non-liner equation system that must be solved is used to validate the reduced
calculation requirements of the solution presented in this Chapter, this in contrast with the clas-
sical approach of [124]. Fig. 4-9 shows the relative size of the non-linear equation systems Seq =
Neq,ip /Neq,cl , where Neq,ip and Neq,cl correspond to the number of equations required in the proposed
procedure and in [124], respectively.

102
 Figure 4-9: Relative size of the non-linear equation systems to find the MPP.

To provide a fair comparison such a figure considers symmetrical TCT arrays, thus n = m, and
a MPP voltage near to 80 % of the open circuit voltage. Moreover, only the half of the modules
have been considered under mismatching conditions, this because all modules mismatched is the
best condition for the validated model in terms of number of equations (two equations), while non
modules mismatched is the worst condition (n + 1 equations). From Fig. 4-9 is noted that the
validated model has to solve an equation system with less than 14 % of the equations required by
the classical solution for n 5. In addition, such a relative size is reduced when the number of
modules increases: for n 80 the size of the equation system in the proposed model is less than 1
% of the one required by the classical approach. Such results put in evidence the strong reduction
of the computational burden achieved with the validated model.

To validate the model accuracy, a TCT array with three rows and two strings (n = 3, m = 2,
named 3x2) of ERDM 85SM/5 modules was considered, whose nominal electrical parameters are:
IscST C = 5.13A, V ocST C = 21.78V , Impp = 4.8A, V mpp = 17.95V , Isc = 0.0013A/o C, and
V oc = 0.07405V /o C.

The tests use experimental data obtained from the laboratory setup presented in Fig. 4-4,
where six I-V characteristics of ERMD85 PV modules with different mismatching conditions were
measured. From such electrical characteristics, the parameters of the explicit model (4.1) were
calculated by means of (4.2)-(4.5) for all the PV modules.

A first configuration, named CF1, was constructed with the experimental data to form the
3x2 array with a deep mismatching profile described in Table 4.3, which values refer to the level
of irradiance that reaches the module: 1.0000 refers to a non-shaded module while 0.4263 refers
to a module with 42.63 % of the maximum irradiance. The data exhibit a maximum irradiance
Gmax = 560.60 W/m2 , therefore in CF1: G1,1 = 560.60 W/m2 for n = 1 and m = 2, while
G3,2 = 322.63 W/m2 for n = 3 and m = 2. Fig. 4-10 presents the I-V and P-V curves of such
CF1 experimental-based TCT array, where three LMPP (power-peaks) and two inflection points are

103
observed.

Table 4.3: Mismatching profile for experimental validation of the model.

CF1 m = 1 m = 2 CF2 m = 1 m = 2
n = 1 0.8359 1.0000 n = 1 1.0000 0.4263
n = 2 0.5530 0.5755 n = 2 0.8359 0.5755
n = 3 0.4263 0.5755 n = 3 0.5530 0.5755

(a) Current vs. voltage curve.

(b) Power vs. voltage curve.

Figure 4-10: MPP calculation for a 4 2 TCT array.

Then, the proposed model was implemented in a Matlab script using the fsolve() function with
the default trust-region-dogleg algorithm, it considering the calculated parameters of (4.1) for each
PV module. The model was processed to estimate the I-V and P-V curves of CF1, which results are
presented in Fig. 4-10, where the LMPP and inflection points where accurately predicted. Moreover,
to validate the model accuracy quantitatively, the Normalized Sum of Squared Errors (NSSE) index
(4.15) was used, where y and ye represent the reference and estimated values, and H represents

104
the number of data. The low N SSE value (N SSE = 0.0807 %) provided by the validated model
guarantees a correct reproduction of the real electrical characteristics. Therefore, the maximum
power achievable by the array can be accurately predicted.

PH 2
k=1 (y(k) ye (k))
N SSE[%] = PH (4.15)
2
k=1 (y(k))

To illustrate the usefulness of the validated model in testing the effect of different positions of
the modules in the TCT array, a second configuration named CF2 was simulated by exchanging the
position of 66.67 % of the PV modules as given in Table 4.3. Fig. 4-10 shows the I-V and P-V
curves of CF2, where the short-circuit current of the array, the inflection points, and the LMPP
have significant changes. The results also show an increment in the maximum power caused by
connecting the modules in CF2 instead of CF1.
Finally, the results presented in Fig. 4-10 put in evidence the correctness of the inflection
points calculation and the accurate reconstruction of the current and power curves. Hence, the
validated model can be used to simulate the power production of TCT arrays without using a
circuital simulator. Moreover, such TCT array simulations can be performed iteratively to predict
the potential energy production of an array in periods of hours, days, months, years, etc. Similarly,
the model can be used to find the best modules position in the TCT array by simulating the possible
cases to contrast the maximum power achievable in each one of them.

4.2.7 Energy evaluation example in TCT configuration

To illustrate the usefulness of the validated model in both the evaluation of energy production and
the comparison of the modules position for long terms, the mismatched CF1 and CF2 arrays were
used to estimate the potential energy production of a TCT power plant. The adopted irradiance
profile, depicted in Fig. 4-11(a), corresponds to a typical winter day at the south of Italy.
Fig. 4-11(b) reports the simulation of such TCT arrays for the 8 hours and 20 minutes of the
irradiance profile, sampled every 30 s (1000 irradiance values), where CF1 produces 0.5225 kW h
while CF2 produces 0.5348 kW h. Such information is essential to calculate the return time of the
investment used to define the economic viability of the solution. Moreover, the simulation reports
that just by changing the position of the PV modules to CF2 the array produces 2.36 % more
energy, which in turns increases the economic viability of the power plant. The simulation of the
irradiance profile (1000 irradiance values) for CF1 and CF2 take 45 minutes in a computer with 8
Gb of memory and a processor Intel Core i7-2 Ghz.
Finally, this example put in evidence the model usefulness in the planning, design and modifica-
tion analysis of TCT photovoltaic power plants.

105
 (a) Irradiance profile.

(b) Power production profile.

Figure 4-11: Simulations for typical winter day, South of Italy, for TCT arrays.

4.3 Modeling of BL arrays

4.3.1 BL configuration analysis

In the analysis of BL configuration the bypass diode is represented by using the Shockley equation
to improve the accuracy of the model; hence, the current of a PV module in the position i of the
column j of the array including a bypass diode is given in (4.16), where IsatBDi,j and BBDi,j
represent the inverse saturation current and the thermal voltage of the bypass diode, respectively.
Fig. 4-12 shows the electrical representation of the ideal single-diode model with the bypass diode.

Ii,j = Isci,j Ai,j exp (Bi,j Vi,j ) IsatBDi,j exp (V /BBDi,j ) (4.16)

106
 Figure 4-12: Ideal single-diode model with bypass diode.

The model of a BL array can be formalized by defining the system of non-linear equations to
be solved for a given operation conditions. Such a system of equations is written to calculate the
voltage of each module in the array, then (4.16) is used to calculate the current of each module; in
that way, all the electric variables of the array will be know.
In a PV array composed by n modules in each string and m strings connected in parallel, it is
necessary to write a system of n m non-linear equations to solve the n m unknown voltages of
the array. Such n m non-linear equations are obtained by applying the Kirchhoff current (KCL)
and voltage (KVL) laws to the array.
An array of n m has n 1 nodes between the modules in each string, where the node 1 is
between the module 1 and module 2 and node n 1 is between modules n 1 and n. For any node
that connects two consecutive strings the KCL is described by (4.17), where Ia,b (Va,b ) represents the
current function of the module a in the string b as a function of its voltage (Va,b ) as shown in
(4.16). a is the number of the node and b is the string at the left of the node. In the odd strings
a only takes even values, while in the even strings a only takes odd values. Such a convention
is assumed to avoid the repetition of nodes in the definition of the non-linear equation system.

Ia,b (Va,b ) + Ia,b+1 (Va,b+1 ) (Ia+1,b (Va+1,b ) + Ia+1,b+1 (Va+1,b+1 )) = 0 (4.17)

In the first and last strings there are nodes formed between two consecutive modules connected
in series (marked with black circles in Fig. 4-1). In such cases the application of KCL results in
(4.18), where c only takes odd number for the first string. For the last string c only takes odd
numbers if m is even and vice versa.

Ic,d (Vc,d ) Ic+1,d (Vc+1,d ) = 0 (4.18)

The connections between the strings create loops composed by the four modules located at the
upper part of the node (marked with white arrows in Fig. 4-1). Applying the KVL to the loop
created by the node a which connects the strings b and b + 1 (4.19) is obtained.

107
 Va1,b1 + Va,b (Va1,b+1 + Va,b+1 ) = 0 (4.19)

The first node that connects an even string with an odd string (e.g. connection between string
2 and string 3 in Fig. 4-1) creates a simple loop of two modules connected in parallel; therefore, the
KVL is described by (4.20), where f only can take even values.

V1,f V1,f +1 = 0 (4.20)

Each string of the BL array is connected in parallel to the power converter (e.g. DC/DC converter
or inverter), which sets the voltage of the array (Va ) to track the GMPP of the system or to guarantee
a fixed array voltage; therefore, the voltage of each string fulfills (4.21), where k goes from 1 to m.

n
X
Va Vi,k = 0, k [1, m] (4.21)
i=1

Considering the general equations presented in (4.17)-(4.21) it is possible to pose a system of

n m non-linear equations required to find the voltages of the array. On the other hand, the
definition of the Jacobian matrix is required to reduce the processing time. To write such a matrix
the derivative of the current of a module with respect to its voltage is required (4.22). Moreover,
the unknown variables of the system (voltages) are organized according to the vector Vvec presented
in (4.23).

dIi,j
= Bi,j Ai,j exp (Bi,j Vi,j ) BBDi,j ABDi,j exp (Vi,j /BBDi,j ) (4.22)
dVi,j
Vvec = [V1,1 V2,1 Vn,1 V1,2 V2,2 Vn,2 V1,m V2,m Vn,m ] (4.23)

The procedure used to define the system of non-linear equations and the Jacobian matrix is
presented in Fig. 4-13. Such a procedure produces a vector with the system of non-linear equations
(F ) and the Jacobian matrix (Ja) by applying the KCL and KVL along the PV array. Then, F and
Ja are used by a numerical method to find the voltages of all the modules in the array.
The non-linear equation system is described by going over all the nodes of the PV array. The
left part of the flow chart is aimed to define the variables, matrix and vectors required to describe
the system of equations and the Jacobian matrix. Then, the decision blocks j = 0 and j = m
identify if the analyzed node belongs to the first or the last string, this to write the equations and

108







  
  





  
  
    (  

"
  #  $)



  
    
       &  




       
  '
    #  $ 
'
    & #  $ 


     
 '
    #  $ 
  
      '
    &#  $ 

  
 
    


 


   "
  #  $+
"
  #  $ *      &  

  
            &   '
    
"
  #  $% '
     '
    
     &              
'
    #  $  '
    & '
    &
'
    & #  $  '
    &

 
     
   
  
 


  
  
 
 
"
  #  $ 


Figure 4-13: Flow chart to write non-linear equation system and Jacobian matrix for BL arrays.

the part of the Jacobian matrix that correspond to the nodes formed by two modules connected in
series. Such nodes are identified with black circles in Fig. 4-1.
The left part of the flow chart is used to describe the equations and Jacobian matrix that
correspond to the nodes that interconnect two strings and the voltage loops formed in the upper
part of those nodes.
Finally, at the end of the process it obtains a vector of the non-linear equation system (F ) and
the Jacobian matrix (Ja). This process can be used by a numerical method to solve the non-linear
equation system of the array.

4.3.2 Simulation results of the BL model

The simulations presented below consider the ERDM 85SM/5 PV module as a reference, while the
P600J diode was selected as a reference for the bypass diode. Such a diode has a voltage drop
of 0.11 V for a forward current of 8 A and an inverse saturation current of 1 A. The model
parameters of the modules were calculated by using (4.5) and (4.3) considering a temperature of

109
25 o C (A = 75.992 A and B = 0.7220 1/V ) while the model parameters of the bypass diode were
calculated from the datasheet information (IsatBD = 1 A and BBD = 6.90 mV ).

4.3.3 Simulation of a small and medium BL array

The circuital models of a small BL array of three 33 (see Fig. 4-1) and a medium array (20x3)
were implemented in Simulink to compare the accuracy and of the proposed model. The I-V and
P-V curves of the small and medium PV arrays are presented in Fig. 4-14(a) and Fig. 4-14(b),
respectively.

15
Current (A)

No Mismatching
10
Mismatching
5
Simulink
0 BL Model
0 20 40 60
Simulink
800 BL Model
Power (W)

600
No
Mismatching
400 Mismatching

200
0
0 20 40 60
Voltage (V)

(a) Small PV array.

15
Simulink
Current (A)

10 BL Model

0
0 100 200 300 400
1500
Power (W)

1000

500

0
0 100 200 300 400
Voltage (V)

(b) Medium PV array.

Figure 4-14: Characteristic curves of a small and medium BL arrays.

For the 33 array two conditions were simulated. The first one considers the system operat-
ing under STC conditions and the second one considers the array operating under mismatching
conditions. The mismatching pattern is presented in Table 4.4 as factors of the STC irradiance.
The values of n and m represent the position of the module in the string and the number of the
string, respectively. In both cases, the agreement between the circuital and proposed model is put

110
in evidence in the low Normalized Sum Squared Errors (N SSE), which are 0.0023 % for STC and
0.0367 % mismatching conditions.

Table 4.4: Mismatching profile of the 3x3 BL array.

m=1 m=2 m=2
n=1 0.90 0.80 0.50
n=2 0.60 1.00 0.40
n=3 0.20 0.10 0.30

The characteristic curves presented in Fig. 4-14(b) show the match between the circuital imple-
mentation and the proposed model in a medium PV array operating under mismatching conditions.
In this case, the short-circuit currents of the modules were generated randomly to simulate a com-
plex mismatching pattern. Again, the low NSSE (0.0226 %) confirms the quality of the proposed
model results for medium PV arrays. For this array the simulation time consumed by Simulink to
calculate 4000 points was 1371.7 s, while the time consumed by the proposed model was 94.4 s (14.5
times faster), which demonstrate the usefulness of the proposed in the reduction of the simulation
time of an array with respect to a circuital implementation.

4.3.4 Energy estimation in a medium BL array

To predict the potential energy that can be produced by a PV array in a given period of time, it is
necessary to calculate the P-V curve of the array for each value of irradiance to calculate the GMPP
of the curve. For BL arrays the inflection points are not available; hence, the method presented in
Section 4.2.5 can not be used.
The values of GMPP are integrated along the time to obtain the potential energy that can be
produced by the array. In this example the energy is calculated for an irradiance record of one
day with samples every minute (Fig. 4-11(a)) considering the 20x3 array presented in the previous
section.
Two different simulations were performed to illustrate the effect of inverting the order of the
modules in the second column of the array. The GMPP vs. time plots for both cases are presented
in Fig. 4-15, where the black line corresponds to the array with the same configuration used for the
previous section (CF1) and the gray line corresponds to the array with the modules of the second
column organized in the opposite way (CF2); this is: the first module (i = 1) is located at the end
of the string (i = 20), the second module (i = 2) takes the position i = n 1, and so on until the
last module (i = 20) takes the position i = 1.
For the BL array in CF1 the energy estimated was 3.263 kW h, while with CF2 the estimated
energy was 3.063 kW h. It is observed that the change of the order of the second string reduces in
6 % the energy obtained from the array in one day, that is why the tools to simulate a PV array in
different configurations are important for designing PV plants.

111
 700

600

500

Power (W)
400

300

200
CF1
100
CF2
0
8:10 10:10 12:10 14:10
Time (hh:mm)

Figure 4-15: GMPP for a 20x3 BL array in two configurations.

4.4 Conclusions

The proposed experiments and simulations confirm the advantages of the TCT and BL models:
low computational burden, high accuracy reproducing the experimental data, and its usefulness to
perform energetic evaluations for viability analysis.
The performed experiments and simulations also highlight that the structure of the models make
simple their implementation in standard programing languages (e.g. C, Matlab, etc.) for using
them in different ways: for example, the models can be used to estimate the energy production of a
TCT or BL array and to rapidly test different configurations, which is useful to perform economic
and technical viability analysis. Similarly, the models can be used in a reconfiguration scheme to
dynamically find the best modules organization.
The proposed models can be further improved by considering the series and parallel resistances
of the single-diode model. But such a condition will generate implicit current/voltage relations that
require the specialized Lambert-W function to be solved [16], which also requires large calculation
times.

The results presented in this Chapter show that different configurations provide different power
profiles that are translated into different amount of harvested energy. The best configuration depends
on the specific mismatching profile of the PV field; therefore, the configuration that best fits with a
particular application should be evaluated considering the specific mismatching profiles.
As shown in Chapter 2, a recent approach presented to increase the maximum power available in
a PV array is to modify dynamically the configuration of the array (RMPPT) looking for to improve
the energy production of PV plants. Such a an approach is analyzed in the next Chapter for the
specif case of the SP configuration considering the model of the whole array, unlike some RMPPT
techniques presented in Chapter 2 where each module is analyzed separately using the single-diode

112
model.

113
114
Chapter 5

Reconguration analysis of
photovoltaic arrays based on
parameters estimation

One of the applications of the array models, with a compromise between accuracy and calculation
time, is the development of model based MPPT techniques. As shown in Chapter 2 the RMPPT
solutions are a new trend in the development of PV systems not only to increase the energy extracted
from the PV generator, but also for the possibility to extract information from each module of the
array, which can be used for diagnostic, monitoring and maintenance procedures. Moreover, the
possibility to disconnect the damaged panels from the array improves the safety of the system.

In a model based reconfiguration algorithm the model parameters of all the modules in the array
are required. Such parameters should be extracted periodically since they change with the irradiance
and temperature as well as the aging and degradation mechanisms. Moreover, the array model needs
to be as fast as possible because it will be used many times in by the PV-RA to determine the next
array configuration.

This Chapter presents a method to determine the PV Series-Parallel (SP) array configuration
that provides the highest Global Maximum Power Point (GMPP). Such a procedure was designed to
only require measurements of voltage and current of each string, which avoids to perform experiments
in each module. The ideal single-diode model parameters of each module in the string are obtained
from the analysis of the I-V characteristics of each string. Using the estimated parameters, all
feasible PV array configurations are evaluated to determine the array configuration that provides
the highest GMPP. Finally, the proposed solution is validated using simulations and experimental
data.

115
 The information and results presented in this Chapter have been published in Simulation Model-
ing Practice and Theory journal in a paper entitled Reconguration analysis of photovoltaic arrays
based on parameters estimation [106].

5.1 Introduction

In the traditional SP and TCT configurations the interconnection of the PV modules are fixed.
However, such connections can be changed through a matrix of switches in order to reduce the effect
of the mismatching conditions [132, 108, 22]. Such systems are named reconfigurable arrays and have
gained popularity in the last years due to the increase achieved in the maximum power available at
the output of the PV generator under mismatching conditions [109, 133].
There are different methods to find the best configuration of the PV array reported in the
literature [125, 132, 108, 22, 134, 110]. In [22] the authors propose a method to reconfigure a TCT
array by balancing, or reducing as much as possible, the difference between average irradiance in each
row. In such a solution the irradiance of each module of the array is estimated to test all feasible
configurations in order to determine which of them provide the highest Global Maximum Power
Point (GMPP). Although the algorithm is simple and can be implemented in a micro-controller, the
number of feasible possibilities, when the number of rows in the array is larger than the number of
columns, is very high with respect to the number of possibilities in SP configuration. Moreover, the
solution requires to measure the voltage and current in each module to estimate the irradiance by
using the ideal single-diode model, but it does not provide a method to calculate the parameters of
the modules.
The system proposed in [108, 109] divides the PV array into a fixed and an adaptive parts.
The fixed part is connected in TCT configuration, while the modules of the adaptive part can be
connected in parallel to any row of the fixed array to compensate the current drops produced by the
mismatched modules. In [108] two algorithms to compensate the mismatching conditions in the fixed
part of the array are presented. The first one assumes that the open-circuit voltages of the adaptive
modules are proportional to their irradiance and connects them one-by-one to the mismatched rows
to balance the rows of the array. The second method is similar to the first one but uses the single-
diode model to estimate the PV current of the fixed and adaptive parts to determine the best
configuration. Nevertheless, the capability of mismatching compensation is limited by the number
of adaptive modules, and the authors do not provide a method to determine the model parameters
of the modules. Besides, the first method requires one voltage sensor in each reconfigurable module
and the second method requires the voltage and current measurements of all modules of the array.
The technique proposed in [110] classifies the modules as shaded or unshaded. The shaded mod-
ules are bypassed by switches and the unshaded ones are connected in series to form strings. If one

116
string is not complete it is connected to a step-up converter to reach the voltage level of the complete
strings. The strings are connected in parallel to feed an inverter. This system requires a reduced
number of switches and the algorithm to control the system is simple. However, the mismatching
compensation is reduced and the energy of the mismatched modules is wasted. Moreover, it requires
the measurements of voltage, current and temperature of each module and it is not clear how to
determine if a module is mismatched or not.
In addition, there are different methods to calculate the parameters of a PV module. Some of
them [135, 136] pose a system of non-linear equations and perform some simplifications in order
to obtain a set of explicit functions to calculate the parameters of the single-diode model from the
experimental data of the complete I-V curve of a PV module.
The works reported in [9, 137, 138, 139, 8] perform simplifications to propose a system of non-
linear equations, which are solved in the operating points provided by the manufacturer datasheet
or using experimental data, to obtain the parameters of the model. However, such parameters may
be different from the real values due to the manufacturer tolerances and environmental disturbances
[140].
Other authors [6, 141, 140, 142] use explicit equations to find some parameters (Iph, Isat and
thermal voltage) and use iterative methods to adjust the values of series and parallel resistance to fit
the model with the information taken from the datasheet. Although such methods could be applied
to PV modules operating in real conditions, it is difficult to use them for real-time applications due
to the complexity of the equations that need to be solved in each iteration. Moreover, the previous
solutions are intended to estimate the parameters from data of each module instead of string or
array data.
The solution proposed in [143] considers a method to calculate the reference values of the pa-
rameters by using experimental and datasheet information, then it uses the equations to update
the parameters values depending on the values of the irradiance and cell temperature. Unless this
method provides the parameters for different environmental conditions with explicit equations, it
would require temperature and irradiance sensors in each module, which increases considerably the
implementation costs of the PV system.
Other approach [144] performs a large number of experiments in a wide range of irradiance
and temperature conditions to obtain the I-V curves of a module. For each I-V curve it solves a
system of five non-linear equations to find the values of the single-diode model parameters. Using
such information the authors train an Artificial Neural Network to predict the parameters values
depending on the irradiance and temperature conditions.
In [145] the authors perform a continuous identification of the parameters of a fourth order
polynomial model of the PV module from few pairs of voltage/current measurements. The authors
fit the model equation to the experimental measurements by modifying the coefficients using the

117
recursive least-square method. Then the Newton-Raphson method is adopted to find the MPP from
the model. Nonetheless, such a method is designed for PV arrays in uniform conditions; therefore,
it is not applicable in mismatching conditions.

