You are on page 1of 6

Optimized-Fuzzy MPPT Controller Using GA for

Stand-alone Photovoltaic Water Pumping System


A. A. S. Mohamed, Alberto Berzoy and Osama Mohammed
Energy Systems Research Laboratory
Department of Electrical and Computer Engineering
Florida International University
Miami, Florida, USA
mohammed@fiu.edu.

Abstract This paper presents a comparative study among and/or loads. MPPT contain hardware (DC-DC converter) and
different Maximum Power Point Trackers for Photovoltaic software (MPPT algorithm).
Water Pumping load. Comprehensive analysis and simulation of Several conventional methods are proposed for MPPT
KC-120-1 PV module (source), Kyocera SD 12-30 solar pump algorithm [1]. Other techniques used artificial intelligent [2-8].
(load) and storage tank (storage) were conducted. Multi-objective In [2], [3] it was claimed that optimization of the control
optimization based on Genetic Algorithms were performed for
technique is done; however no procedure of optimization is
two MPPT techniques: perturb and observe (P&O) and fuzzy
technique. A GA cost function is developed and explained for found. No cost or fitness function was developed. In [4] a
optimization purposes. The fitness function considers the MPPT technique based on GA was presented. The dynamics
irradiance variations of two climates condition (sunny and cloudy of the GA as a controller requires fast and expensive
days), however the algorithm can be easily changed for processors compared with the basic P&O and fuzzy P&O for
considering more cases. A comparative analysis of both implementation. In [5] a fuzzy MPPT algorithm without
techniques was perform before and after optimization based on optimization was studied. An unfair comparative study was
the system energy error and the water flow rate. It is performed between direct coupled system and proposed Fuzzy
demonstrated that GA-optimized Fuzzy algorithm presents a MPPT technique. Obviously the technique seeking the MPPT
more appropriate behavior under the different climatic
is going to be superior. In [6] a MPPT controller based on GA
conditions.
was proposed. However no explanation of the fitness or cost
KeywordsPVPS; MPPT; Genetic Algorithms, Fuzzy function was presented. They just explained the well know
Controller. methodology for the GA. In [7] a MPPT controller based on
GA optimized Neural Networks is presented. In this article
I. INTRODUCTION there is no explanation about the fitness function used. In [8]
In many remote and rural areas, diesel driven pumps are analytical optimization technique for MPPT algorithm is
used for water supply. Diesel pumps consume fossil fuel, proposed however no different climate conditions are taking
affects environment, needs more maintenance, and are less into consideration like all the previous authors [2-8].
reliable. Photovoltaic powered water pumps have received This paper presents a comparative study among different
considerable attention recently due to major developments in MPPT for PVPS. Multi-objective optimization based on
the field of solar cell materials and power electronic systems genetic algorithms was performed for two MPPT techniques:
technology. Photovoltaic water pumping system (PVPS) perturb and observe and fuzzy technique. A GA cost function
consists of the PV array, a pumping subsystem and a storage is developed and explained for optimization purposes. The
element. The pumping subsystem is composed of a motor- fitness function considers the irradiance variations of two
pump set and a power conditioning equipment. The power climates condition (sunny and cloudy days), however the
conditioning has a role to optimize the transfer of energy algorithm can be easily changed for considering more cases.
between the PV array and the motor-pump set. The storage
elements may be batteries for electricity storage or water tank II. MATHEMATICAL MODELING OF PV PUMPING SYSTEM.
for water storage or both. Due to the fact that this system was studied and analyzed
When the PV array is directly connected to the motor, the during a long duration (12 hr), so the average mathematical
operating point will be at the intersection of their I-V curves. equations are used in the system (DC-DC converter and motor-
This point may be anywhere on the curve, and it is seldom to pump set) modeling.
be at the maximum power point (MPP) of PV array, thus it is
not producing the maximum power. As well as, the MPP A. PV Generator Model.
changes as the module temperature and the solar irradiance The PV cell is usually represented by the single or double
change. This mismatching for the maximum power causes diode model. The single diode model will be be used in this
further over-sizing of the PV array and the whole system in study to provide fairly accurate results. The model consists of a
consequence. To mitigate this problem, Maximum Power Point current source (Isc), a diode with ideality factor (n) which varies
Tracker (MPPT) must be added to the system to follow up the from one to two depending on the material and the physical
MPP of PV panel regardless changing climatic conditions construction of the cell [9], a series resistance (Rs) and a
parallel resistance (Rp).

k,((( 
Fig. 1: Equivalent circuit of PV cell.

