See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/344899424

Design and Modelling of the ANFIS based MPPT controller for a Solar
Photovoltaic (SPV) System.

Article in Journal of Solar Energy Engineering · October 2020

DOI: 10.1115/1.4048882

CITATION READS

1 186

3 authors, including:

Ranganai Tawanda Moyo Sibusiso Moyo

Durban University of Technology Durban University of Technology
3 PUBLICATIONS 3 CITATIONS 62 PUBLICATIONS 375 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

STOCHASTIC VOLATILITY View project

Linearisation of Systems of Differential Equations View project

All content following this page was uploaded by Ranganai Tawanda Moyo on 26 June 2021.

The user has requested enhancement of the downloaded file.

 Design and Modeling of the
ANFIS-Based MPPT Controller for
Ranganai T. Moyo1
Department of Mechanical Engineering,
a Solar Photovoltaic System
Durban University of Technology, Maximum power point tracking (MPPT) controllers play an important role in improving the
Durban 4001, South Africa efficiency of solar photovoltaic (SPV) modules. These controllers achieve maximum power
e-mail: [email protected] transfer from PV modules through impedance matching between the PV modules and the
load connected. Several MPPT techniques have been proposed for searching the optimal
Pavel Y. Tabakov matching between the PV module and load resistance. These techniques vary in complexity,
Department of Mechanical Engineering and tracking speed, cost, accuracy, sensor, and hardware requirements. This paper presents the
Institute for Systems Science, design and modeling of the adaptive neuro-fuzzy inference system (ANFIS)-based MPPT
Durban University of Technology, controller. The design consists of a PV module, ANFIS reference model, DC–DC boost con-
Durban 4001, South Africa verter, and the fuzzy logic (FL) power controller for generating the control signal for the
e-mail: [email protected] converter. The performance of the proposed ANFIS-based MPPT controller is evaluated
through simulations in the MATLAB/SIMULINK environment. The simulation results demon-
Sibusiso Moyo strated the effectiveness of the proposed technique since the controller can extract the
Department of Research, maximum available power for both steady-state and varying weather conditions. Moreover,
Innovation and Engagement and Institute for a comparative study between the proposed ANFIS-based MPPT controller and the com-
Systems Science, monly used, perturbation and observation (P&O) MPPT technique is presented. The simu-
Durban University of Technology, lation results reveal that the proposed ANFIS-based MPPT controller is more efficient than
Durban 4001, South Africa the P&O method since it shows a better dynamic response with few oscillations about the
e-mail: [email protected] maximum power point (MPP). In addition, the proposed FL power controller for generating
the duty cycle of the DC–DC boost converter also gave satisfying results for MPPT.
[DOI: 10.1115/1.4048882]

Keywords: maximum power point tracking (MPPT), adaptive neuro-fuzzy inference system
(ANFIS), DC–DC boost converter, solar photovoltaic (SPV) system, perturbation and
observation (P&O) method, efficiency, energy, photovoltaics, simulation, solar

1 Introduction energy management controllers for hybrid SPV systems. The

concept of maximum power point tracking remains an essential
The rapid growth of installed renewable energy sources has
technology for improving the efficiency of PV modules. From
shown a significant change in the energy sector with a move to
several studies, it is evident that the use of solar modules without
replace traditional power generation sources like coal and diesel
MPPT controllers results in energy wastages, which ultimately
with renewable energy sources such as wind and solar [1]. Accord-
results in the need to install more PV modules for the same
ing to the International Renewable Energy Agency (IRENA), the
power requirement [6].
world installed 176 GW of renewable energy capacity in 2020
All MPPT techniques have one objective, that is, to force the PV
with solar and wind having a significant contribution [2]. From
modules always operate at their maximum power point (MPP) for
these statistics, it can be argued that the future is characterized by
any given weather conditions based on the maximum power transfer
a mix of energy technologies with renewable energy sources such
theorem (MPTT). The power output of PV modules is characterized
as solar, wind, and biomass contributing significantly in the new
by non-linear behavior due to the variation of the solar irradiance
global energy economy. Among the mentioned renewable energy
and solar cell temperature. For any given weather condition, the
sources, solar photovoltaic (SPV) technology is considered as the
solar module’s operating point corresponds to a unique point on a
most attractive alternative for power generation [3]. The interest
current–voltage (I–V ) curve. The same unique operating point on
in SPV is growing worldwide due to the continuous price drop of
the I–V curve also corresponds to a point on the power–voltage
both the photovoltaic (PV) modules and solar batteries and the
(P–V ) curve. And the operating point of the PV module must
advances in power electronics [4]. However, SPV systems still
always correspond to the highest value on the P–V curve for the
suffer from relatively low energy conversion efficiency [5]. Cur-
system to generate the maximum power, as shown in Fig. 1.
rently, the research related to SPV systems is concentrated on
However, if the PV module is connected directly to the electrical
solar cell material modification for improving the efficiency,
load, its operating point is dictated by the load connected, that is,
design of efficient maximum power point tracking (MPPT) control-
it can take any value on the P–V curve which might not be the
lers for extracting maximum available power from PV modules,
MPP, depending on the impedance interaction between PV
development of efficient power electronic converters and inverters
modules and the connected load. Therefore, MPPT techniques are
for stand-alone and grid-connected PV systems, as well as
employed to continuously adjust the impedance seen by the PV
solving power stability and quality issues by developing advanced
module to keep the PV module operating at, or close to, its peak
power point under varying solar irradiance, temperature, and load.
Generally, the MPPT controller is composed of a DC–DC power
1
Corresponding author. converter which is controlled by an algorithm to drive the panel’s
Contributed by the Solar Energy Division of ASME for publication in the JOURNAL operating point to the MPP. The MPPT techniques can be
OF SOLAR ENERGY ENGINEERING: INCLUDING WIND ENERGY AND BUILDING ENERGY
grouped into two categories: conventional techniques, such as per-
CONSERVATION. Manuscript received July 13, 2020; final manuscript received
October 9, 2020; published online November 18, 2020. Assoc. Editor: Kevin turbation and observation (P&O) [7–9], incremental conductance
R. Anderson. (InCon) [10,11], and open-circuit voltage (OCV) method [12,13],

