You are on page 1of 22

A Study of Soft Computing Based MPPT

Techniques Used to Achieve Maximum


Power
Abstract: In the previous ten years to increase the efficiency of a photovoltaic (PV) generation
arrangement, an effectual maximum power point tracking (MPPT) procedure is needed. The
major maximum power point (MMPP) cannot be tracked by standard MPPT algorithms when
there is partial shading condition (PSC). A solar panel's power versus voltage output curve
features many minor utmost power points in addition to one major utmost power point (MPPs).
To ensure the tracking of major utmost power point while boosting MPPT's overall efficacy and
performance, soft computing must be integrated into MPPT. Due to the fact that each MPPT
approach has advantages and disadvantages of its own, choosing one is difficult. In order to
evaluate the efficacy of each existing process, this study analyses all implemented ML-
dependent MPPT tactics in terms of convergence speed, steady-state oscillation and efficiency,
steady state error, cost, etc. under PSC conditions.

Key words: PV cell; MPPT; ML; Partial shading Condition; Solar PV array; Soft computing
Technique

Introduction: These days, with the deteriorating amount of non- renewable energy assets, it has
become more important to generate more power from renewable energy resources; one of them
is solar power. It will undoubtedly grow more affordable in the next years and evolve into a
better technology in terms of both cost and usage. Each day more than approx1366W sun energy
incident on the earth. This is a limitless supply of energy that is totally free. The foremost
advantage of solar power compare to other conformist power sources is that the photon light can
be directly transmuted into electrical current with the usage of minutest photovoltaic (PV) solar
cells [1][2]. Since there has an enormous volume of study happening to associate the Sun’s
irradiation progression by emergent PV panel system through the extraordinary translating form.
The prime benefit of solar power is that it is at no cost, accessible to communal community and
accessible in huge amount of contribution as compared to the worth of several remnant gasses
and oils in the earlier decade. Besides it solar power wants ominously lesser man power
expenditures over predictable energy invention technology. The progression of the photovoltaic
production currently has been unpredicted, with numerous achievements in the last several years
in terms of prime solar energy relations. Undoubtedly, solar energy will remain a crucial green
choice in the future. The Indian government made a commitment at the Paris climate summit to
attain 450 GW of renewable energy capability by 2022 due to the advantages of using renewable
energy. By 2020, India will have an installed capacity of around 175 GW, according to the
Ministry of New and Renewable Energy (MNRE), GoI. This comprises 100 Gigawatt of solar
energy, 60.1 Gigawatt of wind energy, 9 Gigawatt of biomass energy, 5 Gigawatt of small
hydropower, and approx 1 Gigawatt of waste-to-electricity as indicated in Fig. 1 [32][33].

Fig 1: Non-conventional energy resources capacity by 2022

The V-I or V-P characteristics graph of PV cell is nonlinear [42][43] and it is varied by the effect
of solar irradiance, temperature and dust effect. Generally it is observed due to impact of Solar
irradiance G PV panel output current will be affected but Vpv will remains almost constant and
due to impact of temperature variation PV panel output will be change but impact of temperature
on voltage is very small [38].The main objective of utmost power point operating is to obtain the
highest power from PV system regardless the environment conditions[44] .Here MPPT
techniques play a vital role that are utilized to harness the extreme power output from PV
system. Numerous studies have been conducted in the earlier time to boost the effectiveness and
power excellency of PV systems [3]. Due to their nonlinear and time-varying I-V and P-V
features with respect to deviation in solar irradiation and temperature, PV arrangements have
squat energy translation efficiency. As a result, the PV arrangements must be run at their utmost
energy point since at the maximum point; a PV panel delivers the greatest electricity and
performs most effectively. In the PV arrangement shown in fig. [2][3], an utmost power point
tracker [3 -5] is typically employed to track the MPP.
Fig 2: P-V curve [2]

Fig 3: V-I curve

1.1 Partial Shading: One PV panel is typically covered by partially hazy skies, soaring animals,
closest trees, and structures. As a result, one or more PV cell receives inadequate solar radiation
which results in decrement in current. Fig. 2 below, the PV characteristic experiences many
power peaks as a consequence of the change in solar irradiation. The PV unit current and PV
energy are directly proportionate to the sun irradiation. Environmental elements like temperature
also may be the resource of shadowing due to having the effect on panel voltage. The ambient
temperature has an inverse relationship with the PV voltage, which is viewed as power loss.
Additionally, as a result of this, the PV unit acts as a load and consumes electricity produced by
other PV unit [39].An incompatible power hammering occurs, which is responsible for load and
battery damage [40].
In order to evade this, a sidestep diode is then added and associated in parallel to every PV unit,
as displayed in Fig. 3 below. Even when a PV unit is shaded, it still enables the sequence current
to flow in that path [39, 40]. On a PV curve, this will result in many power peaks with local
maxima and one maximum point with global maxima, as seen in Fig. 3. PV cells that are shaded
provide lower Ipv due to insufficient sun irradiation [41]. As a result, the shading effect affects
all the photovoltaic cells in a PV panel that are coupled in series, and the current in these cells
will decrease [41]. Therefore, there is a lot of research being done to get the MPP under PSC
[46].

