Solar Energy 153 (2017) 64–76

Contents lists available at ScienceDirect

Solar Energy
journal homepage: www.elsevier.com/locate/solener

Active power control of three-phase grid-connected solar PV systems

using a robust nonlinear adaptive backstepping approach
T.K. Roy, M.A. Mahmud ⇑
Electrical Power & Energy Systems Research Lab (EPESRL), School of Engineering, Deakin University, Geelong, VIC 3220, Australia

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

Article history: This paper presents a robust controller design for three-phase grid-connected solar PhotoVoltaic (PV) sys-
Received 1 November 2016 tems to control active power. The controller is designed based on a nonlinear adaptive backstepping
Received in revised form 26 March 2017 approach and the robustness of the proposed scheme is ensured by considering parametric uncertainties
Accepted 17 April 2017
as well as external disturbances. In the proposed control strategy, all parameters within the grid-
connected solar PV systems are considered as unknown which are then estimated through the adaptation
laws. These estimated parameters along with external disturbances are incorporated into the controller
Keywords:
to ensure the overall stability of the whole system through the formulation of Control Lyapunov
Active power
Control Lyapunov function
Functions (CLFs). An Incremental Conductance (IC) method is used to track the Maximum Power Point
External disturbances (MPP) at which a constant DC-link voltage is maintained and this voltage is used to obtain the reference
Power quality value of the current which is used for active power control. A three-phase grid-connected solar PV is used
Parametric uncertainties to evaluate the performance of the proposed control scheme under different operating conditions. The
Robust adaptive backstepping controller simulations results clearly indicate the robustness of the proposed scheme in terms of injecting active
power into the grid and improving power quality as compared to an existing adaptive backstepping
controller.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction action of the inverter. Maximum Power Point Tracking (MPPT)

The integration of solar PhotoVoltaic (PV) systems is increasing
due to several advantages such as ease of installation, noiseless
operation, safer operation with lower operational costs, and
environment friendly operation without any air pollution
(Vithayasrichareon and MacGill, 2016; Zakzouk et al., 2015;
Tsang and Chan, 2013). Despite these advantages, there are several
challenges for the grid integration of solar PV systems especially
during the peak power generation (Yang et al., 2015). Moreover,
the lifetime of solar PV generators are much less as compared to
the conventional fossil fuel generators (Yang et al., 2015). Thus, it
is essential to utilize the maximum benefit from solar PV systems
which is possible if the maximum generated power can be deliv-
ered into the grid. This can only be achieved by controlling active
power and operating the solar PV system at unity power factor
(Hassaine et al., 2014; Bullich-Massaguá et al., 2017).
It is essential to track the Maximum Power Point (MPP) before
delivering the maximum power into the grid through control
⇑ Corresponding author.
E-mail addresses: tkroy@deakin.edu.au (T.K. Roy), apel.mahmud@deakin.edu.au
(M.A. Mahmud).
techniques are used to track the MPP of solar PV systems through
DC–DC converters under changing atmospheric conditions
(Subudhi and Pradhan, 2013; Zainuri1 et al., 2014; Elobaid et al.,
2015; Faraji et al., 2014; Cecati et al., 2017). The Perturb and
Observe (P&O) algorithm is the most commonly used one for
extracting maximum power from PV arrays due to its simplicity
(Bianconi et al., 2013; Ahmed and Salman, 2015). However, this
method has several limitations, e.g., continuous oscillations around
the operating point and poor tracking result or slower convergence
speed. The Incremental Conductance (IC) method overcomes the
limitations of the P&O method by providing faster convergence
speed under changing atmospheric conditions (Sekhar and
Mishra, 2014; Mei et al., 2011; Elgendy et al., 2016). After obtaining
MPP, it is essential to control active power through switching
actions of inverters.
Linear controllers are commonly used to control the active
power in grid-connected solar PV systems (Selvaraj and Rahim,
2009; Dash and Kazerani, 2011; Rahim et al., 2007; Lauria and
Coppola, 2014). These linear controllers are very useful to serve
their purposes over a fixed set of operating points as these con-
trollers are mainly designed based on linearized models of grid-
connected solar PV systems. For this reason, linear controllers are

http://dx.doi.org/10.1016/j.solener.2017.04.044
0038-092X/Ó 2017 Elsevier Ltd. All rights reserved.
 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76
unable to achieve desired control objectives under a wide variation
of operating points, i.e., under rapidly changing of atmospheric
conditions (Dash and Kazerani, 2011; Chowdhury, 2016).
Nonlinear controllers are used to overcome the limitation of
operating points for grid-connected solar PV systems (Hao et al.,
2013; Kotsopoulos et al., 2003; Mahmud et al., 2014; Mahmud
et al., 2014; Lalili et al., 2011). There are different types of nonlin-
ear controllers in the existing literature of grid-connected solar PV
systems such as Sliding Mode Controller (SMC) (Hao et al., 2013),
Model Predictive Controller (MPC) (Kotsopoulos et al., 2003), Feed-
Back Linearizing (FBL) controller (Mahmud et al., 2014; Mahmud
et al., 2014; Lalili et al., 2011), and backstepping controller (Roy
et al., 2015).
The FBL scheme is an effective way to control active power in
grid-connected solar PV systems which cancels inherent nonlinear-
ities using nonlinear coordinate transformation and transforms the
nonlinear system into a fully or partially linearized one. An exact
FBL scheme is used in Lalili et al. (2011), Zue and Chandra (2009)
and Lalili et al. (2013) to control active power along with an aim
to enhance dynamic stability. However, a grid-connected solar PV
system may not always be exactly linearized and thus, this con-
troller can be implemented by imposing unrealistic some assump-
tions on the dynamical model of the PV system. To overcome these
problems, a partial FBL controller is proposed in Mahmud et al.
(2014) and Mahmud et al. (2014) to control the active power over
a wide range of operating regions. However, these FBL controllers
are very sensitive to the variation of parameters within the grid-
connected solar PV system. Moreover, the implementation of FBL
controllers requires exact parametric values of the system which
is really hard to know in practice. The MPC as proposed in
Kotsopoulos et al. (2003) also provides similar benefits such as fast
dynamic response, accurate reference tracking, and constant
switching frequency. However, model predictive approaches are
highly sensitive to the variations of systems parameters as well
as external disturbances.
The SMCs provide robust performance against the parameter
variations and external disturbances. A SMC is proposed in Kim
(2006) and Kim (2007) to ensure the stability of a three-phase
grid-connected PV system by controlling the current injection into
the grid. Though this SMC has strong robustness and good regula-
tion properties, the output power may significantly reduce due to
chattering phenomenon. A similar control approach is proposed
in Hu et al. (2011), Bakhshi and Sadeh (2016) and Yatimi and
Aroudam (2016) which is mainly based on a time-varying sliding
surface and the main aim is to control the power injection into
the grid. However, the changes in atmospheric conditions are very
fast in solar PV systems which make the selection of the time-
varying sliding surface as extremely difficult.
Adaptive backstepping controllers provide promising solutions
for maintaining the dynamic stability of three-phase grid-
connected solar PV systems while considering the parameters of
the system as completely unknown and dynamically estimating
these parameters through the adaptation laws (Roy et al., 2016).
A Lyapunov function based backstepping controller is proposed
in Thao and Uchida (2013) to control power through the current
injection into the grid. However, the backstepping controller does
not consider neither parametric uncertainties nor external distur-
bances during the design process. An adaptive backstepping con-
troller is used in Roy et al. (2016) where parametric uncertainties
are considered and external disturbances are neglected. However,
the external disturbances have significant effects on the stability
of three-phase grid-connected solar PV systems and therefore, it
is essential to design controller which has the capability to capture
both external disturbances as well as parametric uncertainties.
This paper aims to control the active power injection into the
grid from three-phase grid-connected solar PV systems through
65
the regulation of the respective current using a robust adaptive
backstepping technique. The proposed adaptive control scheme
offers the following advantages as compared to the existing litera-
ture so far discussed in this paper:

