You are on page 1of 9

Solar Energy 170 (2018) 192–200

Contents lists available at ScienceDirect

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

MPPT for photovoltaic system using nonlinear backstepping controller with T


integral action

Muhammad Arsalan, Ramsha Iftikhar, Iftikhar Ahmad , Ammar Hasan, K. Sabahat, A. Javeria
School of Electrical Engineering and Computer Science, National University of Sciences and Technology, Islamabad, Pakistan

A R T I C LE I N FO A B S T R A C T

Keywords: The energy generating capability of a Photovoltaic (PV) system depends on the environmental conditions and the
Photovoltaic (PV) variations in the load connected to it. The performance of controller used for maximum power point tracking
MPPT (MPPT) can improve the efficiency of the PV system. PV array is a nonlinear system, so a nonlinear controller is
Buckboost converter more suitable for MPPT applications. However, the performance of nonlinear controllers completely depend
Backstepping
upon the nonlinear model of the system under consideration. Since all the real world systems are subjected to
Lyapunov stability
vary with time, it is not feasible to dynamically remodel the system as well as the controller all the time. These
variations introduce steady state error in the output which degrades the efficiency of the controller. To reduce
this error, a nonlinear Backstepping controller with integral action has been proposed to improve the perfor-
mance of recently proposed Backstepping controller for MPPT. The study uses a regression plane to generate the
reference peak power voltage for MPPT using non inverting DC – DC Buckboost converter. Global Asymptotic
stability of the whole system has been proved using Lyapunov stability criteria. MATLAB/Simulink is used to test
the performance of the proposed controller under varying irradiance and temperature conditions. To further
validate its performance, we have compared it with modified Perturb and Observe (P&O) algorithm, nonlinear
Backstepping controller and Fuzzy Logic Based nonlinear controller under rapidly varying environmental con-
ditions.

1. Introduction changing this peak power voltage. Consequently, tracking this voltage
is a major task, essential for an efficient PV system.
Every day, Sun bestows limitless energy to our earth that is readily Rigorous research is being carried out in recent times to strive for
available to everyone and everywhere, which can be converted into robust and efficient MPPT algorithms. Mostly, these algorithms either
electricity using PV Cell. PV cells don’t produce any form of pollution try to operate the PV cell at IMPP or VMPP , which eventually enables the
and can be connected directly to the electrical load, but their efficiency system to extract maximum power from it. Similarly, some algorithms
varies with the variation in the operating point of the cell. The char- exploit the fact that only a single MPP can occur in the characteristic
acteristic curves of a PV module are shown in Fig. 1. It can be observed curves of a PV cell and hence use analytical solutions to reach MPP
that the current drawn from the PV array changes with variation in its (Chenni, 2016). We can classify these algorithms or techniques into
output voltage. The efficiency of the array varies accordingly, making following categories:
direct connection of load with the array inefficient. A DC-DC converter
is hence used as an interface between the array and the load, and its • Conventional MPPT Algorithms
duty cycle is varied in such a way that the PV module is operated at • Bio-Inspired MPPT Algorithms
peak power voltage (VMPP ) and maximum current (IMPP ) is drawn from • Fuzzy and Artificial Intelligence (AI) Based Algorithms
it. This ensures that maximum power is drawn from the PV module and • Linear Controllers for MPPT
is operated at Maximum Power Point (MPP) (Ammar Al-Gizi, 2017). • Nonlinear Controllers for MPPT
Therefore, the PV cell should always be operated at its peak power
voltage. However, according to the one diode model of PV cell To operate PV arrays at MPP, hill climbing techniques are used
(Naghmash and Arsalan, 2017), continuous variation in a number of conventionally. These techniques include a number of variants of two
environmental variables such as irradiance and temperature keeps on basic algorithms, namely: Perturb and Observe (P&O) and Incremental


Corresponding author.
E-mail addresses: marsalan.msee15seecs@seecs.edu.pk (M. Arsalan), riftikhar.msee15seecs@seecs.edu.pk (R. Iftikhar), iftikhar.rana@seecs.edu.pk (I. Ahmad),
ammar.hasan@seecs.edu.pk (A. Hasan), skiran.msee16seecs@seecs.edu.pk (K. Sabahat), jahmed.msee16seecs@seecs.edu.pk (A. Javeria).

https://doi.org/10.1016/j.solener.2018.04.061
Received 9 January 2018; Received in revised form 26 April 2018; Accepted 28 April 2018
0038-092X/ © 2018 Elsevier Ltd. All rights reserved.
M. Arsalan et al. Solar Energy 170 (2018) 192–200

Fig. 1. Characteristic curves of a PV module.