Fig 3 .Considering the effect of Partial shading Condition in P-V Curve

1.2 MPPT and its Need:

The highest power from a PV cell can be harnessed by means of a MPPT controller under the
erratic solar insolation and variant temperature The general figure of PV arrangement consisting
of MPPT controller is given in fig 4.To harness the utmost power from solar arrangement
tracking of maximum position can be done by electrically and mechanically [44].For mechanical
tracking PV panel position will be shifted according to solar irradiance power but this one
method is little bit complicated ,second one is to electrical tracking in which an optimum point of
maximum power will be tracked on VI characteristics by using MPPT algorithm.[42-43].
Fig 4: PV system with MPPT Controller

To make sure that the pv structure always performs at utmost point, MPPT is a technology that is
used. A position on a P-V curvature known as MPP is one where dp/dv= 0. According to the
duty cycle, MPPT algorithms determine the corresponding MPP. In the direction of the MPP, the
duty ratio is either increased or decreased. Impedance matching, or matching the output PV
resistance to the load resistance, is the primary goal. The MPPT is compelled by the impedance
matching to haul out the most power from a PV set-up.

The PV system is severely impacted by abrupt variations in temperature and solar insolation. As
a result, the PV structure performs at a relatively low intensity level. Due to this battery and the
load associated to the PV system may be debilitated. Despite harnessing utmost power from a PV
cell, the MPPT approach is essential for safeguarding the battery against deep discharge and
overcharging. Additionally, it increases the MPP's tracking speed and lowers the PSC's power
loss, reduces power losses caused by deceptive mismatches and quickly varying insolation
levels.

MPPT approaches have the appropriate controllers to draw the most of the possible power from
PV structure. To operate PV modules at their highest possible power, several MPPT approaches
are employed. However, the effectiveness of the specific approach depends on its capacity to
track in rapidly varying meteorological conditions.
1.3 Novelty of Work: The mainly severe ecological changes, such as varying irradiance and
PSCs, were considered in terms of algorithms. Complicated problems require more time to give
the solutions. The installation cost will be noticeably greater and tracking performance will be
hampered due to the increased computing task. The conventional methods are the simplest and
quickest ways to obtain MPP when compared to alternative approaches. It has a number of
benefits and a few drawbacks, including:

● MPP oscillation could not be avoided.


● Low efficiency when there is less solar irradiation.
● Rapid irradiance increases cause the algorithm to lose sight of its direction of tracking.
As a result, the MPP tracking will be lost, and the entire process will be unsuccessful.

A literature-based research framework is being built for the PV structure in order to solve these
problems and do detailed monitoring contrasts in terms of tracking duration, tracking efficacy,
and other parameters, as well as a comparison also done to authenticate the efficacy of the
discussed algorithm.

This review paper considers the maximum paper on ML implemented MPPT techniques from
2017-2022.As far as research is concern till now there are many papers on review of AI
dependent solar energy forecasting but not cover the study of ML based maximum PPT
techniques The performance of various approaches is evaluated in a comparable way in tabular
way at the conclusion of each category's discussion, which will help us to select the most useful
option.
This paper gives a survey of different utilized ML implemented MPPT tactics in the previous
research. In this paper, favorable situations and short comings of the conversed methods are
collected, and a general correlation table for every subject of the utilized tactics has been
disscussed. The paper is structured into five segments: Segment II discuss about the MPPT
techniques, Segment III details the PV cell Modeling Segment IV discusses the literature review
of ML algorithm implemented MPPT techniques, Segment V briefs about the comparative
analysis and then conclusion of the paper

2. MPPT Techniques: The hard aspect of this solar power is its dynamic nature, which can
produce variable power and voltage based on the surrounding environments. The characteristics
of solar cell change by elements like wind speed, shade, and sun insolation angle. As a
consequence, for all electrical loads, the generation of maximum power is not assured [46], [47].
MPPT approaches have the appropriate controllers to draw the most power possible from PV
systems. To operate PV unit at their highest possible power, a number of MPPT approaches are
employed. However, the effectiveness of the specific approach depends on its capacity to track in
rapidly changing meteorological circumstances.So, depending on the tracking procedure these
strategies are categorized in fig 4.
Fig3: MPPT Techniques Categorization

The Conventional MPPT procedures [25][26] are open circuit potential, short circuit current
constant voltage, P&O method[5], Hill climbing and Inc Conductance method. These techniques
are simple to design, system have low cost, simple implementation of algorithm but offers high
oscillation around maximum power point and, high steady state error, more tracking time .P&O
method offers high tracking efficiency but shows oscillation around MPP and unable to follow
the utmost power point under erratic environment situation, this problem overcome by Inc-
conductance method, but this algorithm does not work efficiently under PSC.

FLC, Gauss Newton tactic (GNT), ANN, Fabinniocio series dependent MPPT and ML
dependent MPPT [25-27] are Artificial Intelligence dependent MPPT. These all tactics are
capable to work under partial, shading condition and offers high tracking speed but offers high
designing complexity and computational complexity.FLC technique can be implemented without
system knowledge but offers low tracking speed as compare to ANN .ANN require huge amount
of accurate data to train but offers high tracking speed with better tracking efficiency. Fibonacci
and GNT method can track the MPP fastly due to having the capability of updating the searching
range.SMC technology offers easy implementation.ML is the new growing field of AI which
have the capability of self learning and need not to train new data again and again.