easy to integrate with the existing MPPT algorithm while ensur-
ing better performance, e.g., the proposed scheme is used in
conjunction with an existing IC-based MPPT method as dis-
cussed in Mei et al. (2011);

parameter sensitivity problems of FBL techniques are overcome
through the estimation of parameters;

provides robustness against both parametric uncertainties and
external disturbances while eliminating the necessity for the
selection of the sliding surface;

uses the estimated values of unknown parameters and consid-
ers the bound of external disturbances during the implementa-
tion of the proposed scheme and thus, ensures the tracking
accuracy with minimum error (i.e., closed to 0%);

injects power into the grid with improved power quality, i.e.,
with a lower value of Total Harmonic Distortion (THD) which
is usually less than 2.5% though its standard value is around
5%; and

allows to operate the system over a range of operating points
with faster settling time.
In this paper, an IC-based MPPT technique is used to maintain a
constant DC voltage across the DC-link capacitor which is later
used to calculate the reference current. All parameters within the
PV system are considered as unknown and estimated through
the adaptation laws while the disturbances are bounded to ensure
the overall stability of grid-connected PV systems. The overall sta-
bility is analyzed through the formulation of CLFs and the robust-
ness of the proposed scheme against both parametric uncertainties
and external disturbances is analyzed through simulation results
under different operating conditions. All these actions are per-
formed in such a way that the desired control objective is achieved,
i.e., the desired active power is delivered into the grid with a lower
value of the THD. The effectiveness of the proposed scheme is ana-
lyzed on a three-phase grid-connected solar PV system and com-
pared with an existing adaptive backstepping approach (Roy
et al., 2016).
2. Modeling of PV systems and control problem formulation
Fig. 1 shows a Voltage Source Inverter (VSI)-based three-phase
grid-connected solar PV system. The system as shown in Fig. 1
comprises a PV array, DC-link capacitor (C), single-stage three-
phase inverter with six switches (S1 ; S2 ; S3 ; S4 ; S5 ; &S6 ), three con-
necting lines with resistor (R) as well as filter inductor (L), and grid
supply point with three-phase voltages (ea ; eb , and ec ). In this sys-
tem, the MPP tracker is used to maintain a constant DC-link voltage
and the proposed control scheme will ensure the convergence of
this voltage to its desired value. In order to this, it is essential to
Fig. 1. Three-phase grid-connected PV system.
66 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76

obtain the dynamic characteristics of the DC-link capacitor which considers the parametric uncertainties and external disturbances,
can be obtained by applying the KCL at the node where the capac- the control problems need to be formulated by considering these
itor is connected and can be written as follows: factors.
The performance of any controller depends on the parameters
dv dc
C ¼ ipv  idc ð1Þ of the system as well as accuracy of the model. However, it is very
dt hard to know the exact parameters and models of grid-connected
where v dc is the DC-link voltage, ipv is the output current of the solar solar PV systems. Therefore, the parameters of grid-connected
PV array, and idc is the input current to the inverter. Since the aim is solar PV systems can be considered as unknown and the controller
to control the amount of active power injected into the grid, it is can be designed by incorporating these unknown parameters
essential to simplify Eq. (1) based on the active power balance within the model. When the parameters (R; L, and C) within the
between DC- and AC-sides, i.e., P dc ¼ P ac where Pdc is the DC- PV system model are considered as unknown, it can be written as:
power and Pac is the AC power. The expression for AC and DC power
1 R 1
can be written as follows: h1 ¼ ; h2 ¼ ; and h3 ¼
C L L
P dc ¼ v dc idc
where h1 ; h2 , and h3 are unknown parameters. With these unknown
3 ð2Þ
P ac ¼ ðEd Id þ Eq Iq Þ parameters, the dynamical model of a grid-connected PV system as
2 represented by Eqs. (7)–(9) can be rewritten as:
Using the concept of power balance, Eq. (1) can be rewritten as  
3
follows: v_ dc ¼ h1 ipv  Ed Id ð10Þ

 2v dc
dv dc 1 3ðEd Id þ Eq Iq Þ
¼ ipv  ð3Þ
dt C 2v dc I_d ¼ h2 Id þ xIq  h3 Ed þ h3 v dc Sd ð11Þ
The output current of the inverter needs to be controlled in
order to control active power which can clearly be seen from Eq. I_q ¼ h2 Iq  xId þ h3 v dc Sq ð12Þ
(2). From Eq. (2), it can also be seen that the AC current at the out-
The solar PV system exhibits intermittent characteristics due to
put side of the inverter has two components-one is related to the
the uncertainties in solar irradiation and environmental tempera-
active component while other one is for the reactive component.
ture. There exist model inaccuracies due to these uncertainties
Thus, these two components control the amount of active and reac-
along with measurement noises. With these external disturbances,
tive power injection into the grid and for this purpose, the dynamic
the dynamical model as represented by Eqs. (10)–(12) can be writ-
characteristics of these currents need to be obtained. In dq refer-
ten as follows:
ence frame, the dynamics of the current can be written as follows:
 