The current-voltage relationship of the PV cell is given in (1). Fig. 3: Circuit diagram of c'uk converter.
The main components of the MPPT are: the DC-DC



 converter and the MPPT algorithm. This paper presents two

Where: q is the electron charge, k is the boltzman MPPT techniques: conventional P&O and proposed P&O
constant, T is the cell temperature in Kelvin (K). The strategy of based on FLC.
modeling a PV module is not different from modeling a PV A. DC-DC Converter
cell for one module. It uses the same PV cell model and
parameters, but only the module voltage must be divided by the DC-DC converter is used in PV systems to detect and track
number of cells in series [9]. To make a better model, the effect the MPP produced by the PV module. The cuk type converter
of temperature on Isc and Io, was included [1]. is used in this work to provide tracking operation without any
restrictions [9]. The circuit diagram of cuk converter is shown
B. Motor-Pump Set. in Fig. 3. Equations (6), (7) and (8) indicate the
In PV applications, there is a general agreement on superior modeling equations of ideal cuk converter.
efficiency and effectiveness of DC motors in general and PM
DC motors in particular [10]. Therefore, PM DC motors are     

clear choice for PV pumps design. Two types of pumps are
commonly used in PVPS: positive displacement (PD) and
   
centrifugal pumps. The best and suitable choice for
underground water pumping applications is the PD pump, as
this type of pumps works well in higher head applications    
[11]. Thus the PM DC motor coupled with PD pump will be
used in this paper for achieving more efficient, simple and low Where RL, is the equivalent load impedance which is, in
cost PVPS. The average modeling equations of PM DC motor average model, a cubical polynomial equation in voltage [9].
and PD pump are in (2) and (3).
B. Maximum Power Point Tracking Algorithms
  
1) Convetional Perturb and Observe Algorithm.
  
MPPT algorithm is responsible for looking for the MPP
Where Vt is the motor voltage, Ia is the armature current (A), regardless the cell temperature and/or solar irradiance. The
is the motor speed (rad/sec), Ra is the armature resistance (), most commonly used method in practice is the P&O
KE is the voltage constant (V.sec/rad), KT is the torque algorithm. The flow chart shown in Fig. 4 indicates the
constant (N.m/A) and Tp is the pump torque (N.m), which is principle of operation of this method using the converter duty
constant for PD pump [1]. cycle (D) as a control variable.
III. MAXIMUM POWER POINT TRACKER. 2) Proposed Fuzzy Perturb and Observe Algorithm.
MPPT is an interface between the PV generator and the The proposed P&O technique is based on FLC. The
load which allows the system to track the MPP. The Takagi-Sugeno Fuzzy Inference System (TS-FIS) rules are
construction of the MPPT is described in the block diagram of based on the difference of PV power (p) and current (i) as
the proposed PVPS shown in Fig. 2. given in (7) and (8) [12].
(7)
MPPT Algorithm and (8)
Duty Cycle Adjustment

KC-120-1
MPPT Where p(k) is the power at the present control period and p(k-1)
D
PV Module Vo Io is the power of the previous control period. The same
PWM procedure can be applied for the all differences.
Generator ADC
PWM
The proposed fuzzy rules are:

C'uk
Vmeasured Imeasured
PM DC PD
If i(k) > 0 and p(k) > 0 then the Duty is increased.
Converter Motor Pum If i(k) > 0 and p(k) < 0 then the Duty is decreased.
If i(k) < 0 and p(k) > 0 then the Duty is decreased.
Fig. 2: Block diagram of the proposed PVPS. If i(k) < 0 and p(k) < 0 then the Duty is increased.


Begin
P&O
Algorith
m

Measure: Vo(k), Io(k)

Po(k) = Vo(k) x Io(k)


Po= Po(k) Po(k-1)

NO YES
Po> 0

Fig. 6: Surface of the FIS.