Conductance (IC). Both of these techniques try to find the maxima of Bio-inspired algorithms such as Particle Swarm Optimization (PSO),
the Power vs Voltage curve of the PV module shown in Fig. 1. P&O Genetic Algorithm (GA), Shuffled Frog Leaping Algorithm (SFLA) and
algorithm perturbs the output voltage of a PV array (VPV ) and observes Ant Colony Optimization (ACO) can successfully resolve the issue of
any variation in the extracted power. This perturbation is introduced by partial shading (Hugues Renaudineau and Donatantonio, 2015; Mouna
slightly varying the duty cycle of the converter. If increasing the VPV Smida, 2015; Zengrui Yang and Duan, 2017; Sundareswaran and
results in generation of more power, i.e. ΔPower > 0 , then the voltage Sankar, 2016). Most of these techniques follow a similar pattern or
will be further increased. But, if ΔPower < 0 , then the duty cycle will be procedure to achieve optimization. Firstly, a population of different
varied in such a way that VPV is reduced. This process repeats itself and individuals or particles is generated in a region where the solution
the system eventually reaches MPP. However, each perturbation takes exists and each individual represents a distinguished solution which is
some time for the system to react and reach a stable output voltage of initialized randomly. These individuals then interact or socialize with
PV module, increasing the time of convergence to reach MPP. The al- each other to produce their offsprings or new states, which are then
gorithm keeps on searching for the new MPP and hence, introduces compared with their parents using a cost function. The one with better
oscillations in VPV , even if MPP is reached. Hence, overall efficiency of performance becomes the next generation and this procedure continues
the system is reduced (Bashar Zahawi, 2012; Kamarzaman, 2014; till all the individuals converge to the desired solution. When these
Ebrahimi, 2015). algorithms are used, the likelihood of reaching maximum power point
According to Fig. 1, if a PV cell is operated at MPP, then at that is very high, as the initial population is generated randomly with
single point, ΔPower /ΔVPV = 0 . In other words, completely unrelated parameter values (Eberhart, 1995; Logeswaran,
Δ(VPV IPV )/ΔVPV = VPV (ΔIPV /ΔVPV ) + IPV = 0 . Thus, at MPP, the sum of 2013; Kamarzaman, 2014).
incremental conductance (ΔIPV /ΔVPV ) and instantaneous conductance Though Bio-inspired techniques can successfully solve the problem
(IPV / VPV ) is equal to 0 (Bashar Zahawi, 2013, 2012). This is the working of partial shading, yet their performance depends completely on the
principle of IC algorithm. Duty cycle of the converter is varied to search selected parameters as well as the initial conditions. In fact, all of these
for the MPP. IC algorithm is more efficient than P&O because once the techniques require a lot of parameters and processes which increase
MPP is reached; it causes less oscillations about the MPP. Similarly, their steady state time and computational complexity (Kamarzaman,
under rapidly changing weather conditions, IC algorithm outperforms P 2014; ElKhozondar and ElKhozondar, 2016). Moreover, these algo-
&O (Laszlo Mathe and Kerekes, 2013; Lyden, 2015). However, it does rithms do have a tendency to converge on local maxima as well,
require additional control circuitry for its successful operation (Reisi especially if the population is not initialized randomly. Lastly, when
and Moradi, 2013) which makes it complex. these algorithms are used, there is a trade off involved between se-
PV modules are connected in large number in series and parallel lecting convergence accuracy and time of convergence. Increasing po-
combinations to generate more power. It is possible that some of these pulation size and the number of generations/iterations will result in an
modules receive different irradiance due to partial shading. In such an increase in steady state time to reach MPP. Conversely, convergence
event, multiple peaks of local maxima appear in the characteristic accuracy has to be compromised if steady state time is to be reduced
curves of a PV module. Since hill climbing techniques search for a peak (Logeswaran, 2013).
without considering the global response, they are liable to converge to Many MPPT techniques consider the mathematical model of the
one of these local maxima which make these techniques inefficient system to determine the value of control input. But Fuzzy logic (FL)
(Jubaer Ahmed, 2017). Similarly, other conventional methods such as based algorithms rely completely on the human knowledge and in-
Fractional Open-Circuit Voltage (FOCV) and Fractional Short-Circuit formation about that particular system. The inputs and outputs are
Current (FSCC) do not actually track the MPP. Instead, they assume a mapped using a fuzzy if – else rule base, which entirely depends on the
linear relationship between VMPP and open circuit voltage of PV module designer (Ahmad Khateb, 2016). Less information about a system will
(VOC ) or IMPP and short circuit current of PV module (ISC ) . It result in result in designing of a crude control and vice versa. Artificial In-
power loss as there is no such relationship between VMPP and VOC or IMPP telligence (AI) based MPPT algorithms are similar to FLC based algo-
and ISC (Reisi and Moradi, 2013; Farahat and Enany, 2015; Chenni, rithms, since they don’t require the mathematical model of the system,
2016). Ripple Correlation Control (RCC) technique exploits the ripples but they do require a training data set to train the input output relation.
produced in the output voltage of a PV array due to the switching Once deployed, they increase their data set and improve their perfor-
process of power electronics converter. However, it increases the time mance with time (Elobaid, 2015).
of convergence to reach MPP and is also incapable to track MPP at low Fuzzy logic controllers (FLC) with fuzzifier and defuzzifier are used
irradiance (Kamarzaman, 2014; Chenni, 2016; Lyden, 2015). extensively for MPPT (Adly, 2011; Maissa Farhat, 2015). These