Journal of Solar Energy Engineering Copyright © 2020 by ASME AUGUST 2021, Vol. 143 / 041002-1
 Fig. 2 Adaptive neuro-fuzzy inference system architecture [37]

Fig. 1 Current–voltage and power–voltage curves for a PV

module [7]
(backpropagation and least square error algorithms) into a single
technique. To demonstrate the ANFIS architecture, two fuzzy
and based on artificial intelligence (AI) techniques such as artificial IF-THEN rules based on a first-order Sugeno model are considered
neural networks (ANN) [14], fuzzy logic (FL) [15], particle swarm as follows [37]:
optimization (PSO) [16], and adaptive neuro-fuzzy inference system
RULE 1: If x is A1 and y is B1, then f1 = p1x + q1y + r1
(ANFIS) [17]. The P&O and InCon are the widely used conven-
RULE 2: If x is A2 and y is B2, then f2 = p2x + q2y + r2
tional MPPT techniques because of their simple hardware imple-
mentation and sensor requirements as well as low cost. However, where x and y are inputs; Ai and Bi are fuzzy variables; fi represents
as reported in Refs. [18–20], these conventional techniques suffer outputs within fuzzy sets; pi, qi, and ri are design parameters that
from various problems such as slow tracking speed, high fluctua- are determined during the training process of the ANFIS
tions about the MPP, and drift issues involved when there is a system. The ANFIS architecture is shown in Fig. 2. In this figure,
rapidly changing weather condition. Also, conventional techniques a circle indicates a fixed node, whereas a square indicates an adap-
are defined for uniform environmental conditions and they may fail tive node. ANFIS has a five-layer architecture, and the nodes in
to track the global maximum power point (GMPP) during non- each layer have similar functions. Each layer is explained in
linear and partially shaded conditions [21]. Recently, AI-based detail below.
MPPT techniques have been proposed to address problems associ- Layer 1
ated with conventional MPPT techniques. Amongst them, the The outputs of layer 1 are the fuzzy membership grades of the
FL-based MPPT controller is considered as a powerful one due to inputs, which are given by the following equations:
its fast-tracking speed and fewer oscillations as reported in Refs.
[22–25]. The only challenge with this technique is that it heavily O1,i = μAi (x), i = 1, 2 (1)
depends on good knowledge about PV systems. Thus, the efficiency
of the FL-based controller well depends on the appropriate design of O1,i = μBi−2 (y), i=3 (2)
fuzzy rules and memberships functions. The ANN-based MPPT
controller is also considered as a powerful technique because of where x and y are the inputs to the node i, and Ai and Bi are linguis-
its ability to solve complex and non-linear functions. However, tic labels (high or low) associated with this node functions. μAi(x)
ANN-based MPPT controllers have some drawbacks such as the and μBi−2(y) can adopt any fuzzy membership function. For
need for a large amount of training data to ensure accuracy, longer example, if the bell-shaped membership function is employed,
training times, and the complexity in the design of ANN architec- μAi(x) is given as follows:
tures [26]. To solve these limitations, FL can be integrated with
ANNs to form an ANFIS for MPPT. The studies [27–30] have 1
μAi = 
  , i = 1, 2 (3)
proved that ANFIS-based MPPT controllers have a fast dynamic x − ci 2
response and small oscillations about the MPP as compared with 1+ bi
other MPPT techniques. However, it should be noted that most of ai
the ANFIS-based MPPT controllers that have been presented in
the literature rely on the PI(D) controller for generating the duty where ai, bi, and ci are parameters of the membership function.
cycle signal for the DC–DC converter as given in Refs. [17,31– Layer 2
35]. In this study, the FL power controller is proposed for calculation In layer 2, the nodes are fixed nodes. The layer involves fuzzy
of the duty cycle and for providing a control signal to the DC–DC operators, and it uses the AND operator to fuzzify the inputs.
boost converter. Thus, the proposed ANFIS-based MPPT is made They are labeled with π, indicating that they perform as a simple
up of a DC–DC boost converter, ANFIS reference model, and FL multiplier. The output of this layer can be represented as
controller. The ANFIS reference model is trained with a large
O2 , i = wi = μAi (x) ∗ μBi (y), i = 1, 2 (4)
amount of the real data sets to ensure the accuracy and reliability
of the controller. These are the so-called firing strength of the rules.
Layer 3
In layer 3, the nodes are also fixed nodes labeled by N, to indicate
2 The Architecture of the Adaptive Neuro-Fuzzy that they play the normalization role to the firing strengths from the
Inference System previous layer. The output of this layer can be represented by
The ANFIS is a data learning technique that uses FL to transform wi
system inputs into the desired outputs with the use of highly inter- O3 , i = w i = , i = 1, 2 (5)
w1 + w2
connected artificial neural networks, which are weighted to map the
numerical inputs into desired outputs [36]. The ANFIS combines Outputs of this layer are called normalized firing strengths.
the benefits of the two machine learning techniques Layer 4