3
R Ed v dc v_ dc ¼ h1 ipv  Ed Id þ d1 ð13Þ
I_d ¼  Id þ xIq  þ Sd ð4Þ 2v dc
L L L

R v dc I_d ¼ h2 Id þ xIq  h3 Ed þ h3 v dc Sd þ d2 ð14Þ

I_q ¼  Iq  xId þ Sq ð5Þ
L L
where x is the angular frequency, Sd is the switching function in d- I_q ¼ h2 Iq  xId þ h3 v dc Sq þ d3 ð15Þ
frame, and Sq is the switching function in q-frame. These Sd and Sq
where d1 ; d2 , and d3 are the external disturbances which include the
are the control inputs in dq reference frame. When the syn-
measurement noises and model inaccuracies due to solar irradiation
chronously rotating dq frame is aligned with the d-axis of the grid
and environmental temperature. At this point, the desired model to
voltage at steady-state condition, the q-axis component of the grid
design the proposed controller is obtained, i.e., Eqs. (13)–(15) reflect
voltage will be zero, i.e., Eq ¼ 0. In this condition, the active power
the dynamical model and the proposed controller needs to be
delivered to the grid can be written as follows:
designed based on this model. The adaptation laws are used to esti-
3 mate unknown parameters within this model while external distur-
Pac ¼ Ed Id ð6Þ
2 bances are bounded to ensure the robustness while the stability of
the system is ensured through the formulation of CLFs. The proposed
From Eq. (6), it can be seen that active power can be controlled just
controller is designed to overcome the following concerns:
by controlling the d-axis current. Finally, Eqs. (3)–(5) can further be
simplified as follows:
 
Control the DC-link voltage at the desired value to maintain the
1 3 power balance between DC- and AC-sides;
v_ dc ¼ ipv  Ed I d ð7Þ
C 2v dc
Inject a pure sinusoidal current into the grid to maintain the
IEEE-1547 standard, i.e., current with lower THD values; and
R Ed v dc
Deliver maximum active power into the grid under a wide vari-
I_d ¼  Id þ xIq  þ Sd ð8Þ
L L L ation of operating points.

R v dc The following section stresses on the design of the proposed

I_q ¼  Iq  xId þ Sq ð9Þ
L L
From Eq. (7), it can be seen that the DC-link voltage can be regulated
by controlling the d-axis current (Id ) while the two current compo-
nents can be regulated by controlling the switches which can also
be seen from Eqs. (8) and (9). Thus, the complete dynamical model
of three-phase grid-connected systems for controlling active power
can be represented by Eqs. (7)–(9) and the controller needs to be
designed based on this model. Since the proposed control scheme
robust adaptive backstepping controller which will resolve the
aforementioned concerns.
3. Proposed controller design
This section is intended to present the proposed controller
design process along with an overview of the MPPT technique
based on the IC method as proposed in Mei et al. (2011). The
 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76 67

proposed controller design scheme involves different steps which where s is the tuning function which can be defined as follows:
are discussed in the following subsection.  
3
s ¼ c1 e1 ipv  Ed I d ð23Þ
2v dc
3.1. Robust adaptive backstepping controller design
From Eq. (22), it can be seen that the first term on the right side is
The switching control inputs Sd and Sq need to be obtained in negative definite and the second term is also negative in terms of
such a way that the state trajectories can be followed through the estimation error, estimated value, and tuning function. How-
the control actions, i.e., their tracking errors will converge to zero. ever, this second term is tolerated at this step as the estimation of
Here, the reference value for Iq will be zero as the main control h1 is not finalized so far while the last term will be zero if the error
objective is to control active power while maintaining a constant e1 is zero. The dynamic of the virtual control input (a) will be used
DC-link voltage. However, the reference value of Id depends on in the next step which can be calculated as follows:
the amount of active power injection into the grid which is also
a ¼ A þ Bh1  D1 ð24Þ
related to the reference value of the DC-link voltage. The proposed
controller also ensures the convergence of this DC-link voltage where
whose reference value is calculated from the MPPT algorithm.
i_pv 2E_ d v dc ipv 2v dc v€ dcref 2v dc
The following steps are used to obtain the control laws: A ¼ v dc    2
Step 1: In this step, the tracking error dynamics of the DC-link 3Ed 3E2d 3Ed ^h1 3Ed ^h21
voltage is stabilized through designing the d-axis current. To do 2v dc v_ dcref k1
this, the tracking error for the DC-link voltage can be defined as: ðk1 e1  v_ dcref ÞðE_ d ^h1 þ Ed h_ 1 Þ 
3Ed ^h1
e1 ¼ v dc  v dcref ð16Þ   
2 ipv k1 e1 k1 v dc 3
where v dcref is the DC-link reference voltage. Using Eq. (13), the B¼ þ þ ipv  Ed Id
3 Ed Ed ^h1 Ed ^h1 2v dc
dynamics of e1 can be written as:  
  2 ipv k1 e1 k1 v dc
3 D1 ¼  þ þ d1
e_ 1 ¼ h1 ipv  Ed Id þ d1  v_ dcref ð17Þ 3 Ed Ed ^h1 Ed ^h1
2v dc
Step 2: Since a is a virtual control input, a second error variable
An unknown parameter h1 appears in Eq. (17) and it is essential to to represent the tracking of another state, i.e., the d-axis current
define this parameter in term of the estimated value h^1 in order to
tracking error can be defined as:
handle this unknown parameter. An estimator (as discussed later in
this subsection) is used to estimate this unknown parameter for e2 ¼ Id  a ð25Þ
which the estimation error can be defined as ~
h1 ¼ h1  ^
h1 and using Using Eqs. (14) and (24) and incorporating the estimation errors,
this, Eq. (17) can be rewritten as follows: the error dynamics of d-axis current can be written as:
 
