Electric Power Components and Systems

ISSN: 1532-5008 (Print) 1532-5016 (Online) Journal homepage: http://www.tandfonline.com/loi/uemp20

Nonlinear Adaptive Backstepping Controller

Design for Three-Phase Grid-Connected Solar
Photovoltaic Systems

Tushar Kanti Roy, Md Apel Mahmud, Amanullah Maung Than Oo, Ramesh
Bansal & Md Enamul Haque

To cite this article: Tushar Kanti Roy, Md Apel Mahmud, Amanullah Maung Than Oo, Ramesh
Bansal & Md Enamul Haque (2018): Nonlinear Adaptive Backstepping Controller Design for Three-
Phase Grid-Connected Solar Photovoltaic Systems, Electric Power Components and Systems,
DOI: 10.1080/15325008.2018.1431334

To link to this article: https://doi.org/10.1080/15325008.2018.1431334

Published online: 01 Mar 2018.

Submit your article to this journal

Article views: 6

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=uemp20
Electric Power Components and Systems, 0(0):1–18, 2018
Copyright C Taylor & Francis Group, LLC
ISSN: 1532-5008 print / 1532-5016 online
DOI: 10.1080/15325008.2018.1431334

Nonlinear Adaptive Backstepping Controller

Design for Three-Phase Grid-Connected Solar
Photovoltaic Systems
Tushar Kanti Roy,1 Md Apel Mahmud,1 Amanullah Maung Than Oo,1
Ramesh Bansal,2 and Md Enamul Haque1
1
Electrical Power & Energy Systems Research Lab (EPESRL), School of Engineering, Deakin University, Geelong, VIC,
Australia
2
Department of Electrical, Electronic and Computer Engineering, University of Pretoria, Pretoria, South Africa

CONTENTS
Abstract—A cascaded control structure is proposed in this paper for
1. Introduction injecting active and reactive power in a three-phase grid-connected
2. Description and Modeling of Grid-Connected PV Systems solar photovoltaic (PV) system by considering external disturbances.
3. Proposed Controller Design In the proposed cascaded control structure, there are two control
loops—the outer loop voltage controller is used to ensure a contin-
4. Controller Performance Evaluation uous balance in power flow between the PV arrays and electrical
5. Conclusion power grid whereas the inner loop current controller controls the
References output current of the inverter. Moreover, the DC-DC boost converter
is controlled to achieve a constant voltage at the input of the inverter.
Appendix A: Formulas for the Calculation of DC-DC Boost In order to obtain the power balance and extract maximum power,
Converter’s Parameters an incremental conductance (IC) based maximum power point
Appendix B: Alignment Between the Notation of the Symbols in tracking (MPPT) method is used in this paper. The current controller
is designed using a nonlinear adaptive backstepping technique to
Equations (19)–(20) and the Definition of the Park regulate the active and reactive components of the grid current. The
Transformation regulation of these currents towards desired values which in turn con-
Appendix C: Control Lyapunov Function and Stability Analysis trol the active and reactive power delivered into the grid. The overall
stability analysis of the system is performed based on the formula-
tion of control Lyapunov functions (CLFs). Finally, the performance
of the designed controller is tested on three-phase grid-connected
PV systems with single as well as multiple PV units under different
environmental conditions and compared with an existing sliding
mode controller. Simulation results confirm the effectiveness of the
proposed adaptive backstepping control scheme and demonstrate the
superior performance over the sliding mode controller.

Keywords: adaptive backstepping controller, maximum power point
tracking, solar photovoltaic systems, external disturbances, active power,
reactive power
Received 25 October 2016; accepted 12 January 2018
Address correspondence to M. A. Mahmud, Electrical Power & Energy
Systems Research Lab (EPESRL), School of Engineering, Deakin
University, Geelong, VIC 3216, Australia. E-mail:
apel.mahmud@deakin.edu.au
Color versions of one or more of the figures in the article can be found online
at www.tandfonline.com/uemp.
1. INTRODUCTION
Solar energy is one of the most promising energy sources with
distinct advantages such as ease of installation, noise free,
safe operation, and environment friendly without any air pol-
lution [1]–[3]. Therefore, the solar photovoltaic (PV) systems
have been widely installed as grid-connected or stand-alone
power systems. The majority of PV systems (approx. 98%)
are installed as grid-connected systems due to the potential
economic benefits [4].
In most cases, the grid-connected solar PV systems operate
in a manner so that the maximum extracted power from the

1
2 Electric Power Components and Systems, Vol. 0 (2018), No. 0
PV arrays can be delivered into the grid under changing
atmospheric conditions. For this, the grid-connected PV
system has to perform two essential functions: (i) ensure
the operation of PV arrays at maximum power point (MPP)
through the maximum power point tracking (MPPT) tech-
nique and (ii) inject a harmonic free current into the grid by
controlling inverter switches [5].
Various MPPT techniques have been proposed in [6]–[14]
to ensure the operation of PV arrays at MPP under contin-
uously changing atmospheric conditions and most of these
approaches have accuracy problems and inability to work
under partial shading conditions. The perturb and observe
(P&O) is one of the most popular method due to its simple
control structure which is commonly used in solar PV sys-
tems to extract the maximum power where the operating point
of MPPT moves toward the maximum power point by peri-
odically increasing or decreasing the output voltage of PV
arrays [15]–[18]. However, the P&O method has some limi-
tations such as slower response times and oscillations in the
output power of PV arrays. To overcome these problems, an
incremental conductance (IC) method is used in [19], [20] for
its conciseness and higher tracking efficiency which is also
used in this paper to extract maximum power from PV arrays.
The stable operation of a grid-connected PV system is
affected by different disturbances within the system such as
changes in atmospheric conditions, different types of faults,
etc. The stability of grid-connected solar PV systems is main-
tained with the design of different types of controllers which
will regulate the switching signals of the inverter through a
pulse width modulation (PWM) scheme based on cascaded
control loops [21]. The second control loop, which is inner
(current) loop, is used to regulate the active and reactive cur-
rent components of the inverter output for injecting current
into the grid with a power factor close to unity for the delivery
of maximum power [5].
The characteristics of diode currents in a solar cell along
with the switching functions of converters and inverters are
nonlinear. Therefore, grid-connected solar PV systems exhibit
highly nonlinear characteristics and the stable operation of
three-phase grid-connected PV systems under changing atmo-
spheric conditions is a challenging task. Under these circum-
stances, linear controllers are unable to achieve the desired
performances over a wide variation in atmospheric condi-
tions [22]. Thus, the design and implementation of nonlin-
ear controllers are essential as these nonlinear techniques can
effectively tackle the variations of operating points.
There are several nonlinear control techniques such
as sliding mode controller (SMC) [23], model predic-
tive controller [24], and feedback linearizing controller
(FBLC) [25]–[27] which have been proposed to evaluate
the dynamic stability of grid-connected PV systems. Among
these nonlinear controllers, feedback linearization (FBL)
techniques have been widely used. In [25]–[27], a nonlinear
FBL technique is proposed to analyze the dynamic stability
of a three-phase grid-connected PV system. This method
provides a satisfactory performance over a wide range of
operating points. However, it requires exact parameters of
the system for successful implementation. Moreover, FBL
techniques often cancel some useful inherent nonlinearity
within the system during the transformation from nonlinear
system to a fully or partially linearized system.
The wasteful cancelation of nonlinear dynamics in FBL
method can be avoided using a nonlinear backstepping
controller as proposed in [28] which stabilizes the nonlin-
ear terms rather than complete eliminations. The nonlinear
sliding-mode current controller is designed based on the
time-varying sliding surface to ensure the stable operation of
a three-phase grid-connected PV system [29]. Though this
SMC is less sensitive to parameter variations and external
disturbances but the selection of the time-varying sliding
surface is very difficult. A model predictive controller is
proposed in [24] to control a single-phase grid-connected PV
system which has a constant switching frequency and requires
exact information of the system parameters for successful
implementation. It is evident from the existing literature that
most of the controllers are designed only for the operation
of the system at unity power factor, i.e., just for delivering
active power into the grid. Therefore, the reactive power
consumed by loads in distribution networks need to be sup-
plied through other external sources which is very expensive.
Therefore, a controller should be designed in such a way
that the PV system can inject both active and reactive power
into the grid. Though there are some existing literature for
delivering both active and reactive power into the grid from
solar PV systems, there are no clear indications about the
design process of the controllers and external disturbances are
neglected [30], [31].
The purpose of this paper is to design a nonlinear adaptive
backstepping controller to share both active and reactive
power into the grid under any atmospheric conditions. The
DC-DC boost converter in a grid-connected PV system is also
controlled using an adaptive backstepping approach to main-
tain the desired DC voltage at the input of the voltage source
inverter (VSI). External disturbances are incorporated within
the PV system model in order to design the controller. Control
Lyapunov functions (CLFs) are formulated at different stages
of the design process to evaluate the stability of the designed
controller. An existing IC method as discussed in [19], [20] is
used for MPPT. The performance of the designed controller
is evaluated on a single as well as multiple three-phase
grid-connected PV system under standard and changing
atmospheric conditions and compared with an existing SMC.
 Roy et al.: Nonlinear Adaptive Backstepping 3

FIGURE 1. Three-phase grid-connected PV system.