This Chapter introduces a procedure to detect the SP array configuration that provides the
highest maximum power, in uniform or mismatching conditions, requiring current and voltage mea-
surements of strings only. To evaluate all feasible configurations in order to find the one that provides
the highest GMPP, the ideal single-diode model parameters of the modules in each string are ob-
tained. Despite the TCT configuration has been adopted in literature for reconfigurable PV arrays
[108, 22], it exhibits a larger amount of feasible configurations, in comparison with SP arrays, when
the number of rows (number of modules in each column) is higher than the number of columns
(number of modules in each row). This condition suggests that SP arrays require lower calculation
time to find the optimal configuration in high-voltage applications, where larger strings are required
(e.g. grid-connected PV systems). Therefore, this Chapter is focused on SP configuration, but the
method is also applicable to TCT configuration with some modifications.

5.2 Calculation of PV array current by using inflection points

For this analysis, each PV module of the SP array is represented using the ideal single-diode model,
while bypass and blocking diodes are approximated as ideal switches (as in Section 4.2.2). More-
over, the effects on the characteristic curves of both bypass and blocking diodes are lumped in the
parameters of the PV modules.

To simplify the explanation and reading of this Chapter, the expression for the current of a
module in the i-th position of the j-th string is shown in (5.1) as well as the equations used to
calculate the model parameters (5.2)-(5.5). The definition of the model parameters and constants
used in (5.1)-(5.5) were described in Section 4.2.2.

PV module 1,1 Vbld,1 PV module 1,m Vbld,m

I1,1=Ist,1 Ia
V1,1 V1,m

ISC1,1 ID1,1 V1,1 Ist,1 Ist,m

PV module n,1 PV module n,m Va

Vn,1 Vn,m

Figure 5-1: PV Field with m strings of n modules parallel connected.

118
 Ii,j = Isci,j Ai,j exp (Bi,j Vi,j ) (5.1)
Gi,j
Isci,j = IscST C (1 + Isc (Ti,j TST C )) (5.2)
GST C
BST C
Bi,j = (5.3)
1 + V ov (Ti,j TST C )
ln (1 (Impp/IscST C ))
BST C = (5.4)
V mpp V ocST C
Ai,j = IscST C exp (BST C V oc) (5.5)

This Chapter considers one module (group of 18 cells with one bypass diode) of a BP585 panel
[23] for simulations. The main electrical characteristics extracted from the datasheet are: IscST C =
5.0 A, V ocST C = 11.05 V , ImppST C = 4.72 A, V mppST C = 9.0 V , Isc = 0.065 %/K, and
V oc = 80 mV /K.

5.2.1 Inflection points in mismatching conditions

In SP configuration all strings share the same voltage; hence, the objective of the array model, under
both uniform and mismatching conditions, is to calculate the current of each string individually for
a given array voltage. Then, the current of the PV array is obtained by aggregating the currents of
all strings. Therefore, this Section analyzes a single string, and such a procedure must be performed
for each string of the array. Considering that only one string is analyzed, sub-index j is neglected
in the variables and constants; hence, the sub-index specified correspond to the sub-index i only.
Fig. 5-2(a) illustrates one string of 2 PV modules, where parameters Ai and Bi and temperatures
Ti (i [1, 2]) are considered equal (uniform conditions). In uniform conditions both modules receive
the same irradiance (G1 = G2 ), which implies that Isc1 = Isc2 ; therefore, the voltages and currents
of the two modules are equal and depend on the voltage imposed to the string. The I-V curves of
such modules and the string are shown in upper Fig. 5-3(a). It is noted that the characteristic curve
of the string has the same shape of the curve of one module with the voltage scaled by two due to
the modules are connected in series. In this case exists a single MPP in the power versus voltage
curve (bottom part of Fig. 5-3(a)) due to the uniform operating conditions.
To illustrate the effect of the bypass diodes in mismatching conditions, a shade over the module
2,1 is considered. Hence, the short-circuit currents of the modules are different (Isc1 > Isc2 ). When
the string current (Ist ) is lower than Isc2 all the current flows through both modules (see Fig. 5-
2(b)) and their voltages are different but higher than zero, as shown in the red and blue dots in
the top of Fig. 5-3(b). Therefore, both modules are active (bypass diodes are turned off) because
they are delivering power to the load. When Ist = Isc2 the module 2 becomes inactive since its
voltage and power are near to zero. Moreover the bypass diode 2 is turned on, but its current is

119
equal to zero since all the current is flowing through the module 2. For Ist > Isc2 the difference
between Ist and Isc2 flows through the bypass diode 2 as depicted in Fig. 5-2(c), while the current
of module 2 is Isc2 and its voltage is near to zero; therefore, its power is negligible. The I-V curves
of the two modules and the string are presented in the upper part of Fig. 5-3(b), which illustrate the
inflection point that occurs when module 2 becomes active (Ist = Isc2 ). Such an operating point
is also characterized by a discontinuity in the derivative of the current with respect to the voltage
(inflection voltage). The P-V curves of the modules and the string are presented at the bottom of
Fig. 5-3(b), where two MPPs are produced by the mismatching conditions.

PV module 1,1
Ist
V1

Va=V1+V2
ISC1=ISC2 V 1 =V 2 I1=I2=Ist
PV module 2,1

V2

(a) Uniform conditions.

PV module 1,1
Ist
V1

Va=V1+V2
ISC1>ISC2 V1>VPV2 I1=I2=Ist
PV module 2,1

V2

(b) Mismatching conditions with Ist < Isc2 .

PV module 1,1

Ist
V1

V a =V 1
V2=0
ISC1>ISC2 I1=Ist
I2=ISC2
PV module 2,1

V2
ISC2 Ist - ISC2

(c) Mismatching conditions with Ist Isc2 .

Figure 5-2: String of two modules under uniform and mismatching conditions.

In general, the model considers PV strings with n modules organized in decreasing order of short-

120
 6

Current [A]
4

0
0 5 10 15 20

Mod. 1,1
100
Mod. 2,1
Power [W]

Str. 2x1
50
Maximum
Power Point

0
0 5 10 15 20
Voltage [V]
(a) Uniform conditions.
6
Mod. 2,1 active
Current [A]

2 Inflection voltage
Mod. 2,1 inactive
0
0 5 10 15 20
Maximum
Mod. 1,1 Power Points
Mod. 2,1
60
Str. 2x1
Power [W]

40

20 Inflection voltage
0
0 5 10 15 20
Voltage [V]
(b) Mismatching conditions.

Figure 5-3: Characteristic curves of 2x1 PV array. Blue lines: module 1,1. Dashed-red lines: module
2,1. Purple lines: 2x1 string.

121
circuit currents (Isck1,j > Isck,j ). That consideration does not introduce calculation errors since
the PV modules are connected in series. The inflection voltage (V ok,j , k [2, n]) is the minimum
voltage of the j-th string that activates the k-th PV module (or deactivates the k-th bypass diode).
Such a condition is fulfilled when the current of the string is equal to the short-circuit current of the
k-th module (Ist,j = Isck,j ). When the string is operating at an inflection voltage, the associated
bypass diode is active: the modules having a current higher than Isck,j (from i = 1 to i = k 1)
are active, and the remaining modules (from i = k to i = n) are inactive. Hence, to calculate an
inflection voltage it is necessary to add the voltages of all the active elements in the string with
Ii,j = Ist,j = Isck,j as shown in (5.6). The derivation of such expressions is given in [12].

k1
X
V ok,j = V xm,k (5.6)
m=1
 
1 Iscm,j Isck,j + Ak,j
V xm,k = ln
Bm,j Am,j

5.2.2 String and array current calculation

Fig. 5-4 shows an example of the inflection points calculated for one string composed by four
PV modules with different irradiance conditions, where three inflection points appear in the curve.
Those points, along with the short-circuit current and open-circuit voltage, can be used to determine
the number of active modules (nac ) for a given string voltage. For example, between the short-
circuit current and the first inflection point of Fig. 5-4 only one module in the string is active,
this because Ist is lower than Isc1 and higher than the short-circuit current of the other three
modules. Similarly, between the first and second inflections points there are two active modules
since Isc1 > Isc2 > Ist > Isc3 > Isc4 , and so on. Finally, when Isc4 > Ist all PV modules are
active. Such information is used to construct the system of nac non-linear equations given in (5.7) to
calculate the voltages of all active modules for a given voltage of the j-th string, which corresponds
to the array voltage (Va ). Once the nac voltages are calculated the string current can be obtained
by using (5.1) for any active module. The non-linear equation system (5.7) can be solved by using
the improved Newton-Raphson method presented in [12].

nac
X
Va = Vi,j
i=1
I1,j = Ik,j , k [2, nac ] (5.7)

The PV array model described in this section improves the calculation time of the PV array
since the non-linear equation system to be solved has only one non-linear equation for each active

122
 1 active 2 active 3 active 4 active
module modules modules modules
5

String current [A]

3 Shortcircuit
point

2
Inflection
points
1
Opencircuit
point
0
0 5 10 15 20 25 30 35 40
String voltage [V]

Figure 5-4: Inflection points effect on PV strings.

module. Moreover, the calculation of the inflection points also restricts the search range of the
numeric method between the two inflection points at both sides of a given string voltage. These
advantages are useful to perform an evaluation of the I-V curves of different possible configurations
of the PV array and to determine which of them provides the highest GMPP.

5.3 Detection of inflection and open-circuit points from PV

string data

Since the bypass diodes operation defines the number of PV modules active for a given string voltage,
as in Fig. 5-4, the short-circuit, open-circuit and inflection points have to be detected to identify
the modules that contribute to the string voltage in each voltage interval between two consecutive
points of interest (short-circuit point, inflection points and open-circuit point).
The short-circuit point (SCP) is easily detected since it exhibits a PV string voltage equals to
zero, as given in (5.8), where Va (i) and Ist (i) are Ndata experimental measurements of the array
voltage and sting current, respectively. In contrast, the open-circuit point (OCP) corresponds to
the minimum array voltage that reduces the string current to zero (5.9).

ISCP = {max(Ist (i)) : Va (i) = 0, i = 1...Ndata } (5.8)

VOCP = {min(Va (i)) : Ist (i) = 0, i = 1...Ndata } (5.9)

An Inflection Point (INP) is characterized by a large change in the derivative of the cur-
rent/voltage function, as observed in Fig. 5-4. Such a concept is used to detect the INPs from
experimental current/voltage measurements of a PV string: when the magnitude of the derivative

123
(5.10) at the left side of a given voltage is higher than the magnitude of the derivative at the right
side of it, then such a voltage defines an INP. Moreover, in order to find all INPs in the string it
is necessary to scan the entire I-V curve. When a new INP is found, the voltage range to find the
next one is constrained from the voltage of the last INP to VOCP . Such a condition is based on the
fact that each PV module exhibits a single maximum power between two INPs, which permits to
constraint the possible voltage solutions and to calculate the maximum number of INPs to be found
as given in (5.11), where n represents the number of PV modules in the string.

d Ist Ist (i) Ist (i 1)

= (5.10)
d Vst Vst (i) Vst (i 1)
NIN P max = n 1 (5.11)

The algorithm proposed to detect the Interesting Points (IP), which are the SCP, INPs and
OCP, were tested considering the string simulated in Fig. 5-4: since such a string is composed by
four PV modules with mismatching conditions, all the modules exhibit different I-V curves. In such
an example, the SCP is the first IP (IP1 ), the three INPs are IP2 , IP3 and IP4 , and the OCP is
the last IP (IP5 ).

0.2 Inflection points

0
String current derivative [A/V]

0.2

0.4

0.6

0.8

1
Opencircuit
point
1.2

1.4

0 5 10 15 20 25 30 35 40
String voltage [V]

Figure 5-5: Inflection points detection.

Fig. 5-5 shows the current/voltage derivative of the I-V curve of Fig. 5-4 and the voltages where
the IPs were identified. It is clear that INPs and OCP voltages where accurately detected, while
SCP exhibits zero volts. Then, the currents for the IPs are extracted from the current/voltage
measurements, e.g. Fig. 5-4.

124
5.4 Estimation of PV modules parameters

Based on the INP concept it is possible to obtain part of the current and voltage data sets (I-
V curves) of all PV modules in the string and, by using such data sets, to calculate the model
parameters of all the modules. To obtain I-V curves of all PV modules it is necessary to analyze the
string I-V curve from the first to the last IP. Moreover, the number of IPs (NIP ) is given by (5.12),
where NIN P is the number of INPs and the other two points correspond to the SCP and OCP. It is
important to note that NIN P depends on the number of modules operating in different conditions
in the string and its maximum value is given in (5.11), which is obtained when all modules have
different operating conditions.

NIP = NIN P + 2 (5.12)

Between IP1 (SCP) and IP2 (INP1) there could be active one or more PV modules (NamIP,1,2 )
with the same highest short-circuit current and the same I-V characteristic. The voltages set of that
interval is defined as VIP,1,2 . To identify the parameters of the NamIP,1,2 modules the data sets
of currents (Iident,1 ) and voltages (Vident,1 ) of one module only (as discussed in Sections 5.4.3 and
5.4.4) are required, since such NamIP,1,2 modules have the same I-V characteristics. To obtain such
information the first step is to determine the set of voltages generated by the NamIP,1,2 modules,
defined as VM IP,1,2 . Between IP1 and IP2 the condition VM IP,1,2 = VIP,1,2 holds since there is not
active any other module.
The number of modules NamIP,1,2 in this first interval [IP1, IP2] is found by dividing the voltage
span generated in such an interval (max(VM IP,1,2 ) = VM IP,1,2 (Nd,1,2 )) by V ocST C and approximat-
ing the result to the nearest higher integer, where Nd,1,2 is the number of data in VM IP,1,2 . Finally,
the voltage set Vident,1 is obtained by dividing each element of VM IP,1,2 by NamIP,1,2 , where Vident,1
corresponds to the voltage generated by one of the NamIP,1,2 modules. The set Iident,1 is composed
by the current points between IP1 and IP2 since all modules share the same string current.
For the second interval, between IP2 and IP3, there are NamIP,1,2 +NamIP,2,3 active PV modules.
Again, it is required to obtain the voltage set exclusively produced by one of the NamIP,2,3 modules
in order to estimate its parameters.
From the data sets Iident,1 and Vident,1 it is possible to estimate the parameters of the NamIP,1,2
modules active in the first interval (as will be explained in the Sections 5.4.3 and 5.4.4). Then, such
parameters are used to calculate the voltage contribution of the NamIP,1,2 modules to the string
voltage in the second interval VIP,2,3 , i.e. between IP2 and IP3. Such data is used to obtain the
set of voltages VM IP,2,3 exclusively produced by the NamIP,2,3 modules that become active in the
second interval. Then, similarly than in the first interval, NamIP,2,3 is found by dividing the voltage

125
span generated in the second interval (VM IP,2,3 (Nd,2,3 )) by V ocST C and approximating the result to
the nearest higher integer. Finally, the voltage set Vident,2 is obtained by dividing each element of
VM IP,2,3 by NamIP,2,3 , where Vident,2 corresponds to the voltage generated by one of the NamIP,2,3
modules, and Iident,2 is composed by the currents between IP2 and IP3.

In general, for a given string, the sets of voltages between two consecutive IPs (IPk and IPk+1)
is defined in (5.13), where k [1, NIP 1] and the vector VIP contains the voltages of the IPs.
Moreover, the set of voltages produced exclusively by the NamIP,k,k+1 active modules between IPk
and IPk+1 (VM IP,k,k+1 ) is given by (5.14), where k [2, NIP 1], Nam,k is the total number of
active modules up to NamIP,k,k+1 (non-inclusive), and Nd,k,k+1 is the number of data in VIP,k,k+1 .
Nam,k is calculated by using (5.15) where NamIP,n,n+1 is obtained by (5.16) with k [1, NIP 1].
The set of voltages used to identify the parameters of one of the modules active between IPk and
IPk+1 is obtained by dividing each element of (5.14) by (5.16) as shown in (5.17). Finally the set
of currents used for the parameters estimation is given in (5.18), where IIP is a vector with the
currents of the inflections points.

VIP,k,k+1 = {Va (i) : VIP (k) < Va (i) < VIP (k + 1), i = 1 Ndata } (5.13)

N am,k   
X 1 Iscn Ist (i)
VM IP,k,k+1 = VIP,k,k+1 (i) ln , i = 1 Nd,k,k+1 (5.14)

n=1
Bn An
k1
X
Nam,k = NamIP,n,n+1 (5.15)
n=1
 
VM IP,k,k+1 (Nd,k,k+1 )
NamIP,k,k+1 = ceil (5.16)
V ocST C
 
VM IP,k,k+1 (i)
Vident,k = , i = 1 Nd,k,k+1 (5.17)
NamIP,k,k+1
Iident,k = {Ist (i) : IIP (k) < Ist (i) < IIP (k + 1), i = 1 Ndata } (5.18)

Flow chart of the Fig. 5-6 illustrates the proposed process to obtain the IPs and the model
parameters of each PV module in a particular string. The algorithm is presented in two parts: the
left one illustrates the extraction of the IPs, while the right part presents the iterative process used
to obtain the estimation data sets and the model parameters.

From an implementation point of view, it is necessary to detect which physical module is active at
any moment during the voltage scan of the string, this in order to know to which module or modules
correspond the estimated parameters between two consecutive IPs. Such a procedure can be easily
performed by using binary signals to detect if a given module is active or inactive (bypassed).
Considering a microcontroller-based implementation, a digital input signal can be attached to each
PV module to detect if its voltage is different from zero. This implementation avoids the requirement

126

 
  6  7 


 
      
 
   
 
       
     "
     
   


        
          

  
  
  
 

'  "       + ' 3 3 (        +

      
   
, '  " ! '  " *
 3        

*
 '  "       - , ' 3 3 (        1


  
    
 * ' 3 3 (        -
 
       .


           
       /  
3       .

 
3       /

! " # ! 

  
 

      0


  0   *
 '  " 
  
 

      0

          
 1 +   
  0   *
 ' 3 3 ( 

*  1 1           
 1 +  
$   
 $%
! (  1 1
$   
 & $

2   
 

     

      *
 '  " & 2   
 

     
5
            *
 ' 3 3 (  &
'  # !

     
'  # !  
3 ! "
# ! #(   ! (
3 ! 3(

* "
*
) * 


3 ) * ' &

5
5
* ' ! #& 

4


Figure 5-6: Flow chart of the parameters estimation method.

of Analog-to-Digital converters (ADC) for each PV module; instead, ADCs are required for the
current of each string and the array voltage only (all the strings are connected in parallel), while
digital inputs are used to detect the PV modules activation. Moreover, taking into account that
micro-controllers and DSPs provide large amount of digital inputs, such a solution is applicable to
medium of even large PV arrays by interconnecting several micro-controllers or DSPs.

5.4.1 Model parameters and data sets

The model adopted corresponds to the explicit single-diode approximation given in (5.1), which has
been extensively used for modeling PV systems and design of MPP tracking techniques [12, 13, 146,
147]. To simplify the nomenclature, expression (5.1) is rewritten as in (5.19) for a single module.

I = Isc A exp (B V ) (5.19)

127
 Such a model parameters (Isc, A, and B) are valid for a given irradiance and temperature condi-
tions as given in (5.2)-(5.5) [12]. But, taking into account that the estimation process is performed
online to detect the array configuration that provides higher power in the present environmental
conditions, such a model is accurate enough since it provides a realistic calculation of the GMPP for
each possible configuration meanwhile the irradiance and temperature do not change significantly.
When the environmental conditions change, the module parameters must be estimated again.

The parameters of the explicit model must be calculated to accurately reproduce the PV module
current/voltage characteristic. But taking into account that (5.19) is parameterized by three con-
stants, only three points are used to estimate A, B and Isc. Moreover, to guarantee a satisfactory
fitting of the estimated model with the measurements, the short-circuit point of the module SCPm
(Vident,k (1), Iident,k (1)) and the maximum power point of the module MPPm are used. In such a
way, the SCPm and MPPm can be extracted from Vident,k and Iident,k as given in Fig. 5-7, where
SCPm and MPPm are characterized by (0 (V), I0 (A) and (V mpp (V), Impp (A)), respectively.

In addition, to obtain an accurate fitting of the module open-circuit voltage it is required to

calculate the value of the module voltage that occurs at 0 A. But such an information is, in general,
not detectable from Vident,k and Iident,k since the activation of the next PV module in the sequence
(descendant irradiance) occurs at a current higher than 0 A. Therefore, the closest measurements
available are at the last values of the data sets Vident,k and Iident,k : such a point is named VIend,m
and corresponds to (Vident,k (Nd,k,k+1 ) [V], Iident,k (Nd,k,k+1 ) [A]).