MPP can be found by using Optimization-based techniques [25-28], CS, GWO, ABC, ACO and
PSO. A true MPP can be obtained by using PSO technique with very low steady state oscillation.
CS is an optimization technique that draws inspiration from cuckoo birds' parasitic reproduction
method. GWO do the optimization without the knowledge of system.ACO and ABC technique
does not require the sensor to perform the data optimization process.
Using traditional HC utmost point tracking techniques like P&O and IC, the utmost power under
uniform circumstances without PSC may be monitored successfully [6]. Table 1 compares and
contrasts traditional MPPT with ML-based MPPT. The HC MPPT approach to find the true
maximum is complicated by the fact that the power yield from the solar power arrangement
creates numerous utmost points under PSC, comprising one major MPP (GMPP) and many
additional minor peaks as shown in Fig. 3 [7]. As a result, MPPT develops into an algorithm
depend on experimental, meta-heuristic, and evolutionary methodologies. Since traditional HC
MPPT approaches cannot monitor global peaks under PSC and rapidly changing solar
irradiation, it is intended to track global peaks rather than local peaks [8].

Table 1

Comparative analysis of conventional and ML based MPPT:

Parameter Conventional MPPT Technique ML based MPPT

Tracking accuracy Low High

Tracking speed Moderate High

Convergence speed slow Fast

Ability to track under PSC No Yes

Steady-state oscillation Yes Less

Oscillation around MPP Yes Less

Settling time High Less

Complexity Low High

Periodic tuning Yes No

Cost Low High

Computation time Less High

Algorithm complexity Low High

System Design Simple Complex

3. PV Cell Modeling:
Lorenzo in 1994 [4] discussed the equivalent model of PV cell. This type comprises of a parallel
diode connected to a current source. The internal losses of the solar cell brought on by the
current flow are referred to as series resistance (Rs) in this context, and they should be kept to a
minimum for optimal power output. Because the current value is too low and the resistance value
is too high, the (Rp) resistance linked in parallel to the diode refers to the losses brought on by
leakage currents to the ground. Fig. 4's[34] shows the equivalent circuit of PV cell and below
are the equations representing the PV cell's current and voltage. Here, the parameters that change
depending on the properties of the PV cell are the parallel resistance (Rp), ideality factor (n), and
series resistance (Rs). Boltzmann constant (k), electron charge (q), and solar cell temperature are
additional variables (T). IPH and IS stand for the photon current and saturation current,
respectively. [34-37]

Fig 4 [37].PV cell equivalent Model

Ipv=Ipv–Is¿……. (1)
nkt
V= q
[ ln(( Iph+ Is−Ipv)/ Is) – IpvRs ]……………………… (2)

There is a point on PV characteristics at which maximum power will be transferred to the load
and coordinate of this point is (Vmpp, Impp).The optimum value of load resistance at which
maximum power will be obtained is Rapt. The maximum power Pmax can be written as follows:

Pmax=Vmpp*Impp…………………….(3)
By aligning the I-V operating point with the load parameters, MPPT aims to maximize PV panel
output under all circumstances [5]. By controlling the PV panel current or voltage to make the
converter run at the MPP, the provided power can be increased.

4. Literature Review on ML based MPPT


As the most important subset of AI, ML algorithms are designed to study the behavior in data
and are often used to implement certain learning abilities and logical reasoning without using
explicit instructions [8]. In the past few decades, ML algorithms have been proven to be useful in
the fields of renewable energy resources]. Nowadays, the most demanding area of research in
solar energy is solar power forecasting and tracking of maximum power point on P-V cell V-I
characteristics.

The performance parameter which are used to measure the ability of MPP algorithm are the
Mean Error (ME in Equation (2)) , the Mean Square Errors in Equation (3)) , the Standard
Deviation error (σ in Equation (4)) , the Root Mean Square Error (RMSE in Equation (5), means
absolute error (MAE in, Equation (6)), the overall power tracking efficiency (η in Equation (7))
and a quality indicator that provides information about the ability of the ANN to predict the MPP
in Equation (8):
N
1
M=
N
∑ (PowerMPP −PowerPV )……………… (3)
K=1

N
1
MSE=
N
∑ (Po werMPP−PowerPV )2………… (4)
K =1

N
1
σ =√
N
∑ ( PowerMPP¿−μ)2 ¿ ………………… (5)
K=1

N
1
MAE=
N
∑ ¿ PowerMPP−PowerPV ∨¿ ¿…………… (6)
K =1

η=¿ ..(7)
N
1
Q=1−
N
∑ PowerPv /PowerMPP …………………(8)
K =1

N
1 ¿
E=
N
∑ ¿ PowerMPP−PowerPV ∨ PowerMPP ¿…… (9)
K =1

The main goal of this review paper is to give the review of ML implemented MPPT technique in
PV systems to yield optimal power from a PV unit under PSC.

F. Keyrouz et al. [11] devised a Bayesian network-based approach that maximises the solar
system's overall output power while minimising steady-state oscillations and monitoring time. It
makes an effort to lessen the discrepancy between the PV slope and an ideal reference value, and
it controls the on off of a boost DC-DC converter using a PWM generator powered by duty
cycles. The findings demonstrate that the suggested technique has a tracking average of 1.76
seconds overall, compared to 2.57 seconds for WODE and 4.28 seconds for GWO. The
suggested approach was found to have an average overall efficiency of 97.89% compared to
WODE's 95.84% and GWO's 91.24%.