193
M. Arsalan et al. Solar Energy 170 (2018) 192–200

backstepping controller can track it. Similarly, to purify water using


solar energy and desalination process, backstepping controller was
implemented to dissipate the maximum generated power in resistive
load (Hanen Nafaa and Farhat, 2017). The generated heat is then used
to evaporate water which is then condensed and further treated to get
drinkable water.
Backstepping controller is very efficient and robust but its perfor-
mance completely depends on the nonlinear system modelling. The real
world systems are subjected to variation with time, so the equations
used to model them can vary. Hence, as a result, the performance of
backstepping controller can deteriorate. A small modification in con-
Fig. 2. Proposed control methodology. troller, i.e. introduction of integral action can solve this problem.
In this research work, a nonlinear backstepping controller with in-
controllers are flexible enough to be used with a number of conven- tegral action has been proposed for MPPT of PV module, using a non
tional and bio-inspired MPPT algorithms as well (Adly, 2012). In these inverting Buckboost converter. The proposed methodology has been
hybrid algorithms, bio-inspired techniques are used to tune to para- presented in Section 2. Similarly, to generate reference of the peak
meters of the FLC to achieve optimum results (Noureddine, 2017; power voltage, a regression plane has been presented in Section 3 using
Ajiatmo, 2016). When compared with IC and P&O, FL based MPPT the data recorded from the PV array under varying temperature and
algorithms outperforms them because of their robustness (Ammar irradiance conditions. Mathematical modelling of Buckboost converter
Ghalib, 2016). Similarly, when FLC is used, PV array shows oscillations has been presented in Section 4, whereas, the derivation of the pro-
of smaller magnitude which greatly improves its efficiency (Dalila posed controller has been presented in Section 5. Asymptotic stability of
Beriber, 2013). the whole system has also been proved using Lyapunov stability criteria
A different type of fuzzy logic based controller (FLBC) i.e. Takagi- in Section 5. The performance of the system has been analyzed using
Sugeno (TS) fuzzy logic controller was implemented in Naziha Harrabi MATLAB under varying environmental conditions in Section 6 and it
(2017) for MPPT using Buck converter. The study used an algorithm has also been compared with the backstepping controller proposed in
similar to that of IC to reach MPP but the perturbations in duty cycle Naghmash and Arsalan (2017). The comparisons between the analyzed
were introduced in accordance with the fuzzy rules. An artificial neural controllers is summarized in a tabular form in Section 7 and the article
network (ANN) has also been used along with FLC for MPPT of PV is finally concluded in Section 8.
module (Ammar Ghalib, 2017). ANN was used to train a data set con-
taining peak power voltage under varying weather conditions. This data 2. Proposed control methodology
was then used to generate reference voltage for the FLC to reach MPP
under varying environmental conditions. Both Fuzzy and ANN based The working principle of conventional and proposed algorithm is
MPPT algorithms outperform bio-inspired and hybrid algorithms be- quiet different. The proposed system senses the temperature and in-
cause of their robust response, ability to improve their performance by cident irradiance on PV module and provide this information to the
learning on their own and low power consumption. However, compu- regression plane block, as shown in Fig. 2. The regression plane uses a
tational cost, complexity of implementation and tendency to converge relation between VMPP (i.e. Voltage @ MPP) and
at local maxima in case of partial shading are the disadvantages (Zekry, {Irradiance,Temperature} to provides reference peak power voltage
2014; Ali Nasr and Saied, 2012). Similarly, both of these controllers (VPVref ) which can be tracked by the proposed nonlinear controller. The
also require periodic maintenance as the electrical characteristics of PV controller has been derived using the mathematical model of Non-in-
array normally changes with age (ElKhozondar and ElKhozondar, 2016; verting Buckboost converter and generates an output signal μ , which
Lyden, 2015). controls the duty ratio of the PWM signal provided to the converter
Numerous linear controllers can also be used along with different switches.
algorithms for MPPT. The parameters or gains of PID controller were The flow chart diagram of the proposed technique is explained in
optimized in Roshdy Abdelrassoul (2016) to extract maximum power Fig. 3, whereas, the close loop control has been explained in Fig. 4. The
from the PV system. Similarly, in Besheer (2012), the author used ACO regression plane generates the reference x1ref = VPVref which is com-
to optimize the gains of PI controller for stand alone PV systems. The pared with VPV = x1 ( x1: the first state of the converter) to generate an
performance of PID controller was improved using gradient descent error signal, which is supplied to the Integral Backstepping Controller.
optimization method to optimize its gains in Roshdy Abdelrassoul The controller generates the control input μ , which controls the PWM
(2011). Similarly, if reference voltage perturbation is used as the con- signal width and drives the converter to track the reference voltage.
trol parameter for any hill climbing method, then linear controllers can Any variation in temperature or irradiance will generate a new unique
be used to smoothly adjust the duty ratio of power electronics converter reference. In fact, each set of unique temperature and irradiance values
accordingly (Bashar Zahawi, 2012, 2013, 2012). will generate a unique reference in return. Hence, operating the PV
Recently, a lot of research work is being carried out in the field of module on this reference voltage will ensure that maximum power is
nonlinear control. Many nonlinear controllers are also implemented to generated by the system.
track MPP of PV module, including backstepping and sliding mode
control (Naghmash and Arsalan, 2017; Emilio Mamarelis and Petrone, 3. Reference voltage generation by regression plane
2014). A PV cell is itself nonlinear in nature. Similarly, converters are
also modelled with the assumption that they respond linearly to the PV characteristic curve changes for each particular value of tem-
varying duty cycle. However, they also depict nonlinear behavior. The perature and irradiance. A slight variation in any one of them will result
robustness and stability that a nonlinear controller can achieve while in a new curve, hence changing the MPP, as shown in Fig. 1. Multiple
tracking the MPP, cannot be achieved by any other controller. A smart MPPs have been recorded, first by keeping the irradiance level constant
grid connected PV array system is proposed in Aranzazu Martin and at 1000 W/m2 and varying the temperature uniformly from 5 °C to 75 °C.
Cano (2015) to power up telecom towers using nonlinear backstepping Then more points have been recorded by varying irradiance levels from
approach. To generate reference PV array voltage, IC algorithm was 200 W/m2 to 1400 W/m2 at constant temperature of 25 °C. Using linear
proposed in Abderrahim Taouni and Abbou (2016), so that regression, a three dimensional regression plane has been generated
from these data points shown in Fig. 5. Hence, for any temperature and

194
M. Arsalan et al. Solar Energy 170 (2018) 192–200

Fig. 6. Non inverting buckboost converter.