YES YES
D(k)- D(k-1)> 0 D(k)- D(k-1)> 0 IV. OPTIMIZATION BASED ON GENETIC ALGORITHMS

NO NO
The Genetic Algorithm (GA) is a random search
methodology based on population genetics that can be used for
Decrease
duty cycle
Increase duty Decrease Increase duty optimization purposes [14]. Here, the GA is used to optimize
cycle duty cycle cycle
the conventional P&O and the proposed fuzzy algorithm, thus
two different optimizations are applied:
Update
1. Optimization of the duty step change for conventional
History P&O algorithm.
D(k-1) = D(k)
P(k-1) = P(k) 2. Optimization of duty step change, current and power
membership functions value a from Fig. 5.
Fig. 4: Flowchart of the conventional P&O algorithm. The sittings of GA tool box are indicated in table II. The
The FIS inputs are the PV power and PV current algorithm of the proposed fitness function is explained in
differences. Equation (9) shows the fuzzy reasoning procedure Fig. 7. The returned value from the simulation to the GA is
for a zero-order Takagi-Sugeno fuzzy model, where w1 and w2 the error between the Theoretical maximum energy (Wth)
are the products of the weights of the input variables in each and the actual maximum energy (Wact) during 12 hr as given
membership function, and z1 and z2 are the constant values in (10). Where Wth is the area under the instantaneous
assigned to a specific rule [13]. power curve.
(10)
  The proposed GA fitness function is a multi-objective

algorithm due, where the cost function of minimization
Table I shows the rules matrix for the corresponding TS- considered the minimum error in two cases: sunny day (slow
FIS, where DrefStep is the value of the step of duty which can variation of irradiance) and a cloudy day (fast variation of
be added or subtracted to the previous duty reference (D(k-1)) irradiance) as given in (11), (12) and (13).
depending on the rule applied. The output of the fuzzy MPPT     (11)
algorithm is the reference duty cycle. As shown in table I, the
     (12)
universe of discourse is divided into three membership
functions (MF). Triangular MF are used in this control for all      (13)
input variables as shown in Fig. 5, where a is border for the
MF of the current difference or/and the power difference. Fig. The GA generates the step of change of duty cycle. The
6 shows the surface of the proposed FIS, which maps the two scripts inside the fitness function call simulation of the PVPS
fuzzy inputs (p) and (i) with the fuzzy output (D). in two iterations, one for sunny day and the other for cloudy
day. The value of the minimum error of energy is returned
TABLE I. FIS TAKAGI-SUGENO RULE MATRIX. back to the GA and the procedure is repeated again.
i(k) \p(k) NH Z PH
TABLE II. GA SETUP FOR P&O AND FUZZY P&O.
NH DrefStep 0 -DrefStep
Z 0 0 0 GA SetUp P&O Algorithm Fuzzy Algorithm
Number of variables 1 3
PH - DrefStep 0 DrefStep Lower [0.001] [0.001 1e-5 1e-4]
Bounds
Upper [0.0035] [0.0035 9e-5 9e-4]
Population 10 10
Mutation Uniform 80% Uniform 80%
Crossover Single point Single point
Iterations 10 10

Fig. 5: Triangular membership functions.