2. DESCRIPTION AND MODELING OF irradiation and ambient temperature. Moreover, it has a shunt
GRID-CONNECTED PV SYSTEMS resistance Rsh and a series resistance Rs . From this model, the
diode current ID can be written by Shockley equation as
The schematic diagram of a typical three-phase grid-
connected PV system is shown in Figure 1. The three-phase ID = Is eα(Vpv +Rs Ipv )−1 ,
 
(1)
grid-connected PV system consists of a PV array, a DC-DC
boost converter with an input capacitor Cpv , a DC-link capac- where all symbols have their usual meaning which can be seen
itor C, a three-phase inverter, and a filter inductor L which in [5]. By applying the Kirchhoff ’s current law (KCL), the
is connected to the grid via grid voltages ea , eb , and ec . The terminal current of the PV cell can be written as
filter inductor L is used to reduce the ripple components  Vpv + Rs Ipv
Ipv = Iph − Is eα(Vpv +Rs Ipv )−1 −

on the inverter output current due to PWM switching opera- , (2)
Rsh
tion. The DC-link capacitor is used to maintain the PV array
voltage to a certain level for the voltage source inverter (VSI). where Iph is the photo-electric current which depends on the
The three-phase inverter operates in a current controlled mode solar irradiation and temperature which can be shown through
to regulate the reference current for injecting current into the the following equation:
grid and to deliver both active and reactive power into the λ  
grid. The mathematical modeling of a PV array, DC-DC boost Iph = Isc + Ki (Tc − Tre f ) , (3)
1000
converter, and three-phase VSI along with the uncertainty
where Isc is the short-circuit current at the reference temper-
or disturbance modeling as well as the selection of control
ature Tre f , Ki is the cell’s short-circuit current coefficient, λ
objectives are briefly discussed in the following subsections.
is the solar irradiation, Tc is the cells working temperature in
Kelvin. Again, the cells saturated current Is is varied with tem-
2.1. PV Array Model
perature according to the following equation:
A PV cell converts solar energy to electricity by means of
Tc 3 qEAkg Tre1 f − T1c
  
photoelectric phenomenon which is found in certain types of
Is = Irs e , (4)
semiconductor materials such as silicon and selenium. The Tre f
equivalent circuit diagram of a PV cell is shown in Figure 2
where the symbols carry their standard meaning which can
from where it can be seen that it is a current source in par-
also be seen in [5].
allel with an ideal p-n diode that is dependent on the solar
As the output voltage and current of a single PV cell is
very low, a number of PV cells are connected in series and
parallel to obtain the desired voltage and current which form
a PV array. If a PV array contains Nser series and Npar parallel
PV cells, the output current of the PV array Ipv (without the
loss of generality) can be written as

Ipv = Npar Iph − Npar Is eα(Vpv +Rs Ipv )−1

 
 
Npar Vpv Rs Ipv
− + . (5)
FIGURE 2. Equivalent circuit diagram of a PV cell. Rsh Nser Npar
4 Electric Power Components and Systems, Vol. 0 (2018), No. 0

Thus, the power generated by the PV array is given by 2.3. Modeling of VSIs in Three-Phase Grid-Connected
PV Systems with External Disturbances
Ppv = Vpv Ipv . (6) In this subsection, the dynamical model of a VSI in a three-
phase grid-connected PV system as shown in Figure 1 is first
With this PV array model, the dynamical model of a three-
developed in the original three-phase abc-frame. The dynami-
phase grid-connected PV system is developed in the following
cal model of a VSI in the abc-frame can be represented by the
subsection.
following differential equations [5]:
dia Vdc
2.2. Modeling of DC-DC Boost Converter in L = −Ria − ea + (2Sa − Sb − Sc ) (13)
dt 3
Three-Phase Grid-Connected PV Systems with dib Vdc
External Disturbances L = −Rib − eb + (−Sa + 2Sb − Sc ) (14)
dt 3
This subsection is aimed to develop the dynamical model of a dic Vdc
L = −Ric − ec + (−Sa − Sb + 2Sc ) , (15)
DC-DC boost converter which is used for tracking maximum dt 3
power in a three-phase grid-connected PV system. The main where Sa , Sb , and Sc are the input switching signals which
objective of this converter is to achieve a desired voltage at the are the original control inputs of the system. The switches
input side of the VSI. The dynamical model of a DC-DC boost (Sa+ & Sa− ) in Figure 1 correspond to Sa . Similarly, Sb+ &
converter as shown in Figure 1 can be written as follows [32]: Sb− correspond to Sb while Sc+ & Sc− correspond to Sc . These
1 switching signals can be written as follows:
V̇pv = (Ipv − IL ) (7)
Cpv Sa = Sm sin(ωt )
1 Sb = Sm sin(ωt − 120◦ )
I˙L = [−R1 IL + Vpv − (1 − µ)Vdc ] (8)
L1
Sc = Sm sin(ωt + 120◦ ), (16)
1
V̇dc = [(1 − µ)IL − Idc ], (9)
C where Sm is the magnitude of the switching signal which is
basically the modulation index.
where the symbols have their usual meanings and the detailed The dynamical model as represented by Eqs. (13)–(15) is
definitions of these symbols can be found in [32]. The gener- nonlinear due to the switching terms and output current of
alized approach is used to calculate the parametric values of the PV array. The phase voltages at the grid connection point
this DC-DC converter by considering the ripple as 5% with a in Eqs. (13)–(15) for a balanced three-phase system can be
switching frequency of 10 kHz and the detailed formulas are expressed as follows:
provided in Appendix A.
In terms of external disturbances, the dynamical model of ea = Em sin(ωt )
the DC-DC boost converter as represented by Eqs. (7)–(9) can eb = Em sin(ωt − 120◦ )
be rewritten as follows: ec = Em sin(ωt + 120◦ ), (17)
1 where Em is the magnitude of the voltage and ω = 2π f is the
V̇pv = (Ipv − IL ) + d1 (10)
Cpv angular frequency with f as the electrical frequency. Simi-
1 larly, the expressions for the phase currents in Eqs. (13)–(15)
I˙L = [−R1 IL + Vpv − (1 − µ)Vdc ] + d2 (11)
L1 can be expressed as follows:
1
V̇dc = [(1 − µ)IL − Idc ] + d3 , (12) ia = Im sin(ωt )
C
ib = Im sin(ωt − 120◦ )
where d1 , d2 , and d3 are the external disturbances. These
ic = Im sin(ωt + 120◦ ), (18)
external disturbances are modeled by considering the changes
in atmospheric conditions, e.g., changes in solar irradiation where Im is the magnitude of the current.
and environmental temperature. Therefore, the nature of these In practice, it is quite complex to design controller based
disturbances are exogenous as these originate within the sys- on the abc-frame due to increases in the complexities. On the
tem due to the external effects. The output of this DC-DC con- other hand, the currents and voltages in the abc-frame can
verter is fed to the VSI and the dynamical modeling of VSIs easily be transformed to their equivalent DC values in dq-
in a three-phase grid-connected PV system is discussed in the frame which are usually constants and easy to track. The Park
following subsection. transformation is used to transform the three-phase voltages