4. Modeling of non-inverting buckboost converter

Non-Inverting DC – DC Buckboost converter can both step up and


step down the output voltage. The circuit diagram of the converter is
shown in Fig. 6. It is assumed to be operated in continuous conduction
mode (CCM) throughout this paper. Furthermore, this model is derived
while considering ideal semiconductor switches and diodes. Similarly,
the resistance of inductor and capacitor is also assumed to be zero for
the purpose of simplicity. It has two Modes of operation. In Mode1, both
the switches S1 and S2 are closed/turned on and both the diodes D1 and
D2 are reverse biased i.e. they are not conducting. By Kirchoff’s current
and voltage law,

⎧iC1 = iPV −iL


vL = vC1
⎨i = − vC2
⎩ C2 R (2)

In Mode2 , both switches are off and diodes are forward biased, i.e.,
they are conducting. Using Kirchoff’s current and voltage law:

⎧iC1 = iPV
vL = −vC 2
⎨i = i − vC2
Fig. 3. Flow chart. ⎩ C2 L R (3)

By utilizing inductor’s volt second balance and capacitor’s charge bal-


ance, we can write:
dvC1 iPV i
⎧ dt
= − Lu
C1 C1

di v v
L
= CL1 u− CL2 (1−u)
Fig. 4. Close loop control. ⎨ dt
⎪ dvC 2 = iL (1−u)− vC 2
⎩ dt C2 RC2 (4)

Averaging the model over one switching period and assuming x1,x2 and
μ to be the average value of vC1,vC 2 and u, we can write:

⎧ x1 = 〈vC1 〉
⎪ x2 = 〈iL 〉
⎨ x3 = 〈vC 2 〉
⎪ μ = 〈u〉
⎩ (5)

Hence, Eq. (4) takes the form:


iPV x
⎧ x1̇ = − 2μ
C1 C1
⎪ x1 x
x2̇ = L
μ− L3 (1−μ)

x2 x
⎪ x3̇ = (1−μ)− RC3
⎩ C2 2 (6)

This averaged state space model is then used to track the reference peak
power voltage.

Fig. 5. Regression plane.


5. Integral backstepping control

irradiance level, peak power voltage (vPVR) or VMPP can be calculated as: Nonlinear backstepping control is modified by introducing integral
action in it to track the generated reference voltage for PV array. The
vMPP = 332−1.34∗T −0.00964∗I (1) designed controller provides the input μ , that will determine the duty
ratio for the switches of the converter. The reference vMPP generated in
where T is Temperature and I is Irradiance.
Section 2 is termed as x1ref , since it is the first state of the modelled
converter. Assuming ε1 to be the error between actual and required PV
array output voltage, given as:

195
M. Arsalan et al. Solar Energy 170 (2018) 192–200

ε1 = x1−x1ref (7) ε2̇ = x2̇ −β ̇ (24)


The goal is to converge the error signal ε1 to zero. Taking time deriva-
C1 ⎛ i̇ μ̇ i
tive of Eq. (7) and simplifying using Eq. (6), we get β̇ = K1 ε1̇ + PV −x¨1ref + κζ ⎞̇ − 2 C1 ⎛K1 ε1 + PV −x1̇ ref + κζ ⎞
⎜ ⎟ ⎜ ⎟

μ ⎝ C1 ⎠ μ ⎝ C1 ⎠ (25)
i x
ε1̇ = x1̇ −x1̇ ref = PV − 2 μ−x1̇ ref
C1 C1 (8) Simplifying β ̇ using Eqs. (7), (10), (18) and (22), we get
Integral action is added into this error term as shown below: C1 ⎛ ⎛ ε2 i̇ C μ̇
β̇ = ⎜K1 −K1 ε1−
⎜ μ−κζ ⎞ + PV ⎞⎟− 1 (x¨1ref −κε )− β

e1 = ε1 + ζ (9) μ ⎝ ⎝ C1 ⎠ C1 ⎠ μ u (26)
where ζ is given as: Inserting β ̇ from Eq. (26) in Eq. (24), ε2̇ becomes
t
ζ= ∫0 (x1−x1ref ) dt (10) ε2̇ = x2̇ −
C1 ⎛ 2 Kε i̇ C μ̇
−K1 ε1− 1 2 μ−K1 κζ + PV ⎞ + 1 (x¨1ref −κε ) + β
⎜ ⎟
μ ⎝ C1 C1 ⎠ μ u (27)
Let V1 be a positive definite Lyapunov function candidate for
checking the convergence of ε1 to 0. Now, to guarantee convergence of both ε1 and ε2 to zero, a new
composite Lyapunov function VC is defined whose time derivative
1 2 κ
V1 = ε1 + ζ 2 should be negative definite so that our system can reach MPP.
2 2 (11)
1 2
where κ is a positive definite real number. To ensure asymptotic sta- VC = V1 + ε2
2 (28)
bility, time derivative of the Lyapunov function must be negative de-
finite. By taking time derivative of Eq. (11), we get Taking the time derivative of Eq. (28) and using Eq. (23), we get:
V1̇ = ε1 ε1̇ + κζζ ̇ (12) VĊ = V1̇ + ε2 ε2̇ = −K1 ε12−
ε1 ε2
μ + ε2 ε2̇
C1 (29)
Using Eq. (8) and time derivative of Eq. (10), we get

i x ε
V1̇ = ε1 ⎛ PV − 2 μ−x1̇ ref ⎞ + κζ (x1−x1ref )
⎜ ⎟
VĊ = −K1 ε12 + ε2 ⎛ε2̇ − 1 μ⎞
⎜ ⎟