e_ 1 ¼ ð^h1 þ ~h1 Þ ipv 
3
Ed Id þ d1  v_ dcref ð18Þ e_ 2 ¼ ð^h2 þ ~h2 ÞId þ xIq  ð^h3 þ ~h3 ÞðEd  v dc Sd Þ þ d2  A
2v dc
 Bð^h1 þ ~h1 Þ þ D1 ð26Þ
where Id is a stabilizing function for Eq. (18). At this point, the CLF,
in order to achieve v dc ! v dcref (zero tracking error), can be chosen where ~
h2 ¼ h2  ^
h2 and ~
h3 ¼ h3  ^
h3 represent the parameter esti-
as follows: mation errors.
  The tracking error for the q-axis current can be written as:
1 2 1 ~2
W1 ¼ e1 þ h1 ð19Þ e3 ¼ Iq  Iqref ð27Þ
2 c1
where c1 represents the adaptation gain and the convergence speed whose dynamics, in terms of the estimation error, can be written as
follows:
of the estimation depends on the numerical value of this gain. Tak-
ing the time derivative of W 1 and by inserting the value of e_ 1 from e_ 3 ¼ d3  ð^h2 þ ~h2 ÞIq  xId  ð^h3 þ ~h3 Þv dc Sq  I_qref ð28Þ
Eq. (18), it can be written as:

   As overshadowed earlier, the switching control inputs will steer
_ 1 ¼ e1 ^h1 ipv  3 Ed Id  v_ dcref  1 ~h1
W the states (Id and Iq ) to their desired values. In this case, the desired
2v dc c1 value of Id depends on the amount of power while that of for Iq is

  ð20Þ
^h_ 1  c e1 ipv  3
Ed I d þ e1 d1
zero. During the design of an adaptive backstepping controller, the
1
2v dc CLF is used to obtain the conditions for the control laws at which
the designed controller will have the ability to force the states to
The negative semi-definiteness of the CLF ensures the convergence
their desired values. For the three-phase grid-connected solar PV
of the DC-link voltage through the stabilizing function (Id ). Thus, the
system, the following CLF is used to obtain the control inputs Sd
synthetic value of Id needs to be selected in such a way that W _ 160
and Sq .
for faster convergence of the error dynamics and better precision  
tracking. If the synthetic value of Id is a, it can be chosen as follows: W2 ¼ W1 þ
1 2 1 1
e2 þ e23 þ ~h22 þ ~h23 ð29Þ
2 c2 c3
2v 2v
a ¼ dc ipv þ dc^ ðk1 e1  v_ dcref Þ ð21Þ
where c2 and c3 represent two other adaptation gains and the time
3Ed 3Ed h1
derivative of the CLF can be written as follows:
where k1 is a positive constant, i.e., k1 > 0 which is used to tune the
output response to match with the desired trajectories. With this W _ 1 þ e2 e_ 2 þ e3 e_ 3  1 ~h2 ^h_ 2  1 ~h3 ^h_ 3
_ 2¼W ð30Þ
positive constant, Eq. (20) can be written as: c2 c3
_ 1 from Eq. (22), e_ 2 from Eq. (26), and e_ 3
_ 1 ¼ k1 e2  1 ~h1 ð^h_ 1  sÞ þ e1 d1
W ð22Þ
Substituting the values of W
1
c1 from Eq. (28) into Eq. (30); it can be written as:
68 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76

_ 2 ¼ k1 e2 þ e2 ½^h2 Id þ xIq  ^h3 ðEd  v dc Sd Þ  A  ^h1 B

W where sgn is the signum function which can be defined as follows:
1
8
1 _ < þ1 if e > 0
>
þ e3 ð^h2 Iq  xId  ^h3 v dc Sq  I_qref Þ  ~h1 ð^h1  s
c1 sgnðeÞ ¼ 0 if e ¼ 0 ð35Þ
>
:
1 _ 1 _ 1 if e < 0
þ c1 Be2 Þ  ~h2 ð^h2 þ c2 e2 Id þ c2 e2 Iq Þ  ~h3 ½^h3
c2 c3
If Sd and Sq are switching control inputs, Eq. (33) can be written
þ c3 e2 ðEd  v dc Sd Þ þ c3 e3 v dc Sq  þ e1 D1 þ e2 d2 þ e3 d3 ð31Þ
as:
From Eq. (31)), it can be seen that there exist the estimated values _ 2 ¼ k1 e2  k2 e2  k3 e2 þ e1 ðD1  F 1 sgnðe1 ÞÞ þ e2 ðd2
W
of the unknown parameters h ^1 ; ^
h2 and ^
h3 whose effects need to be
1 2 3

eliminated in order to ensure the stability of the whole system.  F 2 sgnðe2 ÞÞ þ e3 ðd3  F 3 sgnðe3 ÞÞ ð36Þ
To do this, the parameter estimation laws can be chosen as follows: Eq. (36) will be negative semi-definite if the external disturbances
^h_ 1 ¼ s  c Be2 are bounded as follows: jD1 j 6 F 1 ; jd2 j 6 F 2 , and jd3 j 6 F 3 .
1 Therefore, the designed switching controllers will stabilize the
^h_ 2 ¼ c ðe2 Id þ e2 Iq Þ ð32Þ whole PV system. The reference value of the current for delivering
2
maximum active power into the grid depends on the DC-link volt-
^h_ 3 ¼ c ½e2 ðEd  v dc Sd Þ þ e3 v dc Sq 
3 age which in turns obtained from the MPPT system. An MPPT algo-
Using these adaptation or estimation laws, Eq. (31) can be rewritten rithm is discussed in the following subsection.
as follows:
3.2. MPPT algorithm
_ 2 ¼ k1 e2 þ e2 ½^h2 Id þ xIq  ^h3 ðEd  v dc Sd Þ  A  ^h1 B
W 1