Sunny and Cloudy radiation
1100
Sunny day
1000 Cloudy day
900

800

Radiation (W/m2)
700

600

500

400

300

200

100

0
6 8 10 12 14 16 18 20
Fig. 7: Flow chart of GA and fitness function for P&O and Hour
fuzzy P&O algorithm. Fig. 9: Solar irradiance data for sunny and cloudy days.
V. SIMULATION RESULTS The proposed PVPS is simulated under different five
The block diagram of the proposed PVPS is shown in Fig. cases:
2. It consists of the KC-120-1 commercial PV module, MPPT 1) The direct coupled system (without MPPT).
and Kyocera SD 12-30 solar pump, which has a maximum 2) The system with the conventional P&O MPPT algorithm.
operating voltage of 30V and a maximum power of 120W. 3) The system with the proposed fuzzy P&O MPPT
MATLAB m-file tool is used for modeling and simulating the algorithm.
whole system. Kyocera KC-120-1 PV module consists of 36 4) The system with the optimized P&O MPPT algorithm.
multi-crystalline silicon solar cells in series and provides 120W 5) The system with the optimized fuzzy P&O MPPT
of nominal maximum power. Fig. 8 indicates MATLAB algorithm.
simulation results for PV module; the discrete points represent
The system in each case is simulated twice; one for sunny day
the given manufacturer datasheet curves [15], while the
and the other for cloudy day. The total actual energy (Wact) and
continuous lines are obtained by simulation. The figure
indicates that MATLAB model gives a good correspondence the total theoretical energy (Wth) are calculated on a 12 hr
between the simulated and the manufacture curves. The series period, and the percentage of error between them is calculated
and parallel resistances are calculated by MATLAB model for by "(14)" .
the chosen module, such that Rs = 4.7m and Rp = 2.9953.
 
By trial and error, and using MATLAB model, it was found
that n = 1.05, is the best value that attains the best match with The flow rate of Kyocera SD 12-30 water pump is
the datasheet curves [1]. proportional to the power delivered and the total dynamic head.
Fig. 9 shows the meteorological data used for this study. When the total dynamic head is 30m, and the power is 120W,
These data have been measured for the PV module located on the flow rate is approximately 10.5L/minute, and the minimum
the roof of a building in EGYPT (Located at North Africa, power required for the pump is 35W [16]; therefore as long as
Latitude: 30 3' 0" N, Longitude: 31 15' 0" E). the output power is higher than 35W, it pumps water with the
The sunny day is on the August 10th, 2010. above flow rate. The total volume of water pumped during the
The cloudy day is on the November 5th, 2010. 12hr period for each system is also calculated.
8 8
1000 W/m2
A. Optimization Results
7 7
2
The results of GA optimization for the conventional P&O
800 W/m
6 6 and the fuzzy P&O algorithm are shown in Fig. 10. It is clear
Module Current (A)
Module Current (A)

25 oC
from the figure that both the two optimization gives
5 5
600 W/m2 approximately the same final optimum fitness value but there is
4 4 a large difference the range of variation between the average
400 W/m2 and the mean values in the two cases. The final optimum
3 50 oC 3
values from GA optimization are indicated in table III.
2 2 200 W/m2
o TABLE III. VARIABLES BEFORE AND AFTER OPTIMIZATION WITH GA
1 75 C 1
Fuzzy
0 0 P&O Fuzzy P&O
0 5 10 15 20 25 30 0 5 10 15 20 25 30 P&O after P&O
before after
Module Voltage (V) Module Voltage (V) before
(a) (b) Step 0.0025 0.00291 0.0025 0.002947
Fig. 8: I-V curves of KC-120-1PV module (a) different a_current_MF - - 4*10-5 7.11 *10-5
temperatures and 1000W/m2, (b) different irradiance and 25 oC. a_power_MF - - 6*10-4 1.73*10-4


Fuzzy and Optimized Fuzzy MPPT
GA Optimization for P&O GA Optimization for Fuzzy P&O 160
Best: 2.85147 Mean: 4.55433 Best: 2.8599 Mean: 12.5472 Fuzzy P&O
15 15 Optimized
140
Best fitness

Module Output Power (W)


Best fitness
Mean fitness Mean fitness 120
12.5 12.5
100

10 10 80
Fitness value

Fitness value
60
7.5 7.5
40

20
5 5
0
0 2 4 6 8 10 12 14 16 18 20 22
2.5 2.5 Module Voltage (V)
Fig. 12: Trace of MPPT tracking on sunny day at 25oC.
0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Generation Generation Direct coupled, P&O, optimized P&O, Fuzzy and Optimized fuzzy
200
Direct Coupled
Fig. 10: GA optimization results for conventional P&O and 180 P&O
P&O Optimized
fuzzy P&O optimization. 160
Fuzzy P&O
140 Fuzzy P&O Opt
B. Simulation Results for Sunny Day.