I0
Impp
MPPm
Module Current (A)

Module Power (W)

SCPm

IIP
VIend,m

0 [V] Vmpp VIP

Module Voltage (V)

Figure 5-7: Points used for parameter estimation.

Finally, the estimation of the PV model parameters is performed using three points only: SCPm ,
MPPm and VIend,m . Such a condition reduces the estimation time in comparison with estimation
processes that require large amount of data, this because the number of terms in the function that
need to be evaluated in each iteration of the optimization method, used to find the parameters, is
proportional to the number of points that the function must be fitted [148].

128
5.4.2 Impact of the model parameters error

From the model current/voltage relation (5.19), the calculation errors in the module current gener-
ated by errors on parameters A, B and Isc are given in (5.20), (5.21) and (5.22), respectively, where
I represents the error on I and A, B and Isc represent the error on the parameters.

I I
I = A , = exp (B V ) (5.20)
A A
I I
I = B , = A V exp (B V ) (5.21)
B B
I I
I = Isc , =1 (5.22)
Isc Isc

Without loss of generality, such relations will be illustrated using one module of a BP585 panel
[23], whose nominal parameters are extracted from the electrical information in the manufacturer
datasheet (presented in Section 5.2): A0 = 0.8941 A, B0 = 1.4060 V1 and Isc0 = 5.0 A. From
the model parameters description (5.2)-(5.5) it is noted that the range of variation of Isc is similar
to the range of I, and A depends exponentially from BST C while B depends linearly from BST C .
Therefore, the magnitude of A is strongly smaller in comparison with B, which is illustrated by
A0 and B0 . From such considerations it is concluded that errors on A (5.20) have a much smaller
impact on I than errors on both B and Isc due to the very small magnitude of A. In addition,
errors caused by B are larger at higher PV voltages, while errors caused by Isc are constant for PV
voltage variations.

6 6 6
Current (A)

Current (A)

Current (A)

4 A0 4 B0 4
Isc0
10% error 10% error 10% error
2 10% error 2 10% error 2
10% error
MPP MPP MPP
0 0 0
0 5 10 0 5 10 0 5 10
Voltage (V) Voltage (V) Voltage (V)

50 50 50
40 40 40
Power (W)

Power (W)

Power (W)

30 30 30
20 20 20
10 10 10
0 0 0
0 5 10 0 5 10 0 5 10
Voltage (V) Voltage (V) Voltage (V)

Figure 5-8: Effect of parameters errors on the PV module current.

Fig. 5-8 shows the errors introduced in I by 10% errors on the model parameters: as previously

129
predicted, errors on A have less impact on the module current calculation than errors on both B
and Isc. In this way, the MPP prediction of the PV module is not significantly affected by errors
on A, while errors on B generate large errors on the MPP voltage and power estimation. Similarly,
errors on Isc generate large errors on the MPP power prediction with small MPP voltage errors.
Such a behavior is illustrated in the following example: if a variation of 20% is introduced to A0
the errors in the Impp prediction are 1.38%, while if the same variation is forced in B0 and ISC0
(separately) the errors in the Impp prediction are 17.64% and 21.38%, respectively.
Therefore, due to the small sensitivity of I to changes on A in comparison with changes on B
and ISC , larger estimation errors on A are expected in comparison with the estimation errors on
B and ISC . Such a condition is due to the estimation algorithm can fit the estimation data with a
larger range of values for A in contrast with the ranges of values for B and ISC .

5.4.3 Parameters estimation using classical estimation algorithms

In literature is proposed to extract one or more model parameters of a PV module by fitting a

non-linear equation with experimental data [149]. In this Section the parameters of the PV model
are estimated by fitting the three selected points from each module I-V curve (SCPm , MPPm and
VIend,m ) with the data generated by means of the current/voltage equation (5.19). Moreover, the
Nonlinear Least Square method is adopted to estimate the parameters.
In addition to the three points selected from the I-V curve, the method requires some initial values
and constraints for the parameters, as well as the current/voltage equation. The initial values for
A and B are taken from the manufacturer datasheet information, while the Isc value is taken from
the registered data of the current/voltage curve at zero volts. The search spaces for the parameter
values are defined with upper and lower limits for each parameter. The A and B parameters are
constrained between the 10 % and 200 % of the nominal values to account for the changes caused
by the temperature, while Isc is constrained between 98 % and 102 % of the initial value since A is
small (micro-Amperes range): for V = 0 V , the relation I = Isc A Isc holds.
The estimation procedure is described in Fig. 5-9. First, the method starts with the initial values
of the parameters. Then, it produces the fitted curve for the current set of parameters. In the third
step, the method adjusts the parameters and determines whether the fit improves. The direction and
magnitude of the adjustment is made by a trust-region algorithm, which is based on the trust-region
concept described in [150]: the set of parameters are evaluated in the model, and if the squared error
decreases within the trust region, then the region is expanded; otherwise the region is contracted.
The trust-region algorithm was selected since lower and upper parameter boundaries are available.
In addition, the convergence criteria are: a maximum squared error of the optimization function
 2
I I 106 , a maximum number of fit iterations (400 iterations), and a maximum number of
 

model evaluations (600 evaluations).

130
 Finally, the property of requiring only three points per module instead of a large amount of
points has been tested: estimating the parameters of four modules using 12 points requires 29 % of
the time used to estimate such parameters from 652 points, which is the number of samples taken
with steps of 0.1 V for the PV string with V ocP = 65.2 V . Such a condition put in evidence the
reduced amount of calculations avoided by using only three data points, obtaining the same results.

vpv Photovoltaic ipv

module
+ Least-square error
_
Nonlinear least squares
PV module method - Trust region
model
pv algorithm

Set of estimated parameters

Set of initial
parameters Input data
Iteration process data

Figure 5-9: Parameters estimation procedure from data.

5.4.4 Parameters estimation using approximated analytical expressions

The parameters of the PV modules can be also calculated explicitly assuming the same small simpli-
fication in the SCPm point used for the Isc initial condition: evaluating (5.19) in SCPm (V = 0 V ),
(5.23) is obtained. But as described in Section 5.4.2, the parameter A is in the range of micro-
Amperes; therefore, the parameter Isc can be approximated by (5.24) since A is much smaller than
Isc, which is in the range of Amperes.

I(0) = Isc A (5.23)

Isc I(0) (5.24)

The other parameters (A and B) are calculated considering the MPPm and VIend,m points. In
MPPm and VIend,m the current of the module is given by (5.25) and (5.26), respectively. Replacing
I1 = Isc Impp and I2 = Isc IV Iend in (5.25) and (5.26), respectively, it is simple to obtain
(5.27) and (5.28). Such expressions, (5.27) and (5.28), form a system of two equations with two
unknowns, from which B and A are obtained as explicit equations (5.29) and (5.30) with V =
VV Iend V mpp.

Impp = Isc A exp (B V mpp) (5.25)

IV Iend = Isc A exp (B VV Iend ) (5.26)

131
 I1 = Isc Impp = A exp (B V mpp) (5.27)

I2 = Isc IV Iend = A exp (B VIV end ) (5.28)

 
1 I2
B= ln (5.29)
V I1
A = I1 exp (B V mpp) (5.30)

According to (5.5), the parameter A depends exponentially with V oc voltage, which in turn
depends on the temperature of the module; hence the parameter A has an exponential relation with
the temperature. Such a relation suggest that the parameters calculation using (5.24), (5.29) and
(5.30) should be used for applications where the maximum ambient temperature is far from the
maximum temperature supported by the PV modules. Such a condition is required since the errors
introduced by the approximation (5.24) increases as the temperature rises. In cases where high tem-
peratures exist, the parameters estimation based on classical estimation algorithms (Section 5.4.3)
may reduce the estimation errors, this because such a solution does not consider any simplification
in the evaluation of equation (5.19) since the approximation (5.24) is only used to set an initial value
for Isc and the estimation algorithm tracks the real value of Isc.
In general the parameters estimation using classical estimation algorithms can be applied for any
temperature range; nevertheless the calculation using the approximated explicit equations is faster
and requires less resources, but it has a restriction in the temperature range that guarantees the
fulfillment of (5.24).

5.4.5 Application example: PV string parameters estimation

A PV string composed by six PV modules (PV1, PV2, PV3, PV4, PV5 and PV6) was considered
to illustrate the proposed solution. The adopted PV modules are half of a BP585 panel, but to take
into account the differences of the temperature and the tolerances in the PV modules, the Ai and Bi
parameters have been generated randomly in the range of 20 % with respect to the nominal values
A0 and B0 presented in Section 5.4.2. Moreover, very different short-circuit current parameters have
been used to represent a deep mismatching condition in the PV string. The PV modules parameters
are given in the second column Table 5.1.
Using the PV modules parameters, the precise current/voltage characteristic of the PV string
was generated, measuring the string voltage and current within the range [0 V, 65.2 V] in steps of
0.1 V, which simulates the online measurements of the identification system. From such a procedure
652 voltage/current data points were obtained, i.e. Va and Ist . The PV string measurements are
depicted in Fig. 5-10 (blue traces), where the string power/voltage characteristic was constructed
from Va and Ist .
From the measured data, the PV string IPs (SCP, INPs and OCP) were detected using the

132
 Table 5.1: PV string parameters
Parameter Value Estimated value NSE (%)
A1 (A) 0.8683 0.8175 0.3420
B1 (V1 ) 1.3238 1.3191 0.0013
Isc1 (A) 4.7000 4.7000 0.0000
A2 (A) 0.9736 1.2082 5.8043
B2 (V1 ) 1.3501 1.3294 0.0234
Isc2 (A) 4.2000 4.1932 0.0003
A3 (A) 0.9623 0.9516 0.0124
B3 (V1 ) 1.3978 1.4092 0.0066
Isc3 (A) 3.1000 3.0991 0.0000
A4 (A) 0.9040 0.8491 0.3690
B4 (V1 ) 1.3302 1.3321 0.0002
Isc4 (A) 2.3000 2.2985 0.0000
A5 (A) 0.9169 0.9469 0.1071
B5 (V1 ) 1.5028 1.5078 0.0011
Isc5 (A) 1.5000 1.4999 0.0000
A6 (A) 0.9106 0.8859 0.0737
B6 (V1 ) 1.3202 1.3164 0.0008
Isc6 (A) 1.4000 1.4025 0.0003

procedure described in Section 5.3, obtaining the results presented in Fig. 5-10 (black circles),
which illustrate the high accuracy of the IP detection algorithm.
Then, the voltage and current data sets related to each PV module, Vident,i and Iident,i , were
obtained from the string measurements and IPs following the procedure previously described in this
Section. Finally, from Vident,i and Iident,i , the estimation data sets [SCPm , MPPm , VIend,m ] were
constructed for each PV module.

6 Measurements
String current [A]

Estimated model
4 IP

2
NSSE = 2.874x103%
0
0 10 20 30 40 50 60 70
String voltage [V]
100
String power [W]

50

NSSE = 9.343x103 %
0
0 10 20 30 40 50 60 70
String voltage [V]

Figure 5-10: PV string measurements and estimated model simulation.

The PV modules parameters were estimated using the estimation algorithm of Section 5.4.3
and the values obtained are presented in the third column of Table 5.1. Such estimated values are

133
close to the reference values, which illustrates the accuracy of the proposed solution. To provide a
quantitative measurement of the estimation accuracy, the Normalized Squared Error (NSE) given in
(5.31) was used, where y and ye represent the reference and estimated values, respectively. The NSE
was calculated for each parameter of all the PV modules, obtaining the fourth column of Table 5.1.
Again, such results put in evidence the satisfactory accuracy of the proposed estimation procedures.

2
(y ye )
N SE [%] = 100 (5.31)
y2

To test the estimated parameters, the current/voltage characteristic of the complete PV string
was reproduced, obtaining the results presented in red traces of Fig. 5-10, which illustrates the
usefulness of the estimated models. To provide a quantitative measurement of the current curve
reproduction, the NSSE (explained in Section 4.2.6) was calculated. The low NSSE value (2.874
103 %) provided by the estimated model guarantees a correct reproduction of the current curve;
therefore the power curve of the PV string can be accurately reproduced (NSSE= 9.343 103 %)
to predict the GMPP.
The predictability of the GMPP makes possible to evaluate the maximum power available in
different strings constructed using the estimated PV modules. To illustrate such a concept, the
power curves of two new strings formed by three of the six previously estimated PV modules have
been calculated: the first string was constructed with PV1, PV3 and PV5, while the second string
was formed by PV2, PV4 and PV6. Fig. 5-11 shows the estimation of the P-V curves of the
new strings using the estimated model (red traces), which are in agreement with the measurements
taken from data generated with the exact string parameters (blue traces), i.e. second column of
Table 5.1. The low NSSE values obtained, 4.483 102 % for the first string (PV1, PV3 and PV5)
and 2.902 104 % for the second string (PV2, PV4 and PV6), put in evidence the usefulness of
the proposed estimation procedure in the test of PV strings with different configuration from the
one used to estimate the modules parameters.
Finally, the estimated models were used to reproduce the electrical characteristics of all the PV
modules, obtaining the results reported in Fig. 5-12, where the satisfactory behavior of the estimated
models and low NSSE are observed. Such precision verifies that the estimated models are accurate
representations of the reference PV modules.

5.5 PV array reconfiguration

The estimated model parameters are intended to detect the PV array configuration that provides
the highest power. Such a procedure is performed using the array model presented in Section 5.2,
which requires a reduced number of calculations in comparison with classical models, providing a

134
 60

String power [W]

40

20
String of PV1PV3PV5:
NSSE = 4.483x102%
0
0 5 10 15 20 25 30 35
String voltage [V]

50 Measurements

String power [W]

Estimated model
40
30
20
10 String of PV2PV4PV6:
NSSE = 2.902x104%
0
0 5 10 15 20 25 30 35
String voltage [V]

Figure 5-11: Estimation of the new strings power characteristics.

5 5 5

Module current [A]

Module current [A]

Module current [A]

4 4 4
3 3 3
2 2 2
1 1 1
NSSE = 7.105x102% NSSE = 3.779x104% NSSE = 3.779x104%
0 0 0
0 5 10 0 5 10 0 5 10
Module voltage [V] Module voltage [V] Module voltage [V]
PV Module characteristic
Estimated model
5 5 5
NSSE = 2.910x106% NSSE = 1.995x104% NSSE = 5.490x105%
Module current [A]

Module current [A]

Module current [A]

4 4 4
3 3 3
2 2 2
1 1 1
0 0 0
0 5 10 0 5 10 0 5 10
Module voltage [V] Module voltage [V] Module voltage [V]

Figure 5-12: PV modules characteristics and estimated models.

fast calculation of the array current [12].

To detect the PV array configuration with highest power, the GMPP of the array configurations
must be compared following a search algorithm. Despite multiple optimization algorithms can be
adopted, i.e. [22, 108, 125, 109], the proposed solution assesses all possible SP configurations of
the array (exhaustive search) that may provide different GMPP values. Such an exhaustive search
algorithm has been successfully tested in reconfiguration of TCT arrays [22].
The number of possible array configurations with potentially different power in an SP structure

135
is given in (5.32) where n is the number of modules in each string and m is the number of strings. In
an equivalent TCT structure, with n rows of m parallel connected modules, the number of feasible
possibilities is calculated using (5.33) [22]. Therefore, for high-voltage requirements, where more PV
modules must be connected in series (more rows than strings), the SP structure exhibits a reduced
number of configurations to test. Such a condition makes SP arrays reconfiguration faster for grid
connected applications in contrast with TCT structures. In any case, the solution proposed in this
Chapter can be extended to arrays in TCT configuration with some modifications in the model and
in the position of sensors connected to the array.

(n m)!
N RSP = m (5.32)
(n!) m!
(n m)!
N RT CT = n (5.33)
(m!) n!

To illustrate the SP array possible configurations, an array composed by two strings of three PV
modules each (3 2) is considered, it exhibiting 10 possible configurations that provide different
power. Such possible configurations are shown in Table 5.2, where the modules connection in the
strings are given as columns: in example, the first configuration has the first string composed by
PV1, PV2 and PV3, while the second string is composed by PV4, PV5 and PV6.

Table 5.2: Possible configurations of a 32 PV array

1 2 3 4 5 6 7 8 9 10
14 13 13 13 12 12 12 12 12 12
kk kk kk kk kk kk kk kk kk kk
25 25 24 24 35 34 34 43 43 53
kk kk kk kk kk kk kk kk kk kk
36 46 56 65 46 56 65 56 65 64

5.5.1 Reconfiguration system

To automatically reconfigure the PV array several switches (switches matrix) are required to change
the strings connections. The number of switches required by SP connections following the structure
presented in the Fig. 5-13 is given by (5.34), while the number of switches requiered by the TCT
structure presented in [22] is given by (5.35).

(m 2) (m 1)
N SwSP = 2 n (m 1) + (n 1)
2
(n m n + 1) (n m n + 2)
+ + (n 2) (n m n + 1) (5.34)
2
N SwT CT = n (n + 1) 2 + 2 m (n m n) (5.35)

136
 Ist1 Ist2

+
1 4 1 2 4 1 2 3 4

S2U S3U S4U S5U S6U

Va

PV1 PV2 PV3 PV4 PV5 PV6

S3D S4D S5D -
1 1 1

Figure 5-13: Switches matrix scheme for the 3 2 PV array reconfiguration.

The switches matrix to reconfigure the adopted 3 2 PV array, depicted in Fig. 5-13, requires
20 single-position switches. In the connection circuit, the upper switches (S2U , S3U , S4U , S5U and
S6U ) are composed by 2, 3 or 4 single-position switches, while the lower switches (S3D , S4D and
S5D ) are single-position switches. Therefore, upper switches are always connected to one terminal,
i.e. one single-position switch connected at time (1, 2, 3 or 4), while the lower switches could be
connected (1) or disconnected (0) only. The positions of the switches (switches states) to obtain the
configurations of Table 5.2 are described in Table 5.3.
In addition, the switches matrix of Fig. 5-13 considers current sensors for each string (Ist,1 and
Ist,2 ). Such sensors are used to obtain measurements of the I-V characteristics of each string of the
array, which are required to estimate the PV modules parameters following the procedures described
in Sections 5.3 and 5.4.

Table 5.3: Switches matrix states for the 3 2 PV array configurations.

Configuration [S2U ] [S3U , S3D ] [S4U , S4D ] [S5U , S5D ] [S6U ]
CF1 [1] [2 , 1] [4 , 0] [4 , 0] [4]
CF2 [1] [4 , 0] [2 , 1] [3 , 0] [4]
CF3 [1] [4 , 0] [3 , 0] [2 , 1] [3]
CF4 [1] [4 , 0] [3 , 0] [4 , 1] [1]
CF5 [4] [1 , 0] [3 , 1] [2 , 0] [4]
CF6 [4] [1 , 0] [2 , 0] [3 , 1] [3]
CF7 [4] [1 , 0] [2 , 0] [4 , 1] [2]
CF8 [4] [2 , 0] [1 , 0] [4 , 1] [2]
CF9 [4] [2 , 0] [1 , 0] [3 , 1] [3]
CF10 [4] [2 , 0] [3 , 1] [1 , 0] [4]

5.5.2 Experimental measurements

To validate the proposed approach, an application example based on experimental measurements

from ERDM85 PV panels [14] is presented. The main datasheet information of such modules are:
IscST C = 5.13 A, V ocST C = 21.78 V , ImppST C = 4.8 A, V mppST C = 17.95 V , Isc = 0.020 %/ K,

137
V oc = 0.34 %/ K and N s = 36. The experimental system 1 presented in Section 1.6 (see Fig.
1-7(a)) was used to collect measurements from PV modules at different operation conditions. Such
real modules characteristics were used to experimentally test the proposed identification procedure.
The experiments were carried out during a clearly sky day in the south-west region of Colombia.
During the experiments the approximated values of the ambient temperature and irradiance were
24 C and 580 W/m2 , respectively.
The experimental measurements were used to construct the current/voltage characteristics of
the 3 2 SP array, for the 10 configurations of Table 5.2. Such a procedure was performed by
implementing the switches matrix of Fig. 5-13 in Matlab, controlling the switches states as described
in Table 5.3.
Without loss of generality, this example considers the switches matrix operating in the first
configuration of Table 5.2, named CF1, in which the current/voltage measurements of the two
strings were generated from the experimental modules characteristics. Such strings measurements
were used to estimate the PV modules parameters following the procedures described in the flow
chart of Fig. 5-6. Fig. 5-14 shows both experimental and estimated power/voltage characteristics
of the first and second strings, where the estimated model provides satisfactory results (NSSE =
0.1078 % and NSSE = 0.0951 %, respectively).

Experimental data
80 Estimated model
First string power [W]

60

40

20
String of: PV1PV2PV3
0
0 10 20 30 40 50 60
PV array voltage [V]
Second string power [W]

80

60

40

20
String of: PV4PV5PV6
0
0 10 20 30 40 50 60
PV array voltage [V]

Figure 5-14: Experimental and estimated strings powear curves.

The estimated model parameters were used to reproduce the experimental PV modules charac-
teristics, obtaining the satisfactory results reported in Fig. 5-15, where the three I-V curves in the
upper and the three I-V curves in the bottom of the figure correspond to the first and second strings,
respectively. Such calculations, characterized by low NSSE values, put in evidence the accurate pre-
diction of the real modules electrical behavior. But it must be pointed out that, in the proposed
procedure, the estimation error is propagated across the estimated modules in the same string, since
the estimated model of the first module (in a string) is used to generate the estimation data for the

138
second module; hence, the errors on the first module parameters increase the estimation error on the
second module. In general, the estimation errors increase with the number of modules previously
estimated for the same string. But, since the proposed procedure estimates the PV modules in
descendent order of irradiance, the modules parameters with higher errors have less contribution to
the PV array power estimation.
The estimated parameters were used to test the 10 possible configurations of the 3 2 PV
array, obtaining the results presented in Table 5.4, where the experimental and estimated GMPP
are reported. From such results it is demonstrated that the estimated models accurately predict the
array configuration that provides the highest power production. Moreover, Table 5.4 also reports the
NSSE for each configuration, which put in evidence the satisfactory results of the proposed solution.
Finally, to illustrate the estimation accuracy, Fig. 5-16 shows the comparison of the experimental
and estimated electrical characteristics of the array configurations that produces the highest (CF2)
and the lowest (CF9) GMPP.