041002-2 / Vol. 143, AUGUST 2021 Transactions of the ASME

 In layer 4, the nodes are adaptive. The output of each node is given by
simply the product of the normalized firing strength and a first-order
polynomial (for a first-order Sugeno model). The output of this Isc
Irs = (10)
layer can be written as follows: e((q Voc )/(n Ns K T)) −1
where Voc is the open-circuit voltage. The module’s saturation
O4 , i = wi fi = wi ( pi x + qi y + ri ), i = 1, 2 (6) current at any given temperature is given by
where w
is the output of layer 3 and pi, qi, and ri are the consequent
3
T
parameters. Io = Irs e[(q Ego ((1/Tn )−(1/T)))/n K] (11)
Layer 5 Tn

In layer 5, there is only one single fixed node labeled with x. where Tn = 298 K; Ego is the bandgap energy. In this study, the
This node performs the summation of all incoming signals. The ART solar module—360 Wp, 39.0 V Si-monocrystalline type
overall output of the model is given by module is used. The parameters of the module at STC are given
  in Table 1.
w i fi
O5 , i =
fi = i
w (7) By using the parameters presented in Table 1, the MATLAB/SIMU-
i i wi LINK model of the solar module is created. After modeling, the
MATLAB/SIMULINK PV module model has to be verified or validated
to check if resembles the characteristics of the actual module as
given in the manufacturer’s datasheet. The model is simulated
3 System Modeling under STCs without connecting any load. STC is a testing condition
3.1 Modeling of the Solar Photovoltaic Module. Photovol- for manufacturers to check the performance of solar modules, and it
taic is a process of converting solar irradiance into electricity specifies that the module must be tested under a solar cell tempera-
using semiconductor materials that exhibit a property known as ture of 25 °C, the irradiance of 1000 W/m2, and an air mass of 1.5
the photovoltaic effect [38]. The general model of a solar cell can (AM1.5) [39]. The peak power point of the simulated MATLAB/SIMU-
be derived from the physical characteristics of a diode, which is LINK model was 360 Wp which is given in the manufacturer’s data-
usually called the single diode model. Figure 3 shows the equivalent sheet. The I–V and P–V curves of the simulated MATLAB model of
circuit for a single diode model. From Fig. 3, the current source, IL, the PV module at 25 °C and different solar irradiance are given in
represents the flow of electrons when solar radiation hits the surface Figs. 4 and 5.
of a solar PV cell. And the diode represents the characteristic beha-
vior of the PN junction of the solar PV cell. The model has two
resistances, namely, series resistance and parallel resistance. The 3.2 Proposed ANFIS-Based MPPT Controller. The pro-
series resistance (Rs) represents current losses due to metal contacts posed ANFIS-based MPPT controller is made up of the ANFIS
within the solar PV cells, and the parallel resistance (RSH) accounts reference model, FL power controller, and a DC–DC boost con-
for current leakages through the resistive path in parallel with the verter as shown in Fig. 6. This MPPT controller is based on the
intrinsic device [17]. Several solar cells have to be connected to fact that by knowing the maximum possible power output of a
form a solar PV module. The output current of the solar PV PV module for a given set of solar irradiance and temperature,
module is given as the real-time MPP of the solar module can be perfectly tracked.
The ANFIS reference model gives out the expected value of the
I = I ph − Io [e((q(V+IRs ))/(nKNs T)) −1] − Ish (8) maximum power output from the PV modules at a specific tempera-
ture and irradiance. At the same irradiance and temperature, the
where Iph is the photo-current; Io is the saturation current; q is the actual power output which is coming from the PV module is mea-
electron charge; V is the output voltage of the PV module; n is sured and compared with the reference value from the ANFIS
the ideality factor of the diode; K is the Boltzmann constant; Ns model. The difference between the two power values is calculated
represents the number of solar cells connected in series; T is the to give an error, which is then fed to the FL power controller to gen-
solar cell temperature; and Ish is the current through the shunt resis- erate a control signal. The signal generated by the FL power control-
tor. The current generated from the incidence of radiation at a given ler is given to the pulse width modulator (PWM). The PWM
temperature is expressed as generates a signal at a high level of frequency to control the duty
cycle of the DC–DC power converter and force PV modules to
 operate at the MPP.
G
I ph = Isc + ki (T−298) (9)
1000
3.2.1 Design of the DC–DC Boost Converter. Maximum
power point tracking algorithms are implemented using highly effi-
where Isc represents the short circuit current; ki is the temperature
cient DC–DC power converters. There are several topologies of
coefficient of the Isc at standard test conditions (STCs); and G is
DC–DC power converters which can be used for maximum
the solar irradiance. The reverse saturation current of the diode is
power point tracking and these include the boost converter, buck
converter, Cùk converter, and SEPIC converter. Because of its

