Electric Power Systems Research 127 (2015) 186–196

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

A generic model of two-stage grid-connected PV systems with

primary frequency response and inertia emulation
Sotirios I. Nanou ∗ , Apostolos G. Papakonstantinou, Stavros A. Papathanassiou
School of Electrical and Computer Engineering, National Technical University of Athens (NTUA), 9 Iroon Polytechniou str., 15780 Athens, Greece

a r t i c l e i n f o a b s t r a c t

Article history: Photovoltaic (PV) stations are increasingly becoming subject to grid code requirements that include
Received 31 January 2015 frequency response and active power control capability. The main goal of this paper is to propose a
Received in revised form 12 May 2015 generic model for a two-stage grid-connected PV system with frequency response capability, suitable
Accepted 8 June 2015
for power system studies. The proposed model includes a suitable control scheme, which provides both
droop and inertial response, as well as the ability to operate at a scheduled active power reserve, enabling
Keywords:
thus the provision of under-frequency response. A linearized small-signal model is developed to assess
PV
the stability of the proposed PV power control loop when the PV generator provides frequency response,
Dynamic model
Grid codes
whereas time-domain simulations are performed in order to quantify the benefits achieved by droop-
Two-stage power conversion system type and inertia frequency controllers, including a discussion on the selection of their parameters. The
Frequency control analysis demonstrates the satisfactory performance of the proposed PV system model, which provides
Inertial response all functionality required by grid codes in the context of active power control.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction is still limited, concerning both its technical implementation and

Present day grid codes impose technical requirements to dis-
tributed generation (DG) stations, including photovoltaic (PV)
plants. Among these, particularly important is the provision of
frequency response, as well as the ability to implement power cur-
tailment commands issued by the network operator [1–3]. It is
imperative that DG models used in power system studies dealing
with future networks with high levels of DG penetration provide
the required functionality in the context of grid code compatibil-
ity.
DG contribution to frequency control is a well-known issue
in the literature. For instance, droop and inertia control schemes
are examined in [4–8] for wind turbines (WTs), whereas similar
approaches can be found in [9–11] for energy storage systems
(ESS) and in [12–16] for PV stations equipped with internal
ESS.
In the case of PV stations without internal ESS, DC link capac-
itors are usually characterized by high charging rates (in the
range of milliseconds) and therefore the challenge of frequency
response is transposed to the PV generator and its control. Liter-
ature on active power control of PV plants without internal ESS
∗ Corresponding author. Tel.: +30 6973931447.
E-mail addresses: sotnanou@central.ntua.gr (S.I. Nanou), apopapak@gmail.com
(A.G. Papakonstantinou), st@power.ece.ntua.gr (S.A. Papathanassiou).
the expected benefits [17]. For single-stage PV systems, bespoke
power control strategies are proposed in [18,19], in order to min-
imize frequency variations in a small power system, employing
a fuzzy logic controller and a minimal order observer, respec-
tively. In [20–22], particular attention is placed on the provision
of droop-type response, while in [21] an emergency controller is
also proposed in order to curtail power after severe over-frequency
events.
Regarding two-stage PV systems, the provision of frequency
response becomes a more challenging task from a control perspec-
tive, as the DC link voltage of the PV inverter is decoupled from
the PV generator voltage, thus providing enhanced flexibility in
operation and control of such systems [23–26]. However, relevant
references on the subject are again limited to droop-type control
[27,28], while authors in [29] propose a method to regulate the
operating point of the PV generator below the maximum power
point (MPP), by controlling the PV array voltage to a specific frac-
tion of its open circuit voltage. However, it is not addressed how
this fraction can be determined in relation to the desired level of
power reserve which might be imposed by the network operator.
Further, a different control method is proposed in [30], which is
based on proportional-integral (PI) control of the grid frequency,
without however addressing how the controller settings can be
adjusted to provide the desired droop or inertial response, or how
to follow external active power commands (e.g. reserve levels), as
stipulated by grid codes.