3 3 3
Module current [A]

Module current [A]

Module current [A]

2 2 2

1 1 1
PV1 PV2 PV3
NSSE = 0.0577 % NSSE = 0.0897 % NSSE = 2.2183 %
0 0 0
0 10 20 0 10 20 0 10 20
Module voltage [V] Module voltage [V] Module voltage [V]
Experimental data
3 3 Estimated models 3
Module current [A]

Module current [A]

Module current [A]

2 2 2

1 1 1
PV4 PV5 PV6
NSSE = 0.0501 % NSSE = 0.0952 % NSSE = 0.4333 %
0 0 0
0 10 20 0 10 20 0 10 20
Module voltage [V] Module voltage [V] Module voltage [V]

Figure 5-15: Experimental and estimated PV modules characteristics.

5.6 Conclusions

A method to determine the SP configuration that provides the highest GMPP, in uniform or mis-
matching conditions, from the measurements of the array voltage and the currents in each string is
presented. The first step of the method is to perform a voltage scanning of the array and measure
the currents in each string to obtain the I-V curve of each string. From such data the IPs (i.e.
SCP, INPs and OCP) are detected to extract the part of the data that correspond to each module.
Then, the parameters of the modules are calculate with a suitable estimation method. When all

139
 Table 5.4: Experimental and estimated GMPP
Configuration GMPP [W] Estimated NSSE [%]
GMPP [W]
CF1 133.02 133.96 0.1266
CF2 154.00 155.57 0.0581
CF3 134.33 136.17 0.0931
CF4 131.40 132.26 0.1191
CF5 133.61 134.29 0.1235
CF6 131.61 132.40 0.1163
CF7 134.11 135.87 0.0950
CF8 135.05 136.91 0.0884
CF9 130.70 131.79 0.1087
CF10 132.83 133.84 0.1226

6 6
PV array Current [A]

PV array Current [A]

4 4

2 2
Experimental data Experimental data
0 Estimated model 0 Estimated model
0 10 20 30 40 50 60 0 10 20 30 40 50 60
PVGMPP
array voltage [V] PV GMPP
array voltage [V]
Estimated GMPP Estimated GMPP
200 200
PV array Power [W]

PV array Power [W]

150 150

100 100

50 NSSE = 0.0581 % 50 NSSE = 0.1087 %

0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
PV array voltage [V] PV array voltage [V]

(a) Configuration CF2 (highest power). (b) Configuration CF9 (lowest power).

Figure 5-16: Experimental and estimated power curves.

parameters are calculated, the method introduced in Section 5.5.1 can be used to find the GMPP of
each configuration to determine the best one for a given mismatching profile and weather conditions.
The model proposed in [12] was selected to evaluate all the feasible configurations due to its
high calculation speed. Nonetheless, it is important to note that such a model assumes ideal bypass
diodes. However, when the bypass diodes are active (PV modules are bypassed) the output voltages
of the modules are negative depending on the current that flows through the bypass diodes. Although
such negative voltages are small, they produce small shifts between the experimental measurements
and the data predicted by the model as analyzed in Chapter 3. Nevertheless, such shifts may not
affect the shape of the curves and the estimation of the best configuration should not be affected.
The solution introduced in this Chapter could be improved by introducing a more detailed model
of the bypass diodes in order to reproduce the small voltage shifts in the I-V curves produced by such
diodes. Nevertheless, it is important to highlight that the parameter calculation of the bypass diodes
is not a trivial task since the measurement of the diodes current are required, but such nodes are not

140
accessible in all the commercial PV panels. Moreover, since the exhaustive search algorithm tests
all the possible configurations, a more optimized search method is required to apply the proposed
solution in middle-size and large-size PV arrays.
Finally, it is important to note that the proposed method can be extended to PV arrays in TCT
or BL configurations. To achieve that goal it is required a model to predict the electrical behavior of
PV arrays in the selected configuration as the ones presented in 4. In addition it would be necessary
to change the number and locations of the current and voltage sensors in the array since all the
modules in each row share the same voltage, but the current of each module may be different.

This Chapter showed one application of the PV array models in the development of MPPT
techniques and a method to extract the model information of each module of the array. Such
information can be used not only by the MPPT control, but also by a monitoring system that
supervises the performance of the whole system to extract important information that can be used
for maintenance and diagnostic purposes.
Monitoring is an important part of the PV system, since it is aimed to detect anomalies in the
operation of the system that can be produced by lack of maintenance of the PV field, a damage in
the hardware of the system (cables, connectors, power converters, etc.) or a degradation in one or
more panels in the array. All those problems are translated into a reduction of both, the energy
production of the PV system and the profitability of the whole system. The following Chapter
presents two model based indicators, that can be implemented in a monitoring system, to detect the
degradation in a PV panel and avoid power losses in the system due to a malfunctioning panel.

141
142
Chapter 6

Diagnostic of PV modules

The mismatching conditions in a PV array not only reduce the performance of the PV system, but
also force the mismatched modules to operate under stress that, along with outdoor operation and
quality problems, may led to the degradation or damages of the cells, modules and/or bypass diodes
[3, 4]. Such degradation or damage is translated into a reduction in the energy production and the
increment of the return of investment time of the PV system. Moreover, the degradation modifies
the parameters that describe each PV module, which entails errors in the applications that use the
model of the array (e.g. the energy estimation).

Considering the importance of the degradation quantification in the PV modules, the diagnostic
methods of PV modules were studied. One the findings was a lack of model based procedures to
estimate the degradation in a PV modules.

In this Chapter, two indicators to detect and quantify the degradation of a PV module are
presented. Such indicators use the single-diode model to represent a PV module without degradation
and combine such information with the experimental measurements (maximum power point voltage
and current, short-circuit current and module temperature) to estimate the increase in the series
resistance and the decrease in the parallel resistance. The estimation of those variations allows to
quantify the module degradation regardless the irradiance and temperature conditions. Simulation
results comparing the proposed indicators with the Fill Factor, which is a traditional indicator to
quantify the electrical quality of a PV module, are presented to illustrate the advantages of the
proposed indicators for different weather conditions including also the measurements uncertainties.
Moreover, experimental results show the usefulness of the proposed indicators.

Part of the content of this Chapter has been included in a paper entitled Model Based Indicators
to Quantify Photovoltaic Module Degradation, which was presented at the 11th International Con-
ference on Modeling and Simulation of Electric Machines, Converters and Systems (ELECTRIMACS
2014) [151]. An extended version of such a paper, with the simulation analysis and experimental

143
results presented in this Chapter, will be submitted to an international journal.

6.1 Introduction

The main objectives of the monitoring and diagnostic systems [152] are: to quantify the energy
production, to detect performance reductions and to identify/quantify premature degradations in
the PV panels. The later is particularly important, since degraded PV modules strongly reduce
the energy harvested from a PV system due to the mismatching effect [12, 17, 153]; moreover, a
large number of failures can not be distinguished from a visual inspection. Without a procedure
for detecting premature degraded PV panels, the customer of a PV system may not identify the
problem in one or more modules in the PV array, which may result in significant difference between
the expected and real return of investment time.
Two groups of diagnostic methods can be defined depending on their objective: diagnostic tech-
niques aimed at ensuring a real 25 years warranty by improving the manufacturing process (quality
standards), and diagnostic techniques to monitor the performance degradation of a PV field [154].
The standards IEC 61215 [155], IEC 61646 [156] and IEC 61730 [157] are widely adopted to
qualify different aspects of the PV modules in the manufacturing process; while other methods,
presented bellow, are aimed to determine the performance degradation in the modules of an operative
PV plant.
IEC 61730 describes the PV module requirements to provide a safe electrical and mechanical
operation; while IEC 61215 and IEC 61646 define procedures and tests to determine (in a short
time) if the crystalline and thin-film PV modules, respectively, are able to withstand prolonged
exposure in a wide range of environmental conditions.
The methods to detect the performance degradation of PV modules, in general, evaluate one
or more indicators along the operation time of the module to compare the actual value with the
one obtained at the beginning of the monitoring process. Depending on the procedure used to
evaluate the indicators, a diagnostic method can be either off-line or on-line [154]. The off-line
methods require the disconnection of the PV module from the PV array and/or specialized in situ
measurements. The on-line methods calculate the indicators during the operation of the PV module.
The traditional measurements performed by the off-line methods are the dark and/or light current
vs. voltage (I-V) curves, which are used to identify part [158, 159, 160, 161] or all [162, 149]
model parameters and the main electrical characteristics of a PV module: voltage (V mpp), current
(Impp) and power (P mpp) in the maximum power point (MPP), short-circuit current (Isc), open-
circuit voltage (V oc) and fill factor (F F ). The estimated model parameters and measured electrical
characteristics are monitored along the time to estimate a reduction in the performance of the PV
module.

144
 In [163] the authors propose a diagnostic method that uses the reference values of the main
electrical characteristics (e.g. datasheet information), as well as light I-V curve, temperature and
detailed irradiance measurements, to generate nine normalized performance sub-indicators that are
multiplied to obtain a general indicator denominated Performance Factor.

Other group of off-line diagnostic methods are aimed to perform a detailed examination of the
PV module surface by using specialized equipment. Although these methods require specialized
hardware, they can determine the location of the failure. Some widely adopted non-invasive methods
are: infrared [164], electroluminescence [165], and ultrasonic [166] imaging, Laser (or Light) beam
induced current [167] and electric reflectrometry [168, 169].

The on-line diagnostic methods require less measurements than the off-line methods. Such an
advantage facilitates the implementation and reduces the costs, but it also reduces the types of
detectable failures and impedes their exact location in the module.

The on-line methods can be divided into two groups. The first group performs a continuous
monitoring of the power and energy produced by a PV module or array to estimate the energy
available from one or multiple irradiance and temperature measurements [170, 171, 172, 173, 174].
By using those energy values it is possible to evaluate the performance of a PV module or array,
in a given window of time, through an indicator denominated Performance Ratio. Monitoring such
indicator makes possible to quantify the degradation in a PV module or array, but it does not provide
any information concerning the failure causing the performance reduction. Besides, these methods
may fail in the estimation of the energy available due to the irradiance sensor can not detect if part
of the PV modules are shaded or not.

The second group of on-line methods does not use the energy to estimate the degradation or
failure in a PV module. So far, the authors have found few on-line diagnostic methods of this
class reported in the literature. Two of them [161, 175] propose to estimate the series resistance
from the experimental measurements to quantify the degradation due to the outdoor operation.
However, these two methods require to perform measurements in the same temperature each time
Rs is estimated; besides, they provide limited information, since the estimated value of Rs depends
on the environment conditions when it is estimated.

Other references [176, 177] are aimed to detect the presence of a degradation or problem in an
array operating in uniform conditions. The diagnostic method presented in [176] is able to detect
a deviation in V mpp with respect to a reference value calculated for a given operating condition.
If such a deviation is permanent, it means that there is a degradation in the module (or array),
otherwise the deviation is produced by a transitory condition (e.g. partial shading). The authors in
[177] propose a method to determine the number of modules in open-circuit and short-circuit within a
string (modules connected in series) from the difference between the actual maximum power and the
power estimated from irradiance measurements. Those last two methods only provide information

145
about the presence or not of a failure without any quantification of its effects into the PV module
or array performance; moreover, such methods assume that all the modules operate under uniform
conditions.
Instead, this Chapter proposes two indicators to quantify the degradation in a PV module.
Such indicators are calculated from four experimental measurements (short-circuit current, voltage
and current in MPP and module temperature) and two points calculated with the single-diode
model, which represents the module without degradation (ideal module). Moreover, simulations
and experimental results comparing the proposed indicators with the Fill Factor (FF) are provided
to illustrate the advantages of the proposed solution.

6.2 Degradation effects in the module I-V curves

During the lifetime of a PV module it is exposed to the environmental conditions of the specific place
where it is installed. Such an exposure may produce different degradation types. From [162, 158]
and [160] it is possible to make a summary of different degradation mechanisms and their effects in
the PV modules. Such a summary is presented in Table 6.1.

Table 6.1: Main PV module degradation mechanisms

Degradation Produced by Effects in the PV
source module
Front surface soil- Lack of maintenance and cleaning. Reduction in Isc
ing.
Optical degrada- Degradation and/or cracks in the transparent Reduction in Isc.
tion (e.g. brown- covers that protect the cells.
ing).
Cell degradation. Daily thermal cycling, humidity that enter Increase in effective
through the small cracks or holes that pro- Rs.
duces oxidation in the terminals, junction box,
etc.
Small breaks in the isolation between the cells Reduction in effec-
and the frame, crystal damages, semiconduc- tive Rh.
tor problems that increase leakage currents.
Mismatched cells. Degradation of Anti-Reflective coating, Reduction in Isc
cracked cells and shadowing. and V oc and/or in-
crease in effective
Rs.
Light Induced Brakes of weak Si links and changes in some Reduction in Isc
Degradation (LID). semiconductors materials due to the initial and V oc, increase
outdoor operation hours. in effective Rs
and/or decrease in
effective Rh.
Potential Induced Grounding problems and high voltages be- Reduction in effec-
Degradation (PID). tween the module frame and the surfaces that tive Rh.
carries the charges in the module like cells and
cables.

146
 All degradation sources produce a reduction in the PV module performance because the shape
of the I-V curve is distorted, which produces a reduction in the maximum power that a PV module
can provide for a given G and T conditions. Two important parameters that are affected by the
degradation are Rs and Rh: Rs can increase between five and ten times its nominal value [159] due
to one or more degradation sources, while Rh may decrase form hundreds to few units of ohms [160],
[159]. Therefore, Rs and Rh are considered in this paper as the main parameters affected by the
degradation mechanisms.
The distortions in the I-V curve under STC produced by increments in Rs and reductions in Rh
are presented in Fig. 6-1. When Rs increases, the slope of the I-V curve close to V oc also increases
(Fig. 6-1(a)); therefore, the voltage of the MPP decreases. The slope of the I-V curve close to Isc
remains approximately constant because Rh does not change and the value of V oc is not modified
since the other parameters ( and Isat) remain without modifications.
A large reduction in Rh produces a decrease in the slope of the I-V curve around Isc (Fig. 6-1(b));
hence, the current in the MPP is reduced. Around V oc the slope of the I-V curves present small
changes; however, there is a significant reduction in V oc.
Current [A]

Current [A]

4 4
RsSTCx6.00 RhSTC/100
2 2 RhSTC/80
RsSTCx5.00
0 RsSTCx4.00 0 RhSTC/60
0 10 20 30 40 50 0 10 20 30 40 50
RsSTCx3.00 Voltage [V] RhSTC/40 Voltage [V]
RsSTCx2.00 RhSTC/20
RhSTC/1
Power [W]

Power [W]

150 RsSTCx1.00 150

100 100
50 50
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Voltage [V] Voltage [V]

(a) Characteristic curves for variations in Rs from 1 to 6 (b) Characteristic curves for variations in Rh from 1 to
times its STC value. 1/100 times its STC value.

Figure 6-1: I-V curves distortion with variations in Rs and Rh.

6.3 Traditional PV module performance indicators

A common practice to evaluate the performance of a PV module is to monitor the FF [149], [162],
[160] and/or the three main points in the I-V curve (Isc, V oc and PM P P ) [178, 179, 180]. The FF
is often used because it collects the effects of different degradation mechanisms.
The FF is defined as the ratio of the power in the MPP (V mpp Impp) and the ideal power of
the the module represented as the product of Isc and V oc. Moreover, it is usually normalized by
using the its value in STC (F FST C ) extracted from the datasheet. The normalized FF (NFF) is

147
calculated by using (6.1).

V mppImpp
V ocIsc FF
NFF = = (6.1)
F FST C F FST C

This indicator is sensible to the position of the measured MPP whith respect to Isc and V oc;
therefore, if any parameter of the PV module is modified by degradation or other problems, the
value of the indicator becomes smaller, otherwise the value should remain close to the STC value
within a given range defined by the variations in the irradiance and temperature of the module.
It is worth nothing that to calculate the FF one current and one voltage sensor (to measure Isc,
V oc, Impp and V mpp) are required as well as the datasheet information for the same points, i.e.
IscST C , V ocST C , ImppST C and V mppST C .

6.3.1 NFF behavior for different degradation conditions

NFF is evaluated for different irradiance and temperature conditions as well as two degradation
mechanisms. The first one considering increments in Rs and the second one introducing reductions
in Rh. The PV panel considered for the analysis is a KC 120-1.
The first step in the simulation process is to calculate the five model parameters of the module
in STC by using the method presented in Appendix A. Such parameters are: Iph = 7.45 A,
Isat = 3.4201e 9 A, = 1.0811, Rs = 0.2435 and Rh = 1.5 k. Then, a set of nine irradiances
(from 0.2 kW/m2 to 1.0 kW/m2 with steps of 0.1 kW/m2 ) and five temperatures (from 12o C to
60o C with steps of 12 o C) is defined. For each different operating condition (45 in total), Iph is
obtained with the approximation Iph Isc (3.5) and Isat is calculated using (3.4); while , Rs and
Rh assume their STC values. Those parameters represent the module without degradation.
The degradation in the module is simulated adding a resistance in series to RsST C (Rs ) and
adding a resistance in parallel to RhST C (Rh ) as shown in Fig. 6-2. Such additional resistances
represent the degradation quantity. Rs takes values form zero (no degradation) to four times RsST C
and Rh assumes values form infinite (no degradation) to 1/80 times RhST C . Those values were
selected considering the variations in Rs and Rh that can be produced by different degradation
effects as shown in [159, 181].

Rs Rs

I
Iph Rh Rh V

Figure 6-2: Single-diode model with degradation represented by additional series and parallel resis-
tances.

148
 The simulated N F F for the different values of Rs and Rh are presented in Fig. 6-3(a) and
Fig. 6-3(b), respectively. In those figures, the abscissas represents the different irradiance levels as
the ratio of the measured Isc and IscST C , this because in a real implementation of the indicators
it is assumed that the irradiance measurement is not available. Moreover, the lines with the same
color represent the value of the indicator with the highest (dashed line) and lowest (continuous line)
simulated temperatures and the same degradation condition.

1.1 1.1

1 1

0.9
0.9
0.8
NFF

NFF
0.8 Rh=
0.7 Rh=75.00
Rs=0.00
0.7 Rh=37.50
Rs=0.24 0.6
Rh=25.00
0.6 Rs=0.49
0.5 Rh=18.75
Rs=0.73
0.5 Rs0.4
=0.97 0.4
0.2 0.6 0.8 1 0.2 0.4 0.6 0.8 1
Isc/IscSTC Isc/IscSTC

(a) Variations in Rs . (b) Variations in Rh .

Figure 6-3: NFF for variations in Rs and Rs and different irradiances and temperatures. Contin-
uous line: minimum temperature, dashed lines: maximum temperature.

It can be observed from Fig. 6-3(a) that NFF has linear behavior with respect to the ratio
Isc/IscST C for different values of Rs . The increments in Rs produce a reduction in the slope of
NFF, while the offset of the NFF is inversely proportional to the temperature.

The variation of NFF with respect to the Isc is non-linear when there is a reduction in Rh (Fig.
6-3(b)). Besides, it can be observed that a reduction in Rh produces a reduction in the offset of the
curves.

The main disadvantage of NFF is that it can assume the same value for different degradation
conditions depending on the temperature and Isc. For example, if in Fig. 6-3(a) N F F = 0.98 and
Isc/IscST C < 0.6, it is not possible to determine if the module is degraded or not, since such a value
can be obtained with (Rs = 0.24 ) or without degradation (Rs = 0 ) depending on the module
temperature. This, because if the module is not degraded but the operating temperature is high
NFF takes values smaller than one. One option to overcome such a disadvantage would be to use
the measurements for high values of Isc/IscST C and to use a temperature sensor for calculating the
NFF always at the same temperature. Nonetheless, such a procedure may complicate the diagnostic
process since it would be difficult to evaluate NFF for different irradiance conditions and the same
temperature in a PV module working outdoors.

149
6.4 Proposed PV module performance indicators

The different degradation mechanisms modify one or more parameters of the PV modules and,
as a consequence, they distort the shape of the I-V curve as show in [159] and Fig. 6-1. Such
distortions moves the location of the MPP in the I-V plane depending on the parameter affected by
the degradation.
The proposed indicators are aimed to quantify the degradation in the module by processing the
shift of the MPP (produced by degradation) with respect to a reference or ideal I-V curve that
represents the module without degradation. The points of the reference I-V curve are calculated with
(1.4) or (1.5) (see Chapter 1) depending on the indicator. Iph and Isat of the ideal module (Iphid
and Isatid ) are calculated using (3.5) and (3.4) (see Chapter 3), respectively; while, the ideality
factor and the resistances of the ideal module assume their STC values (id = ST C , Rsid = RsST C
and Rhid = RhST C ).
The first indicator is focused on the degradation sources that produce reductions in the measured
MPP voltage (V mpp). Such a distortion in the I-V curve can be quantified as the additional
resistance (Rs []) that needs to be connected in series to Rsid to produce the same reduction
in V mppid . The voltage reduction is calculated as the difference between the voltage of the ideal
module (Vid ) evaluated in Impp (measured) and the experimental V mpp (6.2), where Vid is obtained
evaluating (1.5) in the actual operating conditions. The relative value of the Rs with respect to
its RsST C (N Rs ) is obtained with (6.4).