Table 1 Specifications of the ART-solar-360 Wp

Quantity Value

Maximum power, PMPP 360 Wp

The voltage at MPP, VMPP 39.0 V
The current at MPP, IMPP 9.24 A
Voc 47.5 V
Isc 9.71 A
Number of cells, Ns 72
Temperature coefficient of the Isc, ki 0.050%/°C
Temperature coefficient of the Pmax −0.39%/°C
Fig. 3 Equivalent circuit of a single diode model [40]

Journal of Solar Energy Engineering AUGUST 2021, Vol. 143 / 041002-3

 Fig. 4 Power versus voltage curve

Fig. 5 Current versus voltage curve

Fig. 6 ANFIS-based MPPT controller

041002-4 / Vol. 143, AUGUST 2021 Transactions of the ASME

 Fig. 7 DC–DC boost converter

Table 2 Selected parameters of the boost converter (a) ON state

During this state, the MOSFET switch S will be on and the
Parameter Symbol Value diode switch D will be off. The diode will be open-circuited
because the n side of the diode will be at higher voltage com-
Input voltage VMPP 39 V pared with the p side which will be shorted to the ground.
Input current IMPP 9.24 A
The current flows through the inductor and then to the
Duty ratio DMPP 0.35
Inductor L 3.7 mH MOSFET. The inductor stores energy in the form of a mag-
Load resistance R 10 Ω netic field. On the other side of the circuit, the current flows
Output capacitor Co 87.5 μF from the capacitor to the load. The ON state duration is given
Input capacitor Ci 4000 μF by TON = D + T, where D is the duty cycle and T is the
Switching frequency f 20 kHz switching frequency.
(b) OFF state
In this state, the control signal turns off the MOSFET and
simplicity and ability to step up the voltage, the boost converted is the diode is turned on. The inductor discharges and the
adopted in this study. A boost converter is made up of two semicon- current flows through the diode to the filter capacitor and
ductor switches (diode and MOSFET), an inductor and a capacitor the load. The capacitor stores energy in the form of charge.
as shown in Fig. 7. The filter capacitor in the output circuit is assumed to be
The working principle of the boost converter is based on the fact large such that the resistor–capacitor (RC) time constant is
that the inductor L resists sudden changes in input current. Assum- higher than the switching frequency to ensure constant
ing continuous conduction mode of operation, the converter has two output voltage.
states of operation which are given below.

Fig. 8 ANFIS reference model architecture

Journal of Solar Energy Engineering AUGUST 2021, Vol. 143 / 041002-5

 Fig. 9 Surface view of the ANFIS reference model

The boost converter is designed at STC of the solar module. The current relationships of a DC–DC boost converter are given by
solar module specifications are given in Table 1. Thus, the PV
output current, IPV = IMPP, output voltage, VPV = VMPP, and output VMPP
Vo = (12)
power, PPV = PMPP. Therefore, the input/output voltage and 1−D

Io = (1 − D)IMPP (13)

Table 3 Fuzzy rules for the FL power controller where Vo is the output voltage of the converter, VMPP is the input
voltage to the converter, Io is the output current of the converter,
E/CE Very low Low Neutral High Very high and IMPP is the input current to the converter. The relationship
between the load resistance (R) and optimal internal resistance of
Very low VH VH H VL VL the PV module (RMPP) is given as
Low H H H VL L
Neutral H H N L L RMPP
High H H L L VL R= (14)
(1 − DMPP )2
Very high H H L L VL
where R = Vo/Io, RMPP = VMPP/IMPP, and DMPP is the duty cycle

Fig. 10 Membership functions of E

041002-6 / Vol. 143, AUGUST 2021 Transactions of the ASME

 Fig. 11 Membership functions of CE

Fig. 12 Illustration of the P&O MPPT technique [20]