Power (W)
120
The five cases of study are simulated with the irradiance 100
data of the sunny day and the simulation results are indicated in 80
Fig. 11, 12 and 13, and table IV. Fig. 11 shows a comparative 60
study of the MPPT performance for three cases; direct coupled 40
system, using P&O algorithm and using the optimized P&O, 20
while Fig. 12 compares between the other two cases; fuzzy 0
0 2 4 6 8 10 12
algorithm and the optimized fuzzy algorithm. Fig. 13 shows the Hour

power delivered to the motor during the 12hr period for all Fig. 13: Input power to motor for a 12 hr period on sunny day.
cases. And these results are concluded in table III, which
indicates that: The optimized Fuzzy MPPT technique is the best choice for
fast irradiance and fluctuated weather.
There is a large mismatch for the MPP in the direct coupled Both the P&O and the optimized one exhibit loss of control
PVPS system in sunny and cloudy days (E = 29.0269% in operation, but the optimized one is less number of losses
and E = 34.1427% respectively). and restored to the normal faster as shown in Fig. 14.
All other types of control give a very good and very close The optimization technique for the P&O improved the
tracking performance. tracking performance in fast radiations by about 0.1%, but
The conventional P&O MPPT algorithm is the best choice abusive the performance in slow radiations by about
for slow irradiance variations (E = 0.1273%). 0.02%, thus the optimized technique is better in fluctuated
Fuzzy MPPT technique is not recommended in slow weather.
irradiance variations. Fuzzy and optimized fuzzy technique did not exhibit any
160
Direct coupled, P&O and Optimized P&O MPPT loss of control in fast radiation and the optimized one is
Direct Coupled more tracked as shown in Fig. 15.
140 P&O
P&O Optimized
Module Output Power (W)

120

100
I. CONCLUSION
80 This paper presented an optimized fuzzy perturb and
60 observe MPPT control topology for photovoltaic water
40
pumping systems. The whole system was simulated in
20
MATALB software. A comparative study is presented for the
conventional and the proposed technique with and without GA
0
0 2 4 6 8 10 12
Module Voltage (V)
14 16 18 20 22 optimization. A GA cost function was developed and
Fig. 11: Trace of MPPT tracking on sunny day at 25 C. o explained for the optimization purposes. The study considered
slow and fast variations in solar radiations. The presented
technique shows fast response to the fast fluctuations in
C. Simulation Results For Cloudy Day. irradiance level. In addition the presented method maximized
The same figures that discussed before in sunny day the level of energy produced by the PV array and provided a
simulation are simulated again for cloudy day as shown in Fig. greater pumping flow rate in comparison with the other
14, 15 and 16, and the results are summarized in table IV as
discussed techniques especially in the fast variation solar
follow:
radiations.


Direct coupled, P&O and Optimized P&O
160
Direct Coupled
140 P&O
REFERENCES
Module Output Power (W)

P&O Optimized
120
[1] A. A. Mohamed, H. M. B. Metwally, and M. A. Farahat, "Photovoltaic
100 Water Pumping Systems" LAP Lambert Academic Publishing, ISBN
978-3-659-10290-5, April. 2012.
80
[2] Jaziri, S.; Jemli, K., "Optimization of a photovoltaic powered water
60 pumping system," Control, Decision and Information Technologies
Loss of the control
40 (CoDIT), 2013 International Conference on , vol., no., pp.422,428, 6-8
May 2013
20
[3] Ouada, M.; Meridjet, M.S.; Talbi, N., "Optimization photovoltaic
0
0 2 4 6 8 10 12 14 16 18 20 22 pumping system based BLDC using fuzzy logic MPPT control,"
Module Voltage (V) Renewable and Sustainable Energy Conference (IRSEC), 2013
Fig. 14: Trace of MPPT tracking on cloudy day at 25oC. International , vol., no., pp.27,31, 7-9 March 2013
Fuzzy and Optimized Fuzzy MPPT
[4] Dahmane, M.; Bosche, J.; El-Hajjaji, A.; Pierre, X., "MPPT for
160 photovoltaic conversion systems using genetic algorithm and robust
Fuzzy P&O
control," American Control Conference (ACC), 2013, vol., no., pp.6595,
140 Optimized
6600, 17-19 June 2013.
Module Output Power (W)