V mpp = Vid (Impp) V mpp (6.2)

V mpp
Rs = (6.3)
Impp
Rs
N Rs = (6.4)
RsST C

The second indicator is aimed to quantify the reductions in the MPP current produced by the
degradation sources. It takes into account that reductions in Rh generate decrements in the slope of
the curve between Isc and MPP, which is translated into a reduction in the measured MPP current
(Impp). Such a distortion in the I-V curve can be quantified as the additional resistance that needs
to be connected in parallel to Rhid (Rh []) to produce the same reduction in the MPP current of
the ideal module (6.6).
Neglecting Rs allows to approximate the current crossing Rh as the difference between the
current of the ideal module current at V mpp (Iid (V mpp)) and Impp. Therefore, dividing such a
difference by V mpp it can be obtained an estimation of Rh as shown in (6.6). The relative value
of Rh with respect to RhST C (N Rh ) is defined in (6.7).

150
 Impp = Iid (V mpp) Impp (6.5)
V mpp
Rh = (6.6)
Iid (V mpp) Impp
Rh
N Rh = (6.7)
RhST C

To calculate these two indicators it is required to measure Isc, Impp and V mpp as well as the
module temperature to evaluate Vid (Impp) and Iid (V mpp). The datasheet information (IscST C ,
V ocST C , ImppST C and V mppST C , Isc , V oc , N s) is also required to determine the STC values of
the model parameters.

6.4.1 N Rs and N Rh behavior for different degradation conditions

To evaluate the behavior of the proposed indicators for different operating and degradation condi-
tions, the same process and PV panel presented in Section 6.3.1 are used.
N Rs and N Rh for different values of the additional series resistance (Rs ) are presented in
Fig. 6-4; while the behavior of the proposed indicators for different values of the additional parallel
resistance (Rs ) are introduced in Fig. 6-5.
To avoid confusions, it should be noted that Rs and Rh are the additional resistors introduced
in the simulation procedure to represent different degradation sources, while Rs and Rh corre-
spond to the approximations of the variations in the series and parallel resistors calculated from the
measurements and the model of the ideal module.
The proposed indicator to identify increments in Rs (Fig. 6-4(a)) is able to estimate Rs since it is
constant with respect to Isc, irradiance and temperature. Such behavior is an important advantage
because a variation in the indicator value is directly related with an increase in Rs; in such a case
the customer can focus the search of the degradation mechanisms that produce increments in Rs,
which can reduce the identification time of the problem. On the other hand, N Rh does not provide
a useful information for increments in Rs (Fig. 6-4(b)) since the slope of the I-V curve between Isc
and MPP is not significantly modified with variations in Rs.
Fig. 6-5(b) shows that N Rh correctly estimates the relative value of the additional parallel re-
sistance required to produce a reduction in Impp. For example, considering that RhST C = 13.826k
and Rh = 691.29 produces N Rh = 0.05 (blue line Fig. 6-5(b)). Moreover, N Rh has a low
sensitivity to variations in Isc and temperature; such a sensitivity is produced due to Rs is neglected
in the definition of N Rh . For reductions in Rh the value of N Rs does not provide a useful infor-
mation. This because the effects in the slope of the I-V curve between MPP and V oc for reductions
in Rh depend on the irradiance 6-5(a).

151
 4 0.5 Rs=0.00
Rs=0.24
3 0.4 Rs=0.49
Rs=0.73
2

NRh
0.3 Rs=0.97
NRs

1
Rs=0.00 0.2

0 Rs=0.24
0.1
Rs=0.49
1
Rs=0.73
0
0.2 Rs
0.4
=0.97 0.6 0.8 1 0.2 0.4 0.6 0.8 1
Isc/IscSTC Isc/IscSTC

(a) Variations in N Rs . (b) Variations in N Rh .

Figure 6-4: N Rs and N Rh for variations in Rs from 0 to RsST C 4 and different irradiances
and temperatures. Continuous line: minimum temperature, dashed lines: maximum temperature.

40
Rh=
35 0.05
Rh=75.00 Rh=
30 0.045
Rh=37.50 Rh=75.00
0.04
25 Rh=25.00 Rh=37.50
Rh=18.75 0.035
NRh
NRs

20 Rh=25.00
0.03 Rh=18.75
15
0.025
10
0.02
5
0.015
0 0.01
0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1
Isc/IscSTC Isc/IscSTC

(a) Variations in N Rs . (b) Variations in N Rh .

Figure 6-5: N Rs and N Rh for variations in Rh from to RhST C /80 and different irradiances
and temperatures. Continuous line: minimum temperature, dashed lines: maximum temperature.

6.5 Effects of the sensors uncertainties on traditional and

proposed indicators

To perform the calculation of the indicators in a real application (e.g. a distributed MPP tracking
system - DMPPT) it is required to measure the current, voltage and temperature in the PV module.
The current and voltage sensors may be the same used to perform the MPP Tracking (MPPT)
technique, while the temperature sensor needs to be cheap to avoid large increase in the cost of the
system.
In this Section, commercial sensors of current, voltage and temperature are selected to perform a
numerical evaluation of the effects produced by the uncertainties of the sensors in the evaluation of
the indicators. The selected sensors have low cost taking into account that they would be installed

152
in each unit interacting with a PV module (e.g. each unit in a DMPPT system); therefore, the
errors and/or uncertainties are not negligible. The main characteristics of the selected sensors are
presented in Table 6.2.

Table 6.2: Main characteristics of selected sensors

Sensor Model Measurement Error
range
Current L18P010D15 10 A 1%
Voltage Voltage divider and ISO124 20 V 5%
Temperature LM135A -55 to 150 o C 1 oC

The Monte Carlo approach is used to asses the uncertainties effects in the traditional and pro-
posed indicators. The simulation process is similar to the one explained in Section 6.3.1 (irradiances,
temperatures, degradations and PV panel) but introducing the sensors uncertainties every time the
measurements of Isc, Impp, V mpp and T are used in the procedure.
The results of the Monte Carlo procedure with 300 tests are calculated for each one of the
simulated degradation conditions presented in Section 6.3.1 and Section 6.4.1. The simulation results
of N F F for variations in Rs and Rh are presented in Fig. 6-6; while Fig. 6-7 and Fig. 6-8 show the
results of the proposed indicators for increments in Rs and reductions in Rh, respectively.
In general the difference between the results presented by the indicators with respect to the ones
of Section 6.3.1 and Section 6.4.1 are produced due to the uncertainties in the measured values (Isc,
Impp, V mpp and T ) are included into the mathematical expressions to obtain the indicators and
the non-linear equations presented in Section 1.2 and Section 3.2 to calculate the I-V curve without
degradation. In such a process the effects of the uncertainties are propagated along each operation
producing an increase in the indicators span for a given degradation condition.
When the uncertainties in the measurements are included in the simulations, the maximum and
minimum values of NFF become closer for low and high short-circuit currents (see Fig. 6-6) with
respect to the simulation results presented in Section 6.3.1 (Fig. 6-3). Hence, it is even more difficult
to determine if a module is degraded or not, which suggest the necessity of a temperature sensor to
reduce the variation range of NFF.
The uncertainties of the sensors affects N Rs producing difficulties in the degradation detection
for low short-circuit currents, since a particular value of N Rs may be produced for different incre-
ments in Rs (Fig. 6-7). However, taking into account that the variations in N Rs generated by the
sensors uncertainties are symmetrical around the nominal value, the average value can be used to
estimate the increment in Rs. Moreover, it can be observed that for a high short-circuit currents,
the values that N Rs can take for different degradation conditions are different; therefore, it would
be recommended to evaluate this indicator in such a region.
For reductions in Rh, N Rh can provide useful information for low values of Isc, since for mid

153
 1.2 1.2
1.1 1.1
1 1
0.9 0.9
NFF

NFF
0.8 0.8
0.7 0.7
Rh=
Rs=0.00
0.6 0.6 Rh=75.00
Rs=0.24
0.5 0.5 Rh=37.50
Rs=0.49
0.4 0.4 Rh=25.00
Rs=0.73
0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 Rh=18.75
0.8 1
Rs=0.97
Isc/IscSTC Isc/IscSTC

(a) Variations in NFF for increments in Rs. (b) Variations in NFF for reductions in Rh.

Figure 6-6: Monte Carlo analysis of NFF for variations in Rs and Rh as well as different irradiances
and temperatures. Continuous line: maximum values, dashed lines: minimum values.

4
NRs

2 Rs=0.00
Rs=0.24
4 Rs=0.49
0.2 0.4 0.6 0.8 =0.73 1
Rs
Isc/IscSTC
Rs=0.97

Figure 6-7: Monte Carlo analysis of N Rs for increments in Rs and different irradiances and
temperatures. Continuous line: maximum values, dashed lines: minimum values.

and high short-circuit currents N Rh has high sensitivity to the measurement uncertainties. The
the zoom of Fig. 6-8 shows that N Rh has limited sensitivity to the measurement uncertainties at
low Isc values; hence, the average N Rh for low irradiance conditions can be used to estimate the
reduction in Rh.
The behavior of N Rh for increments in Rs and N Rs for reductions in Rh have not been
shown because they do not provide useful information as was explained in Section 6.4.1.

6.6 Experimental results

Traditional and proposed indicators were calculated from experimental data taken in the test bench
2 described in Section 1.6. The data were taken along two clearly sky days during the summer
of 2013 in the south of Italy. The experimental process was to scan the I-V curve of the panel

154
 Rh= 0.14
1000
Rh=75.00 0.12
0.1
800 Rh=37.50
0.08
Rh=25.00 0.06
600 Rh=18.75 0.04
0.02

NRh
400 0.2 0.22 0.24 0.26 0.28 0.3

200

200
0.2 0.4 0.6 0.8 1
Isc/IscSTC

Figure 6-8: Monte Carlo analysis of N Rh for reductions in Rh and different irradiances and
temperatures. Continuous line: maximum values, dashed lines: minimum values.

modifying the duty cycle of the converter and measure the temperature in a central point at the
back of the panel. Such a process was repeated at different hours of the day in order to obtain
different irradiance and temperature conditions. In total 90 I-V curves were recorded, 15 for each
additional resistance connected at the output of the panel, and all the indicators were evaluated for
each curve. An example of the experimental I-V curves obtained with the additional series resistance
and the reference I-V curve is presented in Fig. 6-9.

4
Current [A]

Rs=0.022
2
Rs=0.168
Rs=0.517
1
Rs=1.013
Rs=2.028
0
0 Rs=3.000
5 10 15 20
Voltage [V]

Figure 6-9: Example of reference and experimental I-V curves for the different additional Rs values.
Dashed lines: experimental I-V curves. Continuous lines: reference I-V curves.

The results of N F F and N Rs for the different values of additional series resistance (Rs ) are
presented in Fig. 6-10 and Fig. 6-11, respectively. In such figures the dots represent the experimental
points of the evaluated indicators and the dashed lines represent the trend of the indicators along
the abscissas. The results of N Rh are not presented since they do not provide useful information
for increments in Rs.
Fig. 6-10 represent the experimental NFF values, where the behavior of N F F with respect to

155
Isc/IscST C agrees with the simulation results presented in the previous Sections. It can be deduced
that the panel is degraded because all the experimental values are lower than 0.9, which is the lower
limit of NFF for a panel without degradation (see Fig. 6-6(a)). However, to know such a limit
was necessary to perform the Monte Carlo analysis using the model, otherwise it would be difficult
to establish if the blue dots (for example) are an indication of degradation or a high operating
temperature. In general NFF may be used to detect the presence or not degradation in the panel
when a given threshold is violated (e.g. N F F < 0.8); nevertheless, it is difficult to quantify the
degradation only using the NFF because there is not a reference to compare the measured value.
Moreover, if the degradation detection is limited to a threshold, there is a loss of energy during the
time when the degradation started and the time when NFF violates the threshold.

0.9

0.8

0.7

0.6
NFF

0.5

0.4
Rs=0.022
0.3 Rs=0.168
0.2 Rs=0.517

0.1 Rs=1.013

0.2 Rs0.4
=2.028 0.6 0.8 1
Rs=3.000 Isc/IscSTC

Figure 6-10: Experimental results (dots) of NFF for different increments in Rs and operating con-
ditions. Dashed lines represent the trend of the experimental points.

The experimental N Rs values are represented in Fig. 6-11, where the distribution of the dots
at both sides of the lines show the uncertainties effects in the indicator in agreement with Fig 6-7.
The blue line shows that the PV panel is degraded because with practically no additional resistance
(Rs = 0.022) N Rs 4.3, which means that during the operation time of the panel (around 10
years) Rs has increased around four times its STC value.

Considering such initial degradation, the effective additional resistances with respect the RsST C
are: 4.3, 5.0, 6.4, 8.5, 12.6 and 16.6 times RsST C . The average values of experimental N Rs were:
4.3, 4.7, 6.0, 7.5, 11.1 and 14.6; therefore, N Rs provides a useful estimation of the increments in
Rs.

It is worth noting that a value of N Rs or N Rh provides an indication of the degradation level

in the panel without the imposing a threshold or other reference value. Such a characteristic would
be useful for a monitoring process to generate alarms to the customer, to analyze the trends of the
indicators along the time or to program maintenance or revisions to the PV generator.

156
 Rs=0.022
16
Rs=0.168
14 Rs=0.517

12 Rs=1.013
Rs=2.028
10
Rs=3.000

NRs
8

2
0.2 0.4 0.6 0.8 1
Isc/IscSTC

Figure 6-11: Experimental results (dots) of N Rs for different increments in Rs and operating
conditions. Dashed lines represent the trend of the experimental points.

6.7 Conclusions

Two indicators (N Rs and N Rh ) to detect and quantify the degradation in a PV module, which
compares measurements with a reference I-V curve, have been proposed. They are aimed to estimate
the increase in Rs (N Rs ) and decrease in Rh (N Rh ) from the reductions in V mpp and Impp,
respectively. The points used in the reference I-V curve are calculated using the single-diode model,
which requires the temperature and Isc measurements.
The simulation results presented a comparison between the proposed indicators and a tradi-
tional indicator of the electrical quality of a module (N F F ) for different irradiance and temperature
conditions and considering the sensors uncertainties. In general, the proposed indicators have low
sensibility to the irradiance and temperature conditions but are sensible to the measurements uncer-
tainties. On the other hand, N F F is sensible to both weather conditions and sensors uncertainties.
Moreover, the main disadvantage of N F F is that a given value does not provide enough information
to quantify the module degradation. It is useful only if its value is under a given threshold since
such a condition indicates module is degraded module degradation.
It is important to note that the calculation of the proposed indicators do not require the mea-
surement of V oc. Such characteristic constitute an advantage because it is only necessary to change
the operating point of the module just to one point of zero power (Isc) instead of two points as
required by FF (Isc and V oc). That property represents a reduction in the energy losses due to
the degradation estimation. A disadvantage of the proposed indicators is the measurement of the
module temperature, which increases the system cost. However, an early detection of a problem in
one module that is reducing the energy extraction of all the system (e.g. in a series parallel array)
may represent an important reduction in the energy losses and a compensation of the investment in
the temperatures sensors.

157
 In general, this thesis presented an overview of the MPPT techniques for PV arrays operating
under mismatching conditions that lead to the development of suitable models for PV arrays in
SP, TCT and BL configurations. From the development of those models it was identified their
applicability in the reconfiguration analysis and diagnostic processes. However, the the PV array
models are the base of the simulation process of PV systems in different simulation environments,
which are used for a number of different applications and analysis as was explained in the previous
Chapters. Therefore, there are different applications based on array models that can be developed
to contribute with the strengthening and advance of PV systems.

158
Chapter 7

Conclusions and future works

The mismatching conditions are one of the main causes of the power production reduction in PV
systems, which increases the return of investment time. Such a problem can not be solved with
only one specific method regardless of the specific operating conditions (e.g. with a given MPPT
technique), thus the mismatching problem should be addressed form different points of view in order
to mitigate its effects as much as possible.

For example, in the design process, the surrounding objects should be considered in order to
select the location of the PV modules along the available surface. Then, the mismatching profiles
over the PV array and the models can be used to determine the electrical configuration for the
PV array (or sub-array) that improves the power production. Moreover, the mismatching profiles
characteristics of the particular application should be considered in the selection of the architecture
of the MPPT technique (or techniques) in order fulfill the main objective of the PV system with
a cost-effective solution. Besides, the monitoring and diagnostic process along the operation of the
PV system are important not only to avoid further mismatching due to the modules soiling and
degradation, but also to define maintenance procedures and identify the early degradation of the
system.

For the analysis of the mismatching effects in the different aspects of PV systems mentioned
before, different simulation environments are required: circuit and control simulators (like PSIM or
SIMULINK) for testing and compare MPPT techniques, general purpose languages (like C++ or
Java) that can be used in the designing and energy estimation processes, simulators of power system
transient analysis (like EMTP-RV or etap) to evaluate the energy distributions systems, among
other. For all those environments, suitable models (i.e. tradeoff between accuracy and calculation
burden) of PV arrays are required to reproduce the behavior of the PV array in different operating
conditions, electrical configurations and dimensions. Therefore, the development of more efficient

159
models is a field that deserves further research.

The accuracy of the array model is defined by the circuital representation selected for each
module of the PV field, e.g single-diode model, ideal single-diode model (neglecting Rs and Rh) or
double-diode model. Therefore, it is important to select adequately the accuracy required in the
simulation analysis that is going to be performed since the simulation time varies a lot depending
on the accuracy selected. So that, a wrong selection of the model will waste hardware resources or
turn the simulation time too large for the desired application.

A key point in the use of PV array models is the estimation of the parameters of all the modules
and bypass diodes in the array since they may vary along the time due to degradation and mismatch-
ing conditions. Hence, the PV arrays models should be improved by combining them with methods
to update periodically the parameters. This, to ensure an accurate reproduction of the behavior of
a PV field along its lifetime. It could be useful to have more realistic and reliable models, which
would be able to describe the mid and long term behavior of an array.

The new technologies available in the marked (like hetero-junction, thin-film and organic) should
also be considered in the array modeling process since they require different electrical representa-
tions of the modules to accurately reproduce their electrical characteristics. Therefore, it would be
interesting to extend the procedures presented in Chapters 3 and 4 considering other representations
of the modules, like the double-diode [182] or propose a generalized procedure to model PV arrays
in which the configuration of the array is represented by a matrix and each module is represented by
a non-linear equation. In such a case the electrical configuration of the array and the representation
of each module can be easily changed (to select the accuracy and calculation speed), which could be
useful in the designing process of PV systems or reconfiguration analysis.

One of the application of PV array models is the dynamic reconfiguration analysis, not only
to test and compare RMPPT techniques, but also to develop new model-based reconfiguration or
paradigms. The problem of finding the best configuration of the PV array is a combinatorial opti-
mization problem where the number of possible configuration can be thousands, millions or billions
depending on the size of the system. Therefore, to find a reconfiguration algorithm that determine
the best array configuration with a reduced number of iterations is an open problem that is inter-
esting for middle-size PV arrays. In this particular applications a high-speed PV array model is
required since the time between one configuration and the other is limited and the environmental
conditions may change; therefore, if a model based configuration takes to much time, the time at
which the best configuration is obtained, it may not be valid anymore.

The degradation of the PV modules in PV systems can led to partial or total damage of the PV
modules in the array. Such degradation not only reduces the profitability of the PV system, but

160
also produce significant changes the parameters of the PV modules. Therefore, it is important to
develop monitoring and diagnostic procedures to detect and quantify the degradation.
The degradation indicators proposed in Chapter 6 are aimed to quantify the degradation by the
single-diode model to represent the module without degradation. Such indicators are intended to be
implemented easily in DMPPT architectures (e.g. [72, 183]) where the generator in each DMPPT-U
is a PV panel. However, an extension of such a concept could be explored in order to represent
the PV array without degradation with an appropriate model, in such a case it could be possible to
propose model-based on-line diagnostic indicators for CMPPT and RMPPT architectures as well as
the DMPPT systems that use sub-arrays in each DMPPT-U.
In general, the on-line diagnostic indicators provide instantaneous information of the actual state
of a PV generator (panel, sub-array or array); however, such information could be evaluated for a
high level monitoring system that could perform further processing of the information coming form
the indicators to provide other type of information useful for the owners of the PV systems. For
example, the trends of the indicators can be analyzed to estimate the energy losses produced by the
degraded PV generator or to estimate when a specific panel will break the warranty. Such infor-
mation can help not only in the maintenance of the system, but also in the avoidance of the losses
produced by manufacturing problems.

In conclusion, this thesis gives a state-of-the-art of the MPPT architectures for PV arrays op-
erating under mismatching conditions, which led the development of suitable models of SP, TCT
and BL arrays under uniform an mismatching conditions. Moreover, the usefulness of the models
were tested in the proposed model-based analysis dynamic reconfiguration of PV arrays and the
development model-based indicators for on-line diagnostic applications of PV modules.
In general, the improvement of the overall efficiency of PV systems operating under mismatching
conditions is an open field in which the contributions can be focused on different aspects of the
system. In all the cases, reliable PV array models are the required since they are the base of the
simulation analysis and model-based solutions.

161
162
Appendix A

Estimation of single-diode model

parameter using a Genetic
Algorithm

This Appendix presents a method to calculate the five parameters of the single-diode model for
a photovoltaic module without performing any simplification. A system of non-linear equations,
with the parameters as unknowns, is posed by using five representative points of a reference (or
experimental) I-V curve and the condition of the current derivative in the MPP. Such a non-linear
equation system is rewrite as an optimization problem and solved by using a genetic algorithm.
Simulation results are presented for a mono-crystalline panel to illustrate the utility of the method.
Moreover, the parameters are calculated for 25 different combinations of irradiance and temperature
conditions from reference I-V curves to show the parameters variation with respect to the irradiance
and temperature. Such variations show that all the model parameters need to be adjusted with
respect the environmental conditions to obtain low errors in the current, power and energy prediction.
The content of this Appendix has been included in a paper entitled Parameter Calculation of
Photovoltaic Modules using a Genetic Algorithm, which has been presented in the conference 11th
International Conference on Modeling and Simulation of Electric Machines, Converters and Systems
(ELECTRIMACS 2014) [184].