þ e3 ð^h2 Iq  xId  ^h3 v dc Sq  I_qref Þ þ e1 D1 þ e2 d2 In this paper, an existing MPPT algorithm based on the IC
method is used in order to make sure that the maximum power
þ e3 d3 ð33Þ is extracted from the solar PV unit as well as to obtain a constant
The stability of the whole system will be ensured if the derivative of DC voltage at the input side of the inverter. Since the designed con-
_ 2 6 0. At this stage, the dynam- troller ensures the desired tracking performance of all physical
W 2 is negative semi-definite, i.e., W
properties (e.g., voltage, current, and power), it is essential to
ics of the d-and q-axes current tracking errors can be stabilized by
obtain the desired values of these properties. The reference value
designing Sd and Sq as follows:
of the DC-link voltage can be obtained using an MPPT algorithm
1 ^ which in turn can be used to calculate the reference values of cor-
Sd ¼ ½h I  xIq þ A þ ^h1 B  k2 e2
^h3 v dc 2 d responding AC power. Again, it is worth to note that the main con-
 F 2 sgnðe2 Þ  e1 F 1 sgnðe1 Þ ð34Þ tribution of this paper is to design the controller for the inverter to
regulate the active power.
1 ^
Sq ¼ ½h2 Iq þ xId þ I_qref  k3 e3  F 3 sgnðe3 Þ In this paper, the IC method as proposed in Mei et al. (2011) is
h^3 v dc used to calculate the reference value of the DC-link voltage. The IC
method in Mei et al. (2011) track the maximum power based on
Table 1
Parameters of the three-phase grid-connected PV system.
the power slope, i.e., the derivative of the output power of PV
arrays with respect to the array voltage where the power gener-
Parameter Symbol Value ated by the PV array is given by
400 lF
Ppv ¼ ipv v pv
DC-link capacitor C
DC-link voltage V dc 500 V ð37Þ
Grid impedance R; L 0.2 X, 10 mH
Based on Eq. (37), the power slope dvppvv can be represented as follows:
dP
Switching frequency f inv 20 kHz

Fig. 2. Implementation block diagram of the designed control scheme.

 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76 69

dPpv dipv Di
where Dvppvv represents the incremental conductance and vppvv is the
i
¼ ipv þ v pv ð38Þ
dv pv dv pv instantaneous conductance. In an IC method, the incremental con-
ductance is compared with the instantaneous conductance until
For MPP, the slope will be zero, i.e., dv ppvv ¼ 0 for which Eq. (38) can be
dP
Di
the MPP is achieved, i.e., Dvppvv ¼  vppvv which is done through the
i
simplified as follows:
switching signal of the DC–DC converter. The performance of the
Dipv ipv designed controller is tested on a three-phase grid-connected solar
¼ ð39Þ
Dv pv v pv PV system as discussed in the following section.

Table 2 4. Controller performance evaluation

Numerical values of control parameters.

Parameter Numerical value In this section, the performance of the designed controller is
k1 5 first evaluated on a similar test system as shown in Fig. 1. The
k2 5 PV module in Fig. 1 is a SPR 305-WHT PV module and the detailed
k3 3 description of the PV module along with the corresponding electri-
c1 10
cal characteristics can be found in Roy et al. (2016). The corre-
c2 8
c3 8
sponding parameters of the three-phase grid connected PV
system are listed in Table 1.

Fig. 3. Grid voltage, current, active power delivered into the grid, and THD under standard atmospheric conditions.
70 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76
The PV modules in a grid-connected system can either be in the
tracking or fixed mode in terms of tracking the sunlight. In tracking
mode, the PV modules are integrated with a mechatronic system
which includes some actuators, e.g., DC motors. These actuators
are used to move the PV modules to track the sun in such a way
that these modules capture maximum sunlight (Alexandru,
2013). There are two different types of tracking systems such as
one degree of freedom (single-axis tracking system) and two
degree of freedom (dual-axis tracking system) along with different
tracking mechanisms, e.g., azimuthal, pseudo-azimuthal, equato-
rial or pseudo-equatorial (Alexandru, 2013; Huld et al., 2010;
Semea et al., 2016). On the other, the PV modules do not require
the mechatronic system in the fixed mode (Bione et al., 2004)
and in this paper, the fixed mode is considered for the PV module.
From the implementation block diagram of the designed con-
troller as shown in Fig. 2, it can be seen that the instantaneous grid
voltage and current are transformed from abc frame into dq-frame.
This has been done as the controller is designed in the dq-frame as
discussed in Section II. The online parameter estimators are used
to estimate the unknown parameters of the grid-connected solar
PV system where the estimation initially starts from some randomly
selected initial values. The online parameters estimator continu-
ously uses the control signal, grid voltage, and current in dq-frame
to update the unknown parameters until the steady-state values of
Fig. 4. DC-link voltage tracking and unknown parameter estimations under standard atmospheric conditions.
the desired control objectives are achieved which are the desired
DC-link voltage and the active power for the case as considered in
this paper. White Gaussian Noises (WGNs) are added with all voltage
and current signals in order to demonstrate the robustness of the
designed controller against external disturbances or noises. The
switching control signals are obtained using the estimated parame-
ters, grid voltage, grid current, and DC-link voltage. Since the original
system is in the instantaneous form and the switching signals are
usually the pulses, the control inputs are first transformed from dq
to abc and then through a pulse width modulator which uses the
Pulse Width Modulation (PWM) technique to generate the switching
pulses. All simulations are carried out using MATLAB/SIMULINK
simulation package under different operating conditions.
The performance of the designed controller heavily relies on
control parameters such as the tuning parameters and adaptation
gains. The numerical values of relevant control parameters are
listed in Table 2 where these parameters are obtained for the con-
sidered three-phase grid-connected PV system in order to achieve
the desired control objectives. The performance of the designed
Nonlinear Robust Adaptive Backstepping Controller (NRABC) is
also compared to that of an Existing Nonlinear Adaptive Backstep-
ping Controller (ENABC) as proposed in Roy et al. (2016). The per-
formance of the controller is evaluated by considering the
following three different operating conditions:
 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76

Controller performance at standard atmospheric conditions,

Controller performance with changes in solar irradiation, and

Controller performance with changes in temperature.
An additional case study has also been included in this section
to demonstrate the effectiveness of the designed controller in
terms of maintaining power balance between the generation and
load demand under varying operating conditions. All these case
studies are elaborately discussed in the following.