⎝ C1 ⎠ (30)
⎝ C1 C1 ⎠ (13)
simplifying Eq. (13) using Eq. (7) For V1̇ to be negative definite, let
ε1
i x ε2̇ − μ = −K2 ε2
V1̇ = ε1 ⎛ PV − 2 μ−x1̇ ref + κζ ⎞
⎜ ⎟
C1 (31)
⎝ C 1 C1 ⎠ (14)
For V1̇ to be negative definite, let where K2 is a positive constant, so VĊ can be written as:

iPV x2 VĊ = −K1 ε12−K2 ε22 (32)


− μ−x1̇ ref + κζ = −K1 ε1
C1 C1 (15)
Using Eqs. (6) and (27) in Eq. (31), we get:
where K1 is positive definite. So, V1̇ becomes
x1 x K 2 C1 ε1 K C κζ i ̇ C1 x¨1ref C1 κε1
V1̇ = −K1 ε12 −K2 ε2 = μ− 3 (1−μ) + 1 + K1 ε2 + 1 1 − PV + −
(16) L L μ μ μ μ μ
Rewriting Eq. (15) as μ̇ ε1
+ β− μ
u C1 (33)
C1 ⎛ i
x2 = K1 ε1 + PV −x1̇ ref + κζ ⎞
⎜ ⎟

μ ⎝ C1 ⎠ (17) Solving Eq. (33) for μ̇

Let Eq. (17) be the reference or desired current for inductor, given by μ ⎛ x1 x3 K 2 C1 ε1 ⎞ μ ⎛ K1 C1 κζ i̇
μ̇ = ⎜−K2 ε2− μ + (1−μ)− 1 ⎟ + ⎜−K1 ε2− + PV
β ⎝ L L μ ⎠ β ⎝ μ μ
C1 ⎛ i
β= K1 ε1 + PV −x1̇ ref + κζ ⎞
⎜ ⎟
C1 x¨1ref ⎞ μ
μ ⎝ C1 ⎠ (18) − ⎟ +
⎛ C1 κε1 + ε1 μ⎟⎞

μ ⎠ β ⎝ μ C1 ⎠ (34)
Another error ε2 is defined to track x2 to β
where 0 < μ < 1 and β ≠ 0 Eq. (34) is used to provide μ to control the
ε2 = x2−β (19)
duty cycle of the signal given to switches S1 and S2 . This μ̇ , from Eq.
Rewriting Eq. (19) (34), proves the asymptotic stability of the controller, since the deri-
x2 = ε2 + β (20) vative of Lyapunov function becomes negative definite, which ensures
the convergence of error ε1 to zero.
Putting Eq. (20) in Eq. (8) gives
iPV ε2 + β 6. Simulation & results
ε1̇ = − μ−x1̇ ref
C1 C1 (21)
To verify the performance of the proposed controller, MATLAB/
Substituting β from Eq. (18) in Eq. (21), we get
SIMULINK is used to perform simulations. The parameters of the pro-
ε2 posed controller and the designed converter are mentioned in Table 1.
ε1̇ = −K1 ε1− μ−κζ
C1 (22) Whereas, Table 2 contains the parameters of PV array used for this
Hence, Eq. (12) becomes study. The results are presented in five subsections. The performance of
the proposed controller under varying irradiance and temperature is
ε1 ε2
V1̇ = ε1 ε1̇ + κζζ ̇ = −K1 ε12− μ analyzed in the first two subsections. The next subsection constitutes
C1 (23)
the comparison of the proposed controller with P&O algorithm with
Here, the first term in Eq. (23) is negative definite, but we are not sure improved performance due to the inclusion of PID controller. Similarly,
about the second term. By taking the derivative of Eqs. (18) and (19) the proposed controller is compared with the Backstepping controller
and simplifying the expressions, we have and Fuzzy logic based controller in the next two subsections.

196
M. Arsalan et al. Solar Energy 170 (2018) 192–200

Table 1
Parameters of controller & converter.
Parameter Value

k1 2870.18
k2 1551.65
κ 1510
Input capacitor, C1 67 uF
Inductor, L 11 mH
Output capacitor, C2 480 uF
Load resistor, R 20 Ω
Switching frequency, fs 100 kHz

Fig. 8. Generated power under varying irradiance.


Table 2
Parameters of PV array.
Parameter Value

PV module per string 10


Parallel connected strings 1
No of cells per module 60
Open circuit voltage 363 V
Short circuit current 7.84 A
Voltage at MPP 290 V
Current at MPP 7.35 A
Maximum power per module 213.15 W

Fig. 9. Output waveforms under varying irradiance.

Fig. 7. Tracking of PV module voltage under varying irradiance.

Fig. 10. Tracking of PV module voltage under varying temperature.