and currents from the abc-frame to their equivalent dq com-
ponents. Therefore, the dynamical model needs to be trans-
formed into a synchronous orthogonal frame which is rotating
at the angular frequency (ω = 2π f ) of the power grid [26].
With this dq transformation, Eqs. (13)–(15) can be written as
follows:
R 1 Vdc
I˙d = − Id + ωIq − Ed + Sd
L L L Since the main control objectives for the VSI are to con-
R 1 Vdc
I˙q = − Iq − ωId − Eq + Sq ,
L L L vant equations which also provide some useful insights on the
dq dq dq
where Idq = Tabc iabc , Edq = Tabc eabc , Sdq = Tabc Sabc ; Id and Iq
are d- and q-axes grid currents, respectively; Ed and Eq are
d- and q-axes grid voltages, respectively; Sd and Sq are d-
and q-axes switching control inputs, respectively; and T is the
Park transformation matrix. The alignment between the nota-
tion of the symbols in Eqs. (19)–(20) and the definition of the
Park transformation is shown in Appendix B. Equations (19)–
(20) represent the dynamical model of a three-phase grid-
connected PV system.
Since the output power of the PV system depends on
the solar irradiation and ambient temperature which have
intermittent characteristics and in turn cause uncertainties
within the dynamical model of PV systems. Moreover, the
parameters of a PV system are not exactly known and there
are parametric uncertainties as well. These uncertainties
must be considered during the controller design process to
ensure reliable and robust operation. When these external
uncertainties are added into the dynamic model as described
by Eqs. (19)–(20), the dynamics model will be as follows:
R 1 Vdc
I˙d = − Id + ωIq − Ed + Sd + d4
L L L
R 1 Vdc
I˙q = − Iq − ωId − Eq + Sq + d5 ,
L L L
where d4 and d5 are the external disturbances. These distur-
bances are mainly due to the noises in the measurements of
the grid current and voltages while the natures are exogenous.
Equations (21)–(22) represent the complete dynamical model
of a VSI in a three-phase grid-connected PV system with
external disturbances in the dq frame.
The controller for the DC-DC converter can be designed
based on the dynamical model as represented by Eqs. (10)–
(12) while for the VSI can be designed based on the model
as represented by Eqs. (21)–(22). However, the control objec-
tives need to be defined for these models before designing the
controllers for DC-DC converter and VSI which are discussed
in the following subsection.
Roy et al.: Nonlinear Adaptive Backstepping 5
2.4. Selection of Control Variables
The VSI in a grid-connected PV system usually requires a
constant DC voltage at the input. However, the output voltage
of the PV unit fluctuates due to changes in atmospheric condi-
tions. The DC-DC converter maintains a constant DC voltage
at its output side which is actually the input of the VSI. Thus,
the control variable for the DC-DC converter is the DC-link
voltage (Vdc ).
(19)
(20) trol active and reactive power, it is essential to derive the rele-
variables which need to be controlled. The apparent power S
which is delivered into the grid can be written as
S = P + jQ, (23)
where P is the active power and Q is the reactive power deliv-
ered from the inverter into the grid. For a three-phase system,
the active and reactive power can be written as
3 
P= Ed Id + Eq Iq (24)
2
3 
Q= Ed Iq − Eq Id . (25)
2
In a synchronously rotating dq reference frame, the average
value of Eq can be assumed to be zero at steady-state condi-
tion. Therefore, Eqs. (24)–(25) can be simplified as
2P
Id =
3Ed
2Q
Iq = . (26)
3Ed
In Eq. (26), Ed is the grid voltage which is maintained and
controlled by the network. Therefore, the active and reactive
(21) power delivered into the grid depends on the d- and q-axes
currents where the only way of controlling active and reactive
(22) power is to control the currents Id and Iq injected into the grid.
Based on the mathematical models of the DC-DC converter
in Eqs. (11)–(13) and VSI as described by Eqs. (21)–(22), the
nonlinear adaptive backstepping controller design procedure
is elaborately discussed in the following section.
3. PROPOSED CONTROLLER DESIGN
This section comprises three parts: (i) the design of an
adaptive backstepping controller for the DC-DC boost con-
verter, (ii) an adaptive backstepping controller design for the
VSI, and (iii) the design of an MPPT system based on an
existing IC method as used in [19]. The first two parts are
the main contribution of this paper. The detailed controller
design procedures for maintaining the desired output voltage
6 Electric Power Components and Systems, Vol. 0 (2018), No. 0

through DC-DC converter, controlling the active and reactive where k1 is a positive constant which is used to tune the output
power injection into the grid, and extracting maximum power response. Using Eq. (32) into Eq. (31), it can be written as
from the PV unit using an MPPT technique are elaborately follows:
discussed in the following subsections.
Ẇ = −k e2 − 1 d˜ (d˙ˆ − τ ),
1 1 1 γ1 1 1 1 (33)
3.1. Adaptive Controller Design for DC-DC Boost where τ1 = γ1 e1 is a tuning function. Since IL is a stabilizing
Converters in Grid-Connected PV applications function, the synthetic value (α) of IL can be obtained from
In this subsection, the main control objective is to achieve a Eq. (32) and written as follows:
desired DC voltage at the output of the DC-DC converter. The α = Ipv − Cpv (V̇pv (re f ) − dˆ1 − k1 e1 ). (34)
following steps are used to elaborately discuss the design pro-
cedure of an adaptive backstepping controller for a DC-DC The stability of IL will be analyzed in the following step for
boost converter in solar PV applications: which it is essential to obtain the derivative of α as follows:
Step 1: The tracking error corresponding to the output α̇ = A + Cpv k1 d1 , (35)
voltage of the solar PV unit can be written as follows:
where
e1 = Vpv − Vpv (re f ), (27)
A = I˙pv − Cpv (V̈pv (re f ) − d˙ˆ1 ) + k1 (Ipv − IL ) − Cpv k1V̇pv (re f ).
where Vpv (re f ) is the reference voltage of the PV system which
Step 2: The zero tracking error will be achieved if IL = α.
will be obtained from the MPPT system. The dynamic of e1
Thus, the error variable between IL and α can be written as
can be obtained by taking its derivative as follows:
follows:
1
ė1 = (Ipv − IL ) + d1 − V̇pv (re f ) . (28) e2 = IL − α (36)
Cpv
The estimation error for the unknown disturbance d1 can be whose dynamics can be written as follows:
incorporated into Eq. (28) and written as 1
ė2 = [−R1 IL + Vpv − (1 − µ)Vdc ] + d2 − A − Cpv k1 d1 .
1 L1
ė1 = (Ipv − IL ) + d˜1 + dˆ1 − V̇pv (re f ), (29) (37)
Cpv
In terms of estimation errors of external disturbances d1 and
where d˜1 = d1 − dˆ1 is the estimation error with dˆ1 as its d2 , Eq. (37) can be written as
estimated value. In this step, IL is the stabilizing function or
1
virtual control variable to stabilize ė1 . For this purpose, the ė2 = [−R1 IL + Vpv − (1 − µ)Vdc ] + d˜2
control Lyapunov function (CLF) is selected as follows: L1
+ dˆ2 − A − Cpv k1 (d˜1 + dˆ1 ), (38)
1 1 ˜2
W1 = e21 + d , (30)
2 2γ1 1 where d˜2 = d2 − dˆ2 is the estimation error with dˆ2 as its esti-
mated value. In this case, the CLF can be chosen as follows:
where γ1 is an adaptation gain parameter. The detailed expla-
nation about the CLF for the stability analysis is provided in 1 1 ˜2
W2 = W1 + e22 + d (39)
Appendix C. Now, the derivative of W1 can be written in the 2 2γ2 2
following simplified form: whose derivative can be written in the following final form

1

1 after using the values of Ẇ1 and ė2 :
Ẇ1 = e1 (Ipv − IL ) + dˆ1 − V̇pv (re f ) − d˜1 (d˙ˆ1 − γ1 e1 ). 
Cpv γ1 2 1
Ẇ2 = −k1 e1 + e2 [−R1 IL + Vpv − (1 − µ)Vdc ]
(31) L1

The stability of ė1 will be ensured if Ẇ1 in Eq. (31) is negative 1
+ dˆ2 − A − Cpv k3 dˆ1 − d˜1 (d˙ˆ1 − τ1 + γ1 e2Cpv k3 )
definite (Ẇ1 < 0) or semi-definite (Ẇ1 ≤ 0) for which the γ1
first term in the right side of Eq. (31)) can be written as 1 ˜ ˙ˆ
follows: − d2 (d2 − γ2 e2 ). (40)
γ2
1
(Ipv − IL ) + dˆ1 − V̇pv (re f ) = −k1 e1 , (32) The stability of the error dynamics (ė2 ) can only be ensured if
Cpv
Ẇ2 < 0 or Ẇ2 ≤ 0. For this purpose, the synthetic value (α1 )
 Roy et al.: Nonlinear Adaptive Backstepping 7

of the stabilizing function (Vdc ) can be written as follows: whose dynamics can be written as follows:

L1

Vpv R1
 1
α1 = − IL + dˆ2 − A − Cpv k1 dˆ1 + k2 e2 . ė3 = [(1 − µ)IL − Idc ] + d3 − M
1 − µ L1 L1 C
−N − M1 d1 − N1 d2 − B1 µ̇. (45)
(41)
In terms of estimation errors, Eq. (45) can be rewritten as
Using Eq. (41), the derivative of the CLF in Eq. (40) can be follows:
simplified as follows: 1
ė3 = [(1 − µ)IL − Idc ] + d˜3 + dˆ3 − M − N
1 ˜  ˙ˆ 1  C
d1 d1 − τ11 − d˜2 d˙ˆ2 − τ2 ,

Ẇ2 = −k1 e21 − k2 e22 − −M1 (d˜1 + dˆ1 ) − N1 (d˜2 + dˆ2 ) − B1 µ̇, (46)
γ1 γ2
(42) where d˜3 = d3 − dˆ3 is the estimation error with dˆ3 as its esti-
mated value. At this point, the final CLF can be selected as
where τ11 = τ1 + γ1 e2Cpv k3 and τ2 = γ2 e2 are another two
tuning functions. At this step, the control input (µ) for the DC- 1 1 ˜2
W3 = W2 + e23 + d . (47)
DC converter appears and it is not determined as the external 2 2γ3 3
disturbance (d3 ) is still not appeared during the design pro- Substituting the values of Ẇ2 from Eq. (42) and ė3 from
cess as well as the stability of V̇dc is not analyzed yet. There- Eq. (46), the time derivative of W3 can be written as:
fore, it is essential to proceed further with the design process.

1
Usually, the design process for an adaptive backstepping con- 2 2
Ẇ3 = −k1 e1 − k2 e2 + e3 [(1 − µ)IL − Idc ] + dˆ3 − M − N
C
troller continues until the stability of all states are ensured.
1
− M1 dˆ1 − N1 dˆ2 − B1 µ̇ − d˜1 (d˙ˆ1 − τ11 + γ1 e3 M1 )

Therefore, it is essential to obtain the derivative of α1 in order
to complete the final step which can be written as follows: γ1
1 ˜ ˙ˆ 1
− d2 (d2 − τ2 + γ2 e3 N1 ) − d˜3 (d˙ˆ3 − γ3 e3 ). (48)
α̇1 = M + N + M1 d1 + N1 d2 + B1 µ̇, (43) γ2 γ3
The adaptive control law for the DC-DC converter can be
where
chosen as follows (based on Eq. (48)) in order to make the



R1 Vpv (1 − µ)Vdc R1 third term negative definite or semi-definite on the right side
M = − IL + − − + L1 k2
L1 L1 L1 1−µ of Eq. (48):
1 
N= (Ipv − IL ) + B2 − L1 k2 A 1 1
Cpv (1 − µ) µ̇ = [(1 − µ)IL − Idc ] + dˆ3 − M − N
B1 C
1
B2 = (L1 d˙ˆ2 − L1 Ȧ + L1 k1 d˙ˆ1 )
1−µ
− M1 dˆ1 − N1 dˆ2 + k3 e3 . (49)
1
M1 = + L1 k1 k2
1−µ
Similarly, the adaptation laws can be chosen as follows in
R1
N1 = L1 k2 − order to eliminate the influence of d˜1 , d˜2 , and d˜3 in Ẇ3 :
1−µ
1 d˙ˆ = τ − γ e M
1 11 1 3 1
B1 = (R1 IL + Vpv + L1 d˙ˆ2 − L1 Ȧ
(1 − µ)2 d˙ˆ2 = τ2 − γ2 e3 N1
+ L k d˙ˆ + L k e ).
1 1 1 1 2 2 d˙ˆ3 = γ3 e3 . (50)

The derivation of the proposed control law for the DC-DC
boost converter along with the overall stability analysis and
derivation of the adaptation laws to estimate all unknown dis-
turbances are discussed in the following step.
Step 3: Since α1 is a virtual control variable to stabilize
the dynamic of the error variable corresponding to the DC-
link voltage, the third error variable to represent the tracking
of the DC-link voltage can be defined as follows:
e3 = Vdc − α1 (44)
These adaptation laws are used to estimate external distur-
bances without imposing any bound while guaranteeing the
convergence of all physical properties of the system, e.g., a
constant DC-link voltage.
Using Eqs. (61) and (50), Eq. (48) can be simplified as
follows:
Ẇ3 = −k1 e21 − k2 e22 − k3 e23 ≤ 0, (51)
which will be negative definite or semi-definite for any value
of the error, i.e., the proposed adaptive control scheme can
8 Electric Power Components and Systems, Vol. 0 (2018), No. 0

be used to stabilize the dynamics relevant to the DC-DC con- obtain the control laws Sd and Sq :
verter. As mentioned earlier, the VSI is used to control the 1 2 1 2 1 ˜2 1 ˜2
active and reactive power injection into the grid while consid- W4 = e + e + d ++ d (59)
2 4 2 5 2γ4 4 2γ5 5
ering the effect of external disturbances and at the same time
stabilizing the grid-connected PV system. The adaptive con- whose derivative can be written as follows:
troller design for a VSI in a three-phase grid-connected PV Ẇ = e ė + e ė − 1 d˜ d˙ˆ − 1 d˜ d˙ˆ .
4 4 4 5 5 γ4 4 4 γ5 5 5
(60)
system is discussed in the following subsection.
By inserting the values of ė4 from Eq. (57) and ė5 from
3.2. Adaptive Controller Design for VSIs in Eq. (58) into Eq. (60), it can be written as
Grid-Connected PV applications
 
R Ed Vdc
Ẇ4 = e4 − Id + ωIq − + Sd + dˆ4 − I˙dre f + k4 e4
The objective of this subsection is to derive the switching con- L L L
 
trol inputs Sd and Sq in such a way that Id and Iq can track their R 1 Vdc
+ e5 − Iq − ωId − Eq + Sq + dˆ5 − I˙qre f + k5 e5
predefined reference values, i.e., their tracking errors converge L L L
˙
dˆ − γ e − d˜ γ −1 d˙ˆ − γ e ,
 
to zero and the inverter can inject a desired current into the − k e − k e − d˜ γ
2
4 4
2
5 5
−1
4 4 4 4 4 5 5 5 5 5
grid as this ensures the delivery of both desired active and (61)
reactive power into the grid.
First, let us define the d-axis current tracking error as where k4 and k5 are positive constants which are used to tune
the output responses of the system. In order to stabilize the
e4 = Id − Idre f . (52) error dynamics in Eqs. (55) and (58), the switching control
laws Sd and Sq can be designed as follows:
The time derivative of e4 can be written as
R ωL Ed L L ˆ L
ė4 = I˙d − I˙dre f . (53) Sd = Id − Iq + + I˙dre f − d4 − k4 e4
Vdc Vdc Vdc Vdc Vdc Vdc
(62)
The substitution of the value of I˙d from Eq. (21) into
R ωL Eq L L ˆ L
Eq. (53) yields Sq = Iq + Id + + I˙are f − d5 − k5 e5
Vdc Vdc Vdc Vdc Vdc Vdc
Vdc
ė4 = − RL Id + ωIq − L1 Ed + L d
S + d4 − I˙dre f . (54) (63)

In terms of the estimation error, Eq. (54) can be rewritten as and the adaptation laws to estimate external disturbances d4
and d5 can be chosen as follows:
ė4 = − RL Id + ωIq − L1 Ed + Vdc
S + dˆ4 + d˜4 − I˙dre f (55)
L d
dˆ˙4 = γ4 e4 (64)
with d˜4 = d4 − dˆ4 where dˆ4 is an estimate of the unknown ˙
dˆ = γ e . (65)
5 5 5
external disturbance.
Now define the q-axis current tracking error as These adaptation laws are used to estimate external distur-
bances without imposing any bound while guaranteeing deliv-
e5 = Iq − Iqre f . (56) ery of desired active and reactive power into the grid.
With these adaptation laws and switching control laws,
By inserting the value of I˙q from Eq. (22), the dynamics of e2
Eq. (61) can be simplified as
can be written as
Vdc
Ẇ4 = −k4 e24 − k5 e25 ≤ 0, (66)
ė5 = − RL Iq − ωId − L1 Eq + L q
S + d5 − I˙qre f . (57)
which will be negative definite or semi-definite for any error
Similarly in terms of the estimation error, Eq. (57) can be in Eq. (66). Thus, the switching control laws will stabilize
written as the dynamics related to the VSI while injecting the desired
amount of active and reactive power into the grid. An exist-
ė5 = − RL Iq − ωId − L1 Eq + Vdc
L q
S + dˆ5 + d˜5 − I˙qre f (58)
ing MPPT as discussed in the following subsection is used to
with d˜5 = d5 − dˆ5 where dˆ5 is another estimate of the extract the maximum power from the PV unit.
unknown external disturbance. In backstepping control
scheme, a Lyapunov function usually used for driving the con- 3.3. MPPT Controller Design Algorithm
ditions on the control law that will drive the state trajectory The MPPT using an IC method as discussed in [19] is based
to the equilibrium point. Therefore, the CLF is considered to on the derivative of the output power of PV arrays with respect
 Roy et al.: Nonlinear Adaptive Backstepping 9