http://dx.doi.org/10.1016/j.epsr.2015.06.011
0378-7796/© 2015 Elsevier B.V. All rights reserved.
 S.I. Nanou et al. / Electric Power Systems Research 127 (2015) 186–196
Fig. 1. Active power regulation in FSM [1].
A sufficiently detailed PV system model, providing all required
functionality in terms of active power control and frequency con-
trol, is still missing from the literature. In this paper, such a model is
introduced for a two-stage grid-connected PV system without ESS,
which is suitable for power system studies. The required frequency
response is achieved by regulation of the PV generator output
power via the DC/DC converter. The proposed control scheme per-
mits the following alternative operating modes: (a) conventional
maximum power point tracking (MPPT), (b) droop and (c) iner-
tial frequency response, (d) operation at a given power reserve,
which in turn enables under-frequency response. The frequency
controller is integrated in the DC/DC converter controller of [31],
which permits regulation of the generated PV power to a given
reference.
To evaluate the dynamic response of the proposed PV system
model, a small isolated network is used as a study-case system,
comprising a diesel unit (DU), a full-power converter wind turbine
(FCWT), a PV plant and local load. The disturbances considered
include load changes as well as stochastic variations of the wind
speed of the connected WT.
The paper is organized as follows: The main frequency response
requirements stipulated by grid codes are briefly discussed in Sec-
tion 2. The modeling approach adopted for each generating unit is
presented in Section 3. The proposed control schemes for the DC/DC
converter and PV inverter are outlined in Section 4. Time-domain
simulations are presented in Section 5 and the main conclusions
are summarized in Section 6. Parameter values are provided in
Appendix.
2. Frequency response requirements
Fig. 1 depicts a typical frequency response characteristic, from
the new ENTSO-E Code for Generators [1]. The objective of this
requirement is for each generating unit to automatically adjust
active power output in case of frequency deviations. According to
Fig. 1, two operating modes are identified:
• “Frequency sensitive mode-over-frequency (FSM-O)”: the gen-
erating unit is expected to curtail active power proportionally
(droop-type control) to the frequency increase f fn
, where fn is
the nominal system frequency. This operating mode is limited by
|p |
the minimum regulating level − Pmax2 that the station is allowed
to operate, where Pmax is the maximum capacity.
• “Frequency sensitive mode-under-frequency (FSM-U)”: If under-
frequency events occur, the generating unit is expected to release
additional active power up to its maximum capacity Pmax . The
resulted under-frequency response depends on the operating and
187
|p1 |
ambient conditions of the station. The active power range Pmax
in Fig. 1 varies from 1.5 to 10% [1].
The frequency dead-band depicted in Fig. 1 lies in the range of
0–500 mHz, whereas the droop constant R varies from 2 to 12%
[1]. Both parameters are provided by the network operator. Gen-
erating units are also required to receive external active power
set-points by the network operator via a suitable control interface
and adjust their power output within specified time intervals and
error tolerances. A similar requirement is also foreseen in the Ger-
man feed-in tariff law for PV stations [3], which must be able to
limit their active power output down to 70% of the installed PV
capacity.
Apart from the conventional droop-type control, the ENSTO-
E Network Code foresees also the ability of a generating unit to
provide synthetic inertial response in order to limit the rate of
change of frequency (ROCOF) following major system disturbances.
For power plants without inherent inertial response capability (e.g.
inverter-interfaced units), an amendment to the power plant con-
troller is required in order to emulate inertial response [1].
The aforementioned requirements are used as a reference in the
rest of this paper, in order to develop a PV system model offer-
ing a wide range of alternative active power control modes, in the
context of grid code compatibility.
3. System modeling and control
As already mentioned in Section 1, the proposed PV system
model will be exploited in the paper to investigate the potential
for frequency regulation offered by PV plants without internal ESS.
For this purpose, a small isolated network is selected as the study-
case system, comprising a DU, a FCWT, a PV plant and local load.
In the following, the main modeling guidelines adopted for each
generating unit are briefly presented.
The PV generator is modeled by the standard single-diode equiv-
alent circuit [32,33], whose accuracy suffices for the purpose of
this study. Its equations along with typical parameter values can
be found in [32]. The two-stage power conversion system of the PV
plant is depicted in Fig. 2. It comprises two parallel 3-phase DC/AC
inverters connected to the MV network through a three-winding
transformer [34], along with multiple DC/DC converters connected
in parallel at the DC side [25]. In this work, to simplify converter
modeling and reduce computational burden, a single PV array and
DC/DC converter is assumed for each DC/AC inverter. As it is further
explained in Section 4, the DC/DC converter regulates the operat-
ing point of the PV array, while the PV inverter controls the DC link
voltage to its reference value. Parameter values are given in Table 1
of Appendix.
The simple DU model employed in this study is illustrated in
block diagram form in Fig. 3 [35]. The diesel engine and the valve
actuator servomechanism are represented by first-order lags, with
time constants TD and TSM , respectively. Inputs to the synchronous
generator model are the mechanical power pD and the terminal
voltage vt , whereas outputs are the generator speed, ωD , and the
electrical output power pe . Parameters of the speed governor are
the droop constant R and the integral (isochronous) control gain Ki ,
which eliminates the steady-state frequency error. An IEEE Type
I automatic voltage regulator and excitation system is assumed
and the standard 6th-order dq model available in Matlab/Simulink
library is employed for the synchronous generator. Parameter val-
ues are given in Table 2 of Appendix.
The typical configuration of a FCWT based on a multi-pole per-
manent magnet synchronous generator (PMSG) is illustrated in
Fig. 4. The PMSG is controlled by the generator side converter which
implements the MPPT strategy, whereas the grid side converter
188 S.I. Nanou et al. / Electric Power Systems Research 127 (2015) 186–196

Fig. 2. Two-stage PV power conversion system.

regulates the DC link voltage. Additional details for the overall WT
control scheme can be found in [5]. The PMSG model available
in Matlab/Simulink library is used. Parameter values are given in
Table 3 of Appendix.
Since the main focus of this paper is on the frequency response
of the system, all high-frequency components related to the switch-
ing operation of the power converters are neglected and averaging
techniques were applied in order to obtain time-averaged models
[36]. For the DC/DC converter, the full-bridge topology is consid-
ered (Fig. 2) [26], using a suitable average-value model, following
the modeling approach of [33,36]. The DC/AC converter is described
by the fundamental frequency model of [32,33].
4. PV control scheme
4.1. Control strategy of the DC/DC converter
The controller of the DC/DC converter is shown in Fig. 5. Two
control loops are implemented: The inner loop comprises a PI com-
pensator which regulates the PV array voltage vpv to its reference
value v∗pv , by controlling the duty cycle of the DC/DC converter, dpv .
A gain scheduling strategy is applied for the PI regulator, with the
PV output current ipv as the scheduling variable [37]. This selection
is based on the fact that the dynamic performance of the system
depends strongly on solar irradiance, which affects the PV current