6.1. Test under varying irradiance
6.2. Test under varying temperature
To test the proposed controller under rapidly varying conditions,
the initial Irradiance is kept at 1000 W/m2 , which is abruptly changed to In this test, the initial temperature is first maintained at 30 °C which
800 W/m2 after 0.05 s. Similarly, after 0.1 s, it is changed to 600 W/m2 . is then increased to 45 ° C and then to 60 ° C after regular intervals of
During this experiment, the temperature of PV module has been 0.5 s respectively. The irradiance is kept at 1000 W/m2 throughout this
maintained ar constant temperature of 25 °C. The regression plane experiment. The controller yet again successfully tracks the reference
successfully generates the tracking peak power voltage which is suc- voltage, as shown in Fig. 10. The steady state error between the re-
cessfully tracked by the controller, as shown in Fig. 7. It is also clearly ference and the actual VPV is zero. Yet, a small overshoot of 7.44 V
visible that the system tracks the reference with zero steady state error. (2.6% of steady state value) was observed, as shown in Fig. 10. Simi-
However, an overshoot of 13.8 V (4.7% of steady state value) was ob- larly the controller is robust enough to maximize the power by reaching
served in the response. Similarly, in Fig. 8, the variation in power as a MPP with in less than 1 ms. The generated power under varying tem-
result of variation in irradiance is depicted. Both the generated power perature and the output power transmitted to the load by the converter
(by PV array) and the power available at load, i.e. output power are are shown in Fig. 11. Whereas, the converter output voltage and current
shown in Fig. 8. The PV module reaches at MPP within 2 ms with al- waveforms are shown in Fig. 12. Again, the efficiency of the converter
most negligible ripple. Moreover, the converter successfully transmits is observed to be more than 95%.
maximum power to the load with more than 95% efficiency. Converter
output waveforms ar shown in Fig. 9.1
6.3. Comparison with perturb & observe algorithm

Conventional P&O is modified by the introduction of PID controller


(Anto and Asumadu, 2016) to improve its performance and is then
1
To test the proposed controller more rigorously in a realistic environment, real power compared with the proposed controller. The comparison is done under
electronics switching devices, with forward voltage drops and resistances have been
considered for simulation purpose. Due to which, some power is lost in the converter. The
both varying temperature and irradiance conditions, mentioned in
controller does not have any information regarding these changes as it was derived using previous two subsections. The proposed controller clearly outperforms
the ideal model of Buckboost converter. the modified P&O algorithm. The comparison of the generated power

197
M. Arsalan et al. Solar Energy 170 (2018) 192–200

Fig. 11. Generated power under varying temperature.


Fig. 15. PV array voltage comparison under varying irradiance.

Fig. 12. Output waveforms under varying temperature.


Fig. 16. PV array voltage comparison under varying irradiance.

Fig. 13. Power comparison under varying temperature.

Fig. 17. PV array voltage comparison under varying temperature.

Fig. 14. Power comparison under varying irradiance.

under varying temperature is shown in Fig. 13. Similarly, the power Fig. 18. PV array voltage comparison under varying irradiance.
generation comparison under varying irradiance is shown in Fig. 14. It
can be observed that the modified P&O has improved its performance 6.4. Comparison with backstepping controller
when compared with the conventional P&O as is much more robust. But
it still takes considerable amount of time to reach MPP whenever the When backstepping controller is implemented for the application of
array is subjected to massive parameter variations. Moreover, these MPPT (Naghmash and Arsalan, 2017), a small steady state error has
comparisons also validate the performance of the proposed controller as been noticed in both the scenarios of varying temperature as well as
it is not only tracking the PV array voltage but also achieving MPPT irradiance. Integral backstepping on the other hand completely di-
robustly. minishes any error, and perfectly tracks the peak power voltage. The
results for varying temperature are shown in Fig. 15, whereas, Fig. 16

198
M. Arsalan et al. Solar Energy 170 (2018) 192–200

Table 3
Comparison.
Method RT (ms) ST 5% Criteria (ms) ST 2% Criteria (ms) SSE (V) Overshoot (V) Ripples (V)

Backstepping 2.42 3.1 3.8 1 0 0.65


P&O 2.22 19 NA 6 74 19
FLBC 2.17 3 8.4 0.5 15.4 3.6
IBS 2.17 2.9 3.2 0 13.94 0.63

shows the comparison at varying irradiance. Moreover, it can be ob- Similarly, the reference generation under partial shading may require
served that the proposed controller is also more robust as its rise time is the integration of other techniques such as PSO and ACO to successfully
2.2 ms, whereas, it is 2.5 ms with the backstepping controller. Similarly, generate maximum power.
a small overshoot of 8.2 volts (2.8%) can also be observed in case of
integral backstepping controller. Conflict of interest