FIGURE 3. Implementation block diagram of an adaptive backstepping controller in a three-phase grid-connected PV system.

to voltage. According to the output power equation of PV controller which uses the estimated values of external dis-
dP
arrays as represented by Eq. (6), the power slope dVpvpv can be turbances along with other physical properties (e.g., voltages
expressed as and currents) on the DC-side. The disturbances are estimated
in such a way that the controller for the DC-DC converter
dPpv dI achieves the desired output DC voltage which can also be seen
= Ipv + Vpv dVpvpv . (67)
dVpv from Figure 3. Finally, the output of this converter controller
dPpv
is converted into switching pulses using the pulse width mod-
At MPP, dVpv
= 0 for which Eq. (67) yields ulation (PWM) technique.
From Figure 3, it can be seen that the instantaneous phase
Ipv Ipv voltages at the grid connection point and currents are trans-
=− , (68)
Vpv Vpv formed into dq-frame through abc-dq transformation. Exter-
I I
nal disturbances are then incorporated and estimated using an
where Vpvpv is the IC and Vpvpv is the instantaneous conductance. online estimator. The estimated values of these disturbances
The IC method is designed based on the slope of the power along with other physical properties are used to deliver the
curve of PV arrays. The slope of the PV array power curve is desired amount of active and reactive power into the grid.
zero at MPP whatever the solar irradiation and temperature on The switching control signals are obtained using the estimated
the PV cell, negative at the right of the MPP, and positive at external disturbances, grid voltage, grid current, and DC-link
the left of the MPP which can be obtained by considering the voltage. Finally, the switching signals of the nonlinear adap-
following conditions: tive controller are transformed into instantaneous three-phase
r At MMP, Ipv = − Ipv components using the dq-abc transformation as the original
Vpv Vpv
r At the left of MMP, Ipv > − Ipv system is in the instantaneous form. These instantaneous sig-
Vpv Vpv
nals are feed into a pulse width modulator which uses the
r At the right of MMP, Ipv < − Ipv .
Vpv Vpv PWM technique to generate the switching pulses. The perfor-
The IC method tracks the MPP of the PV array by com- mance of the designed controller is evaluated in the following
paring incremental conductance with instantaneous conduc- section.
tance and the MPP tracker regulates the PWM control switch-
ing signal of the DC-DC converter until the condition Ipv +
dI 4. CONTROLLER PERFORMANCE EVALUATION
Vpv dVpvpv = 0 is satisfied.
The implementation block diagram of the designed con- The performance of the designed controller is evaluated on a
troller is shown in Figure 3. From Figure 3, it can be seen three-phase grid connected system with single PV unit as well
that there are two controllers: one for the DC-DC boost con- as multiple PV units by considering different operating sce-
verter and another for the VSI. The control signal for the narios. For single and multiple PV applications, the control
DC-DC converter is generated through the designed adaptive parameters are kept same. The parameters for the NABC and
10 Electric Power Components and Systems, Vol. 0 (2018), No. 0
PI NABC ESMC
Kp = 2, Ki = 120 γ1 = 8, γ2 = 10, k1dc = 8, k2dc = 10,
γ3 = 6, k1 = 12, ρdc = 0.05
k2 = 9, k3 = 7
TABLE 1. Control parameters for the DC-DC boost converter.
ESMC are calculated based on the satisfaction of different the-
oretical conditions as discussed during the controller design
process. At the same time, the parameters for the PI controller
are selected based on the Ziegler–Nichols method. The control
parameters for the DC-DC boost converter and VSI are pro-
vided in Tables 1 and 2, respectively. The performance of the
designed nonlinear adaptive backstepping controller (NABC)
is compared with an existing sliding mode controller (ESMC).
In this section, the following criteria are used to evaluate the
performance of the designed controllers while comparing with
an ESMC:
r Tracking error for the injection of desired amount of cur-
rent as well as power (both active and reactive power),
r Fluctuations in the power responses, and
r Power quality in terms of THD in the current injected
into the grid.
, and the inductance is 25 mH. The following
Based on these evaluation criteria, the performance of
the designed controller is evaluated in the following sub-
sections where all simulations are carried out using MAT-
LAB/SIMULINK simulation package under different operat-
ing conditions. Moreover, all results are presented to analyze
the stability of the grid-connected PV system as the controls
laws (for both DC-DC converter and VSI) derived based on
the satisfaction of the CLF.
4.1. Controller Performance Evaluation on a
Three-Phase Grid-Connected PV System with
Single PV unit
In this subsection, the performance of the designed con-
troller is evaluated on a simple three-phase grid-connected
PV system which is shown in Figure 1. The three-phase grid-
connected PV system includes an SPR 305-WHT PV module
along with the DC-DC converter and VSI. The correspond-
ing electrical characteristics of this solar model are given in
Table 3. The reference DC-link capacitor voltage is considered
as 500 V. The PV array is connected to a DC-DC converter to
PI NABC ESMC
Kp = 1.6, Ki = 36 γ4 = 5, γ5 = 10, kd = 5, kq = 8,
k4 = 50, k5 = 45.7 ρd = 0.06, ρq = 0.05
TABLE 2. Control parameters for the VSI.
Parameter Symbol Value
Maximum power Pmax 100 kW
Open circuit voltage Voc 64.2 V
Short circuit current Isc 5.96 A
Maximum power voltage Vmax 54.7 V
Maximum power current Imax 5.58 A
Number of series connected module per string Nser 5
Number of parallel strings Npar 66
TABLE 3. Electrical specification of SPR 305-WHT PV array [33].
maintain a desired DC-link voltage of 500 V while the MPP
tracker uses an IC technique to track the maximum power at
the terminal of the PV array. The output of the DC-DC con-
verter is fed to a DC-link capacitor where a constant voltage
500 V is always maintained in order to balance the power flow
within the system. The three-phase inverter converts that 500
V DC voltage to 260 V AC with a leading/lagging power fac-
tor depending on the types of loads. A 100 kVA 260 V/20
kV three-phase coupling transformer is used to connect the
inverter to the utility grid. The utility grid consists of a 20
kV distribution feeders and 120 kV equivalent transmission
system. The value of DC-link capacitor is 6 µF, the line resis-
tance is 0.19
three cases are considered to evaluate the performance of the
designed controller as compared to that of an existing SMC as
proposed in [29].
r Case 1: Controller performance under standard
atmospheric conditions
In this case study, the simulation is carried out for the
standard atmospheric conditions where the atmospheric con-
ditions (solar irradiation and temperature of the environment)
are 1 kW/m2 and 25o C, respectively. The main control objec-
tive is to inject maximum power (100 kW) into the grid at a
power factor of 0.85. The corresponding grid voltage and cur-
rent responses are shown in Figure 4(a).
From Figure 4(a), it can be seen the output current of the
inverter with the designed controller has negligible harmonic
content (black line). However, the inverter output current with
an SMC is similar to a sine wave but it contains significant
harmonics (red line). Since the system is operating at a power
factor other than unity, the grid current would not be in phase
with grid voltage which can also be seen from Figure 4(a). In
this situation, both active and reactive power will be delivered
into the grid which is shown in Figures 4(b) and 4(c), respec-
tively. As the power factor is 0.85, the active power delivered
into the grid will be 85 kW and reactive will be 52.5 kVAr.
From Figures 4(b) and 4(c), it can be seen that the proposed
controller (black line) can control the power injection into the
grid more effectively than the SMC (red line). This effective
 Roy et al.: Nonlinear Adaptive Backstepping 11