A.1 Introduction
Given the variability of the PV modules with respect to irradiance and temperature it is important
to have accurate and comprehensive models for PV modules as well as parameterization procedures
to use the models in different applications.

163
 The single-diode model uses five parameters that can be calculated following different procedures
available in literature. Some procedures consider that all parameters varies with the irradiance
and/or temperature [117, 8], while other procedures assumes that three parameters are constant
and two parameters depends on the irradiance and temperature [6, 185, 186, 116].

The methods for calculating the parameters use different approaches. Some of them perform sim-
plifications to obtain analytic expressions to calculate all the parameters [7, 135] from the datasheet
information and/or experimental I-V curves. Other methods, introduced in [137, 9, 138], perform
less simplifications to increase the accuracy and obtain a system of non-linear equations, which is
solved in the operating points provided by the manufacturer datasheet to obtain the parameters of
the model. Other authors [141, 140, 6, 142, 185, 186] use explicit equations to find some parame-
ters (e.g. PV current, inverse saturation current and thermal voltage) and use iterative methods to
adjust the values of remaining ones (e.g. series and parallel resistance) to fit the model with the
information taken from the datasheet.

Just one reference [116] proposes a method to obtain the five parameters form the three points of
the I-V curve in Standard Test Conditions (STC) provided by the datasheet. Such a method poses a
system of five non-linear equations without any simplification to find the five parameters of the model.
The solution of the non-linear equation system is formalized as an optimization problem solved by the
Generalized Reduced Gradient algorithm. Although the errors obtained in the simulations presented
in [116] are small, this procedure assumes that the values of three parameters (ideality factor, series
resistance and parallel resistances) are constant for different irradiance and temperature conditions.
Such an assumption could introduce errors in the evaluation of a PV module along the time because
the degradation produced by the outdoor operation may modifies one or more parameters as shown
in Chapter 6.

This Appendix presents a method to calculate the five parameters of the single-diode model of
a PV module for a given irradiance and temperature conditions by posing a system of six non-
linear equations without simplifications. Such a system of equations is formalized considering the
information of experimental I-V curves or information available in the free database System Advisor
Model (SAM) provided by the U.S. Department of Energy. Moreover, its solution is posed as an
optimization problem, which is solved by using a genetic algorithm.

A.2 Proposed method to calculate the parameters

The expression that relates the current and voltage of a PV module represented with the single-diode
model was presented in Section 1.2. Such an expression is re-wrote here to simplify the explanation
(A.1).

164
 ! !
V + I Rs V + I Rs
I = Iph Isat exp N skT
1 (A.1)
q
Rh

From the information found, so far, in literature it is not clear if there are some parameters that
are constant for different irradiance and temperature conditions. That is why the method presented
in this Appendix proposes to estimate the five parameters to fit (A.1) with five key points of the
I-V curve of a given module. The five I-V points used (defined in [187]) are presented bellow and
illustrated in Fig. A-1 (taken from [187]):

Short-circuit: V = 0 V , I = Isc A

MPP: V = V mpp V , I = Impp A

Open-circuit V = V oc V , I = 0 A

Middle open-circuit V = V oc/2 V , I = Ix A

Mid point between MPP and open-circuit: V = (V mpp + V oc)/2 V , I = Ixx A

Figure A-1: Five key points of an I-V curve proposed in Kings Model.

A system of six non-linear equations is posed by using the five key points proposed in [187] and
the condition fulfilled by the I-V curve in the MPP point. Functions f1 to f5 ((A.2)-(A.6)) are
obtained by evaluating (A.1) in the five key points presented in Fig. A-1 (taken from [187]). In such
equations IphG,T , IsatG,T , G,T , RsG,T and RhG,T represents the parameters values for a given
irradiance (G) and temperature (T ) conditions. f6 is obtained from the derivative of the power in
the MPP (A.7) as presented in [116].

   
Isc RsG,T Isc RsG,T
f1 : 0 = IphG,T IsatG,T exp 1 Isc (A.2)
G,T V t RhG,T

165
    
V oc/2 + Ix RsG,T
f2 : 0 = IphG,T IsatG,T exp 1
G,T V t
V oc/2 + Ix RsG,T
Ix (A.3)
RhG,T
   
V mpp + Impp RsG,T
f3 : 0 = IphG,T IsatG,T exp 1
G,T V t
V mpp + Impp RsG,T
Impp (A.4)
RhG,T
   
(V mpp + V oc)/2 + Ixx RsG,T
f4 : 0 = IphG,T IsatG,T exp 1
G,T V t
(V mpp + V oc)/2 + Ixx RsG,T
Ixx (A.5)
RhG,T
   
V oc V oc
f5 : 0 = IphG,T IsatG,T exp 1 (A.6)
G,T V t RhG,T
 
V mpp+ImppRsG,T
G,T V t IsatG,T RhG,T exp G,T V t
f6 : 0 =   
V mpp+ImppRsG,T
G,T V t RsG,T + RhG,T + IsatG,T RhG,T RsG,T exp G,T V t
Impp
+ (A.7)
V mpp

The system of non-linear equations composed by f1 to f6 can not be solved by traditional methods
like Newton-Raphson, since the values of the parameters are in different ranges (e.g. Iph is in the
order of amperes, while Isat is in the micro-amperes). Therefore, the system of non-linear equations
to be solved is ill-conditioned and needs to be solved with alternative methods.
The method used in this document is to pose the solution of the non-linear equations system as an
optimization problem to minimize a cost function that considers the functions f1 to f6 . The proposed
method uses a genetic algorithm to minimize (A.8) due to the complexity and high non-linearity of
the system.

F C(IphG,T , IsatG,T , G,T , RsG,T , RhG,T ) =

6
X
fi2 (IphG,T , IsatG,T , G,T , RsG,T , RhG,T ) (A.8)
i=1

There are additional information that can be provided to the optimization algorithm in order
to help it in the search of a solution. The first information is that the value of Iph is bigger than
Isc, while the other parameters are positive. These information is provided as lower bounds for the
optimization algorithm as form of inequalities (A.9). It is important to note that the lower bound
of is not defined as zero to avoid the infinite values in the initial population.

IphG,T > Isc

166
 IsatG, T > 0

G, T > 0.1

RsG, T > 0

RhG, T > 0 (A.9)

Other information useful for the optimization algorithm is the initial range of the parameters.
Such range is calculated from the parameters calculated in a given irradiance and temperature
conditions by using the analytical procedure presented in [7]. From such values the possible variation
range of each parameter is defined based on information found in literature [188, 159, 189]. In this
document, the initial range of the solution search is presented in A.10 where IphG,T , IsatG,T , G,T ,
RsG,T and RhG,T are calculated with (A.11)-(A.15) [7].

Isc > IphG,T > 1.1 Isc

0.1 Isat > IsatG,T > 10 Isat

0.5 > G,T > 1.5

0.1 Rs > RsG,T > 10 Rs

0.1 Rh > RhG,T > 0.1 Rh (A.10)

Iph Isc (A.11)

V oc
V
Isat = Iph e t (A.12)
V oc V oc/T
=   (A.13)
N s V t Iph
Isc
T3 Egap
kT 2
V mpp(V mpp2V t)

2 2
V mpp (2 Impp Iph) e (V t)
+ 2 V mpp V mpp
x = W 2
V t Isat Vt ( V t)
V mpp + Rs Impp
Rs = (A.14)
Vt
V mpp + Rs Impp
Rh =  V mpp+RsImpp  (A.15)
Iph Impp Isat e V t 1

A.3 Simulation results

A genetic algorithm was selected to solve the optimization problem represented by (A.8)-(A.10).
Such algorithm is implemented in Matlab and it is inside of the Global Optimization toolbox. The
final parameters set to the genetic algorithm, after some trial and error process, are summarized in

167
Table A.1.

Table A.1: Genetic Algorithm settings

Parameter Set value(s) Parameter Set value(s)
Population type Double vector Population size 500
Creation function Constraint dependent Initial range See (A.10)
Scaling function Rank Selection function Stochatic uniform
Elite count 2 Crossover fraction 80 %
Mutation function Constraint dependent Crossover Scattered
Migration direction Forward Interval 20
Initial penalty 10 Penalty factor 100
Lower bound Iph = Isc, Isat = 0, Number of variables 5
= 0.1, Rs = 0, Rh = 0

One crystalline silicon panel (BP585 [23]) was selected as an example to illustrate the application
of the method and the variation of the parameters for different irradiance and temperature condi-
tions. The main electrical characteristics of the selected panel are presented in Table 1.1 (Chapter
1.

The five parameters were calculated for different irradiances and temperature conditions. The
five key point of the I-V curve were calculated by using the equations proposed in [187] and the
coefficients provided by SAM database.

Table A.2 presents the parameters for a different irradiance and a fixed cells temperature of 25
o
C. Table A.3 shows the parameters for different temperatures and a fixed irradiance of 1 kW/m2 .
In both tables the last column represents the Normalized Sum Squared Error (NSSE) calculated for
the five key points of each I-V curve predicted by the model and the ones obtained following Kings
procedure. The matching between the I-V curve reproduced from the calculated parameters and the
reference points calculated with the model proposed by [187], for the conditions presented in Tables
A.2 and A.3, are presented in Fig. A-2 and Fig. A-3.

Table A.2: BP585 (2002) parameters for different irradiances and fixed temperature
G [W/m2 ] Iph [A] Isat [A] Rs [] Rh [] NSSE [%]
1000 5.0290 1.964e-08 1.233 0.192 1114.127 6.8401e-3
800 4.002 9.751e-10 1.068 0.264 289.424 2.9662e-3
600 3.001 1.994e-08 1.237 0.151 237.129 1.0329e-2
400 2.007 1.612e-08 1.220 0.150 649.716 4.2868e-3
200 1.001 5.917e-09 1.157 0.300 5007.149 8.6262e-3

168
 Table A.3: BP585 (2002) parameters for fixed irradiance and different temperatures
T [o C] Iph [A] Isat [A] Rs [] Rh [] NSSE [%]
25 5.029 1.964e-08 1.233 0.192 1114.127 6.8401e-3
35 5.018 8.410e-08 1.241 0.187 293.188 4.0348e-3
45 5.005 5.809e-08 1.129 0.222 355.614 6.0976e-4
55 5.008 7.590e-07 1.217 0.183 264.042 1.1517e-3
65 5.005 3.286e-06 1.244 0.182 206.302 4.8927e-3

5.5
5 Model
King
4.5
4
3.5
Current [A]

3
2.5
2
1.5
1
0.5
0
0 5 10 15 20 25
Voltage [V]

Figure A-2: I-V curves obtained form calculated parameters for different irradiances.

5.5
5
4.5
4
3.5
Current [A]

3
2.5
2
1.5
1 Model
0.5 King

0
0 5 10 15 20 25
Voltage [V]

Figure A-3: I-V curves obtained form calculated parameters for different temperatures.

169
A.4 Parameters variations with different irradiance and tem-
perature conditions

Some documents found in literature assume that only Iph and Isat are affected by irradiance and
temperature variations, while , Rs and Rh are constant [185, 116, 190]. However, the resistivity
of the semiconductor, the conductors and semiconductor-metal contacts may be affected by the
operating conditions (temperature and/or irradiance). Such relations are not clearly documented in
the literature but may affect the accuracy of a PV panel model, which may be propagated in the
prediction of a PV array current, power and energy production.

This Section presents the parameters estimated by the proposed method for 25 combinations
of five different irradiances and five different temperatures to illustrate the variability that may be
present in the parameters for different environmental conditions. Fig. A-4(a) to Fig. A-4(e) show
the parameters values for irradiance going from 0.2 to 1 kW/m2 and cell temperatures going from
25 to 65 o C.

Iph is the only parameter that shows a regular surface. This responds to the fact that Iph varies
according to irradiance and temperature as shown in (A.16), where IphST C is Iph in STC, TST C is
the temperature in STC, G is the irradiance and GST C is G in STC.

Form Fig. A-4(b) it can be observed that, in general, the inverse saturation current (Isat)
increases with the temperature and has low variations with respect to the irradiance. Such a behavior
matches with the traditional expression used calculate Isat form the module temperature (A.17).

G
Iph = (IphST C + Isc (T TST C )) (A.16)
GST C
 3 
T

Egap
Egap
Isat = IsatST C e kTST C kT (A.17)
TST C

In the surfaces of and Rs (Fig. A-4(c) and A-4(d), respectively) it can not be identified a clear
dependency with respect to the irradiance and temperature; however, there are significant variations
in the parameters values for different operating conditions. On the other hand, in the surface of Rh
it can be observed that the parameter has a relatively flat value for mid and high irradiance and
temperatures and its value increases significantly when G and T decreases.

Finally, the I-V curve for each operating condition was calculated using the parameters obtained
with the proposed method. The NSSE error for the current with respect to the five key points of
the I-V curve provided by Kings model. Such NSSE errors are presented in Fig. A-4(f). It can be
observed that the error is always lower than 0.3 %; therefore, there is a match between the model
and the reference values for a wide range of operating conditions.

170
 6
x 10

6
6

4 4

Isat [A]
Iph [A]

2 2

0 0

0.8 70 0.8 70
60 0.6 60
0.6 50 50
0.4 40 0.4 40
30 30
0.2 20 2 0.2 20
G [kW/m2] T [oC] G [kW/m ] o
T [ C]

(a) Iph for different irradiance and temperature condi- (b) Isat for different irradiance and temperature conditions.
tions.

1.4 0.4

1.3 0.3
Rs []

1.2 0.2

1.1 0.1

1 0

0.8 70 0.8 70
0.6 60 60
50 0.6 50
0.4 40 0.4 40
30 30
0.2 20 0.2 20
G [kW/m2] T [oC] G [kW/m2] T [oC]

(c) for different irradiance and temperature conditions. (d) Rs for different irradiance and temperature conditions.

6000 0.03

4000 0.02
NSSE [%]
Rh []

2000 0.01

0 0
0.2
0.4 20 0.8 70
30 60
0.6 40 0.6 50
0.8 50 0.4 40
60 30
70 2 0.2 20
T [oC] G [kW/m ] T [ C]
o
G [kW/m2]

(e) Rh for different irradiance and temperature conditions. (f) Current N SSE for the I-V curve in different irradiance
and temperature conditions.

Figure A-4: Parameters values for different irradiance and temperature conditions.

171
A.5 Remarks
The main objective of the presented method is to obtain the parameters from the five main points
of an experimental I-V without performing any simplification. A system of six non-linear equations
is posed from the five known points of the I-V curve and the derivative condition in the MPP. Such
system of non-linear equations is re-posed as an optimization problem, which is solved through a
Genetic Algorithm.
The method was tested for a mono-crystalline PV panel (BP585 (2002)) for 25 different irradiance
and temperature conditions. The low NSSE errors obtained in all the curves show the agreement
between the predicted I-V curve and the reference points obtained form the Kings model [187] and
the coefficients in SAM database.
Unlike other works reported in literature where the parameters , Rs and Rh are considered
constant for all irradiance and temperature conditions, in this document the five parameters are
calculated from the I-V curve in each operating condition in order to analyze their variability with
respect to irradiance and temperature. The graphics of the parameters shows that Iph and Isat
have behaviors similar to the ones described by the traditional equations reported in literature.
, Rs and Rh present significant variations according to the weather conditions but it can not be
identified a regular pattern. The graph of Rh shows that its magnitude increases, in an exponential
way, when irradiance and temperature decrease. Such results suggest that , Rs and Rh are affected
by irradiance and temperature and should not be considered constant to improve the accuracy of
the models.
The presented method can be used to characterize a PV panel, from experimental measurements
or from the SAM database information, by calculating the model parameters for a wide range of
irradiance and temperature conditions. Once the panel has been characterized it is possible to
train an artificial neural network (ANN) reproduce the parameters values for any given operating
condition [117]. Such an ANN can be used to correct the parameters values to improve the accuracy
of an PV panel and/or array model.
Due to soiling, aging, degradation, etc. the effective irradiance perceived by a PV panel may
be lower than the one measured by an irradiance sensor. Moreover, in a PV array the irradiance
measured by the sensor may be different from the irradiance revived by a PV panel distant form the
sensor, which introduces accuracy errors. Considering the last disadvantages, a possible improvement
to the ANN proposed in [117] could be to replace the irradiance input with the measured short-
circuit current of the panel. In such a case only the effective irradiance in each panel is considered
and there is a reduction in the initial investment due to the elimination of the irradiance sensor or
sensors.

172
Appendix B

Publications

The main results of my Ph.D. research have been or are intended to be published in journal pa-
pers and conference proceedings. The next journal papers and conference papers are published or
accepted for publication, and they contribute to the divulgation of my research results.

Journal Papers

(Published) J.D. Bastidas-Rodriguez, E. Franco, G. Petrone, C.A. Ramos-Paja, G. Spagnuolo, A

model of photovoltaic fields in mismatching conditions featuring an improved calculation speed,
Electric Power Systems Research, Vol. 96, pp. 81-90, 2013.

(Published) J.D. Bastidas-Rodriguez, C.A. Ramos-Paja, A.J. Saavedra-Montes, Reconfiguration

analysis of photovoltaic arrays based on parameters estimation, Simulation Modelling Practice and
Theory, Vol. 35, pp. 50-68, 2013.

(Published) J.D. Bastidas-Rodriguez, E. Franco, G. Petrone, C.A. Ramos-Paja, G. Spagnuolo, Max-

imum Power Point Tracking Architectures for Photovoltaic Systems in Mismatching Conditions: A
Review, IET Power Electronics, Vol. March, pp. 1-18, 2014.

(Published) J.D. Bastidas-Rodriguez, C.A. Ramos-Paja, A.J. Saavedra-Montes, Experimental Val-

idation of a Model for Photovoltaic Arrays in Total-Cross-Tied Configuration, Dyna, Vol. 80, no.
182, pp. 191-199, 2013.

(Published) J.D. Bastidas-Rodriguez, C.A. Ramos-Paja, E. Franco, Modeling and parameter cal-
culation of photovoltaic fields in irregular weather conditions, Ingeniera, Vol. 17, no. 1, pp. 37-48,
2012.

173
(Published) J.D. Bastidas-Rodriguez, C.A. Ramos-Paja, A. Trejos-Grisales, Mathematical Model
of Bridge-Linked Photovoltaic Arrays Operating Under Irregular, Tecno L
ogicas, pp. 223-235, 2013.

Conference Papers

(Published) C.A. Ramos-Paja, G. Spagnuolo, G. Petrone, M. Vitelli, J.D. Bastidas, A multivariable

MPPT algorithm for granular control of photovoltaic systems, 2010 IEEE International Symposium
on Industrial Electronics (ISIE), Bari-Italy, July 2011.

(Published) J.D. Bastidas, C.A. Ramos-Paja, E. Franco, G. Spagnuolo, G. Petrone, Modeling of

photovoltaic fields in mismatching conditions by means of inflection voltages, 2012 Workshop on
Engineering Applications (WEA), Bogota-Colombia, May 2012.

(Published) C.A. Ramos-Paja, J.D. Bastidas, A.J. Saavedra-Montes, F. Guinjoan-Gispert, M. Goez,

Mathematical model of total cross-tied photovoltaic arrays in mismatching conditions, 2012 IEEE
4th Colombian Workshop on Circuits and Systems (CWCAS), Barranquilla-Colombia, November
2012

(Published) J.D. Bastidas, G. Petrone, C.A. Ramos-Paja, G. Spagnuolo, Photovoltaic modules diag-
nostic: an overview, 39th Annual Conference of the IEEE Industrial Electronics Society (IECON),
Vienna-Austria, November 2013.

(Published) J.D. Bastidas, G. Petrone, C.A. Ramos-Paja, G. Spagnuolo, Parameter Calculation of

Photovoltaic Modules using a Genetic Algorithm, 11th International Conference on Modeling and
Simulation of Electric Machines, Converters and Systems (ELECTRIMACS 2014), Valencia-Spain,
May 2014.

(Published) J.D. Bastidas, G. Petrone, C.A. Ramos-Paja, G. Spagnuolo, Model Based Indicators
to Quantify Photovoltaic Module Degradation, 11th International Conference on Modeling and
Simulation of Electric Machines, Converters and Systems (ELECTRIMACS 2014), Valencia-Spain,
May 2014.

174
Bibliography

[1] IEA, Technology Roadmap Solar photovoltaic energy, International Energy Agency, Tech. Rep., 2010.
[Online]. Available: http://www.iea.org/papers/2010/pv roadmap.pdf

[2] , TRENDS IN PHOTOVOLTAIC APPLICATIONS, Survey report of selected IEA countries between
1992 and 2011, International Energy Agency, Tech. Rep., 2012. [Online]. Available: http://www.iea-pvps.org/

[3] M. Munoz, M. Alonso-Garca, N. Vela, and F. Chenlo, Early degradation of silicon PV modules and guaranty
conditions, Solar Energy, vol. 85, no. 9, pp. 22642274, Sep. 2011.

[4] K. Kato, PVRessQ! PV Module Failures Observed in the Field, in Photovoltaic Module Reliability Workshop
2012, 2012, pp. 124.

[5] R. Messenger and J. Ventre, Photovoltaic systems engineering. CRC, 2004.

[6] M. Villalva, J. Gazoli, and E. Filho, Comprehensive Approach to Modeling and Simulation of Photovoltaic
Arrays, IEEE Trans. Power Electron., vol. 24, no. 5, pp. 11981208, May 2009.