6.5. Comparison with fuzzy logic based controller The authors declared that there is no conflict of interest.

The performance of proposed controller has been further validated References


by comparing it with that of Fuzzy logic based nonlinear controller for
MPPT, proposed in Bhanu and Sanam (2017). Fig. 17 compares the two Abderrahim Taouni, M.A., Abbou, Ahmed, 2016. Mppt design for photovoltaic system
using backstepping control with boost converter. In: 2016 International Renewable
controllers under varying temperature condition. Similarly, Fig. 18 and Sustainable Energy Conference (IRSEC). IEEE, pp. 469–475.
shows the response under varying irradiance. Fuzzy logic based non- Adly, I., 2011. Maximum power point tracker for a pv cell using a fuzzy agent adapted by
linear controller also shows promising results. The response is robust the fractional open circuit voltage technique. In: 2011 IEEE International Conference
on Fuzzy Systems (FUZZ-IEEE 2011). IEEE, pp. 1918–1922.
but oscillations of larger magnitude around the MPP can be observed, Adly, B., 2012. An optimized fuzzy maximum power point tracker for stand alone pho-
which are negligible in the case of the proposed controller. Similarly, in tovoltaic systems: ant colony approach. In: 2012 7th IEEE Conference on Industrial
case of integral backstepping controller, smaller overshoot is observed. Electronics and Applications (ICIEA). IEEE, pp. 113–119.
Ahmad Khateb, M.N., 2016. Fuzzy logic controller based sepic converter for maximum
power point tracking. IEEE Trans. Ind. Appl. 50, 2349–2358.
7. Comparison between analyzed techniques Ajiatmo, R., 2016. A hybrid fuzzy logic controller-firefly algorithm (flc-fa) based for mppt
photovoltaic (pv) system in solar car. In: 2016 IEEE International Conference on
Power and Renewable Energy. IEEE, pp. 606–610.
The analyzed techniques presented in the research work vary the Ali Nasr, M.Z.M.T.M.A.M., Saied, M.H., 2012. A survey of maximum ppt techniques of pv
duty cycle of the converter to control the output voltage of PV array. systems. In: 2012 IEEE Energytech. IEEE, pp. 1–17.
Ammar Al-Gizi, A.C., 2017. Improving the performance of pv system using genetically-
The results are presented in Table 3. The table is formulated by com- tuned flc based mppt. In: 2017 International Conference on Optimization of Electrical
paring the analyzed techniques on the basis of Rise Time (RT), Settling and Electronic Equipment (OPTIM) & ACEMP. IEEE, pp. 642–647.
Time (ST) (2% and 5% criteria), Steady State Error (SSE), Overshoot Ammar Ghalib, S.J., 2016. Study of flc based mppt in comparison with p&o and inc for pv
systems. In: 2016 International Symposium on Fundamentals of Electrical
and Ripples in VPV measured in Volts (peak to peak). Both Integral
Engineering (ISFEE). IEEE, pp. 1–6.
Backstepping (mentioned as IBS in Table 3) and FLBC responded swiftly Ammar Ghalib, S.J., 2017. The use of ann to supervise the pv mppt based on flc. In: THE
to the control signal as they exhibit the lowest RT. Though the differ- 10th International Symposium on Advanced Topics in Electrical Engineering. IEEE,
pp. 703–708.
ence is not very significant in comparison with the remaining techni-
Anto, P.Y.O.E.K., Asumadu, J.A., 2016. Pid control for improving p&o-mppt performance
ques. However IBS not only achieves steady state more rapidly, when of a grid-connected solar pv system with ziegler-nichols tuning method. In: IEEE 11th
compared with other techniques, but it also attained zero steady state Conference on Industrial Electronics and Applications (ICIEA). IEEE, pp. 1842–1847.
error. VPV of PV array, also depicts the least magnitude of ripples in its Aranzazu Martin, F.S., Cano, 2015. Backstepping control of smart grid-connected dis-
tributed photovoltaic power supplies for telecom equipment. IEEE Trans. Energy
waveform. The only drawback is the overshoot voltage which is again Convers. 30, 1496–1504.
less than 5% of the steady state value. The biggest overshoot and the Bashar Zahawi, M.E., 2012. Evaluation of perturb and observe mppt algorithm im-
largest ripples were observed in the case of P&O. These ripples can be plementation techniques. In: 6th IET International Conference on Power Electronics,
Machines and Drives (PEMD 2012). IET, pp. 1–6.
reduced but the trade off will be between ripple magnitude and the rise Bashar Zahawi, M.E., 2012. Assessment of perturb and observe mppt algorithm im-
time or robustness of the algorithm. Similarly, ripples of larger mag- plementation techniques for pv pumping applications. IEEE Trans. Sustain. Energy 3
nitude are also noticeable in the case of FLBC. Hence, the results pre- (1), 21–33.
Bashar Zahawi, M.E., 2013. Assessment of the incremental conductance maximum power
sented in Table 3 validates the superiority of the proposed controller point tracking algorithm. IEEE Trans. Sustain. Energy 4 (1), 108–117.
over others. Besheer, A., 2012. Ant colony system based pi maximum power point tracking for stand
alone photovoltaic system. In: 2012 IEEE International Conference on Industrial
Technology. IEEE, pp. 693–698.
8. Conclusion and future work Bhanu, P.H.C., Sanam, S.K., 2017. A fuzzy logic based mppt method for solar power
generation. In: International Conference on Intelligent Computing and Control
Systems (ICICCS). IEEE, pp. 1182–1186.
In this paper, nonlinear Integral Backstepping controller has been Chenni, A.S.R., 2016. A survey of maximum peak power tracking techniques used in
proposed for the application of MPPT using non-inverting Buckboost photovoltaic power systems. In: 2016 Future Technologies Conference (FTC). IEEE,
converter. The peak power voltage reference has been generated using pp. 430–443.
Dalila Beriber, A.T., 2013. Mppt techniques for pv systems. In: 4th International
regression plane and global asymptotic stability has been proved using
Conference on Power Engineering, Energy and Electrical Drives. IEEE, pp.
Lyapunov stability criteria. To validate the performance of the proposed 1437–1442.
controller, it has been compared with modified P&O, nonlinear Eberhart, J.K.R., 1995. A new optimizer using particle swarm theory. In: Proceedings of
Backstepping controller and Fuzzy Logic based nonlinear controller. the Sixth International Symposium on Micro Machine and Human Science, 1995.
IEEE, pp. 39–43.
The proposed controller performed well in comparison with all of them Ebrahimi, M.J., 2015. General overview of maximum power point tracking methods for
and it completely removes the steady state error as well. Although the photovoltaic power generation systems. In: 30th International Power System
controller is exceptional in its performance, but it is completely de- Conference (PSC2015). Niroo Research Institute, Tehran, Iran.
ElKhozondar, K.M.H.J., ElKhozondar, R.J., 2016. A review study of photovoltaic array
pendant on the authenticity of the regression plane. Since, the PV arrays maximum power tracking algorithms. Renew.: Wind, Water, Solar.
are subjected to wear and tear in real world, so maintenance of re- Elobaid, A.M., 2015. Artificial neural network-based photovoltaic maximum power point
gression plane is necessary for the controller’s optimal performance. tracking techniques: a survey. IET Renew. Power Gener. 9, 1043–1063.