FIGURE 4. System responses under the standard atmospheric

condition.
FIGURE 5. THD in grid current under standard atmospheric
condition.
operation is obvious from Figures 4(b) and 4(c) as the inverter
is delivering more stable output active (85 kW) and reactive
(52.5 kVAr) power into the grid with the proposed controller inverter as shown in Figure 7(a). From Figure 7(a), it can be
without any fluctuation whereas the SMC is not able to deliver seen that the grid current varies significantly in accordance
desired power to the grid and contains some low-order har- to the variation of solar irradiance. However, the designed
monic distortion due to the chattering phenomenon. Under controller (black line) can inject current into the grid with
this condition, the total harmonic distortion (THD) is shown lower harmonics than the SMC (red line). The corresponding
in Figure 5 from where it can be seen that the design controller active and reactive power sharing capabilities are shown in
has less harmonics as compared to the existing SMC. Figures 7(b) and 7(c), respectively. It can be also observed
r Case 2: Controller performance under changing that their amplitudes change whenever the solar radiation
atmospheric conditions is varied. However, active and reactive power responses are
more stable with the designed controller as these responses
Practically, the atmospheric conditions (solar irradiation quickly settle down to their steady-state values. The corre-
and temperature) are not fixed for which the output voltage, sponding THD is shown in Figure 8 from where it can be seen
current, and power of the PV array change significantly. In
this condition, the performance of the designed controller is
verified under changing solar irradiation. At t = 0.6 sec, the
solar irradiance step downs from 1000 W/m2 to 450 W/m2 ,
maintains this level for 0.125 sec, and again steps up from
450 W/m2 to 1000 W/m2 at t = 0.725 sec. The MPPT will
track the maximum power though there are variations in
atmospheric conditions which can be seen from Figure 6.
Therefore, the amount of power delivered into the grid will
be changed as the output current of the PV array will change FIGURE 6. Maximum power point tracking capability of the
which in turn causes the changes in the output current of the MPPT algorithm under changing atmospheric conditions.
12 Electric Power Components and Systems, Vol. 0 (2018), No. 0

FIGURE 9. System responses under changing atmospheric

condition and at unity power factor.

that the THD is slightly higher than the previous case but still
much better than the SMC.
r Case 3: Controller performance under changing
FIGURE 7. System responses under changing atmospheric
condition. atmospheric conditions at unity power factor
In this case, the simulation is carried out for a unity power
factor under changing atmospheric conditions. When the load
power factor is unity, no reactive power will be delivered into
the grid as the voltage and current will be in phase with each
other. The corresponding grid voltage and current as well as
power responses are shown in Figures 9(a) and 9(b), respec-
tively. From Figure 9(a), it is clear that although the SMC is
able to generate the desired current but still there is phase dif-
ference between the grid voltage and current (red line) as it
cannot respond to the changes in a quicker. On the other hand,
the voltage and current are exactly in phase for the designed
controller (black line). Moreover, from Figure 9(b), it can be
seen that the existing controller is not able to supply har-
monic free 100 kW power to the grid due to the chattering
phenomenon but it is almost 100 kW with the proposed con-
troller. The THD for this operating condition can be seen in
Figure 10.
The estimations of external disturbances for all cases are
shown in Figure 11 from where it can be seen that the designed
scheme accurately estimates the unknown disturbances.

4.2. Controller Performance Evaluation on a

FIGURE 8. THD in grid current under changing atmospheric
condition.
Three-Phase Grid-Connected PV System with
Multiple PV units
In this subsection, the performance of the designed controller
is evaluated by considering multiple PV units where three PV
 Roy et al.: Nonlinear Adaptive Backstepping 13

FIGURE 10. THD in grid current under changing atmo-

spheric condition and at unity power factor.

units are connected to the grid through DC-DC boost convert-

ers and VSIs as shown in Figure 12. From Figure 12, it can be
seen that the capacities of the first PV unit (PV1), second PV
unit (PV2), and third PV unit (PV3) are 100 kW, 50 kW, and
25 kW, respectively. When multiple PV units are connected
together to a grid connection point, overall power quality of
the system degrades. For example, there will low harmonics
when only a single PV unit is connected and the effects of har-
monics will be severe for multiple PV units due to the inter-
actions among these units.
At the beginning of this subsection, the interactions among
different PV units in a three-phase grid-connected system
are analyzed by considering simple but finely tuned propor-
tional integral (PI) controllers for both DC-DC converters
and VSIs. With these PI controllers, the interactions due to
the integration of multiple PV units in a grid-connected PV
system are shown in Figure 13 in terms of harmonics in the
overall current injected into the grid. From Figure 13(a), it
can be seen that the harmonics are less when a single PV unit
is used to supply power into the grid. However, the harmonics
are increasing with the integration of more PV units which
can easily be seen from Figures 13(b) and 13(c). These
interactions are shown when all PV units are operating under
standard atmospheric conditions. However, the situations are FIGURE 11. Estimations of external disturbances.
14 Electric Power Components and Systems, Vol. 0 (2018), No. 0

FIGURE 12. Schematic diagram of a grid-connected PV sys-

tem with multiple PV units.

even the worst in case of changing atmospheric condition. In

such conditions, the designed controllers are very useful to
achieve the desired control objectives with lower harmonics.
In this subsection, the designed controllers are used in a
decentralized way to evaluate the performance by considering
both standard and changing atmospheric conditions. At the
beginning, it is assumed that all PV units are running at stan-
dard atmospheric conditions till t = 1 sec and under changing
atmospheric condition from t = 1 sec to t = 1.15 sec due to FIGURE 14. Maximum power point tracking capability of the
the drops in solar irradiation from 1000 W/m2 to 600 W/m2 . MPPT algorithm in a grid-connected system with multiple PV
After t = 1.15 sec, all PV units resume their operations under units.

FIGURE 15. System responses for a grid-connected system

FIGURE 13. Effects of dynamic interactions among different with multiple PV units (under both standard and changing
PV units in a grid-connected system. atmospheric conditions).