M.K Behra et al [12] proposed a real time maximum power forecasting model in SLFN
trained by modified extreme learning machine (ELM) technique, whose weights are updated by
PSO technique and their performance matrices are compared with accessible models like back
propagation (BP) forecasting model. For 15 min time duration RMSE, MAPE (%), MAE for
ELM 0.0362, 2.9417 and 0.0294 and for BP ANN it is 0.0472 3.62 and 0.0363.

K S Tey [13] et al proposed a SEPIC and result simulate on MATLAB. Results show that
the suggested approach is 99% accurate at tracking the GMPP within 2 s and 0.1 s at responding
to load variations.

Du Yan et al [14] proposed a localized MPPT algorithm which avoids the periodic
training of parameters by using supervised machine learning algorithms that support vector
machine (SVM) and extreme learning machine (ELM). The simulation results demonstrate the
suggested method's superiority over the conventional MPPT design. Simulation is done on
Matlab/Simulink and MPP ratio is 99.5%.

Takuri et al [15] also used SVM to foresee the ideal reference voltage of a PV set-up under
variable solar irradiance temperature and load condition. A proportional integral derivative
controller will then utilise this reference voltage as a reference. The Generalized regression
neural network (GRNN) and SVM are two machine learning methods, and their effects on the
system's performance are contrasted. SVM algorithm offers RMSE in Vref(V) is 0.0023 and
RMSE in Pmax (W) is 0.0278.

By conducting three trials in a variety of meteorological circumstances, Nakumble et al.


[16] demonstrate Nine ML-based MPPT approaches. The performances of the DT, Multivariate
Linear Regression (MLR), Gaussian Process Regression (GPR), Weighted K-Nearest Neighbors
(WK-NN), Linear Discriminant Analysis (LDA), Bagged Tree (BT), Naive Bayes classifier
(NBC), SVM, and Recurrent Neural Network (RNN) are validated.Simulation is done on
Matlab/Simulink and according to his research WK-NN algorithm shows best performance in all
sense.

Mounil et. all [17] proposed ML based multivariate linear regression algorithm in pre
existing P&O method and increase the efficiency of hybrid model up to 99.8% and decrease the
convergence time to achieve the Maximum power point as compare to conventional method.
Author considered three input parameters to train the models that was solar irradiance,
temperature and Humidity and used the python for simulation therefore the number of iteration
was 1000 corresponding to 83 hour of training of ML model which is quite long time period to
train the model.
Kah Yung Yap et al [18] proposed, a in depth comparison of categorization and
performance between six major AI-based MPPT techniques that are artificial intelligence (AI),
fuzzy logic control (FLC), artificial neural network (ANN), genetic algorithm (GA), swarm
intelligence (SI), machine learning (ML).

Debottam Mukherjee et all [19] proposed the ML based PV power production forecasting
inclusive and exclusive of MPPT controllers. The data set different machine learning techniques
are applied.Coarse Tree, with an RMSE of 1.675, is the best model for forecasting solar power
generation with an MPPT controller, whereas Rational Quadratic Gaussian Process Regression
(RQGPR), with an RMSE of 1.628, is the best model for forecasting solar power generation
without an MPPT controller.

In order to follow the utmost power point (MPP) of solar cell unit, Ruhi et al[20]
projected a supervised machine learning algorithm dependent on a Neural network to find out the
MPP for smaller data, supporting the diverse train data set. This technique uses Bayesian
regularization method to train the model which is best for a smaller data set.

Christos Kalogerakis et all [21] proposed Q -learning based, a global maximum power
point tracking (GMPPT) technique under PSC.This technique offers searching time of GMPP
under various shading patterns reduced by 80.5-98.3% but offers less MPPT efficiency as
compared to the PSO algorithm. The convergence time and quantity of hunting steps is more for
the initial partial shading pattern but it starts to reduce after learning from different patterns.

For optimization and to increase MPPT efficiency in photovoltaic systems, M. Farayola et


al. [22] proposed the use of a linear regression machine learning algorithm with interactions
(LIR) technique. The results were compared to those of conventional ANFIS, bootstrap
aggregation (bagged), and boosted tree ensemble regression as bagged-ANFIS and boosted-
ANFIS results under various weather conditions. The RMSE for 92 samples using the ANFIS,
LIR, Bagged, and Boosted Tree approaches are 1.2324e-5, 5.5339e-7, 3.3703e-1, and 9.1700e-1,
respectively. PV efficiency for ANFIS, Bagged, Boosted and LIR under normal operating
condition is 73.28%, 72.39%, 71.87% and 73.24%, under PVUSA test condition are 102.13%
97.21% 100.4% 102.18% and under standard test condition are 100.12% 95.94% 98.43%
100.16%.

Mahesh et al [23] presents a Decision Tree regression ML algorithm to calculate the


MPPT for an isolated PV system. The simulation result is compared with β-MPPT, CS and ANN
results and result shows efficiency more than 93.99% under static solar irradiance and
temperature condition with a quick output in 0.16 sec and the system settles in 0.27 sec. The
response and settlement times for the -MPPT, CS, and PANN techniques are 0.06, 0.21, and 0.16
seconds, respectively.

Rafeeq et all [24] present a deep learning model by utilizing back propagation neural
network (BPNN) that helps to achieve the utmost power point. The authors used the MATLAB
NN-Tool to construct a Levenberg-Marquardt back propagation method to record the utmost
highest point. The proposed system is capable of achieving 98% efficiency and MSE is 26.6393
at epoch 9.