Fig. 3. Simplified DU model in block diagram form. Fig. 4. Full-power converter WT (FCWT).
 S.I. Nanou et al. / Electric Power Systems Research 127 (2015) 186–196
Fig. 5. DC/DC converter controller.
[38]. The controller design of the inner control loop is discussed in
detail in [37].
The voltage reference v∗pv is determined either by the MPPT
controller, where the conventional incremental conductance (INC)
strategy is applied [32], or by the outer power control loop of Fig. 5
(power control mode), where a PI compensator regulates the PV
array power ppv to a reference value p∗pv :
p∗pv = (1 − r ∗ ) · P̂max − p∗freq
where r* is the power reserve command imposed by the network
operator, P̂max is the estimate of the maximum available (MPP)
power and p∗freq is the output of the frequency controller.
The P̂max estimator block in Fig. 5, explained in detail in [31],
applies linear and quadratic approximations of the P–V curve, using
measurements of the PV generator terminal voltage and current. Its
purpose is to estimate the PV power actually available, when the
PV generator does not operate in MPPT mode.
When the power control mode is activated, the PV array power
ppv can be controlled to the reference value p∗pv , either by modulat-
ing the PV voltage vpv at values below the MPP voltage Vmpp (current
source region of the I–V characteristic [38]), or by increasing the PV
voltage above Vmpp (voltage source region). In this study, operation
in the current source region is selected, due to the improved per-
formance of the P̂max estimator in this region, as further explained
in [31]. Nevertheless, as it will be shown in Section 5.4, both reg-
ulation modes are tested in order to better illustrate the effect on
the operating point of the PV array when the PV system operates
in FSM.
To fully comply with the frequency response requirements
presented in Section 2, the proposed frequency controller com-
prises droop (proportional-P) and inertial (derivative-D) terms, to
produce a frequency-dependent PV power reference component,
p∗freq :
p∗freq = f̂ /Rpv + 2Hpv ḟˆ
where Rpv is the droop constant and Hpv the equivalent (synthetic
or virtual) inertia gain. Estimates of the system frequency f̂ and the
189
Fig. 6. Block diagram of the closed-loop PV and DC/DC converter system.
ROCOF ḟˆ are provided by the synchronous reference frame (SRF)
phase locked loop (PLL) employed for the PV inverter controller.
Following the approach of [8], a filtered derivative of the system
frequency ḟˆ can be obtained, utilizing the q-axis component of the
grid voltage vgq provided by the PLL, based on the following relation
[8]:
kipll kipll s
ḟˆ = vgq = 2 f (3)
2 s + skppll + kipll
where kppll , and kipll are the proportional and integral gains of the
PLL controller. This approach makes the inertial response of the PV
plant more immune to the internal dynamics of the PLL, as com-
pared to simply using the derivative of the estimated frequency f̂ ,
since a left half-plane zero exists in the closed-loop transfer func-
tion of the PLL:
skppll + kipll
f̂ = f (4)
s2 + skppll + kipll
The transition between the MPPT and power control modes,
depicted in Fig. 5, is controlled by a hysteresis block, to avoid unnec-
(1) essary switching between operating modes: If the instantaneous
power reference p∗pv becomes less than 95% of the maximum power
P̂max , the controller switches from MPPT to power control, whereas
the control switches back to MPPT if p∗pv exceeds 99% of P̂max .
4.2. Stability analysis for the PV power control loop
To assess the stability of the PV power control loop when the
PV generator operates in FSM, a suitable small-signal frequency-
domain model is needed. Following the approach of [33,37], the
linearized model of Fig. 6 is obtained in block diagram form, where
the differential equations governing the PV system response are
perturbed and linearized around an equilibrium point. All system
variables are normalized on MPP values at Standard Test Conditions
(STC), provided in Table 1 of Appendix. The small-signal PV voltage
vpv varies in relation to the duty cycle dpv of the DC/DC converter
according to the small-signal transfer function Gv-d [37]:
Gv−d,n s2L ipv0 /dpv0 + nvdc0
Gv−d (s) = =−  2 (5)
Gv−d,d s 4ci L − s2L /rpv0 + vdc0 /vpv0
2
where ipv0 , and vpv0 are the pre-disturbance values of the PV gener-
ator current and voltage respectively, rpv0 is the dynamic resistance
vpv
(slope ipv
) of the PV generator [38], and vdc0 and dpv0 are the
pre-disturbance values of the voltage output and duty cycle of the
DC/DC converter. Parameters  ci , and  L represent the equivalent
time constants of the intermediate capacitance Cint and inductor
(2) Ldc (Fig. 2), given by:
2
Cint Vmpp,STC
ci = (6)
2Vmpp,STC Impp,STC
190 S.I. Nanou et al. / Electric Power Systems Research 127 (2015) 186–196

Fig. 7. Stability analysis of the PV power control loop of Fig. 6. Bode plots of (a) magnitude and (b) phase of Gol assuming operation at r* = 20% (ppv /Pmpp = 0.8), and variation
of (c) phase margin and (d) controller bandwidth in relation to the operating point of the PV array.