FIGURE 16. THD in the grid current for a grid-connected
system with multiple PV units.
standard conditions. Therefore, the output powers of all PV
units will be changed at t = 1 sec continue till t = 1.15 sec.
The MPPT systems of all PV units efficiently track the maxi-
mum power which can be seen from Figure 14.
The DC-DC converters for all PV units are controlled to
achieve the DC voltages as 500 V at their output sides. At the
same time, the VSIs are controlled at a power factor of 0.85
to deliver 148.75 kW active power and 92.19 kVAr reactive
power to the grid as the total input power to the VSI is 175
kVA. The designed controller injects the current correspond-
ing to this power factor which can be seen from Figure 15(a).
From Figure 15(a), it can be seen that the designed NABC
comprises less harmonics as compared to the ESMC. The
designed NABC also delivers the desired power (both active
and reactive) into the grid which can be Figures 15(b) and
15(c) from where it can be observed there are less fluctua-
tions when the designed controller is used. The fluctuations
in power are clearly visible with the ESMC as it is extremely
difficult to select the sliding surface for a grid-connected PV
system with multiple PV units. The designed controller not
only supplies the desired amount of active and reactive power
into the grid but also improves the power quality which can
Roy et al.: Nonlinear Adaptive Backstepping 15
be clearly seen from Figure 16. From Figure 16, it can be
seen that the value of THD is 5.89% when the ESMC is used
though the maximum allowable THD (according to the IEEE
standard) in the grid current is 5%. However, the THD in the
grid current is 4.10% when the designed controller is used and
thus, it meets the IEEE standard. Finally, it can be said that the
designed controller is equally applicable for grid-connected
PV systems with multiple PV units.
From the above analyses, it is obvious that the designed
controller can enhance the dynamic stability of a three-phase
grid-connected PV system by sharing both active and reac-
tive power into the grid in a better way than the existing SMC
under various atmospheric and operating conditions.
5. CONCLUSION
A two loops cascaded control structure is used to control
the DC-link voltage and current injection into the grid. The
designed adaptive backstepping controller generates robust
switching signals to ensure the delivery of both active and
reactive power into the grid. The external disturbances are
considered as uncertainties and the stability of the whole grid-
connected PV system is also guaranteed through the formula-
tion of control Lyapunov functions and time-domain simula-
tions. From the simulation results, it is clear that the proposed
controller provides a very satisfactory performance in terms
of maintaining steady-state operation under various operating
conditions. The proposed controller is capable to work under
any changing atmospheric conditions by supplying both active
and reactive power into the grid. Future work will be devoted
on developing a robust adaptive backstepping controller that
can overcome the influence of parametric uncertainties includ-
ing external disturbances in a PV model for a large-scale oper-
ation and implementation on a laboratory-based system.
REFERENCES
[1] C. W. Hsu “Using a system dynamics model to assess the effect
of capital subsidies and feed-in tariffs on solar PV installation,”
Appl. Energy, vol. 100, pp. 205–217, December 2012.
[2] P. Vithayasrichareon, and I. F. MacGill “Valuing large-scale
solar photovoltaics in future electricity generation portfolios
and its implications for energy and climate policies,” IET
Renew. Power Gener., vol. 10, no. 1, pp. 79–87, January 2016.
[3] K. M. Tsang and W. L. Chan “Three-level grid-connected pho-
tovoltaic inverter with maximum power point tracking,” Energy
Conver. Manage., vol. 65, pp. 221–227, January 2013.
[4] A. B. R. Boue, R. G. Valverde, F. d. A. Ruz-Vila, and J.
M. Torrelo-Ponce “An integrative approach to the design
methodology for 3-phase power conditioners in photovoltaic
grid-connected systems,” Energy Conver. Manage., vol. 56,
pp. 80–95, April 2012.
16 Electric Power Components and Systems, Vol. 0 (2018), No. 0
[5] M. A. Mahmud, H. R. Pota, and M. J. Hossain “Dynamic sta-
bility of three-phase grid-connected photovoltaic system using
zero dynamic design approach,” IEEE J. Photovolt., vol. 2, no.
4, pp. 564–571, October 2012.
[6] M. A. A. M. Zainuri1, M. A. M. Radzil, A. C. Soh, and N.
A. Rahim “Development of adaptive perturb and observe-fuzzy
control maximum power point tracking for photovoltaic boost
dc-dc converter,” IET Renew. Power Genera., vol. 8, no. 2, pp.
183–194, March 2014.
[7] L. M. Elobaid, A. K. Abdelsalam, and E. E. Zakzouk “Artifi-
cial neural network-based photovoltaic maximum power point
tracking techniques: A survey,” IET Renew. Power Genera., vol.
9, no. 8, pp. 1043–1063, November 2015.
[8] T. L. Kottas, Y. S. Boutalis, and A. D. Karlis “New maximum
power point tracker for PV arrays using fuzzy controller in
close cooperation with fuzzy cognitive networks,” IEEE Trans.
Energy Conver., vol. 91, pp. 793–803, August 2006.
[9] T. Noguchi, S. Togashi, and R. Nokamota “Short-current
pulse-based maximum-power-point tracking method for mul-
tiple photovoltaic and converter module system,” IEEE Trans.
Energy Conver., vol. 49, no. 1, pp. 217–223, February 2002.
[10] U. Zimmermann and M. Edoff “A maximum power point
tracker for long-term logging of PV module performance,”
IEEE J. Photovolt., vol. 2, no. 1, pp. 47–55, January 2012.
[11] E. Koutroulis and F. Blaabjerg “A new technique for tracking
the global maximum power point of PV arrays operating under
partial-shading conditions,” IEEE J. Photovolt., vol. 2, no. 2,
pp. 184–190, April 2012.
[12] F. Bouchafaa, D. Beriber, and M. S. Boucherit “Modeling
and control of a gird connected PV generation system,” 18th
Mediterranean Conference on Control and Automation (MED),
pp. 315–320, Marrakech, Morocco, 23–25 June 2010.
[13] R. Faraji, A. Rouholamini, and H. R. Naji “FPGA-based real
time incremental conductance maximum power point tracking
controller for photovoltaic systems,” IET Renew. Power Gen-
era., vol. 7, no. 5, pp. 1294–1304, May 2014.
[14] P. C. Sekhar and S. Mishra “Takagi-Sugeno fuzzy-based incre-
mental conductance algorithm for maximum power point track-
ing of a photovoltaic generating system,” IET Renew. Power
Genera., vol. 8, no. 8, pp. 900–914, November 2014.
[15] K. Ishaque, Z. Salam, and G. Lauss “The performance of
perturb and observe and incremental conductance maximum
power point tracking method under dynamic weather condi-
tions,” Appl. Energy, vol. 119, pp. 228–236, April 2014.
[16] E. Bianconi, J. Calvente, R. Giral, E. Mamarelis, G. Petrone, C.
A. R. Paja, and M. Vitelli “Perturb and observe MPPT algo-
rithm with a current controller based on the sliding mode,”
Int. J. Elect. Power Energy Syst., vol. 44, pp. 346–356, January
2013.
[17] J. Ahmed and Z. Salman “An improved perturb and observe
(P&O) maximum power point tracking (MPPT) algorithm
for higher efficiency,” Appl. Energy, vol. 150, pp. 97–108,
July 2015.
[18] Y. Yang and F. P. Zhao “Adaptive perturb and observe MPPT
technique for grid-connected photovoltaic inverters,” Int. Conf.
on Power Electronic and Engineering Application, pp. 468–
473, Shenzhen, China, 24–25 December 2011.
[19] C. H. Lin, C. H. Huang, Y. C. Du, and J. L. Chen “Max-
imum photovoltaic power tracking for the PV array using
the fractional-order incremental conductance method,” Appl.
Energy, vol. 88, pp. 4840–4847, December 2011.
[20] B. Liu, S. Duan, F. Liu, and P. Xu “Analysis and improvement
of maximum power point tracking algorithm based on incre-
mental conductance method for photovoltaic array,” Proc. of
IEEE Int. Conf. on Power Electron. Drive Syst., pp. 637–641,
Bangkok, Thailand, 27–30 November 2007.
[21] F. Blaabjerg, R. Teodorescu, M. Liserre, and A. V. Timbus
“Overview of control and grid synchronization for distributed
power generation systems,” IEEE Trans. Indust. Electr., vol. 53,
no. 5, pp. 1398–1409, October 2006.
[22] J. Selvaraj and N. A. Rahim “Multilevel inverter for grid-
connected PV system employing digital pi controller,” IEEE
Trans. Indust. Electr., vol. 56, no. 1, pp. 149–158, January
2009.
[23] I. S. Kim “Sliding mode controller for the single-phase grid-
connected photovoltaic system,” Appl. Energy, vol. 83, pp.
1101–1115, October 2006.
[24] A. Kotsopoulos, J. L. Darte and M. A. M. Hendrix “Predic-
tive DC voltage control of single-phase PV inverters with small
dc link capacitance,” Proc. IEEE Int. Symp. Ind. Electron., pp.
793–797, Rio de Janeiro, Brazil, 9–11 June 2003.
[25] M. A. Mahmud, H. R. Pota, M. J. Hossain, and N. Roy
“Robust partial feedback linearization stabilization scheme for
three-phase grid-connected photovoltaic systems,” IEEE Trans.
Power Deliv., vol. 29, no. 3, pp. 1221–1230, June 2014.
[26] M. A. Mahmud, H. R. Pota, M. J. Hossain, and N. K. Roy
“Robust partial feedback linearization stabilization scheme
for three-phase grid-connected photovoltaic systems,” IEEE J.
Photovolt., vol. 4, no. 1, pp. 423–431, January 2014.
[27] D. Lalili, A. Mellit, N. Lourci, B. Medjahed, and E. M. Berk-
ouk “Input output feedback linearization control and variable
step size MPPT algorithm of a grid-connected photovoltaic
inverter,” Renew. Energy, vol. 36, no. 12, pp. 3282–3291,
December 2011.
[28] A. E. Fadili, F. Giri, and A. E. Magri “Reference voltage opti-
mizer for maximum power point tracking in triphase grid-
connected photovoltaic systems,” Int. J. Elect. Power Energy
Syst., vol. 60, pp. 293–301, September 2014.
[29] I. S. Kim “Robust maximum power point tracking using slid-
ing mode controller for the three-phase grid-connected pho-
tovoltaic system,” Sol. Energy, vol. 81, pp. 405–414, March
2007.
[30] F. L. Albuquerque, A. J. Moraes, G. C. Guimaraes, S. M. R.
Sanhueza, and A. R. Vaz “Optimization of a photovoltaic sys-
tem connected to electric power grid,” IEEE/PES Transmis-
sion and Distribution Conf. and Exposition, pp. 645–650, Sao
Paulo, Brazil, 8–11 November 2004.
[31] L. Hassaine, E. Olias, J. Quintero, and M. Haddadi “Digital
power factor control and reactive power regulation for grid-
connected photovoltaic inverter,” Renew. Energy, vol. 34, pp.
315–321, January 2009.
[32] T. Alnejaili, S. Drid, D. Mehdi, L. Chrifi-Alaoui, R. Belarbi,
and A. Hamdouni “Dynamic control and advanced load man-
agement of a stand-alone hybrid renewable power system for
remote housing,” Energy Conver. Manage., vol. 105, pp. 377–
392, November 2015.
[33] B. Ehlal and G. Ahmed “Maximum power point tracker
of photovoltaic generation system based on INC algorithm
 Roy et al.: Nonlinear Adaptive Backstepping 17

control with integral regulator,” Proc. International Conference for these voltages can be expressed by the following equa-
on Electrical Engineering, Batna, Algeria, 16–17 November tion:
2014.
cos ωt − 2π cos ωt + 2π