Case 1: Controller performance under standard atmospheric
conditions
In this case study, the system is simulated at standard atmo-
spheric conditions where the value of solar irradiation is consid-
ered as 1 kW/m2 while the environmental temperature as 25 °C.
The main purpose of the designed controller is to inject maximum
active power (100 kW) into the grid. This will happen if the voltage
and current are in phase with each other. The grid voltage and cur-
rent responses of the system are shown in Fig. 3(a).
Fig. 5. Grid current, active power delivered into the grid, and THD with changes in solar irradiation.
71
Fig. 3(a) clearly indicates that the voltage and current are in
phase with both NRABC and ENABC. From Fig. 3(a), it can also be
seen the output current of the inverter is very close to a pure sinu-
soidal waveform (black line) with negligible harmonic content
while the designed controller is used. On the other hand, the out-
put current (red line) has significant harmonic components when
an ENABC is used though it looks very similar to that of the
designed controller. The improved performance of the designed
controller in terms of improving power quality will be obvious
from Fig. 3(c) and (d) which show the THD in the current with both
controllers. From Fig. 3(c), it can be seen that the THD in the grid
current is 2.12% with the designed controller whereas it is 2.26%
as shown in Fig. 3(d) when ENABC is used. Therefore, it can be said
that designed NRABC outperforms ENABC in terms of injecting
active power into the grid with lower THD.
The power tracking performance of the designed controller can
be seen from Fig. 3(b). Since the controller is designed only to deli-
ver active power into the grid, the total output power of the PV
array, i.e., 100 kW will be delivered into the grid as shown in
Fig. 3(b). From Fig. 3(b), it can also be seen that the designed
72 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76
NRABC controller has less fluctuations(black line) as compared to
the ENABC (red line). The main reason behind these lower fluctua-
tions with the designed NRABC is the inclusion of external distur-
bances during its design process. Since the ENABC does not
consider the effect of external noises or disturbances, there are
more fluctuations in the active power delivered into the grid. In
this paper, the DC-link voltage is considered as 500 V and the
designed controller accurately tracks this voltage with lower fluca-
tions as compared to the ENABC which can be seen from Fig. 4(a).
In this case study, the initial values of estimated parameters are
considered as ^h1 ð0Þ ¼ 2400; ^
h2 ð0Þ ¼ 80, and ^
h3 ð0Þ ¼ 80. The adapta-
tion laws are used to obtain the steady-state values of these param-
eters. The convergence of the estimated parameters is shown in
Fig. 4(b) to (d) from where it can be observed that the final steady-
state values of these parameters are different from their initial val-
ues. These indicate that the designed controller effectively estimate
the unknown parameters while achieving the control objectives.

Case 2: Controller performance with changes in solar
irradiation
During the practical operation of a solar PV system, the solar
irradiation continuously changes which in turn significantly affects
Fig. 6. DC-link voltage tracking and unknown parameter estimations with changes in solar irradiation.
the output power of PV arrays. In this case study, the changes in
solar irradiation is considered to show the effectiveness of the
designed controller. In this situation, the solar irradiation suddenly
drops from 1000 W/m2 to 500 W/m2 at t ¼ 1 s and continues to
maintain this irradiation level till t = 1.3 s. Finally, the irradiance
level settle back to 1000 W/m2 after t ¼ 1:3 s. Under this circum-
stance, the output power of the PV array as well as the power
delivered into the grid will be changed during the period t ¼ 1 s
to t ¼ 1:3 s which can be seen from Fig. 5(b). The designed con-
troller has the ability to inject desired amount of active power
(black line) into the grid with less overshoots as compared to the
ENABC (red line). The output current of the inverter also changes
due to the changes in power. The output current of the inverter
is shown in Fig. 5(a) from where it can clearly be seen that the cur-
rent varies significantly. However, the designed controller (black
line) can inject a better sinusoidal current into the grid than the
existing controller (red line). In this situation, the THD will be more
as compared to the previous case study. Fig. 5(c) shows the THD in
the grid current when the designed controller is used and the
numerical value of the THD is 2.20%. However, the grid current
contains more harmonics if the ENABC is used which can be seen
from Fig. 5(d). In this case, the value of THD is 2.74%. Thus, it can
be said that the grid current is more sinusoidal with the designed
 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76
controller as compared to the ENABC. The corresponding DC-link
voltage tracking performance of the designed controller is shown
in Fig. 6(a) from where it is obvious that the designed controller
has much better tracking capability than the existing controller.
In this case study, the initial values of the unknown parameters
are changed as compared to previous case study and these values
are considered as h ^1 ð0Þ ¼ 1500; h^2 ð0Þ ¼ 15, and h
^3 ð0Þ ¼ 80. With
these initial values, the estimated parameters are shown in Fig. 6
from where it can be seen that the estimated values are exactly
equal to the original values even for their different initial values.
However, the estimations are disturbed due to the changes in solar
irradiations which can also be seen from Fig. 6(b)–(d). Thus, it can be
said that the designed controller is capable to estimate the parame-
ters for any initial values while maintaining the robust tracking per-
formance for the three-phase grid-connected PV system.

Case 3: Controller performance with changes in atmospheric
temperature
In this case study, the performance of the designed controller is
verified when the temperature varies. At the beginning, it is
Fig. 7. Grid current, active power delivered into the grid, and THD with changes in atmospheric temperature.
73
assumed that the system is running at standard temperature, i.e.,
at 25 °C. At t ¼ 0 s, the temperature starts to gradually decrease
from 25 °C and reaches to 17 °C at t ¼ 0:7 s. From t ¼ 0:7 s to
t ¼ 1:35 s, the temperature starts increasing gradually from 17 °C
to 40 °C. Finally, the temperature come back to 25 °C at t ¼ 1:5 s.
In this situation, the grid current is shown in Fig. 7(a). However,
the effects of the changes are not clearly reflected in Fig. 7(a) due
to the longer time period. Fig. 7(b) shows the power injected into
the grid when the aforementioned changes in temperature occur
within the system. From Fig. 7(b), it can be seen that the changes
in temperature affect the power injected into the grid as the output
power of the PV array is influenced by these changes. The improve-
ment in power quality is shown through the THD as shown in Fig. 7
(c) and (d) from where it is clear that the grid current has lower
THD with designed controller. The DC-link voltage tracking is
shown in Fig. 8(a) which clearly shows that the NRABC has better
tracking capability as compared to the ENABC.
In this case, the initial values of unknown parameters are con-
sidered as h ^1 ð0Þ ¼ 2800; h
^2 ð0Þ ¼ 35, and h
^3 ð0Þ ¼ 135 which are
completely different from two previous cases. Fig. 8(b)–(d) clearly
indicate that the designed controller has the ability to estimate
74 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76

unknown parameters even with changes in temperature and well

as for different initial values of the unknown parameters. There-
fore, it can be said that the designed controller provides robust per-
formance against unknown parameters as well changes in
temperature and solar irradiation.