[7] J. Accarino, G. Petrone, C. a. Ramos-Paja, and G. Spagnuolo, Symbolic algebra for the calculation of the
series and parallel resistances in PV module model, in 2013 International Conference on Clean Electrical
Power (ICCEP). IEEE, Jun. 2013, pp. 6266.

[8] W. De Soto, S. Klein, and W. Beckman, Improvement and validation of a model for photovoltaic array
performance, Solar Energy, vol. 80, no. 1, pp. 7888, Jan. 2006.

[9] D. Chan and J. Phang, Analytical methods for the extraction of solar-cell single- and double-diode model
parameters from I-V characteristics, IEEE Trans. Electron Devices, vol. 34, no. 2, pp. 286293, Feb. 1987.

[10] E. V. Paraskevadaki and S. A. Papathanassiou, Evaluation of MPP Voltage and Power of mc-Si PV Modules
in Partial Shading Conditions, IEEE Transactions on Energy Conversion, vol. 26, no. 3, pp. 923932, 2011.

[11] R. M. Corless, G. H. Gonnet, D. E. G. Hare, D. J. Jeffrey, and D. E. Knuth, On the LambertW function,
Advances in Computational Mathematics, vol. 5, no. 1, pp. 329359, Dec. 1996.

[12] G. Petrone and C. Ramos-Paja, Modeling of photovoltaic fields in mismatched conditions for energy yield
evaluations, Electric Power Systems Research, vol. 81, no. 4, pp. 10031013, 2011.

[13] T. Shimizu, M. Hirakata, T. Kamezawa, and H. Watanabe, Generation control circuit for photovoltaic mod-
ules, IEEE Transactions on Power Electronics, vol. 16, no. 3, pp. 293300, 2001.

[14] ERDM Solar, Datasheet ERDM SOLAR 75SM 80SM5 85SM. [Online]. Available:
http://www.erdmsolar.com/shop/images/stories/file/SM580.pdf

[15] H. Patel and V. Agarwal, MATLAB-Based Modeling to Study the Effects of Partial Shading on PV Array
Characteristics, IEEE Transactions on Energy Conversion, vol. 23, no. 1, pp. 302310, Mar. 2008.

[16] G. Petrone, G. Spagnuolo, and M. Vitelli, Analytical model of mismatched photovoltaic fields by means of
Lambert W-function, Sol. Energy Mater. and Sol. Cells, vol. 91, no. 18, pp. 16521657, Nov. 2007.

175
[17] J. D. Bastidas, E. Franco, G. Petrone, C. A. Ramos-Paja, and G. Spagnuolo, A model of photovoltaic fields
in mismatching conditions featuring an improved calculation speed, Electric Power Systems Research, vol. 96,
pp. 8190, 2013.

[18] N. Femia, G. Lisi, G. Petrone, G. Spagnuolo, and M. Vitelli, Distributed Maximum Power Point Tracking
of Photovoltaic Arrays: Novel Approach and System Analysis, IEEE Transactions on Industrial Electronics,
vol. 55, no. 7, pp. 26102621, 2008.

[19] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, Optimization of Perturb and Observe Maximum Power
Point Tracking Method, IEEE Transactions on Power Electronics, vol. 20, no. 4, pp. 963973, 2005.

[20] T. Esram and P. L. Chapman, Comparison of Photovoltaic Array Maximum Power Point Tracking Techniques,
IEEE Transactions on Energy Conversion, vol. 22, no. 2, pp. 439449, 2007.

[21] S. Kjaer, Evaluation of the hill climbing and the incremental conductance maximum power point trackers for
photovoltaic power systems, IEEE Transactions on Energy Conversion, vol. 27, no. 4, pp. 922 929, 2012.

[22] G. Velasco-Quesada, F. Guinjoan-Gispert, R. Pique-Lopez, M. Roman-Lumbreras, and A. Conesa-Roca, Elec-

trical PV Array Reconfiguration Strategy for Energy Extraction Improvement in Grid-Connected PV Systems,
IEEE Transactions on Industrial Electronics, vol. 56, no. 11, pp. 43194331, 2009.

[23] BP Solar, Datasheet BP 585. [Online]. Available: http://www.oksolar.com/pdfiles/Solar Panels bp 585.pdf

[24] Kyocera, Datasheet KC120-1. [Online]. Available: http://www.kyocerasolar.com/assets/001/5180.pdf

[25] A. S. Bouhouras, A. G. Marinopoulos, D. P. Labridis, and P. S. Dokopoulos, Installation of PV systems in

GreeceReliability improvement in the transmission and distribution system, Electr. Power Syst. Res., vol. 80,
no. 5, pp. 547555, May 2010.

[26] J. Bastidas-Rodriguez, E. Franco, G. Petrone, C. Ramos-paja, and G. Spagnuolo, Maximum power point
tracking architectures for photovoltaic systems in mismatching conditions : a review, IET Power Electronics,
pp. 118, 2014.

[27] A. Goetzberger and V. Hoffmann, Cap 10. Efficiency and Performance of PV Systems, in Photovoltaic Solar
Energy Generation, 2005, pp. 147161.

[28] H. Patel and V. Agarwal, Maximum Power Point Tracking Scheme for PV Systems Operating Under Partially
Shaded Conditions, IEEE Transactions on Industrial Electronics, vol. 55, no. 4, pp. 16891698, Apr. 2008.

[29] P. Tsao, S. Sarhan, I. Jorio, and N. Semiconductor, Distributed max power point tracking for photovoltaic
arrays, in 34th IEEE Photovoltaic Specialists Conference (PVSC), 2009, pp. 002 293002 298.

[30] R. Faranda, S. Leva, and V. Maugeri, MPPT techniques for PV Systems: Energetic and cost comparison, in
IEEE Power and Energy Society General Meeting - Conversion and Delivery of Electrical Energy in the 21st
Century, 2008, pp. 16.

[31] W. Xiao, Evaluating maximum power point tracking performance by using artificial lights, 30th Annual
Conferen of the IEEE Industrial Electronics Society, vol. 3, pp. 28832887, 2004.

[32] D. Hohm and M. Ropp, Comparative study of maximum power point tracking algorithms using an experi-
mental, programmable, maximum power point tracking test bed, in Conference Record of the Twenty-Eighth
IEEE Photovoltaic Specialists Conference, 2000, pp. 16991702.

[33] D. Ahmadi, S. A. Mansouri, and J. Wang, Circuit topology study for distributed MPPT in very large scale PV
power plants, in Applied Power Electronics Conference and Exposition (APEC), vol. 1, no. d. IEEE, 2011,
pp. 786791.

[34] S. Poshtkouhi, J. Varley, and R. Popuri, Analysis of distributed peak power tracking in photovoltaic systems,
in International Power Electronics Conference (IPEC). IEEE, 2010, pp. 942947.

176
[35] A. Bidram, A. Davoudi, and R. S. Balog, Control and Circuit Techniques to Mitigate Partial Shading Effects
in Photovoltaic Arrays, IEEE Journal of Photovoltaics, vol. PP, no. 99, pp. 115, 2012.

[36] K. Ishaque and Z. Salam, A review of maximum power point tracking techniques of PV system for uniform
insolation and partial shading condition, Renewable and Sustainable Energy Reviews, vol. 19, pp. 475488,
Mar. 2013.

[37] H. J. Hovel, Semiconductors and Semimetals Volume 11 Solar Cells, R. K. Willardson and A. C. Beer, Eds.
Academic Press, 1975, vol. 11.

[38] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, Power Electronics and Control Techniques for Maximum
Energy Harvesting in Photovoltaic Systems. CRC Press, 2013.

[39] E. Bianconi, J. Calvente, R. Giral, E. Mamarelis, G. Petrone, C. A. Ramos-Paja, G. Spagnuolo, and M. Vitelli,
Perturb and Observe MPPT algorithm with a current controller based on the sliding mode, International
Journal of Electrical Power & Energy Systems, vol. 44, no. 1, pp. 346356, 2013.

[40] E. Bianconi, J. Calvente, R. Giral, E. Mamarelis, G. Petrone, C. Ramos-Paja, G. Spagnuolo, and M. Vitelli, A
fast current-based mppt technique employing sliding mode control, Industrial Electronics, IEEE Transactions
on, vol. 60, no. 3, pp. 11681178, 2013.

[41] Y. Levron and D. Shmilovitz, Maximum power point tracking employing sliding mode control, Circuits and
Systems I: Regular Papers, IEEE Transactions on, vol. 60, no. 3, pp. 724732, 2013.

[42] G. Petrone, G. Spagnuolo, and M. Vitelli, Distributed maximum power point tracking: challenges and commer-
cial solutions, Automatika: Journal for Control, Measurement, Electronics, Computing and Communications,
vol. 2, no. 53, pp. 128141, 2012.

[43] C. A. Ramos-paja, A. J. Saavedra-Montes, and M. Vitelli, Distributed Maximum Power Point Tracking With
Overvoltage Protection For PV Systems, Dyna, vol. 80, no. 178, pp. 141150, 2013.

[44] L. Linares, R. W. Erickson, S. MacAlpine, and M. Brandemuehl, Improved Energy Capture in Series String
Photovoltaics via Smart Distributed Power Electronics, in Twenty-Fourth Annual IEEE Applied Power Elec-
tronics Conference and Exposition, 2009, pp. 904910.

[45] J. Zhang, A. Djabbari, and G. Lisi, Method And System For Selecting Between Centralize And Distributed
Maximum Power Point Tracking In An Energy Generating System. WO2009/140551, pp. 169, 2009.

[46] J. Johnson, Overview of Arc-Faults and Detection Challenges Arcing in PV Systems, Sandia National
Laboratories, Tech. Rep., 2011. [Online]. Available: http://energy.sandia.gov

[47] J. Strauch, Arc Fault PV Industry Activities , and Testing and Modeling Current PV Industry Activities in
Arc Fault, Sandia National Laboratories, Tech. Rep., 2011. [Online]. Available: http://energy.sandia.gov

[48] P. Lei, Y. Li, and J. Seem, Sequential ESC Based Global MPPT Control for Photovoltaic Array with Variable
Shading, IEEE Transactions on Sustainable Energy, vol. 2, no. 99, pp. 11, 2011.

[49] B. N. Alajmi, K. H. Ahmed, S. J. Finney, and B. W. Williams, A Maximum Power Point Tracking Technique
for Partially Shaded Photovoltaic Systems in Microgrids, IEEE Transactions on Industrial Electronics, vol. 60,
no. 4, pp. 15961606, 2013.

[50] R. Alonso, P. Ibaez, V. Martinez, E. Roman, and A. Sanz, An innovative perturb, observe and check algorithm
for partially shaded PV systems, in 13th European Conference on Power Electronics and Applications, EPE
09, 2009, pp. 18.

[51] G. Carannante, C. Fraddanno, M. Pagano, and L. Piegari, Experimental Performance of MPPT Algorithm
for Photovoltaic Sources Subject to Inhomogeneous Insolation, IEEE Transactions on Industrial Electronics,
vol. 56, no. 11, pp. 43744380, 2009.

177
[52] E. Koutroulis and F. Blaabjerg, A New Technique for Tracking the Global Maximum Power Point of PV Arrays
Operating Under Partial-Shading Conditions, IEEE Journal of Photovoltaics, vol. 2, no. 2, pp. 184190, 2012.

[53] Y.-H. Ji, D.-Y. Jung, J.-G. Kim, J.-H. Kim, T.-W. Lee, and C.-Y. Won, A Real Maximum Power Point Tracking
Method for Mismatching Compensation in PV Array Under Partially Shaded Conditions, IEEE Transactions
on Power Electronics, vol. 26, no. 4, pp. 10011009, 2011.

[54] L. Zhou, Y. Chen, K. Guo, and F. Jia, New approach for MPPT control of photovoltaic system with mutative
scale dual carrier chaotic search, IEEE Transactions on Power Electronics, vol. 26, no. 99, pp. 10381048,
2011.

[55] R. Ramaprabha, M. Balaji, and B. L. Mathur, Maximum power point tracking of partially shaded solar
PV system using modified Fibonacci search method with fuzzy controller, International Journal of Electrical
Power & Energy Systems, vol. 43, no. 1, pp. 754765, 2012.

[56] N. a. Ahmed and M. Miyatake, A novel maximum power point tracking for photovoltaic applications under
partially shaded insolation conditions, Electric Power Systems Research, vol. 78, no. 5, pp. 777784, 2008.

[57] Y.-h. Liu, S.-C. S.-c. Huang, J.-w. J.-W. Huang, and W.-c. W.-C. Liang, A Particle Swarm Optimization-Based
Maximum Power Point Tracking Algorithm for PV Systems Operating Under Partially Shaded Conditions,
IEEE Transactions on Energy Conversion, vol. 27, no. 4, pp. 10271035, 2012.

[58] K. Ishaque, Z. Salam, M. Amjad, and S. Mekhilef, An Improved Particle Swarm Optimization (PSO)Based
MPPT for PV With Reduced Steady-State Oscillation, IEEE Transactions on Power Electronics, vol. 27,
no. 8, pp. 36273638, 2012.

[59] K. Ishaque, Z. Salam, A. Shamsudin, and M. Amjad, A direct control based maximum power point tracking
method for photovoltaic system under partial shading conditions using particle swarm optimization algorithm,
Applied Energy, vol. 99, pp. 414422, 2012.

[60] Syafaruddin, E. Karatepe, and T. Hiyama, Artificial neural network-polar coordinated fuzzy controller based
maximum power point tracking control under partially shaded conditions, IET Renewable Power Generation,
vol. 3, no. 2, pp. 239253, 2009.

[61] R. Dougal and a.P. Iotova, Parallel-Connected Solar PV System to Address Partial and Rapidly Fluctuating
Shadow Conditions, IEEE Transactions on Industrial Electronics, vol. 56, no. 5, pp. 15481556, 2009.

[62] E. E. V. Solodovnik, S. Liu, S. Member, and R. A. Dougal, Power Controller Design for Maximum Power
Tracking in Solar Installations, IEEE Transactions on Power Electronics, vol. 19, no. 5, pp. 12951304, 2004.

[63] E. Karatepe, S. Syafaruddin, and T. Hiyama, Simple and high-efficiency photovoltaic system under non-uniform
operating conditions, IET Renewable Power Generation, vol. 4, no. 4, pp. 354368, 2010.

[64] [Online]. Available: http://www.solarmagic.com/en/

[65] [Online]. Available: http://www.tigoenergy.com/it/products

[66] [Online]. Available: http://www.eiqenergy.com/index.html

[67] R. Hester, C. Thornton, and S. Dhople, High efficiency wide load range buck/boost/bridge photovoltaic mi-
croconverter, in Twenty-Sixth Annual IEEE Applied Power Electronics Conference and Exposition (APEC),
2011, pp. 309313.

[68] R. W. J. Erickson, S. Croft, S. Everson, and A. Schultz, Thin-Film Photovoltaic Power Element With Integrated
Low-Profile High-Efficiency DC-DC Converter. WO2010/093702A2, pp. 142, 2010.

[69] D. Perreault, R. Pilawa-Podgurski, C. Sullivan, and A. Latham, Power Processing Methods And Apparatus
For Photovoltaic Systems. WO2011/163437, pp. 198, 2011.

178
[70] S. Ang and C. Burgers, Keith, Distributed Maximum Power Point Tracking Converter. WO2009/035995, pp.
129, 2009.

[71] A. Ledenev and M. Porter, Robert, Systems For Highly Efficient Solar Power. WO2009/051853, pp. 1103,
2009.

[72] E. Roman, R. Alonso, P. Ibanez, S. Elorduizapatarietxe, and D. Goitia, Intelligent PV Module for Grid-
Connected PV Systems, IEEE Transactions on Industrial Electronics, vol. 53, no. 4, pp. 10661073, 2006.

[73] E. Roman, V. Martinez, J. Jimeno, R. Alonso, P. Ibanez, and S. Elorduizapatarietxe, Experimental results of
controlled PV module for building integrated PV systems, Solar Energy, vol. 82, no. 5, pp. 471480, 2008.

[74] R. Kadri, J.-P. Gaubert, and G. Champenois, Nondissipative String Current Diverter for Solving the Cascaded
DCDC Converter Connection Problem in Photovoltaic Power Generation System, IEEE Transactions on Power
Electronics, vol. 27, no. 3, pp. 12491258, 2012.

[75] A. I. Bratcu, I. Munteanu, S. Bacha, D. Picault, and B. Raison, Cascaded DCDC Converter Photovoltaic
Systems: Power Optimization Issues, IEEE Transactions on Industrial Electronics, vol. 58, no. 2, pp. 403
411, 2011.

[76] Y. Du and D. D.-C. Lu, Battery-integrated boost converter utilizing distributed MPPT configuration for
photovoltaic systems, Solar Energy, vol. 85, no. 9, pp. 19922002, 2011.

[77] G. Petrone, G. Spagnuolo, and M. Vitelli, TEODI: A new technique for Distributed Maximum Power Point
Tracking PV Applications, in IEEE International Conference on Industrial Technology, vol. 0, no. 1, 2010,
pp. 982987.

[78] M. A. De Rooij, J. S. Glaser, O. G. Mayer, and S. F. S. El-Barbari, Quasi-ac, photovoltaic module for unfolder
photovoltaic inverter. EP2166638A2, pp. 115, 2009.

[79] S. Arditi, Systems And Methods For Detecting and Correcting A Suboptimal Operation Of One Or More
Inverters In A Multi-Inverter System. WO2011/119587, pp. 123, 2011.

[80] A. E. Shaver and R. M. Newdoll, Distributed Maximum Power Point Tracking System, Structure An Process.
WO2008/119034, pp. 158, 2008.

[81] T. Shimizu, O. Hashimoto, and G. Kimura, A novel high-performance utility-interactive photovoltaic inverter
system, IEEE Transactions on Power Electronics, vol. 18, no. 2, pp. 704711, 2003.

[82] H. W. Lee, K. H. Koh, K. Y. Suh, and K. H. Suk, Method For Compensating For Partial Shade In Photovoltaic
Power System. WO2005/112551, pp. 117, 2005.

[83] H. C. J. Buthker, K. De Waal, and H. J. Bergveld, A photovoltaic unit, a dc-dc converter therefor, and a
method of operating the same. EP2280469, pp. 135, 2011.

[84] A. Stratakos and A. Ikiannikov, System, Method, Module, And Energy Exchanger for Optimizing Output of
Series-Connected Photovoltaic And Electrochemical Devices. WO2012/024536, pp. 1110, 2012.

[85] L. F. L. Villa, T.-P. Ho, J.-C. Crebier, and B. Raison, A Power Electronics Equalizer Application for Partially
Shaded Photovoltaic Modules, IEEE Transactions on Industrial Electronics, vol. 60, no. 3, pp. 11791190,
2013.

[86] Y. Nimni and D. Shmilovitz, A returned energy architecture for improved photovoltaic systems efficiency, in
IEEE International Symposium on Circuits and Systems (ISCAS), 2010, pp. 21912194.

[87] E. Villanueva, P. Correa, J. Rodrguez, S. Member, and M. Pacas, Control of a Single-Phase Cascaded H-Bridge
Multilevel Inverter for Grid-Connected Photovoltaic Systems, IEEE Transaction on Industrial Electronics,
vol. 56, no. 11, pp. 43994406, 2009.

179
 [88] G. Petrone, C. A. Ramos-Paja, G. Spagnuolo, and M. Vitelli, Granular control of photovoltaic arrays by
means of a multi-output Maximum Power Point Tracking algorithm, Progress in Photovoltaics: Research and
Applications, 2012.

[89] C. Ramos-Paja, a. Romero, R. Giral, J. Calvente, and L. Martinez-Salamero, Mathematical analysis of hybrid
topologies efficiency for PEM fuel cell power systems design, International Journal of Electrical Power &
Energy Systems, vol. 32, no. 9, pp. 10491061, 2010.

[90] L. Nousiainen and T. Suntio, Dual-mode current-fed semi-quadratic buck-boost converter for transformer-
less modular photovoltaic applications, in 14th European Conference on Power Electronics and Applications
(EPE), 2011, pp. 110.

[91] O. Abdel-Rahim, M. Orabi, and M. Ahmed, Development of high-gain high-efficiency grid-connected inverter
for PV Module, in 2nd IEEE International Symposium on Power Electronics for Distributed Generation
Systems (PEDG), 2010, pp. 368373.

[92] J. Reynaud, O. Gantet, P. Aloisi, B. Estibals, and C. Alonso, A novel distributed photovoltaic power architec-
ture using advanced Li-ion batteries, in 14th International Power Electronics and Motion Control Conference
(EPE/PEMC), 2010, pp. S96S912.

[93] J. Erickson, Robert Warren, Integrated Photovoltaic Module. WO2010/132369, pp. 139, 2010.

[94] R. W. J. Erickson, S. Pisklak, and T. PLUM, Low Profile Power Conversion System For Rooftop Photovoltaic
Power Systems. WO2010/153106, pp. 135, 2011.

[95] W. Yao, M. Gao, Z. Ren, M. Chen, and Z. Qian, Improvement of performance and flexibility for photovoltaic
module using individual DC/DC converter, in IEEE 6th International Power Electronics and Motion Control
Conference, vol. 3, 2009, pp. 441444.

[96] S. Rai, U. Kumar, and R. Naik, A novel technique for photovoltaic maximum power point tracking system,
in IEEE European Conference on Power Electronics and Applications, 2005, pp. 18.

[97] H. Kim, J. Kim, H. Kim, K. Lee, J. Kim, D. Yoo, and D. Shin, A high efficiency photovoltaic module integrated
converter with the asymmetrical half-bridge flyback converter, Solar Energy, vol. 84, no. 8, pp. 13761381,
2010.

[98] S. Dhople, J. Ehlmann, A. Davoudi, and P. Chapman, Multiple-input boost converter to minimize power losses
due to partial shading in photovoltaic modules, in IEEE Energy Conversion Congress and Exposition (ECCE),
2010, pp. 26332636.