120 [5] Aashoor, F.A.O.; Robinson, F.V.P., "Maximum power point tracking of
100 photovoltaic water pumping system using fuzzy logic controller," Power
Engineering Conference (UPEC), 2013 48th International Universities' ,
80 vol., no., pp.1,5, 2-5 Sept. 2013
60 [6] Hadji, S.; Krim, F.; Gaubert, J-P, "Development of an algorithm of
40
maximum power point tracking for photovoltaic systems using genetic
algorithms," Systems, Signal Processing and their Applications
20 (WOSSPA), 2011 7th International Workshop on , vol., no., pp.43,46, 9-
0
11 May 2011
0 2 4 6 8 10 12 14 16 18 20 22
Module Voltage (V) [7] Ramaprabha, R.; Gothandaraman, V.; Kanimozhi, K.; Divya, R.;
o Mathur, B.L., "Maximum power point tracking using GA-optimized
Fig. 15: Trace of MPPT tracking on cloudy day at 25 C. artificial neural network for Solar PV system," Electrical Energy
Systems (ICEES), 2011 1st International Conference on , vol., no.,
Direct coupled, P&O, optimized P&O, Fuzzy and Optimized fuzzy
200 pp.264,268, 3-5 Jan. 2011
Direct Coupled
180 P&O
[8] Ouachani, I.; Rabhi, A.; Tidhaf, B.; Zouggar, S.; Elhajjaji, A.,
P&O Optimized "Optimization and control for a photovoltaic pumping system,"
160
Fuzzy P&O Renewable Energy Research and Applications (ICRERA), 2013
140 Fuzzy P&O Opt International Conference on , vol., no., pp.734,739, 20-23 Oct. 2013
Power (W)

120
[9] M. A. Farahat, H. M. B. Metwally, and A. A. Mohamed," Optimal
100 Choice and Design of Different Topologies of DC-DC Converter Used
80 in PV Systems, At Different Climatic Conditions In EGYPT,
60
Renewable Energy Journal, REJ Vol. 43, pp. 393-402, January. 2012.
40
[10] Applebaum J., "The Quality of Load Matching in a Direct-Coupling
Photovoltaic System", IEEE Transactions on Energy Conversion, EC-
20
2(4), pp. 534-541, 1987.
0
0 2 4 6 8 10 12 [11] Posonski R., "Photovoltaic Water Pumps- An Attractive Tool for Rural
Hour
Drinking Water Supply ", Solar Energy, Vol. 58, pp. 155-163, 1996.
Fig. 16: Motor input power for a 12 hr period on cloudy day. [12] Berzoy, A.; Baethge, E.; Restrepo, J.; Viola, J., "Fuzzy Control System
for Maximum Power Point Tracking in Solar Panels Based on DC-DC
TABLE IV. SUMMARIZED SIMULATION RESULTS. Converter PI Current Control," Andean Region International Conference
Sunny Irradiance Cloudy Irradiance
(ANDESCON), 2012 VI , vol., no., pp.119,122, 7-9 Nov. 2012
Wth = 988.9554 (Wh) Wth = 664.2035 (Wh) [13] J. S. Jang, C. T. Sun, and E. Mizutani, Neuro-Fuzzy and Soft Computing.
Wact (Wh) %E Volume (L/day) Wact (Wh) %E Volume (L/day) Toronto Canada: Prentice Hall, 1997.
Direct Coupled 701.8919 29.0269 4474.3 437.4267 34.1427 2521.9 [14] Randy H., Sue H Haupt Practical Genetic Algoriths US John Wiley
and Sons. INC, 1998.
P&O MPPT 987.696 0.1273 6065 661.8121 0.3600 3680.3 [15] Kyocera KC120-1 Multi-crystalline Photovoltaic Module Datasheet,
Optimized P&O 2001 (downloaded from www.kyocerasolar.com).
987.5129 0.1459 6063.8 662.4725 0.2606 3684.1
MPPT
[16] Kyocera Solar Inc. Solar Water Pump Applications Guide 2001
Fuzzy MPPT 987.4623 0.1510 6063.6 662.7673 0.2162 3687.5 (downloaded from www.kyocerasolar.com
Optimized Fuzzy
987.4773 0.1495 6063.5 662.7771 0.2148 3688
MPPT



You might also like