at MPP at STC. Since the range of the duty cycle is between 0 and 1, DC–DC boost converter are given in Table 2.
the load resistance must be equal or greater than the optimal internal
resistance of the PV module (R ≥ RMPP). Using the load resistance 3.2.2 Design of the ANFIS Reference Model. The ANFIS
of 10 Ω and assuming a lossless converter (Po = PPV), the output reference model has two input variables (solar irradiance and tem-
voltage of the converter is determined as perature) and one output variable (reference maximum power
output). The temperature and irradiance values for the particular
Vo = Po R = 60 V (15) site used in this study have been obtained from the PVGIS
website. By simulating the MATLAB/SIMULINK model of the PV
The duty cycle at MPP at STC is determined as module (presented in Table 1) without connecting any load, the
VPV reference maximum power output values are generated. The
DMPP = 1 − = 0.35 (16) input/output data sets are then used to train the ANFIS reference
V0
model. The model was trained with the help of 145 data sets and
The PV module’s voltage varies with the current and to minimize using triangular membership functions. Figure 8 shows the
the ripples, the minimum value of the inductor has to be designed ANFIS reference model architecture with five membership func-
for 1% of current ripples (ΔIPV) at a high-frequency value of tions for solar irradiance and three membership functions for
20 kHz as given below [40] temperature.
Figure 9 shows the structure of the surface view plot of the
VPV × DMPP ANFIS reference model, and the mapping between inputs and
L≥ = 3.7 mH (17)
2 × ΔIPV × f outputs is demonstrated.
Additionally, the minimum value of the input capacitor has to be
3.2.3 Design of the Power Controller. The power controller is
designed for 1% of the voltage ripples as given below [40]
based on FL, and it is used to generate the control signal for the con-
DMPP verter. The signal is generated based on the error between the actual
C≥ = 87.5 μF (18) power output of the PV module and the reference power output
2 × ΔVo × f × R
given by the ANFIS reference model. The FL power controller
In this study, the input capacitor Ci is incorporated to reduce the has two input variables (error E and change in error CE) and one
ripples of the input voltage as well as to deliver alternating output variable (duty cycle increment ΔD). Triangular membership
current to the inductor. The selected design parameters of the functions were used for each variable in the design of the FL

Journal of Solar Energy Engineering AUGUST 2021, Vol. 143 / 041002-7

 Fig. 13 At STC, with and without the proposed MPPT controller

controller. Five membership functions were chosen for each vari- perturbation is then calculated and compared with the previous
able and defined as: Very Low (VL), Low (L), Neutral (N), High power output. The difference in the two power output values
(H), and Very High (VH). The ranges of the variables are given (ΔP) is calculated and if the ΔP is greater than zero, the perturbation
as E (−100 to 100), CE (−1 to 1), and D (−0.1 to 0.1). is kept in that direction. When ΔP becomes less than zero, the per-
The controller is designed with 25 fuzzy rules shown in Table 3. turbation is reversed and this process is repeated until the MPP is
Rows and columns represent the input variables (E) and (CE), and reached. Figure 12 illustrates the P&O MPPT technique.
the output variable (ΔD) is located at the intersection of the row and In this paper, the P&O MPPT technique is for comparison with
the column (Figs. 10 and 11). the proposed ANFIS-based MPPT technique. The P&O MPPT
technique is explained in detail in Refs. [9,19,41–44].
3.2.4 Design of the Perturbation and Observation MPPT
Controller. The P&O MPPT technique is the widely used MPPT
method for improving the efficiency of SPV modules. In this tech-
nique, a perturbation is first introduced to the operating voltage of
the PV module. The PV module’s power output after the

Fig. 14 Under varying solar irradiance, with and without the proposed MPPT controller

041002-8 / Vol. 143, AUGUST 2021 Transactions of the ASME

 Fig. 15 At STC, comparing with the P&O MPPT controller

4 Simulation Results and Discussion The proposed MPPT controller is simulated at STC (1000 W/m2
and 25 °C) and then compared with the simulation of the same
As stated before, if the PV module is connected directly to the
circuit but without the MPPT controller. Figure 13 shows the PV
load, its operating point is rarely at the MPP. MPPT techniques
module power output with the proposed MPPT (black line) and
are employed to match the internal resistance of the PV module
without the MPPT controller (white line). In both graphs, the
to the load resistance for maximum power to be transferred
PV power output rises sharply from zero up to about the MPP
to the load. To analyze the dynamic behavior of the proposed
(360 W) after a time of 20 ms. For the system with the MPPT
ANFIS-based MPPT controller, different scenarios were considered
controller, the power output settles at MPP since the MPPT control-
and simulated in the MATLAB/SIMULINK environment.
ler will be continuously forcing the PV module to operate at
Scenario 1: At STC, with and without the proposed MPPT
the MPP. For the system without the controller, the PV power
controller

Fig. 16 Under varying solar irradiance, comparing with the P&O MPPT controller

Journal of Solar Energy Engineering AUGUST 2021, Vol. 143 / 041002-9

 Fig. 17 Membership functions of ΔD

Fig. 18 Zoomed view of the oscillations of the P&O MPPT controller

output drops and settles at around 200 W. This is because the load MPP. However, it should be noted that between 0.4 s and 0.8 s,
resistance does not match the internal resistance of the module and the system without the controller operates around the MPP. This
there is no external circuit which can force the PV to find its MPP. is because, during that period, the optimal internal resistance of
Scenario 2: Under varying irradiance, with and without the pro-
posed MPPT
In this case, the proposed controller is evaluated with an operat- Table 4 Data sets of solar irradiance and temperature
ing temperature of 25 °C and sudden changes in solar irradiance
Solar irradiance (W/m2) Temperature (°C)
(1000 W/m2, 850 W/m2, 500 W/m2, and 50 W/m2).
Figure 14 shows the PV power output curves for the two systems. 633 30.6
It can be noted that the proposed MPPT controller presents a good 440 25.2
performance under varying solar irradiance since it perfectly tracks 222 20.1
the MPP for different solar irradiances levels. For the PV module 30 15.2
without the MPPT controller, the system does not operate at the