2
Ldc Impp,STC 4.3. Control strategy of the PV inverter
L = (7)
2Vmpp,STC Impp,STC
The PV inverter controller, illustrated in Fig. 8, is based on a
where Vmpp,STC , and Impp,STC are the base values of the PV array volt- conventional vector controller of the output current [32]. The mod-
age and current at STC conditions. Based on the small-signal model ulation indices m∗˛ˇ are provided by the inner PI current controller
shown in Fig. 6, standard frequency-domain techniques can now be which regulates the inverter output current ig , utilizing the phase
applied in order to assess the stability of the PV power control loop angle  PLL provided by the three-phase SRF-PLL. The outer feedback
of Fig. 5 under different operating and ambient conditions. More loop provides the active power reference p∗g , by regulating the DC
specifically, the open loop transfer function of the PV and DC/DC link voltage vdc to its reference value (1 p.u.). For tuning the current
converter system Gol is given by:
   
1 Gv−d,n kpv (sv + 1) vpv0
Gol (s)=kpp 1+ ipv0 +
sp Gv−d,n kpv (sv + 1) − sv Gv−d,d rpv0

(8)

The detailed Bode magnitude and phase diagrams for the open
loop transfer function Gol are illustrated in Fig. 7(a) and (b), for two
different irradiance levels, assuming that the PV generator initially
operates with a power reserve level of r* = 20% (ppv /Pmpp = 0.8). The
control gains kpp and  p of the outer control loop are set equal to
10 and 0.01 s respectively, whereas the gain scheduling technique
of [33] is adopted for the inner PI voltage controller, where the
proportional gain kpv varies in inverse proportion to the measured
PV generator current ipv . The time constant  v is set equal to 0.01 s.
Based on Fig. 7(a) and (b) it is evident that, even though low solar
irradiance levels impair to a certain extent the transient response
of the PV system, both the gain crossover frequency and the phase
margin remain sufficiently high.
To evaluate the dynamic performance of the PV system at a
wider range of potential operating points, at a given solar irradi-
ance, additional plots are presented in Fig. 7(c) and (d), where the
phase margin and the controller bandwidth are plotted against the
PV array power ppv , normalized on its respective MPP value, assum-
ing operation in the current source region of the I–V characteristic.
It is noteworthy that PV power variations do not alter the dynamic
response characteristics of the power control scheme, confirming
that the PV plant is capable of modulating effectively its power
output to the levels imposed by the frequency controller of Fig. 5. Fig. 8. Control scheme for the PV inverter.
 S.I. Nanou et al. / Electric Power Systems Research 127 (2015) 186–196
Fig. 9. Study-case system.
and DC voltage controllers, standard frequency-domain techniques
were applied, with the objective of achieving a phase margin of at
least 45◦ , while maintaining high gain crossover frequencies.
5. Time-domain simulations
5.1. Study-case system
To evaluate the dynamic response of the PV system under grid
frequency disturbances, the small isolated network of Fig. 9 has
been selected as a study case. The system includes a DU, a FCWT
and a PV plant. All units are connected via step-up transformers
Fig. 10. (a) System frequency, (b) DU active output power, (c) PV array power and (d) PV array voltage, following a 20% step decrease of the load at t = 1 s and an equal increase
at t = 15 s, for alternative frequency control approaches (Rpv = 5%, Hpv = 15 s). (For interpretation of the references to color in this figure citation in text, the reader is referred
to the web version of this article.)
191
(TFs) to the medium voltage (MV) level, feeding the local load. The
initial load demand is 1.9 MW and the MPP power and voltage of
each PV array are Pmpp = 300 kW and Vmpp = 553 V respectively. All
basic system data are given in Tables 1–3 of Appendix.
5.2. Response to load changes
The objective of this section is to assess the frequency response
of the system under consecutive load changes. The overall sys-
tem response is presented in Fig. 10, where step variations of the
load occur (20% decrease at t = 1 s and an equal increase at t = 15 s).
The WT operates at constant wind speed, the PV system at con-
stant solar irradiance and the power reserve command r* is zero
(MPPT control in steady state). The droop constant and the virtual
inertia gain of the frequency controller are Rpv = 5% and Hpv = 15 s,
respectively.
The response of the system frequency in Fig. 10(a) is obtained
using the different frequency control approaches presented in
Fig. 5. If operation in FSM-O is suspended, a maximum frequency
excursion of approximately 0.8 Hz occurs following the load rejec-
tion disturbance at t = 1 s (blue curve). Droop control alone (green
curve) achieves a substantial reduction of the maximum frequency
deviation and provides adequate damping to the dominant elec-
tromechanical mode of the autonomous system. When inertia
control is applied (red curve), both the maximum frequency and the
ROCOF are notably reduced. The combined droop and inertia (PD
type) control (black curve) expectedly provides best results, as the
maximum frequency excursion is reduced by 60%. The improved
frequency response is attributed to the reduction in the PV power
output following the rejection of the load, stipulated by the fre-
quency response characteristic of Fig. 1, which in turn results in
a smoother transient of the DU power (Fig. 10(b)). The PV array
power ppv and operating voltage vpv are presented in Fig. 10(c) and
(d) respectively, normalized on the respective MPP values, in order
192 S.I. Nanou et al. / Electric Power Systems Research 127 (2015) 186–196
Fig. 11. (a) System frequency, (b) PV array power, (c) PV array voltage, for the same
disturbance as in Fig. 10. Droop-type PV frequency controller, with different droop
parameter values.
to illustrate more clearly the effect of operation in FSM-O on the
operating point of the PV array.
The dynamic behavior of the system is different during the
under-frequency event following the reconnection of the load at
t = 15 s. In this case, operation in FSM-U is entirely ineffective
because each PV array initially operates at MPP and therefore it
cannot provide additional active power to compensate for the load
increase (p1 = 0 in Fig. 1). Inertia control is also ineffective in the
first phase after the event, where the frequency decreases and the
ROCOF is negative. However, it is subsequently activated, during
the frequency restoration phase (positive ROCOF), thus improving
the damping of the system. Nevertheless, the maximum frequency
excursion remains unaffected.
As it can be seen in Fig. 10(c) and (d), the operating point of the
PV array deviates significantly from the MPP when entering FSM-O,
since large reductions of the PV voltage are necessary to effect sub-
stantial changes of the output power. Nevertheless, these voltage
deviations are made possible by the two-stage power conversion
system, which decouples the DC link voltage of the PV inverter from
Fig. 12. (a) System frequency, (b) PV array power, (c) PV array voltage, for the same