[99] S. Poshtkouhi and O. Trescases, Multi-input Single-inductor DC-DC Converter for MPPT in Parallel-
Connected Photovoltaic Applications, in Twenty-Sixth Annual IEEE Applied Power Electronics Conference
and Exposition (APEC), vol. 4, 2011, pp. 4147.

[100] T. Noguchi, S. Togashi, and R. Nakamoto, Short-current pulse-based maximum-power-point tracking method
for multiple photovoltaic-and-converter module system, IEEE Transactions on Industrial Electronics, vol. 49,
no. 1, pp. 217223, 2002.

[101] Syafaruddin, E. Karatepe, and T. Hiyama, Performance enhancement of photovoltaic array through string and
central based MPPT system under non-uniform irradiance conditions, Energy Conversion and Management,
vol. 62, pp. 131140, 2012.

[102] M. Miyatake, M. Veerachary, F. Toriumi, N. Fujii, and H. Ko, Maximum Power Point Tracking of Multiple
Photovoltaic Arrays: A PSO Approach, IEEE Transactions on Aerospace and Electronic Systems, vol. 47,
no. 1, pp. 367380, 2011.

180
[103] E. Karatepe, T. Hiyama, M. Boztepe, and M. C
olak, Voltage based power compensation system for photovoltaic
generation system under partially shaded insolation conditions, Energy Conversion and Management, vol. 49,
no. 8, pp. 23072316, 2008.

[104] , Power controller design for photovoltaic generation system under partially shaded insolation conditions,
in International Conference on Intelligent Systems Applications to Power Systems (ISAP), 2007, pp. 16.

[105] C. Hargis, Voltage Compensation. WO2011/076707, pp. 140, 2011.

[106] J. D. Bastidas-Rodriguez, C. A. Ramos-Paja, and A. J. Saavedra-Montes, Reconfiguration analysis of photo-

voltaic arrays based on parameters estimation, Simulation Modelling Practice and Theory, vol. 35, pp. 5068,
2013.

[107] Bitron, Endana - Optimization, monitoring and safety for photovoltaics, pp. 12, 2013. [Online]. Available:
http://www.bitronenergy.com/documents/endana.pdf

[108] D. Nguyen and B. Lehman, An Adaptive Solar Photovoltaic Array Using Model-Based Reconfiguration Algo-
rithm, IEEE Transactions on Industrial Electronics, vol. 55, no. 7, pp. 26442654, 2008.

[109] M. Z. M. A. Shamseldein, M. Kazerani, and M. Salama, Reconfigurable Photovoltaic Structure.

WO2012/006723, pp. 155, 2012.

[110] M. Alahmad, M. A. Chaaban, S. K. Lau, J. Shi, and J. Neal, An adaptive utility interactive photovoltaic
system based on a flexible switch matrix to optimize performance in real-time, Solar Energy, vol. 86, no. 3,
pp. 951963, 2012.

[111] H. Patel and V. Agarwal, MATLAB-Based Modeling to Study the Effects of Partial Shading on PV Array
Characteristics, IEEE Trans. Energy Convers., vol. 23, no. 1, pp. 302310, 2008.

[112] H. Tsai and C. Tu, Development of generalized photovoltaic model using MATLAB/SIMULINK, in Proceed-
ings of the World Congress on Engineering and Computer Science 2008, 2008, pp. 16.

[113] K.-H. Chao, S.-H. Ho, and M.-H. Wang, Modeling and fault diagnosis of a photovoltaic system, Electr. Power
Syst. Res., vol. 78, no. 1, pp. 97105, Jan. 2008.

[114] A. R. Di Fazio and M. Russo, Photovoltaic generator modelling to improve numerical robustness of EMT
simulation, Electr. Power Syst. Res., vol. 83, no. 1, pp. 136143, Feb. 2012.

[115] R. Ramabadran, Matlab based modelling and performance study of series connected SPVA under partial
shaded conditions, J. Sustain. Dev., vol. 2, no. 3, pp. 8594, 2009.

[116] V. Lo Brano and G. Ciulla, An efficient analytical approach for obtaining a five parameters model of photo-
voltaic modules using only reference data, Applied Energy, vol. 111, pp. 894903, Nov. 2013.

[117] E. Karatepe, M. Boztepe, and M. Colak, Neural network based solar cell model, Energy Conversion and
Management, vol. 47, no. 9-10, pp. 11591178, Jun. 2006.

[118] J. A. Duffie and W. A. Beckman, Solar Engineering of Thermal Processes, 3rd ed. Wiley, 2006.

[119] Free Software Foundation Inc., GSL - GNU Scientific Library. [Online]. Available:
http://www.gnu.org/software/gsl/

[120] C. A. Ramos-paja and J. D. Bastidas, Experimental Validation of a Model for Photovoltaic Arrays in Total-
Cross-Tied Configuration, Dyna, vol. 80, no. 182, pp. 191199, 2013.

[121] J. Bastidas-Rodriguez, C. Ramos-Paja, and A. Trejos-Grisales, Mathematical Model of Bridge-Linked Photo-

voltaic Arrays Operating Under Irregular, Tecno L
ogicas, pp. 223235, 2013.

[122] J. D. Bastidas, C. A. Ramos-Paja, E. Franco, G. Spagnuolo, and G. Petrone, Modeling of photovoltaic fields
in mismatching conditions by means of inflection voltages, in 2012 Workshop on Engineering Applications
(WEA), 2012, pp. 16.

181
[123] C. a. Ramos-Paja, J. D. Bastidas, a. J. Saavedra-Montes, F. Guinjoan-Gispert, and M. Goez, Mathematical
model of total cross-tied photovoltaic arrays in mismatching conditions, in 2012 IEEE 4th Colombian Workshop
on Circuits and Systems (CWCAS). Ieee, Nov. 2012, pp. 16.

[124] Y.-J. Wang and P.-C. Hsu, An investigation on partial shading of PV modules with different connection
configurations of PV cells, Energy, vol. 36, no. 5, pp. 30693078, 2011.

[125] M. Z. Shams El-Dein, M. Kazerani, M. M. A. Salama, M. Z. S. El-dein, and S. S. Member, An Optimal Total
Cross Tied Interconnection for Reducing Mismatch Losses in Photovoltaic Arrays, Sustainable Energy, IEEE
Transactions on, vol. 4, no. 1, pp. 99107, 2012.

[126] D. Picault, B. Raison, S. Bacha, J. de la Casa, and J. Aguilera, Forecasting photovoltaic array power production
subject to mismatch losses, Solar Energy, vol. 84, no. 7, pp. 13011309, Jul. 2010.

[127] S. Silvestre, a. Boronat, and a. Chouder, Study of bypass diodes configuration on PV modules, Applied
Energy, vol. 86, no. 9, pp. 16321640, 2009.

[128] J. Hudson, L. Vasilyev, J. Schmidt, and G. Horner, Economic impacts and approaches to address hot-spot
defects in photovoltaic devices, in 2010 35th IEEE Photovoltaic Specialists Conference. IEEE, Jun. 2010, pp.
001 706001 709.

[129] M. Kolhe, S. Kolhe, and J. Joshi, Economic viability of stand-alone solar photovoltaic system in comparison
with diesel-powered system for India, Energy Economics, vol. 24, no. 2, pp. 155165, 2002.

[130] G. Bakos and M. Soursos, Technical feasibility and economic viability of a grid-connected PV installation for
low cost electricity production, Energy and Buildings, vol. 34, no. 7, pp. 753758, 2002.

[131] a. Al Tarabsheh, I. Etier, and A. Nimrat, Energy Yield of Tracking PV Systems in Jordan, International
Journal of Photoenergy, vol. 2012, pp. 15, 2012.

[132] Y. Zhao, L. Yang, and B. Lehman, Reconfigurable solar photovoltaic battery charger using a switch matrix,
in 2012 IEEE 34th International Telecommunications Energy Conference (INTELEC), 2012, pp. 17.

[133] M. Z. M. A. Shamseldein, M. Kazerani, and M. Salama, System, Method And Computer Program For Reducing
Mismatch In A Photovoltaic Structure. WO2011/143739, pp. 130, 2011.

[134] Z. Cheng, Z. Pang, Y. Liu, and P. Xue, An adaptive solar photovoltaic array reconfiguration method based
on fuzzy control, in 2010 IEEE 8th World Congress on Intelligent Control and Automation, Jul. 2010, pp.
176181.

[135] N. Seddaoui, L. Rahmani, and A. Chauder, Parameters extraction of photovoltaic module at reference and
real conditions, in EFEEA10 International Symposium on Environment Friendly Energies in Electrical Ap-
plications, 2010, pp. 16.

[136] A. Orioli and A. Di Gangi, A procedure to calculate the five-parameter model of crystalline silicon photovoltaic
modules on the basis of the tabular performance data, Applied Energy, vol. 102, pp. 11601177, Feb. 2013.

[137] D. Sera, R. Teodorescu, and P. Rodriguez, PV panel model based on datasheet values, in 2007 IEEE Inter-
national Symposium on Industrial Electronics, Jun. 2007, pp. 23922396.

[138] A. Chatterjee, A. Keyhani, and D. Kapoor, Identification of Photovoltaic Source Models, IEEE Transactions
on Energy Conversion, vol. 26, no. 3, pp. 883889, 2011.

[139] G. Farivar and B. Asaei, A New Approach for Solar Module Temperature Estimation Using the Simple Diode
Model, IEEE Transactions on Energy Conversion, vol. 26, no. 4, pp. 11181126, Dec. 2011.

[140] U. Eicker, Solar Technologies for Buildings. Chichester, UK: John Wiley & Sons, Ltd, Jun. 2003.

182
[141] K. Ishaque, Z. Salam, and H. Taheri, Modeling and simulation of photovoltaic (PV) system during partial
shading based on a two-diode model, Simulation Modelling Practice and Theory, vol. 19, no. 7, pp. 16131626,
Aug. 2011.

[142] F. Adamo, F. Attivissimo, A. Di Nisio, and M. Spadavecchia, Characterization and Testing of a Tool for
Photovoltaic Panel Modeling, IEEE Transactions on Instrumentation and Measurement, vol. 60, no. 5, pp.
16131622, May 2011.

[143] A. Chouder, S. Silvestre, N. Sadaoui, and L. Rahmani, Modeling and simulation of a grid connected PV system
based on the evaluation of main PV module parameters, Simulation Modelling Practice and Theory, vol. 20,
no. 1, pp. 4658, Jan. 2012.

[144] E. Karatepe, M. Boztepe, and M. C

olak, Development of a suitable model for characterizing photovoltaic
arrays with shaded solar cells, Solar Energy, vol. 81, no. 8, pp. 977992, Aug. 2007.

[145] W. Xiao, M. Lind, W. Dunford, and A. Capel, Real-Time Identification of Optimal Operating Points in
Photovoltaic Power Systems, IEEE Transactions on Industrial Electronics, vol. 53, no. 4, pp. 10171026, Jun.
2006.

[146] A. Bellini and S. Bifaretti, MPPT algorithm for current balancing of partially shaded Photovoltaic modules,
Industrial Electronics (ISIE), pp. 933938, 2010.

[147] S. Kjaer, Evaluation of the Hill Climbing and the Incremental Conductance Maximum Power Point Trackers
for Photovoltaic Power Systems, IEEE Transactions on Energy Conversion, vol. PP, no. 99, pp. 18, 2012.

[148] C. Kelley, Iterative methods for optimization. Raleigh, North Carolina: Society for Industrial and Applied
Mathematics, 1987.

[149] S. Kaplanis and E. Kaplani, Energy performance and degradation over 20 years performance of BP c-Si PV
modules, Simulation Modelling Practice and Theory, vol. 19, no. 4, pp. 12011211, Apr. 2011.

[150] A. Conn, N. Gould, and P. Toint, Trust Region Methods. Society for Industrial and Applied Mathematics,
2000.

[151] J. Bastidas-Rodriguez, E. Franco, G. Petrone, and G. Spagnuolo, Model Based Indicators To Quantify Pho-
tovoltaic Module Degradation, in 11th International Conference on Modeling and Simulation of Electric Ma-
chines, Converters and Systems (ELECTRIMACS 2014), no. May, 2014, pp. 4954.

[152] C. R. Osterwald, Standards , Calibration , and Testing of PV Modules and Solar Cells, in Practical Handbook
of Photovoltaics. Elsevier Ltd, 2012, ch. III-2, pp. 10451070.

[153] M. Orozco-Gutierrez, J. Ramirez-Scarpetta, G. Spagnuolo, and C. Ramos-Paja, A technique for mismatched

PV array simulation, Renewable Energy, vol. 55, pp. 417427, 2013.

[154] J. Bastidas-Rodriguez, G. Petrone, C. Ramos-Paja, and G. Spagnuolo, Photovoltaic modules diagnostic: an

overview, in 39th Annual Conference of the IEEE Industrial Electronics Society, 2013, pp. 16.

[155] IEC - International Electrotechnical Commission, IEC 61215 Second edition: Crystalline silicon terrestrial
photovoltaic (PV) modules - Design qualification and type approval, pp. 185, 2005.

[156] , IEC 61646 Second edition: Thin-film terrestrial photovoltaic (PV) modules - Design qualification and
type approval, pp. 136, 2008.

[157] , Photovoltaic (PV) module safety qualification - Part 1: Requirements for construction, pp. 139, 2013.

[158] P. Hacke, R. Smith, K. Terwilliger, S. Glick, D. Jordan, S. Johnston, M. Kempe, and S. Kurtz, Testing and
Analysis for Lifetime Prediction of Crystalline Silicon PV Modules Undergoing Degradation by System Voltage
Stress, IEEE Journal of Photovoltaics, vol. 3, no. 1, pp. 246253, Jan. 2013.

183
[159] E. van Dyk and E. Meyer, Analysis of the effect of parasitic resistances on the performance of photovoltaic
modules, Renewable Energy, vol. 29, no. 3, pp. 333344, Mar. 2004.

[160] B. Sopori, P. Basnyat, S. Devayajanam, S. Shet, V. Mehta, J. Binns, and J. Appel, Understanding light-
induced degradation of c-Si solar cells, in Photovoltaic Specialists Conference (PVSC), 2012 38th IEEE, 2012,
pp. 11151120.

[161] D. Sera, L. Mathe, T. Kerekes, R. Teodorescu, and P. Rodriguez, A low-disturbance diagnostic function inte-
grated in the PV arrays MPPT algorithm, in IECON 2011 - 37th Annual Conference of the IEEE Industrial
Electronics Society, no. 2. Ieee, Nov. 2011, pp. 24562460.

[162] E. L. Meyer and E. E. Van Dyk, Assessing the Reliability and Degradation of Photovoltaic Module Performance
Parameters, IEEE Transactions on Reliability, vol. 53, no. 1, pp. 8392, 2004.

[163] S. Sellner, J. Sutterl

uti, L. Schreier, and S. Ransome, Advanced PV module performance characterization
and validation using the novel Loss Factors Model, in 2012 38th IEEE Photovoltaic Specialists Conference
(PVSC), 2011, pp. 29382943.

[164] T. Markvart and L. Casta

ner, Sollar Cells Materials, Manufacture and Operation, 1st ed., T. Markvart and
L. Casta
ner, Eds. Elsevier Ltd., 2005.

[165] T. Kirchartz, A. Helbig, W. Reetz, M. Reuter, J. H. Werner, and U. Rau, Reciprocity between electrolumi-
nescence and quantum efficiency used for the characterization of silicon solar cells, Progress in Photovoltaics:
Research and Applications, vol. 17, no. 6, pp. 394402, 2009.

[166] D. L. King, M. A. Quintana, J. A. Kratochvil, D. E. Ellibee, and B. R. Hansen, Photovoltaic module perfor-
mance and durability following long-term field exposure, Progress in Photovoltaics: Research and Applications,
vol. 8, no. 2, pp. 241256, 2000.

[167] J. Salinger, V. Benda, and Z. Machacek, Optimal Resolution of LBIV / LBIC Methods for Diagnostics of Solar
Cell Homogeneity, in 26th International Conference on Microelectronics, no. Miel, 2008, pp. 35.

[168] T. Takashima, J. Yamaguchi, M. Ishioa, and A. I. Science, Fault Detection by Signal Response in PV Module
Strings, in 2008. PVSC 08. 33rd IEEE Photovoltaic Specialists Conference, no. 1, 2008, pp. 812.

[169] T. Takashima, J. Yamaguchi, K. Otani, K. Kato, and M. Ishida, Experimental Studies of Failure Detection
Methods in PV Module Strings, in Conference Record of the 2006 IEEE 4th World Conference on Photovoltaic
Energy Conversion, no. 1, 2006, pp. 22272230.

[170] R. Eke and A. Senturk, Monitoring the Performance of Single and Triple Junction Amorphous Silicon Modules
in Two Building Integrated Photovoltaic (BIPV) Installations, Applied Energy, vol. 109, pp. 154162, 2013.

[171] G. Makrides, B. Zinsser, G. E. Georghiou, M. Schubert, and J. H. Werner, Evaluation of Grid-Connected

Photovoltaic System Performance Losses in Cyprus, in 7th Mediterranean Conference and Exhibition on Power
Generation, Transmission, Distribution and Energy Conversion (MedPower 2010), 2010, pp. 17.

[172] A. Kamei, S. Yoshida, H. Takakura, and T. Minemoto, Ten years outdoor operation of silicon based photovoltaic
modules at central latitude of Japan, Renewable Energy, pp. 1-5, 2013.

[173] T. Ishii, T. Takashima, and K. Otani, Long-term performance degradation of various kinds of photovoltaic
modules under moderate climatic conditions, Progress in Photovoltaics: Research and Applications, vol. 19,
no. 2, pp. 170179, 2011.

[174] K. O. Davis, S. R. Kurtz, D. C. Jordan, and J. H. Wohlgemuth, Multi-Pronged Analysis of Degradation Rates of
Photovoltaic Modules and Arrays Deployed in Florida, Progress in Photovoltaics: Research and Applications,
vol. 21, no. 4, pp. 702712, 2012.

184
[175] D. Polverini, G. Tzamalis, and H. Mullejans, A validation study of photovoltaic module series resistance
determination under various operating conditions according to IEC 60891, Progress in Photovoltaics: Research
and Applications, no. October 2011, pp. 650660, 2012.

[176] B.-k. Kang, S. Member, S.-t. Kim, and S. Member, Diagnosis of Output Power Lowering in a PV Array by
Using the Kalman-Filter Algorithm, IEEE Transactions on Energy Conversion, vol. 27, no. 4, pp. 885894,
2012.

[177] N. Gokmen, E. Karatepe, B. Celik, and S. Silvestre, Simple Diagnostic Approach for Determining of Faulted
PV Modules in String Based PV Arrays, Solar Energy, vol. 86, no. 11, pp. 33643377, Nov. 2012.

[178] International Electrothecnical Commision. [Online]. Available:

http://webstore.ansi.org/SdoInfo.aspx?sdoid=40

[179] IEEE Standards Board, IEEE Std. 1262-1995. IEEE Recommended Practice for Qualification of Photovoltaic
(PV) Modules, pp. 128, 1995.

[180] D. King, B. Hansen, J. Kratochvil, and M. Quintana, Dark Current-Voltage Measurements on Photovoltaic
Modules as a Diagnostic or Manufacturing Tool, in 26th IEEE Photovoltaic Specialists Conference, no. Septem-
ber, 1997, pp. 1125-1128.

[181] S. Pingel, O. Frank, M. Winkler, S. Daryan, T. Geipel, H. Hoehne, and J. Berghold, Potential Induced
Degradation of solar cells and panels, in Photovoltaic Specialists Conference (PVSC), 2010 35th IEEE, 2010,
pp. 28172822.

[182] A. Hovinen, Fitting of the solar cell IV -curve to the two diode model, Physica Scripta, vol. 1994, no. T54,
p. 175, 1994. [Online]. Available: http://stacks.iop.org/1402-4896/1994/i=T54/a=043

[183] Solaredge, SolarEdge Power Optimizer Module Add-On, pp. 12, 2013.

[184] J. Bastidas-Rodriguez, G. Petrone, C. Ramos-Paja, and G. Spagnuolo, Parameter Calculation Of Photovoltaic

Modules Using A Genetic Algorithm, in 11th International Conference on Modeling and Simulation of Electric
Machines, Converters and Systems (ELECTRIMACS 2014), no. May, 2014, pp. 3742.

[185] J. J. Soon, K.-s. Low, and S. Member, Photovoltaic Model Identification Using Particle Swarm Optimization
With Inverse Barrier Constraint, IEEE Transactions on Power, vol. 27, no. 9, pp. 39753983, 2012.

[186] H. Qin and J. Kimball, Parameter determination of photovoltaic cells from field testing data using particle
swarm optimization, in 2011 IEEE Power and Energy Conference at Illinois (PECI), no. 4, 2011, pp. 47.

[187] D. L. King, W. E. Boyson, and J. A. Kratochvill, Photovoltaic Array Performance Model, Sandia National
Laboratories, Albuquerque, Tech. Rep. December, 2004.

[188] L. Peng, Y. Sun, Z. Meng, Y. Wang, and Y. Xu, A new method for determining the characteristics
of solar cells, Journal of Power Sources, vol. 227, pp. 131136, Apr. 2013. [Online]. Available:
http://linkinghub.elsevier.com/retrieve/pii/S0378775312011962

[189] M. Bashahu and P. Nkundabakura, Review and tests of methods for the determination of the solar cell junction
ideality factors, Solar Energy, vol. 81, no. 7, pp. 856863, 2007.

[190] K. Ishaque and Z. Salam, An improved modeling method to determine the model parameters of photovoltaic
(PV) modules using differential evolution (DE), Solar Energy, vol. 85, no. 9, pp. 23492359, 2011.

185