041002-10 / Vol. 143, AUGUST 2021 Transactions of the ASME

 Fig. 19 Under varying solar irradiance, varying temperature comparing with the P&O MPPT controller

the solar module is almost equal to the load resistance of the circuit. which makes the proposed MPPT technique efficient for any
At 25 °C, 500 W/m2, the optimal internal resistance of the PV given solar irradiance level and temperature (Fig. 18).
module is given by Scenario 5: Under varying solar irradiance, varying temperature
V2 and comparing with the P&O MPPT controller
RMPP = MPP = 8.5 Ω The performance of the proposed ANFIS-based MPPT is also
PMPP
evaluated under varying solar irradiance and temperature by com-
Which is almost equal to the load resistance of 10 Ω. Because of paring it with the P&O MPPT technique. In general, the efficiency
that, the PV module can operate at almost MPP without the inclu- of PV modules decreases with an increase in solar cell temperature.
sion of any controller to control its operating point. Four real environmental data sets of solar irradiance and tempera-
Scenario 3: At STC, comparing with the P&O MPPT controller ture are used. Table 4 shows these data sets.
The performance of the proposed ANFIS-based MPPT controller Figure 19 shows the power output curves of the two controllers.
is also evaluated by comparing it with the P&O MPPT technique at The proposed ANFIS-based MPPT shows a better response for
STC. It can be noted from Fig. 15 that for both MPPT controllers, changing solar irradiance and temperature. For the P&O MPPT
the PV module power output rises sharply up to the MPP with the controller, the accuracy and tracking speed is slower and this
rise time of 20 ms. For the system with the ANFIS-based MPPT results in power wastages since the SPV system will not be operat-
controller, the power output settles and maintains that value. ing at the MPP. The reason for this poor performance is because
However, for the system with the P&O MPPT controller, the the P&O technique relies on fixed steps to update the duty
power output oscillates about the MPP before settling down. The cycle of the DC–DC boost converter and it takes time for this con-
power output of the P&O technique settles at the MPP after 80 troller to locate the new MPP for rapidly changing environmental
ms as shown in Fig. 15. conditions.
Scenario 4: Under varying solar irradiance, comparing with the
P&O MPPT controller
At this point, the performance of the proposed MPPT controller is
evaluated by comparing it with the P&O MPPT controller under
varying solar irradiance levels (1000 W/m2, 850 W/m2, 500 W/m2,
and 50 W/m2). 5 Conclusion
Both controllers exhibit satisfactory tracking performance, but The design, modeling, and evaluation of the proposed ANFIS-
the degree of accuracy is different as shown in Fig. 16. The ANFIS- based MPPT controller were presented in this paper. By knowing
based MPPT controller displays a fast response to sudden changes the maximum possible power output of a PV module for a given
in solar irradiance levels with small oscillations about the MPPT. set of solar irradiance and temperature, the real-time MPP of the
For the P&O MPPT controller, the tracking speed is lower and solar module was thoroughly tracked. The components and the sub-
oscillations are much higher. The P&O MPPT controller also exhib- systems of the proposed MPPT controller were modeled and simu-
its the drift phenomenon (caused by the incorrect decision to either lated in MATLAB/SIMULINK environment. The proposed MPPT
decrease or increase the duty cycle for fast-changing irradiance controller was evaluated by comparing it with a circuit without
levels). It must be also noted that the efficiency of the P&O the MPPT controller as well as with the P&O MPPT technique.
MPPT technique is very poor for lower solar irradiance level (at Simulation results reveal that the proposed ANFIS-based MPPT
500 W/m2) as illustrated in Fig. 16 (between 40 and 60 ms). This can effectively track the maximum power point of PV modules
is because the P&O method uses a fixed step to either decrease or under different weather conditions with the same level of consis-
increase the duty cycle (non-adaptive) but for the ANFIS-based tency. The proposed FL power controller which was used to gener-
MPPT controller, ΔD changes (from −0.1 to 0.1 as shown in ate the control signal to the boost converter also gave satisfying
Fig. 17) depending on the error given to the FL power controller, results for MPPT.