disturbance as in Fig. 10. Inertia PV frequency controller, with different virtual inertia
values.
the PV array voltage. Further, the time-domain simulations demon-
strate also the satisfactory dynamic performance of the PV power
control scheme of Fig. 5, thus confirming the results obtained by
the frequency-domain analysis performed in Section 4.2.
5.3. Effect of droop and virtual inertia values
To examine the response of the droop-type frequency controller
in more detail, additional time-domain simulations are presented
in Fig. 11, assuming different values of the droop constant Rpv . From
Fig. 11(a) it is evident that low droops lead to improved response
characteristics during over-frequency events. As it can be seen in
Fig. 11(b) and (c), the PV station contributes more actively to fre-
quency control when low droops are used, reducing its output
power and operating voltage in transient conditions.
The system response using solely the inertia frequency con-
troller (without droop) is demonstrated in Fig. 12, for different
values of the synthetic inertia constant Hpv . Increased Hpv values
 S.I. Nanou et al. / Electric Power Systems Research 127 (2015) 186–196
Fig. 13. (a) System frequency, (b) PV array power and (c) PV array voltage, for the
same disturbance as in Fig. 10. Combined droop and inertia PV frequency controller,
with different levels of PV power reserves (Rpv = 5%, Hpv = 15 s).
are more effective in reducing both ROCOF and frequency excur-
sions, without a notable deterioration in the damping of the system,
at least in the Hpv range examined here. Hence, the proposed
implementation of the inertia controller is not compromised by
the performance of the PLL, thus permitting the adoption of quite
high Hpv values.
Besides the effect on the PV generator power in transient condi-
tions, the basic limitation in selecting high gains for the frequency
controller is associated with the large transient depression of the
PV array voltage. In reality, practical limits are imposed on the vari-
ation of the DC/DC converter duty cycle, mainly due to the presence
of parasitic elements in the actual circuit topology, which inevitably
affect the actual voltage transformation ratio and impair the DC/DC
converter efficiency [39]. Hence, depending on the topology, design
and implementation of the DC/DC converter, the applicable limits
to the duty cycle and voltage transformation ratio will constrain
the selection of the frequency controller gains. In this study, a min-
imum PV generator voltage around 30% of the nominal MPP voltage
has been indicatively assumed.
193
Fig. 14. (a) System frequency, (b) PV array power and (c) PV array voltage,
for the same disturbance as in Fig. 10. Regulation in the voltage source region
(Vmpp < vpv < Voc ). Combined droop and inertia frequency controller, with a 20% PV
power reserve command (Rpv = 5%, Hpv = 15 s).
5.4. Operation with PV power reserve
A PV system without internal storage may provide under-
frequency response only if sufficient power reserves are maintained
in normal operation (p1 in Fig. 1), before entering FSM-U. Such
a situation might occur in practice if the PV station is operated at
a reduced power output, due to a curtailment command issued by
the network operator (Fig. 2), a functionality already foreseen by
recent network codes [1,3]. To explore this possibility, the same dis-
turbances are simulated in Fig. 13, assuming different levels of the
power reserve command r* . In all cases, the combined PD frequency
control mode is applied.
The provision of improved under-frequency response
is clear in Fig. 13(a), as increased power reserves are main-
tained, while the dynamic performance of the proposed controller
is very good. It is observed that the maximum frequency drop can
be reduced by up to 40% when a 20% power reserve is available.
However, since the achieved improvement depends on the relation
of the available reserve to the magnitude of the load disturbance,
194 S.I. Nanou et al. / Electric Power Systems Research 127 (2015) 186–196
the optimum selection of frequency regulation parameters (droop,
inertia and power reserve) is not unique, but depends on the
particular characteristics of the network under study.
It is noteworthy that high power reserve levels, besides their
obvious economic implications, may lead to low PV generator vol-
tages when operating in FSM-O. As it can be observed in Fig. 13(c),
for r* = 20%, the PV generator voltage is reduced to about 30% of
its MPP value, indicating that reduced frequency controller gains
might be needed if the regulation limits of the DC/DC converter are
reached.
As mentioned in Section 4.1, regulation of the PV array voltage
above Vmpp (in the voltage-source region of the I–V characteris-
tic) is also possible in order to reduce the output power in FSM-O.
Time-domain simulations adopting this policy are shown in Fig. 14,
for the same load disturbances as in Fig. 10, assuming that the
system operates with a power reserve command of r* = 20%. Com-
paring Figs. 13 and 14, even though the desired frequency response
Fig. 15. (a) Solar irradiance, (b) system frequency (c) PV station power, for stochas-
tic solar irradiance fluctuations [19], variable steady state reserves maintained by
the PV station and combined droop and inertia PV frequency controller (Rpv = 5%,
Hpv = 15 s).
characteristics are obtained with much smaller PV voltage excur-
sions when operating above Vmpp , the frequency controller appears
relatively less effective. This is attributed to the increased error
of the P̂max estimator, which has been developed for operation
in the current source region of the I–V characteristic [31]. Hence,
the amount of power reserve actually maintained before entering
FSM is less than 20%, as it can be seen in Fig. 14(b). Nevertheless,
applying another estimator, which would be more accurate in the
voltage source region [21], could possibly lead to improved results,
comparable to those of Fig. 13.
The ability of the PV station to operate satisfactorily at a sched-
uled power reserve command r* in the presence of continuous
solar irradiance variations is demonstrated in Fig. 15, where the
30-min time-series of Fig. 15(a) has been applied [19]. As shown in
Fig. 15(c), the power output of the PV station successfully responds
to the power reserve command r* , thus maintaining the stipulated
primary reserve during normal operation. A slight improvement in
system frequency response can also be observed in Fig. 15(b), as a
result of the action of the combined droop and inertia frequency
controller during fast frequency excursions. Nevertheless, due to
the relatively slow solar irradiance fluctuations, the overall impact
of PV power fluctuations on system frequency is small.
Fig. 16. (a) Wind speed (b) system frequency, (c) PV array power, for stochastic wind
speed variations of the WT, variable steady state reserves maintained by the PV array
and combined droop and inertia PV frequency controller (Rpv = 5%, Hpv = 15 s).
 S.I. Nanou et al. / Electric Power Systems Research 127 (2015) 186–196 195