Section V

Comparative Analysis of ML Based MPPT technique:

MPPT Technique Hardware/Software DC-DC Performance Parameter Result


Simulation Converter

Bayesian Neural Matlab/Simulink PID with Tracking Average Time, Tracking average Time=1.76s
Network[11] Boost Efficiency Efficiency=97.89%

SLFFN with ELM Matlab/Simulink Boost RMSE,MAPE,MAE 15 min time duration RMSE, MAPE
Algorithm[12] (%) ,MAE for MELM are 1.4440
0.0144 0.0178

Differential Evolution PowerSim SEPIC Accuracy in tracking ,Tracking GMPP in 2 seconds with 99%
based MPPT[13] PV-UE125MF5N Time accuracy and react to load changes
in 0.1 seconds.

SVM and ELM[14] Matlab/Simulink Boost MPP Ratio ELM accuracy=94.52%


SPR-305E-WHT-D
SVM accuracy=92.33%

Support Vector Machine Matlab/Simulink PID RMSE SVR algorithm offers RMSE in
learning[15] Mitsubishi TD185MF5 Controller Vref (V) is 0.0023 and RMSE in
panel with Boost Pmax (W) is 0.0278.
9 ML based MPPT Matlab/Simulink Boost RMSE and R Squared Error RMSE:
techniques[16] Soltech 1STH-215-P DT = 0.42
WK-NN = 0.37
MLR = 0.44
LDA = 0.48
BT = 0.73
GPR = 0.4
NBC = 0.51
SVM = 0.14
RNN = 0.36
And
Train duration:
DT = 0.91 s
WK-NN = 0.78 s
MLR = 6.17 s
LDA = 2.32 s
BT = 2.35 s
GPR = 5.04 s
NBC = 8.56 s
SVM = 1.1178 s
RNN = 8.9 s

Multivariate Linear Python Buck Efficiency and Error threshold Efficiency=


Regression model with
P&O method[17] 99.8%

Error
Threshold=0.5
%

AI,FLC,ANN,SI,ML, Matlab/Simulink Boost Tracking time ,Steady state Tracking time 0.60s
GA based controller (SPR-305E-WHT-D) oscillation
[18] Steady-state
oscillation (%)
±1.5

Coarse Tree with Matlab/Simulink Boost RMSE Shows the RMSE


MPPT Controller 1.675 for coarse tree
&RQGPR without method and 1.628
MPP controller[19] for RQGPR
algorithm

Bayesian Matlab/Simulink Boost Efficiency, Efficiency=99.794%


Regularization NN ZM-A-M-100 MSE
with P&O method[20] MSE= 2.87*10^-3

Q learning Matlab/Simulink Boost Convergence Time ,Efficiency Convergence


based time
algorithm 80.5-90.3%
[21] reduced,
Efficiency=99.3-99.6%

LIR Method[22] Psim, 1STH-215-P module Cuk RMSE,PV efficiency RMSE 5.5339e-7 and PV efficiency
converter 73.24%,102.18% and 100.16%
under NOCT,PST and STC.

DT Regression Matlab/Simulink Boost Efficiency, Settling Time efficiency>93.99%


MPPT[23] Settling time 0.27 sec and rise time
0.16sec

BPNN-DL[24] Matlab/Simulink Boost Accuracy Accuracy 98%


Table:2

Comparative analysis based on Performance parameter:

MPPT Track Computation Tracking Ability to track Input Output Grid Cost Periodic
Technique Speed al Accurac under PSC Parameter parameter Suppor Tuning
Complexity y ted

Bayesian Neural Fast Less High Yes Voltage, Pmax Yes


Network[11] Current

SLFFN with - High High Yes Voltage Pmax Yes High


MELM Current
Algorithm[12] High

Differential Fast High High Yes Voltage,Curr ΔD ,ΔP Low Moderate


Evolution based ent
MPPT[13]

SVM and Fast High High Yes Voltage Optimal Yes Low No
ELM[14] Step Size

Support Vector fast Less High Yes Voltage/ Vref High


Machine Current
learning[15] No

9 ML based WK-NN is Less SVM and Yes except RNN Voltage and PV Low
MPPT faster WK-NN algorithm based current efficiency No
techniques[16] Tracking MPPT
accuracy
> 97%

Multivariate Slow High High Yes Temperature, Pref Yes Mod NO


Linear Humidity, erate
Regression model Solar
with irradiance
P&O method[17]

AI,FLC,ANN,SI, Moderate High High Yes Voltage and Pmax yes NO


ML,GA based current
controller
[18]

Coarse Tree with Fast high high Yes solar Vmax and yes No
MPPT Controller insolation, Imax
& RQGPR PV panel
without temperature
MPP and ambient
controller[19] temperature
Bayesian Less high Not consider Temperature Impp Yes No
Regularization and Solar
NN Irradiance,Pa
with P&O nel current
method[20]

Q learning Fast Less yes Vpv Duty cycle yes No


based
algorithm
[21]

LIR Method[22] Fast High High Yes Solar Imax No


Irradiance
and
Temperature

DT Regression Fast High Moderate Yes Current Duty cycle No


MPPT[23] and PM

BPNN-DL[24] Fast High high yes ΔT,ΔG Vref Yes No

Table 3:

Comparative analysis based on Pros and Cons:

MPPT Technique Findings Merits Demerits

Bayesian Neural Network[11] The controller maintains best Convergence speed for the Temperature variation is not
possible features to diminish the GMPPT is increased. considered.
steady state oscillations and rise
times of the output power with
the help of Bayesian fusion.