    
[34] M. A. Mahmud, M. J. Hossain, H. R. Pota, and M. S. Ali “Gen- Ed 2 cos ωt 3 3
=
3 − sin ωt − sin ωt − 2π − sin ωt + 2π
 
eralized Lyapunov function for stability analysis of intercon- Eq 3 3
nected power systems,” AUPEC 2011, pp. 1–6, Brisbane, Aus- ⎡ ⎤
ea
tralia, 25–28 September 2011.
× eb ⎦ .
⎣
ec
The instantaneous values of all currents (ia , ib , and ic ) are
APPENDIX A: FORMULAS FOR THE known. The relationship between the dq-frame and abc-frame
CALCULATION OF DC-DC BOOST CONVERTER’S for these currents can be expressed by the following equation:
PARAMETERS
cos ωt − 2π cos ωt + 2π


    
Id 2 cos ωt 3 3
If the switching input for the DC-DC boost converter is µ =
3 − sin ωt − sin ωt − 2π − sin ωt + 2π
 
Iq 3 3
and RL represents the load resistor, the minimum value of ⎡ ⎤
the inductor, for a switching frequency fs , can be written as ia
follows: × ⎣ ib ⎦ .
ic
µ (1 − µ)2 RL
Lmin = .
2 fs
APPENDIX C: CONTROL LYAPUNOV FUNCTION
If Vin is the input voltage to the DC-DC boost converter and AND STABILITY ANALYSIS
Vo is the output, the relationship between the input and output The Lyapunov function is used to analyze the stability of
voltages can be written as follows: a dynamical system. Lyapunov stability theory includes two
Vin methods: (i) Lyapunov’s first method and (ii) Lyapunov’s
Vo = . direct method. The first method is used to analyze the stability
1−µ
of linear dynamical systems while the direct method for non-
If V o
represents the ripple factor, the value of the capacitor linear dynamical systems. The control Lyapunov functions are
Vo
can be written as follows: usually Lyapunov functions with control inputs to analyze the
stability of the system by applying the control inputs.
D 1 If the states of the system is represented by x, a positive
C= × Vo .
RL fs V definite function W (x) can be said to be a Lyapunov function
o

when its derivative is negative definite or semi-definite, i.e.,

W (x) < 0 or W (x) ≤ 0 [34]. The Lyapunov or control Lya-
APPENDIX B: ALIGNMENT BETWEEN THE punov functions are also unknown as the energy function as
NOTATION OF THE SYMBOLS IN the energy of a stable system is always positive definite.
EQUATIONS (19)–(20) AND THE DEFINITION OF From the definition of Lyapunov functions or energy func-
THE PARK TRANSFORMATION tions, it is obvious that the derivative of W (x) needs to be
The instantaneous values of all switching signals (Sa , Sb , and negative definite or semi-definite which means that the con-
Sc ) are known. The relationship between the dq-frame and trol action will reduce the energy of each state. If the control
abc-frame for these switching signals can be expressed by the action is capable to reduce the energy of the system, it is pos-
following equation: sible to bring the energy of the system to zero and in this case,
the system will be in a stable condition [34].
cos ωt − 2π cos ωt + 2π


    
Sd 2 cos ωt 3  3 
=
Sq 3 − sin ωt − sin ωt − 2π3
− sin ωt + 2π3
⎡ ⎤ BIOGRAPHIES
Sa
× ⎣ Sb ⎦ . Tushar Kanti Roy received the B. Sc. degree from
Sc Rajshahi University of Engineering and Technology, Rajshahi,
Bangladesh, in 2008 in Electrical & Electronic Engineering
The instantaneous values of all voltages (ea , eb , and ec ) are and Master of Engineering by Research from the University of
known. The relationship between the dq-frame and abc-frame New South Wales, Canberra, Australia, in 2012 in Electrical
18 Electric Power Components and Systems, Vol. 0 (2018), No. 0
Engineering. He is currently working toward the Ph.D. degree
in the School of Engineering, Deakin University, Australia.
His research interests include nonlinear control theory and
applications, excitation control, microgrid, renewable energy,
and power electronics.
Md Apel Mahmud received his Ph.D. in Electrical Engineer-
ing with the best thesis award from the University of New
South Wales, Australia in December 2012 and a Bachelor
Degree in Electrical Engineering from Rajshahi University of
Engineering and Technology (RUET) in 2008 with the Univer-
sity Gold Medal. Currently he is working as a senior lecturer
in Electrical & Renewable Energy Engineering at Deakin Uni-
versity, Australia where he also worked as a lecturer. He also
worked as a lecturer in Electrical & Electronic Engineering at
Swinburne University of Technology, a research fellow at the
University of Melbourne, and a research publication fellow at
the University of New South Wales. Over the last seven years,
he has published more than 150 research articles including
three high impact journal papers and he has attracted more
than 1 million dollars internal and external grants. According
to Google Scholar, his research articles have received more
than 1200 citations with an H-index of 19. He was also listed
as program committee members of several IEEE conferences.
His research interests include various topics of electric power
engineering.
Amanullah Maung Than Oo received his Masters
and Ph.D. qualifications in electrical engineering from
the University of Melbourne and Victoria University,
Australia, respectively. He is currently Deputy Head of
School of the School of Engineering at Deakin University and
Course Director of the Electrical & Electronics Engineering
Discipline. He has made significant research contributions
in the area of electrical power engineering and renewable
energy, engineering education and published more than 250
scholarly articles in the peer reviewed high impact journals,
books, and conference proceedings. He has supervised more
than 10 Ph.D. students to completion and is currently super-
vising several Ph.D. students in the area of electrical power
engineering and renewable energy engineering. He has been
actively working with several international and national pro-
fessional communities and industries. He is sought worldwide
to deliver keynote addresses and presentations at workshops
and conferences. Aman has research interests and expertise in
renewable energy & energy storage system integration, smart
grid communication, power system stability and control,
energy management and efficiency, protection and security of
smart grids, sustainable operation and control of microgrids
as well as in engineering education.
Ramesh Bansal has more than 25 years of teaching, research
& industrial experience. Currently he is Professor and Group
head (Power) in the Department of Electrical, Electronic and
Computer Engineering at University of Pretoria, South Africa.
In previous postings he was with the University of Queens-
land, Australia; the University of the South Pacific, Fiji; Birla
Institute of Technology and Science, Pilani, India; and Civil
Construction Wing, All India Radio. He has published over
250 papers in journals and conferences. Prof. Bansal is an
Editor of several reputed journals including IET-Renewable
Power Generation and Electric Power Components and Sys-
tems. He is a Fellow, and CPEng. IET-UK, Fellow Engi-
neers Australia, Fellow Institution of Engineers (India), and
Senior Member-IEEE. He has diversified research interests
in the areas of renewable energy and conventional power
systems which includes wind, PV, hybrid power systems,
distributed generation, grid integration of renewable energy,
power systems analysis (reactive power/voltage control, sta-
bility, faults and protection), smart grid, FACTS, and power
quality.
Md Enamul Haque graduated in Electrical and Electronic
Engineering from Rajshahi University of Engineering Tech-
nology (RUET), Bangladesh, in 1995. He received the M.Eng.
degree in electrical engineering from University Technology
Malaysia, Malaysia in 1998, and the Ph.D. degree in electrical
engineering from University of New South Wales (UNSW),
Sydney, Australia, in 2002. He has worked for King Saud
University, Saudi Arabia, United Arab Emirates University,
and University of Tasmania, Australia. He is currently work-
ing as a senior Lecturer in Renewable and Smart Energy
Systems in the School of Engineering, Deakin University,
Australia. His research interests include control of variable
speed wind turbines, grid integration impact of renewable
energy sources, performance enhancement of power system
using STATCOM and energy storage, power management of
microgrid with wind, photovoltaic, plug-in hybrid vehicle, fuel
cell and energy storage, and power electronics applications in
renewable energy and power systems.