Case 4: Controller performance in terms of maintaining

power balance

In this case study, the three-phase grid-connected PV system is

modified by connecting a variable three-phase load at the output of
the inverter and the single line diagram of the modified PV system
is shown in Fig. 9. The maximum power capacity of the load is con-
sidered as 20 kW with a power factor of 1. At the beginning of this
simulation, it is considered that the PV unit is operating at stan-
dard atmospheric conditions till t = 2 s, i.e., the output power from
the PV unit is 100 kW and the load demand during this period is
10 kW. Therefore, the active power delivered into the grid will be
90 kW from t ¼ 0 s to t ¼ 2 s in order to maintain the power bal-
ance. From Fig. 10, it can be seen that the output power of the
PV unit is 100 kW, the load demand is 10 kW, and the power deliv-
Fig. 9. Grid-connected PV system with load.
ered into the grid is 90 kW. Therefore, it can be said that there is
power balance within the system till t ¼ 2 s.
At t ¼ 2 s, the solar irradiation changes from 1000 W/m2 to
500 W/m2 for which the corresponding output power from the
PV unit is 70 kW and this situation continues till t ¼ 2:5 s while
the load is still 10 kW till t ¼ 3 s. Therefore, the load consumes
10 kW and the remaining 60 kW is delivered into the grid as shown

Fig. 8. DC-link voltage tracking and unknown parameter estimations with changes in atmospheric temperature.
 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76
Fig. 10. Power balance between PV unit, load, and the grid under different operating conditions.
in Fig. 10. From Fig. 10, it can be seen that there are some oscilla-
tions at the instant of t ¼ 2 s as the solar irradiation changes at this
instance. The designed NRABC flattens these fluctuations in a much
better way as compared to the ENABC. The PV units return to its
standard operation at t ¼ 2:5 s and delivers a load of 10 kW. At this
point, the power delivered to the grid increases from 60 kW to
90 kW till t ¼ 3 s as the load varies from 10 kW to 15 kW after this
period. Therefore, the amount of active power delivered into the
grid after t ¼ 3 s is 85 kW. Therefore, it can said that there are fluc-
tuations at t ¼ 2:5 s due to changes in the solar radiation from
500 W/m2 to 1000 W/m2 and at t ¼ 3 s due to the changes in the
load demand. From Fig. 10, it can be seen that the power balance
is maintained by both NRABC and ENABC. However, the fluctua-
tions in power are clearly visible when the ENABC is used while
these are highly damped with the design NRABC.
5. Conclusion
A new control approach has been proposed to control the active
power in a grid-connected solar PV system while the parameters of
the system are unknown and the environmental conditions change
75
abruptly. The developed adaptation laws are robust as these laws
can estimate the unknown parameters for their different initial
values as well as under different atmospheric conditions. These
adapted parameters are incorporated into the controller with an
aim to deliver maximum active power into the grid while main-
taining a constant DC-link voltage. Therefore, it can be said that
the designed controller is robust against parametric uncertainties
due to its capability to estimate unknown parameters under any
situation. The parameter adaptation feature of the designed con-
troller overcome the problem of knowing the exact parameters of
the system.
From simulation results, it is clear that the designed robust
adaptive backstepping controller provides a very satisfactory per-
formance in terms of maintaining steady-state operation under
various operating conditions as compared to the existing con-
troller. The simulation results also indicate that the designed
robust adaptive backstepping controller improves the power qual-
ity of the system by reducing the THD of the injected current into
the grid. Moreover, the designed controller is capable to maintain
the power balance within the grid-connected PV system under dif-
ferent operating conditions. Therefore, another important feature
76 T.K. Roy, M.A. Mahmud / Solar Energy 153 (2017) 64–76