SLFFN with MELM The implementation of the Foresee accuracy increased with Short term maximum solar power
Algorithm[12] modified ELM algorithm, whose less mean square error. can be tracked.
weights are updated using
various particle swarm Computational complexity is
optimization (PSO) methods and high.
their performance matrices
compared with BP model. Small data set is considered.

Differential Evolution based Ability to follow GMPP and react Less parameters so easy Simulation and Experiment
MPPT[13] more quickly to changes in load; regulation is possible. performed for constant
when an optimization technique Faster Response against load temperature.
is used, it may look for the variation.
GMPP across a wider operating Algorithm is free from initial
zone by employing a single point dependency to track
ended primary-inductor maximum power point.
converter. Avoid the repeated tracking of
solution for same particle

SVM and ELM[14] A localized MPPT algorithm An automatic location Decision of optimal step size
proposed for different weather classification system designed to depends on the degree of
conditions and results is harness the maximum power. fluctuation of solar irradiance.
compared with conventional
MPPT algorithm.
Support Vector Machine Method is used to predict an ideal The robustness of the system is An accurate data acquisition
learning[15] reference voltage in all offered by using a PID controller. system required high quality
conditions and the result is sensors.
compared with GRNN
implemented networks.

9 ML based MPPT DT, MLR, GPR, WK-NN, Fine Tuning of error is obtained Small data set is considered.
techniques[16] LDA,BT, NBC, SVM and RNN by using PID controller.
are compared in different shading No need of sensor
condition and WK-NN perform
best.

Multivariate Linear Hybrid Technique offers better The proposed model overcomes Training time is very high,
Regression model with efficiency and lesser time under over fitting. approximately 83 hours.
P&O method[17] variable solar irradiance Efficacy and the Accuracy of the
condition. proposed strategy is not affected
by the modification in input data
or small fluctuations

AI,FLC,ANN,SI,ML,GA based Comparative analysis is A comprehensive comparison of Design Complexity and


controller done ,good convergence but popular AI-dependent MPPT computational complexity is very
[18] costly techniques and large data techniques for the solar power high.
set required system is given.

Coarse Tree with Compare the RMSE of proposed Ml algorithm shows the better Unreal data of solar irradiance,
MPPT Controller algorithm with prediction result of maximum PV panel temperature and
& RQGPR without Bagged Tree ,Mattern 5/2 power with the MPPT controller. ambient temperature is used.
MPP controller[19] GPR and shows less RMSE

Bayesian Best suited for smaller data and Data set proposed which is free A theoretical result has been
Regularization NN shows improved efficiency, from over fitting problems. proposed by simulation.
with P&O method[20] reduce misjudgment and avoid
power loss at MPP as compare to
P&O method

Q learning Compared the result with It does not need the information MPPT efficiency is less than the
based PSO techniques and shows the of the operational characteristics PSO MPPT efficiency.
algorithm Lesser convergence time. of the PV units.
[21] Number of search steps reduced
after shading pattern learning.

LIR Method[22] Compared the result with ANFIS, Technique offers better Large and precise training data
Bagged tree and boost technique maximum efficiency under sets are required.
and showed the better result of diverse weather conditions.
LIR technique in term of
efficiency and RMSE

DT Regression MPPT[23] Compared the result with β Tracking efficiency >93.93% Partial Shading effect is not
MPPT, CS and ANN. Tracking time= 0.16 sec. considered.

BPNN-DL[24] Prediction of optimum reference when the solar panels are linked Cost of the proposed system is
voltage under variable load and to the boost converter under high.
weather conditions. variable load situations, it
maximizes the output power from
the solar grids.
Conclusion: One of the current major issues in the field of renewable energy research is
maximizing power extraction from solar PV systems to increase efficiency. In this regard, it is
determined that the MPPT controller idea is important since it maximizes the output power
supplied by the solar PV module. Many studies offering the thorough study of various MPPT
approaches have previously been published. In order to maximize the potential of ML to capture
the most solar energy, scientists and engineers may use this review to study the features of
various solar MPPT approaches and identify whether ML can enhance their results..In summary,
how to find out the optimum MPPT in the solar power model is one of the main research
directions in the future.

Reference:

1. Rabaia MKH, Abdelkareem MA, Sayed ET, Elsaid K, Chae KJ, Wilberforce T, et al.
Environmental impacts of solar energy systems: A review. Sci Total Environ
2021;754:141989. https://doi. org/10.1016/j.scitotenv.2020.141989.
2. Motahir S, El Hammoumi A, El Ghzizal A. The most used MPPT algorithms: Review and
the suitable low-cost embedded board for each algorithm. J Clean Prod 2020;246:118983.
https://doi. org/10.1016/j.jclepro.2019.118983.
3. Kumar M, Ban DK, Kim J. Photo-induced pyroelectric spikes for neuromorphic sensors.
Mater Lett 2018;225:46–9. https://doi. org/10.1016/j.matlet.2018.04.106.
4. Lorenzo E. Solar electricity: engineering of photovoltaic systems. Earthscan/James &
James; 1994.
5. Kumar N, Hussain I, Singh B, Panigrahi BK. Framework of Maximum Power Extraction
from Solar PV Panel Using Self Predictive Perturb and Observe Algorithm. IEEE Trans
Sustain Energy 2018;9:895–903. https://doi.org/10.1109/ TSTE.2017.2764266.
6. A. Pandey, N. Dasgupta, and A. K. Mukerjee, “A simple single-sensor MPPT solution,”
IEEE Transactions on Power Electronics, vol. 22, no. 2, pp. 698–700, Mar. 2007.
7. H. Q. Wang, L. Vinayagam, H. Jiang, Z. Q. Cai, and H. Q. Li, “New MPPT solar
generation implemented with constant-voltage constant current DC/DC converter,” in
2016 51st International Universities Power Engineering Conference (UPEC), Coimbra,
2016, pp. 1–6.
8. Yagli Gokhan Mert, Yang Dazhi, Srinivasan Dipti. Automatic hourly solar forecasting
using machine learning models. Renew Sustain Energy Rev 2019;105:487–98.
9. Ahmad Tanveer, Chen Huanxin. A review on machine learning forecasting growth trends
and their real-time applications in different energy systems. Sustain Cities Soc
2020;54:102010.
10. Fouilloy Alexis, Voyant Cyril, Notton Gilles. Solar irradiation prediction with machine
learning: forecasting models selection method depending on weather variability. Energy
2018;165:620–9.
11. F. Keyrouz, “Enhanced Bayesian based MPPT controller for PV systems,” IEEE Power
and Energy Technology Systems Journal, vol. 5,no. 1, pp. 11-17, Mar. 2018.
12. M. K. Behera, I. Majumder, and N. Nayak, “Solar photovoltaic power forecasting using
optimized modified extreme learning machine technique,” Engineering Science and
Technology, vol. 21, pp. 428-438,Jun. 2018.
13. K. S. Tey, S. Mekhilef, M. Seyedmahmoudian et al., “Improved differential evolution-
based MPPT algorithm using SEPIC for PV systems under partial shading conditions and
load variation,” IEEE Transactions on Industrial Informatics, vol. 14, no. 10, pp. 4322-
4333, Jan. 2018.
14. Du, Y.; Yan, K.; Ren, Z.; Xiao, W. Designing localized MPPT for PV systems using
fuzzy-weighted extreme learning machine.Energies 2018, 11, 2615.
15. Takruri, M.; Farhat, M.; Barambones, O.; Ramos-Hernanz, J.A.; Turkieh, M.J.; Badawi,
M.; AlZoubi, H.; Sakur, M.A. Maximum power point tracking of PV system based on
machine learning. Energies 2020, 13, 692. [CrossRef]
16. Nkambule, M.S.; Hasan, A.N.; Ali, A.; Hong, J.; Geem, Z.W. Comprehensive Evaluation
of Machine Learning MPPT Algorithms for a PV System Under Different Weather
Conditions. J. Electr. Eng. Technol. 2020, 16, 411–427.
17. Mounil Memayaa, C. Balakrishna Moorthyb, Sahitya Tahilianic, Siddarth
Sreenid,”Machine learning based maximum power point tracking in solar energy
conversion systems”, International Journal of Smart Grid and Clean Energy”, doi:
10.12720/sgce.8.6.662-669.
18. Kah Yung Yap, Charles R. Sarimuthu, and Joanne Mun-Yee Lim,” Artificial Intelligence
Based MPPT Techniques for Solar Power System: A review”, JOURNAL OF MODERN
POWER SYSTEMS AND CLEAN ENERGY, VOL. 8, NO. 6, November 2020.
19. Debottam Mukherjee,Samarat Chakarabotry,Pabitra Kumar ,” Machine Learning based
Solar Power Generation Forecasting with and without MPPT Controller”,”IEEE
International Conference for convergence in Engineering 2020”.
20. Ruhi Sharmin, Sayeed Shafayet Chowdhury, Farihal Abedin, and Kazi Mujibur Rahman,”
Implementation of MPPT Technique of Solar Module with Supervised Machine
Learning”,Oct 2021, https://doi.org/10.48550/arXiv.2110.00728.
21. Christos Kalogerakis , Eftichis Koutroulis and Michail G. Lagoudakis,” Global MPPT
Based on Machine-Learning for PV Arrays Operating under Partial Shading Conditions”,
Appl. Sci. 2020, 10, 700; doi:10.3390/app10020700.
22. Adedayo Farayola,Ali N Hasan,Ahmad Ali,Optimization Of Pv Systems Using Linear
Interactions Regression And Anfis Mppt Techniques”,2018 IEEE PES/IAS
PowerAfrica,DOI: 10.1109/PowerAfrica.2018.8521064.
23. P. Venkata Mahesh,S. Meyyappan and RamaKoteswara Rao Alla ,”Maximum power
point tracking using decision-tree machine-learning algorithm for photovoltaic