5.5. Response to stochastic wind speed variations Appendix. Study case system parameter values

In this section, the system response is simulated assuming the See Tables 1–3.
stochastic wind speed profile shown in Fig. 16(a), which was gener-
ated using the model of [11,40]. The goal is to assess the frequency Table 1
PV system parameters.
response of the PV station in the presence of continuous wind
power fluctuations in the network of Fig. 9, while the solar irra- Parameter Symbol Value
diance remains constant. PV array current at MPP (STC) Impp,STC 874 A
The provision of frequency response by the PV system is clear in PV array voltage at MPP (STC) Vmpp,STC 578 V
Fig. 16(b), particularly following the strong wind gust at t = 11 s. If no PV system installed capacity Ppv,n 2 × 500 kWp
power reserve is maintained, the PV plant mainly operates in FSM- DC/DC inductance Ldc 4 mH
DC/DC capacitance Cdc 40 mF
O, as it is unable to provide additional power during frequency dips.
Intermediate capacitance Cint 10 mF
The provision of under-frequency response is possible at increased Rated DC link voltage Vdc 700 V
reserve levels (Fig. 16(c)), thus leading to a smoother frequency Filter inductance Lf 25.5 ␮H
response. However, the differences are not significant enough to Transformer voltage ratio TFpv 20/0.4/0.4 kV
justify the curtailment of PV energy for this purpose, while oper-
ation at reduced output power may lead to unrealistically low PV
generator voltages in transient conditions. Table 2
Diesel generator parameters.