Journal of Solar Energy Engineering AUGUST 2021, Vol. 143 / 041002-11

Acknowledgment [19] Kamala Devi, V., Premkumar, K., Bisharathu Beevi, A., and Ramaiyer, S., 2017,
“A Modified Perturb & Observe MPPT Technique to Tackle Steady State and
The authors would like to thank the National Research Founda- Rapidly Varying Atmospheric Conditions,” Sol. Energy, 157, pp. 419–426.
tion of South Africa (NRF) and the Durban University of Technol- [20] Kumar, A., Chaudhary, P., and Rizwan, M., 2015, “Development of Fuzzy Logic
Based MPPT Controller for PV System at Varying Meteorological Parameters,”
ogy for their financial support. 2015 Annual IEEE India Conference (INDICON), New Delhi, India, Dec.
17–20, pp. 1–6.
[21] Lyden, S., and Haque, M. E., 2015, “Maximum Power Point Tracking Techniques
for Photovoltaic Systems: A Comprehensive Review and Comparative Analysis,”
Renew. Sustain. Energy Rev., 52, pp. 1504–1518.
Conflict of Interest [22] Ben Salah, C., and Ouali, M., 2011, “Comparison of Fuzzy Logic and Neural
There are no conflicts of interest. Network in Maximum Power Point Tracker for PV Systems,” Electric Power
Syst. Res., 81(1), pp. 43–50.
[23] Gupta, A., Kumar, P., Pachauri, R. K., and Chauhan, Y. K., 2014, “Performance
Analysis of Neural Network and Fuzzy Logic Based MPPT Techniques for Solar
PV Systems,” 2014 6th IEEE Power India International Conference (PIICON),
References Delhi, India, Dec. 5–7, IEEE, pp. 1–6.
[24] Chim, C. S., Neelakantan, P., Yoong, H. P., and Teo, K. T. K., 2011, “Fuzzy
[1] Owusu, P. A., and Asumadu-Sarkodie, S., 2016, “A Review of Renewable Logic Based MPPT for Photovoltaic Modules Influenced by Solar Irradiation
Energy Sources, Sustainability Issues and Climate Change Mitigation,” Cogent and Cell Temperature,” 2011 UkSim 13th International Conference on
Eng., 3(1), pp. 1–14. Computer Modelling and Simulation, Cambridge, UK, Mar. 30–Apr. 1, IEEE,
[2] I.-I. R. E. Agency. “Renewables Account for Almost Three Quarters of New pp. 376–381.
Capacity in 2019,” IRENA, https://www.irena.org/newsroom/pressreleases/ [25] Menniti, D., Pinnarelli, A., and Brusco, G., 2011, “Implementation of a Novel
2020/Apr/Renewables-Account-for-Almost-Three-Quarters-of-New-Capacity- Fuzzy-Logic Based MPPT for Grid-Connected Photovoltaic Generation
in-2019, Accessed 2020. System,” 2011 IEEE Trondheim PowerTech, Trondheim, Norway, June 19–23,
[3] Ahmadi, M. H., Ghazvini, M., Sadeghzadeh, M., Nazari, M. A., Kumar, R., IEEE, pp. 1–7.
Naeimi, A., and Ming, T., 2018, “Solar Power Technology for Electricity [26] Jiang, L. L., Nayanasiri, D. R., Maskell, D. L., and Vilathgamuwa, D. M., 2015,
Generation: A Critical Review,” Energy Sci. Eng., 6(5), pp. 340–361. “A Hybrid Maximum Power Point Tracking for Partially Shaded Photovoltaic
[4] Malinowski, M., Leon, J., and Abu-Rub, H., 2017, “Solar Photovoltaic and Systems in the Tropics,” Renew. Energy, 76, pp. 53–65.
Thermal Energy Systems: Current Technology and Future Trends,” Proc. IEEE, [27] Khaehintung, N., Sirisuk, P., and Kurutach, W., 2003, “A Novel ANFIS
105(11), pp. 1–15. Controller for Maximum Power Point Tracking in Photovoltaic Systems,” The
[5] Hossain, J., and Mahmud, A., 2014, Renewable Energy Integration: Challenges Fifth International Conference on Power Electronics and Drive Systems
and Solutions, Springer Science & Business Media, New York, p. 1. (PEDS), Singapore, Nov. 17–20, pp. 833–836, Vol. 2.
[6] Rosu-Hamzescu, M., and Oprea, S., 2013, “Practical Guide to Implementing [28] Khosrojerdi, F., Taheri, S., and Cretu, A., 2016, “An Adaptive Neuro-Fuzzy
Solar Panel MPPT Algorithms,” Microchip Technology Inc. Inference System-Based MPPT Controller for Photovoltaic Arrays,” 2016 IEEE
[7] Sharma, D., and Purohit, G., 2012, “Advanced Perturbation and Observation Electrical Power and Energy Conference (EPEC), Ottawa, ON, Canada, Oct.
(P&O) Based Maximum Power Point Tracking (MPPT) of a Solar Photo- 12–14, pp. 1–6.
Voltaic System” 2012 IEEE India International Conference on Power [29] Noman, A. M., Addoweesh, K. E., and Alolah, A. I., 2017, “Simulation and
Electronics (IICPE), Delhi, India, Dec. 6–8. Practical Implementation of ANFIS-Based MPPT Method for PV Applications
[8] Sweidan, T. O., and Widyan, M. S., 2017, “Perturbation and Observation as Using Isolated Ć uk Converter,” Int. J. Photoenergy, 2017, p. 3106734.
MPPT Algorithm Applied on the Transient Analysis of PV-Powered DC Series [30] Sarhan, M. A., Ding, M., Chen, X., and Wu, M., 2017, “Performance Study of