199
M. Arsalan et al. Solar Energy 170 (2018) 192–200

Emilio Mamarelis, G.S., Petrone, Giovanni, 2014. Design of a sliding-mode-controlled Modelling, Identification and Control (ICMIC 2015). IEEE, pp. 1–6.
sepic for pv mppt applications. IEEE Trans. Ind. Electron. 61, 3387–3398. Naghmash, I.A., Arsalan, Muhammad, 2017. Backstepping based non-linear control for
Farahat, A.N.M.A., Enany, M.A., 2015. Assessment of maximum power point tracking maximum power point tracking in photovoltaic system. Sol. Energy 159, 134–141.
techniques for photovoltaic system applications. J. Renew. Sustain. Energy. Naziha Harrabi, M.S., 2017. Peak power tracker based on t-s fuzzy model for photovoltaic
Hanen Nafaa, S.L., Farhat, Maissa, 2017. A pv water desalination system using back- system. In: 2017 25th Mediterranean Conference on Control and Automation (MED).
stepping approach. In: 2017 International Conference on Green Energy Conversion IEEE, pp. 547–552.
Systems (GECS). IEEE, pp. 1–5. Noureddine, D., 2017. Flc based gaussian membership functions tuned by pso and ga for
Hugues Renaudineau, G.P., Donatantonio, Fabrizio, 2015. A pso-based global mppt mppt of photovoltaic system: a comparative study. In: Proceedings of the 6th
technique for distributed pv power generation. IEEE Trans. Ind. Electron. 62, International Conference on Systems and Control. IEEE, pp. 317–322.
1047–1058. Reisi, S.A.R., Moradi, M.H., 2013. Classification and comparison of maximum power
Jubaer Ahmed, Z.S., 2017. An accurate method for mppt to detect the partial shading point tracking techniques for photovoltaic system: a review. Renew. Sustain. Energy
occurrence in pv system. IEEE Trans. Industr. Inf. 13, 2151–2161. Rev. 433–443.
Kamarzaman, C.W.T.N.A., 2014. A comprehensive review of maximum power point Roshdy Abdelrassoul, Y.A., 2011. Optimized pid controller for both single phase inverter
tracking algorithms for photovoltaic systems. Renew. Sustain. Energy Rev. 585–598. and mppt sepic dc/dc converter of pv module. In: 2011 IEEE International Electric
Laszlo Mathe, D.S., Kerekes, Tamas, 2013. On the perturb-and-observe and incremental Machines & Drives Conference (IEMDC). IEEE, pp. 1036–1041.
conductance mppt methods for pv systems. IEEE J. Photovolt. 3 (3), 1070–1078. Roshdy Abdelrassoul, Y.A., 2016. Genetic algorithm-optimized pid controller for better
Logeswaran, A.S.T., 2013. A review of maximum power point tracking algorithms for performance of pv system. In: 2016 World Symposium on Computer Applications and
photovoltaic systems under uniform and non-uniform irradiances. In: 4th Research. IEEE, pp. 18–22.
International Conference on Advances in Energy Research 2013, ICAER 2013. Sundareswaran, S.P., Sankar, Peddapati, 2016. Development of an improved p&o algo-
ELSEVIER, pp. 228–235. rithm assisted through a colony of foraging ants for mppt in pv system. IEEE Trans.
Lyden, M.H.S., 2015. Maximum power point tracking techniques for photovoltaic sys- Ind. Inform. 12, 187–200.
tems: a comprehensive review and comparative analysis. Renew. Sustain. Energy Zekry, M.E.E.T.A.Y.A.A., 2014. Intelligent techniques for mppt control in photovoltaic
Rev. 1504–1518. systems: a comprehensive review. In: 4th International Conference on Artificial
Maissa Farhat, O., 2015. A stable flc-based mppt technique for photovoltaic system. In: Intelligence with Applications in Engineering and Technology. IEEE, pp. 17–22.
2015 IEEE International Conference on Industrial Technology (ICIT). IEEE, pp. Zengrui Yang, Z.X., Duan, Qichang, 2017. Analysis of improved pso and perturb & ob-
890–895. serve global mppt algorithm for pv array under partial shading condition. In: 2017
Mouna Smida, A.S., 2015. Genetic based algorithm for mppt for grid connected pv sys- 29th Chinese Control And Decision Conference (CCDC). IEEE, pp. 549–553.
tems operating under partial shaded conditions. In: 7th International Conference on

200

You might also like