Parameter Symbol Value

6. Conclusions Nominal power PDU,n 2 MW
Droop RDU 5%
In this paper, a model was introduced for a two-stage grid- Integral gain Ki 5
connected PV system with flexible active power control and Diesel engine time constant TD 0.5 s
Valve actuator servomechanism time constant TSM 0.05 s
frequency response capabilities, suitable for power system studies. Diesel engine inertia time constant HD 2.5 s
A power control scheme was proposed and applied to the DC/DC Transformer voltage ratio TFDU 20/6.6 kV
converter, which enables operation at a scheduled level of power
reserve and permits provision of droop-type and inertial response
from the PV station, via modulation of the PV array voltage. A lin- Table 3
earized small-signal model was also developed to assess the effect FCWT parameters.
of PV operating conditions on the stability of the proposed power Parameter Symbol Value
controller when the PV generator operates in FSM. It was shown
WT installed capacity PWT,n 1 MW
that the dynamic response characteristics of the closed-loop system
WT inertia constant HWT 2.5 s
are mainly dependent on solar irradiance, whereas the effect of PV Transformer voltage ratio TFWT 20/0.69 kV
power variations at given ambient conditions is insignificant. Con-
sequently, the proposed power control scheme is able to effectively
modulate the PV power output for frequency response purposes,
References
under wide variations in ambient and operating conditions.
To assess the effectiveness of alternative frequency controller [1] ENTSO-E, Network Code for Requirements for Grid Connection Applicable to
implementations presented in this paper, a small isolated net- All Generators, ENTSO-E, Belgium, 2013.
work was selected as a study case and various grid disturbances [2] Technical Guideline – Generating Plants Connected to the Medium-Voltage
Network, BDEW, Germany, 2008.
were investigated through simulation. The results obtained from [3] J. von Appen, M. Braun, T. Stetz, K. Diwold, D. Geibel, Time in the Sun: the
the time-domain simulations demonstrate the potential for a sub- challenge of high PV penetration in the German electric grid, IEEE Power Energy
stantial contribution of the PV system to frequency control. The Mag. 11 (2013) 55–64.
[4] I.D. Margaris, S.A. Papathanassiou, N.D. Hatziargyriou, A.D. Hansen, P. Sørensen,
droop-type controller contributes to the reduction of maximum Frequency control in autonomous power systems with high wind power pen-
frequency excursions, while the inertia controller increases the etration, IEEE Trans. Sustain. Energy 3 (2012) 189–199.
apparent system inertia. The combined droop and inertia control [5] J.F. Conroy, R. Watson, Frequency response capability of full converter wind tur-
bine generators in comparison to conventional generation, IEEE Trans. Power
provides best results, leading to the effective containment of large Syst. 23 (2008) 649–656.
frequency excursions and improving the overall damping of the [6] J. Morren, J. Pierik, S.W.H. de Haan, Inertial response of variable speed wind
system. The effect of droop and virtual inertia parameter selection turbines, Electr. Power Syst. Res. 76 (2006) 980–987.
[7] S.I. Nanou, G.N. Patsakis, S.A. Papathanassiou, Assessment of communication-
was also investigated, confirming that PV station operation is not
independent grid code compatibility solutions for VSC-HVDC connected
particularly sensitive to variations of the values of those parameters offshore wind farms, Electr. Power Syst. Res. 121 (2015) 38–51.
within a reasonable range. [8] M.F.M. Arani, E.F. El-Saadany, Implementing virtual inertia in DFIG-based wind
power generation, IEEE Trans. Power Syst. 28 (2013) 1373–1384.
The main limitation of PV stations without internal ESS is their
[9] M.R. Aghamohammadi, H. Abdolahinia, A new approach for optimal sizing
inherent inability to provide under-frequency response, unless of battery energy storage system for primary frequency control of islanded
power reserves are maintained during normal operation. In this Microgrid, Int. J. Electr. Power Energy Syst. 54 (2014) 325–333.
study, a reserve level of 10–20% suffices in order for the PV system to [10] G. Delille, B. François, G. Malarange, Dynamic frequency control support by
energy storage to reduce the impact of wind and solar generation on isolated
have a positive effect on frequency regulation, although the results power system’s inertia, IEEE Trans. Sustain. Energy 3 (2012) 931–939.
would probably not justify the cost of output power curtailments, [11] M. Torres, L.A.C. Lopes, Frequency control improvement in an autonomous
solely for frequency control purposes. power system: an application of virtual synchronous machines, in: Proceedings
of the 8th IEEE International Conference on Power Electronics and ECCE Asia
(ICPE & ECCE), May 30–June 3, Jeju, Korea, 2011, pp. 2188–2195.
[12] N. Sa-ngawong, I. Ngamroo, Intelligent photovoltaic farms for robust frequency
Acknowledgment stabilization in multi-area interconnected power system based on PSO-based
optimal Sugeno fuzzy logic control, Renew. Energy 74 (2015) 555–567.
[13] M. Datta, T. Senjyu, A. Yona, T. Funabashi, C.H. Kim, A frequency-control
The work of Sotirios Nanou is supported by IKY Fellowships of approach by photovoltaic generator in a PV–diesel hybrid power system, IEEE
Excellence for Postgraduate Studies in Greece–Siemens Program. Trans. Energy Convers. 26 (2011) 559–571.
196 S.I. Nanou et al. / Electric Power Systems Research 127 (2015) 186–196