can be noted as the designed controller allows the system to oper- Kotsopoulos, A., Darte, J.L., Hendrix, M.A.M., 2003. Predictive DC voltage control of
single-phase pv inverters with small dc link capacitance. In: Proc. IEEE Int.
ate over an extended operating region.
Symp. Ind. Electron., pp. 793–797.
Future work will contribute to development of a similar plat- Lalili, D., Mellit, A., Lourci, N., Medjahed, B., Berkouk, E.M., 2011. Input output
form by considering the tracking mode of PV modules. The consid- feedback linearization control and variable step size mppt algorithm of a grid-
eration of the tracking mode of PV modules will allow to design connected photovoltaic inverter. Renew. Energy 36 (12), 3282–3291.
Lalili, D., Mellit, A., Lourci, N., Medjahed, B., Boubakir, C., 2013. State feedback
controller for mechanical devices. Future work will also be devoted control and variable step size MPPT algorithm of three-level grid-connected
on a laboratory-based system to implement the designed con- photovoltaic inverter. Sol. Energy 98, 561–571.
troller in real-time applications using a Hardware-In-Loop (HIL) Lauria, D., Coppola, M., 2014. Design and control of an advanced PV inverter. Sol.
Energy 110, 533–542.
system. Mahmud, M.A., Pota, H.R., Hossain, M.J., Roy, N., 2014. Robust partial feedback
linearization stabilization scheme for three-phase grid-connected photovoltaic
References systems. IEEE Trans. Power Deliv. 29 (3), 1221–1230.
Mahmud, M.A., Pota, H.R., Hossain, M.J., Roy, N.K., 2014. Robust partial feedback
linearization stabilization scheme for three-phase grid-connected photovoltaic
Ahmed, J., Salman, Z., 2015. An improved perturb and observe (p&o) maximum
systems. IEEE J. Photovolt. 4 (1), 423–431.
power point tracking (MPPT) algorithm for higher efficiency. Appl. Energy 150,
Mei, Q., Shan, M., Liu, L., 2011. A novel improved variable step-size incremental-
97–108.
resistance MPPT method for PV systems. IEEE Trans. Ind. Electron. 58 (6), 2427–
Alexandru, C., 2013. A novel open-loop tracking strategy for photovoltaic systems.
2434.
Sci. World J. 2013 (205396), 12.
Rahim, N.A., Selvaraj, J., Krismadinata, C.C., 2007. Hysteresis current control and
Bakhshi, R., Sadeh, J., 2016. Voltage positive feedback based active method for
sensorless mppt for grid-connected photovoltaic system. In: Proc. of IEEE Int.
islanding detection of photovoltaic system with string inverter using sliding
Symp. Industrial Electronics, pp. 572–577.
mode controller. Sol. Energy 137, 564–577.
Roy, T.K., Mahmud, M.A., Hossain, M.J., Oo, A.M.T., September 2015. Nonlinear
Bianconi, E., Calvente, J., Giral, R., Mamarelis, E., Petrone, G., Paja, C.A.R., Vitelli, M.,
backstepping controller design for sharing active and reactive power in three-
2013. Perturb and observe MPPT algorithm with a current controller based on
phase grid-connected photovoltaic systems. In: Australasian Universities Power
the sliding mode. Int. J. Electr. Power Energy Syst. 44, 346–356.
Engineering Conference, pp. 1–6.
Bione, J., Vilela, O.C., Fraidenraich, N., 2004. Comparison of the performance of PV
Roy, T.K., Mahmud, M.A., Oo, A.M.T., Haque, M.E., 2016. Robust nonlinear adaptive
water pumping systems driven by fixed, tracking and V-trough generators. Sol.
backstepping controller design for three-phase grid-connected solar
Energy 76 (6), 703–711.
photovoltaic systems with unknown parameters. In: 2016 IEEE Power and
Bullich-Massaguá, E., nalba, M.A.-P., Sumper, A., Boix-Aragones, O., 2017. Active
Energy Society General Meeting (PESGM), pp. 1–5.
power control in a hybrid PV-storage power plant for frequency support. Sol.
Sekhar, P.C., Mishra, S., 2014. Takagi-Sugeno fuzzy-based incremental conductance
Energy 144, 49–62.
algorithm for maximum power point tracking of a photovoltaic generating
Cecati, C., Khalid, H.A., Tinari, M., Adinolfi, G., Graditi, G., 2017. DC nanogrid for
system. IET Renew. Power Gener. 8 (8), 900–914.
renewable sources with modular DC/DC LLC converter building block. IET Power
Selvaraj, J., Rahim, N.A., 2009. Multilevel inverter for grid-connected PV system
Electron. 10 (5), 536–544.
employing digital PI controller. IEEE Trans. Indust. Electron. 56 (1), 149–158.
Chowdhury, M.A., 2016. Dual-loop H1 controller design for a grid-connected single-
Semea, S., Štumbergera, B., Hadžiselimović, M., 2016. A novel prediction algorithm
phase photovoltaic system. Sol. Energy 139, 640–649.
for solar angles using second derivative of the energy for photovoltaic sun
Dash, P.P., Kazerani, M., 2011. Dynamic modelling and performance analysis of a
tracking purposes. Sol. Energy 137, 201–211.
grid-connected current-source inverter based photovoltaic system. IEEE Trans.
Subudhi, B., Pradhan, R., 2013. A comparative study on maximum power point
Sustain. Energy 2 (4), 443–450.
tracking techniques for photovoltaic power systems. IEEE Trans. Sustain. Energy
Elgendy, M.A., Atkinson, D.J., Zahawi, B., 2016. Experimental investigation of the
4 (1), 89–98.
incremental conductance maximum power point tracking algorithm at high
Thao, N.G.M., Uchida, K., 2013. Control the photovoltaic grid-connected system
perturbation rates. IET Renew. Power Gener. 10 (2), 133–139.
using fuzzy logic and backstepping approach. In: 9th Asian Control Conference,
Elobaid, L.M., Abdelsalam, A.K., Zakzouk, E.E., 2015. Artificial neural network-based
pp. 1–8.
photovoltaic maximum power point tracking techniques: a survey. IET Renew.
Tsang, K.M., Chan, W.L., 2013. Three-level grid-connected photovoltaic inverter
Power Gener. 9 (8), 1043–1063.
with maximum power point tracking. Int. J. Energy Convers. Manage. 65, 221–
Faraji, R., Rouholamini, A., Naji, H.R., 2014. FPGA-based real time incremental
227.
conductance maximum power point tracking controller for photovoltaic
Vithayasrichareon, P., MacGill, I.F., 2016. Valuing large-scale solar photovoltaics in
systems. IET Renew. Power Gener. 7 (5), 1294–1304.
future electricity generation portfolios and its implications for energy and
Hao, X., Xu, Y., Liu, T., Huang, L., Chen, W., 2013. A sliding-mode controller with
climate policies. IET Renew. Power Gener. 10 (1), 79–87.
multiresonant sliding surface for single-phase grid-connected VSI with an LCL
Yang, Y., Koutroulis, E., Sangwongwanich, A., Blaabjerg, F., 2015. Minimizing the
filter. IEEE Trans. Power Electron. 28 (5), 2259–2268.
levelized cost of energy in single-phase photovoltaic systems with an absolute
Hassaine, L., Olias, E., Quintero, J., Barrado, A., 2014. Power control for grid
active power control. IEEE Energy Conversion Congress and Exposition, pp. 28–
connected applications based on the phase shifting of the inverter output
34.
voltage with respect to the grid voltage. Int. J. Electr. Power Energy Syst. 57,
Yatimi, H., Aroudam, E., 2016. Assessment and control of a photovoltaic energy
250–260.
storage system based on the robust sliding mode MPPT controller. Sol. Energy
Hu, J., Shang, L., Zhu, Z.Q., 2011. Direct active and reactive power regulation of a
139, 557–568.
grid-connected DC/AC converters using sliding mode control approach. IEEE
Zainuri1, M.A.A.M., Radzil, M.A.M., Soh, A.C., Rahim, N.A., 2014. Development of
Trans. Power Electron. 26 (1), 210–222.
adaptive perturb and observe-fuzzy control maximum power point tracking for
Huld, T., Cebecauer, T., Šúri, M., Dunlop, E.D., 2010. Analysis of one-axis tracking
photovoltaic boost dcdc converter. IET Renew. Power Gener. 8 (2), 183–194.
strategies for PV systems in Europe. Prog. Photovolt.: Res. Appl. 18 (3), 183–194.
Zakzouk, N.E., Elsaharty, M.A., Abdelsalam, A.K., Helal, A.A., Williams, B.W., 2015.
Kim, S., 2006. Sliding mode controller for the single-phase grid-connected
Improved performance low-cost incremental conductance PV MPPT technique.
photovoltaic system. Appl. Energy 83, 1101–1115.
IET Renew. Power Gener., 1–14
Kim, I.S., 2007. Robust maximum power point tracking using sliding mode
Zue, A.O., Chandra, A., October 2009. State feedback linearization control of a grid
controller for the three-phase grid-connected photovoltaic system. Sol.
connected photovoltaic interface with MPPT. In: Proc. of IEEE EPEC.
Energy 81, 405–414.