Motor,” 8th International Renewable Energy Congress (IREC), Amman, Neural Network and ANFIS Based MPPT Methods For Grid Connected PV
Jordan, Mar. 21–23, pp. 1–6. System,” Proceedings of the 2017 VI International Conference on Network,
[9] Kamran, M., Mudassar, M., Fazal, M. R., Asghar, M. U., Bilal, M., and Asghar, Communication and Computing, Kunming, China, December.
R., 2018, “Implementation of Improved Perturb & Observe MPPT Technique [31] Duwadi, K.,2015, Design of ANN based MPPT for Solar Panel..
With Confined Search Space for Standalone Photovoltaic System,” J. King [32] Worku, M., and Abido, M., 2016, “Grid Connected PV System Using ANFIS
Saud Univ.—Eng. Sci., 32(1), pp. 432–441. Based MPPT Controller in Real Time,” International Conference on Renewable
[10] Putri, R. I., Wibowo, S., and Rifa’i, M., 2015, “Maximum Power Point Tracking Energies and Power Quality (ICREPQ 16), Madrid, Spain, May 4–6.
for Photovoltaic Using Incremental Conductance Method,” Energy Procedia, 68, [33] Al-Majidi, S. D., Abbod, M. F., and Al-Raweshidy, H. S., 2019, “Design of an
pp. 22–30. Efficient Maximum Power Point Tracker Based on ANFIS Using an
[11] Safari, A., and Mekhilef, S., 2011, “Incremental Conductance MPPT Method for Experimental Photovoltaic System Data,” Electronics, 8(8), p. 858.
PV Systems,” 24th Canadian Conference on Electrical and Computer [34] Hussein, H., Aloui, A., and AlShammari, B., 2018, “ANFIS-Based PI Controller
Engineering(CCECE), Niagara Falls, ON, Canada, May 8–11, pp. 000345– for Maximum Power Point Tracking in PV Systems,” Int. J. Adv. Appl. Sci., 5(2),
000347. pp. 90–96.
[12] Das, P., 2016, “Maximum Power Tracking Based Open Circuit Voltage Method [35] Abido, M., Khalid, M. S., and Worku, M. Y., 2015, “An Efficient ANFIS-Based
for PV System,” Energy Procedia, 90, pp. 2–13. PI Controller for Maximum Power Point Tracking of PV Systems,” Arabian
[13] Ch, S. B., Kumari, J., and Kullayappa, T., 2011, “Design and Analysis of Open J. Sci. Eng., 40(9), pp. 2641–2651.
Circuit Voltage Based Maximum Power Point Tracking for Photovoltaic [36] Al-Hmouz, A., Shen, J., Al-Hmouz, R., and Yan, J., 2011, “Modeling and
System,” Int. J. Adv. Sci. Technol., 2, pp. 51–60. Simulation of an Adaptive Neuro-Fuzzy Inference System (ANFIS) for Mobile
[14] Dzung, P. Q., Le Dinh, K., Hong Hee, L., Le Minh, P., and Nguyen Truong Dan, Learning,” IEEE Trans. Learn. Technol., 5(3), pp. 226–237.
V., 2010, “The New MPPT Algorithm Using ANN-Based PV,” International [37] Hakim, S. J. S., Razak, H. A., Allemang, R., De Clerck, J., Nieziecki, C., and
Forum on Strategic Technology, Ulsan, South Korea, Oct. 13–15, pp. 402–407. Blough, J. R, 2012, Topics in Modal Analysis I, 5th ed., Vol. 5, Springer, New
[15] Mahamudul, H., Saad, M., and Ibrahim Henk, M., 2013, “Photovoltaic System York, pp. 399–405.
Modeling With Fuzzy Logic Based Maximum Power Point Tracking [38] Bhatia, S., 2014, Advanced Renewable Energy Systems (Part 1 and 2), CRC
Algorithm,” Int. J. Photoenergy, 2013, p. 762946. Press, New York.
[16] Liu, Y., Huang, S., Huang, J., and Liang, W., 2012, “A Particle Swarm [39] Bücher, K., 1997, “Site Dependence of the Energy Collection of PV Modules,”
Optimization-Based Maximum Power Point Tracking Algorithm for PV Sol. Energy Mater. Sol. Cells, 47(1–4), pp. 85–94.
Systems Operating Under Partially Shaded Conditions,” IEEE Trans. Energy [40] Kollimalla, S. K., and Mishra, M. K., 2014, “A Novel Adaptive P&O MPPT
Convers., 27(4), pp. 1027–1035. Algorithm Considering Sudden Changes in the Irradiance,” IEEE Trans.
[17] Aldair, A. A., Obed, A. A., and Halihal, A. F., 2018, “Design and Implementation Energy Convers., 29(3), pp. 602–610.
of ANFIS-Reference Model Controller Based MPPT Using FPGA for [41] Nedumgatt, J. J., Jayakrishnan, K. B., Umashankar, S., Vijayakumar, D., and
Photovoltaic System,” Renew. Sustain. Energy Rev., 82, pp. 2202–2217. Kothari, D. P., 2011, “Perturb and Observe MPPT Algorithm for Solar PV
[18] D’Souza, N. S., Lopes, L. A., and Liu, X., 2010, “Comparative Study of Variable Systems-Modeling and Simulation,” 2011 Annual IEEE India Conference,
Size Perturbation and Observation Maximum Power Point Trackers for PV Hyderabad, India, Dec. 16–18, pp. 1–6.
Systems,” Electric Power Syst. Res., 80(3), pp. 296–305.

041002-12 / Vol. 143, AUGUST 2021 Transactions of the ASME

View publication stats