[14] R. Bhatt, B. Chowdhury, Grid frequency and voltage support using PV systems
with energy storage, in: Proceedings of the North American Power Symposium
(NAPS), August 4–6, Boston, MA, 2011, pp. 1–6.
[15] N. Kakimoto, S. Takayama, H. Satoh, K. Nakamura, Power modulation of pho-
tovoltaic generator for frequency control of power system, IEEE Trans. Energy
Convers. 24 (2009) 943–949.
[16] D. Yan, S. Jianhui, S. Yong, A unified power controller for photovoltaic generators
in microgrid, in: Proceedings of the 4th International Conference on Electric
Utility Deregulation and Restructuring and Power Technologies (DRPT), July
6–9, Weihai, Shandong, 2011, pp. 1121–1125.
[17] A. Hoke, D. Maksimović, Active power control of photovoltaic power systems,
in: Proceedings of the 1st IEEE Conference on Technologies for Sustainability
(SusTech), August 1–2, Portland, OR, 2013, pp. 70–77.
[18] M. Datta, T. Senjyu, A. Yona, T. Funabashi, A fuzzy based method for leveling
output power fluctuations of photovoltaic–diesel hybrid power system, Renew.
Energy 36 (2011) 1693–1703.
[19] T. Senjyu, M. Datta, A. Yona, C.H. Kim, A control method for small utility con-
nected large PV system to reduce frequency deviation using a minimal-order
observer, IEEE Trans. Energy Convers. 24 (2009) 520–528.
[20] P.P. Zarina, S. Mishra, P.C. Sekhar, Exploring frequency control capability of a PV
system in a hybrid PV-rotating machine-without storage system, Int. J. Electr.
Power Energy Syst. 60 (2014) 258–267.
[21] H. Xin, Y. Liu, Z. Wang, D. Gan, T. Yang, A new frequency regulation strategy
for photovoltaic systems without energy storage, IEEE Trans. Sustain. Energy 4
(2013) 985–993.
[22] Y. Bae, T.K. Vu, R.Y. Kim, Implemental control strategy for grid stabilization of
grid-connected PV system based on German grid code in symmetrical low-to-
medium voltage network, IEEE Trans. Energy Convers. 28 (2013) 619–631.
[23] P. Veena, V. Indragandhi, R. Jeyabharath, V. Subramaniyaswamy, Review of grid
integration schemes for renewable power generation system, Renew. Sustain.
Energy Rev. 34 (2014) 628–641.
[24] L. Chen, A. Amirahmadi, Q. Zhang, N. Kutkut, I. Batarseh, Design and implemen-
tation of three-phase two-stage grid-connected module integrated converter,
IEEE Trans. Power Electron. 29 (2014) 3881–3892.
[25] H. Choi, M. Ciobotaru, V.G. Agelidis, High gain DC/DC converter for the grid
integration of large-scale PV systems, in: Proceedings of the IEEE International
Symposium on Industrial Electronics (ISIE), May 28–31, Hangzhou, China, 2012,
pp. 1011–1016.
[26] S.B. Kjaer, J.K. Pedersen, F. Blaabjerg, A review of single-phase grid-
connected inverters for photovoltaic modules, IEEE Trans. Ind. Appl. 41 (2005)
1292–1306.
[27] A.F. Okou, O. Akhri, R. Beguenane, M. Tarbouchi, Nonlinear control strategy
insuring contribution of PV generator to voltage and frequency regulation,
in: Proceedings of the 6th IET International Conference on Power Electronics
Machines and Drives (PEMD), March 27–29, Bristol, UK, 2012, pp. 1–5.
[28] R.G. Wandhare, V. Agarwal, Novel control scheme for high power centralized
PV-grid system to realize functionalities of AVR and governor as in conventional
generators, in: Proceedings of the 37th IEEE Photovoltaic Specialists Conference
(PVSC), June 19–24, Seattle, WA, 2011, pp. 2460–2465.
[29] V.A.K. Pappu, B. Chowdhury, R. Bhatt, Implementing frequency regulation capa-
bility in a solar photovoltaic power plant, in: Proceedings of the North American
Power Symposium (NAPS), September 26–28, Arlington, TX, 2010, pp. 1–6.
[30] L.D. Watson, J.W. Kimball, Frequency regulation of a microgrid using solar
power, in: Proceedings of the 26th IEEE Annual Applied Power Electronics Con-
ference and Exposition (APEC), March 6–11, Fort Worth, TX, 2011, pp. 321–326.
[31] S. Nanou, A. Papakonstantinou, S. Papathanassiou, Control of a PV generator to
maintain active power reserves during operation, in: Proceedings of the 27th
European Photovoltaic Solar Energy Conference and Exhibition (EU PVSEC),
September 24–28, Frankfurt, Germany, 2012, pp. 4059–4063.
[32] A. Yazdani, A.R. Di Fazio, H. Ghoddami, M. Russo, M. Kazerani, J. Jatskevich, et al.,
Modeling guidelines and a benchmark for power system simulation studies
of three-phase single-stage photovoltaic systems, IEEE Trans. Power Deliv. 26
(2011) 1247–1264.
[33] S.I. Nanou, S.A. Papathanassiou, Modeling of a PV system with grid code com-
patibility, Electr. Power Syst. Res. 116 (2014) 301–310.
[34] ABB Solar Inverters for Photovoltaic Systems – Brochure, 2014, http://www.
abb.com/solarinverters
[35] S.A. Papathanassiou, M.P. Papadopoulos, Dynamic characteristics of
autonomous wind–diesel systems, Renew. Energy 23 (2001) 293–311.
[36] R.W. Erickson, D. Maksimovic, Fundamentals of Power Electronics, 2nd ed.,
Springer, Norwell, MA, 2001.
[37] S. Nanou, S. Papathanassiou, G. Vokas, Small signal analysis and gain scheduling
control of a photovoltaic DC/DC converter, in: Proceedings of the 27th European
Photovoltaic Solar Energy Conference and Exhibition (EU PVSEC), September
24–28, Frankfurt, Germany, 2012, pp. 3579–3583.
[38] W. Xiao, W.G. Dunford, P.R. Palmer, A. Capel, Regulation of photovoltaic voltage,
IEEE Trans. Ind. Electron. 54 (2007) 1365–1374.
[39] A. Davoudi, J. Jatskevich, Realization of parasitics in state-space average-value
modeling of PWM DC–DC converters, IEEE Trans. Power Electron. 21 (2006)
1142–1147.
[40] P.M. Anderson, A. Bose, Stability simulation of wind turbine systems, IEEE Trans.
Power Appar. Syst. PAS-102 (1983) 3791–3795.