systems”,Clean Energy, 2022, Vol. 6, No. 5, 762–775,Advance access publication 28
October 2022 Research Article //doi.org/10.1093/ce/zkac057
24. Rafeeq Ahmed K, Sayeed F, Logavani K, et al. Maximum power point tracking of PV
grids using deep learning. International Journal of Photoenergy, 2022, 2022:11123251–
11123257.
25. R. Ahmad, A. F. Murtaza, and H. A. Sher, “Power tracking techniques for efficient
operation of photovoltaic array in solar applications- A review,” Renewable and
Sustainable Energy Reviews, vol. 101, pp. 82–102, Mar. 2019.
26. M. A. G. de Brito, L. Galotto, L. P. Sampaio, G. D. A. e Melo, and C.A. Canesin,
“Evaluation of the main MPPT techniques for photovoltaic applications,” IEEE
Transactions on Industrial Electronics, vol. 60, no.3, pp. 1156–1167, Mar. 2013.
27. M. Seyedmahmoudian, B. Horan, T. K. Soon, R. Rahmani, A. M. T. OoS. Mekhilef, and
A. Stojcevski, “State of the art artificial intelligence-based MPPT techniques for
mitigating partial shading effects on PV systems-A review,” Renewable and Sustainable
Energy Reviews, vol.64, pp. 435–455, Oct. 2016.
28. H. Islam, S. Mekhilef, N. B. M. Shah, T. K. Soon, M. Seyedmah-mousian, B. Horan, and
A. Stojcevski, “Performance evaluation of maximum power point tracking approaches and
photovoltaic systems,”Energies, vol. 11, no. 2, pp. 365, Feb. 2018.
29. Hill, J.S.: ‘ Global solar market demand expected to reach 100 Gigawatts in 2017, says
solarpower Europe’, 2017. Available at https://cleantechnica.com/2017/10/27/global-
solar-market-demand-expected-reach-100-gw-2017-solarpower-europe/, accessed 05
November 2017.
30. International Energy Agency: ‘ Snapshot of global photovoltaic markets 2017’, 2017.
Available at http://www.iea-pvps.org/fileadmin/dam/public/report/statistics/IEA-PVPS-A
Snapshot of Global P V-1992-2016-1.pdf, accessed 05 November 2017.
31. Kamarzaman, N.A., Tan, C.W.: ‘A comprehensive review of maximum power point
tracking algorithms for photovoltaic systems’, Renew. Sustain. Energy Rev., 2014, 37, pp.
585– 598.
32. Takun, P., Kaitwanidvilai, S., Jettanasen, C.: ‘ Maximum power point tracking using
fuzzy logic control for photovoltaic systems’. Proc. International Multiconference of
Engineers and Computer Scientists (IMECS), Hong Kong, 2011, vol. 2, pp. 1– 5.
33. Asim, N., Sopian, K., Ahmadi, S., et al: ‘A review on the role of materials science in solar
cells’, Renew. Sustain. Energy Rev., 2012, 16, (8), pp. 5834– 5847.
34. Kamarzaman, N.A., Tan, C.W.: ‘A comprehensive review of maximum power point
tracking algorithms for photovoltaic systems’, Renew. Sustain. Energy Rev., 2014, 37, pp.
585–598.
35. Bendib, B., Belmili, H., Krim, F.: ‘A survey of the most used MPPT methods:
conventional and advanced algorithms applied for photovoltaic systems’, Renew. Sustain.
Energy Rev., 2015, 45, pp. 637–648.
36. Verma, D., Nema, S., Shandilya, A.M., et al: ‘Maximum power point tracking (MPPT)
techniques: recapitulation in solar photovoltaic systems’, Renew. Sustain. Energy Rev.,
2016, 54, pp. 1018–1034.
37. Tsang, K.M., Chan, W.L.: ‘Maximum power point tracking for PV systems under partial
shading conditions using current sweeping’, Energy Convers. Manage., 2015, 93, pp.
249–258.
38. Jordehi, A.R.: ‘Maximum power point tracking in photovoltaic (PV) systems: a review of
different approaches’, Renew. Sustain. Energy Rev., 2106, 65, pp. 1127–1138
39. Jun, P., Feng, W.C.H.: ‘Research of photovoltaic charging system with maximum power
point tracking’. Proc. Ninth Int. Conf. of Electronic Instruments, Beijing, China,
September 2009, pp. 478–481
40. Toshihiko, N., Shigenori, T., Ryo, N.: ‘Short-current pulse based maximum-power-point
tracking method for multiple photovoltaic and-converter module system’, IEEE Trans.
Ind. Electron., 2002, 49, (1), pp. 217–223
41. A. K. Podder, N. K. Roy, and H. R. Pota, “MPPT methods for solar PV systems: a critical
review based on tracking nature,” IET Renewable Power Generation, vol. 13, no. 10, pp.
1615-1632, Jul. 2019.

42. H. D. Maheshappa, J. Nagaraju and M. V. K. Murthy, "An improved maximum power


point tracker using a step-up converter with current locked loop", Renew. Energy, vol. 13,
no. 2, pp. 195-201, Feb. 1998.

43. T. Radjai, L. Rahmani, S. Mekhilef and J. P. Gaubert, "Implementation of a modified


incremental conductance MPPT algorithm with direct control based on a fuzzy duty cycle
change estimator using d-SPACE", Sol. Energy, vol. 110, pp. 325-337, Dec. 2014.

44. J. A. Roger, "Theory of the direct coupling between d. C. Motors and photovoltaic solar
arrays", Sol. Energy, vol. 23, no. 3, pp. 193-198, 1979.

45. Liu Y-H, Chen J-H, Huang J-W (2015) A review of maximum power point tracking
techniques for use in partially shaded conditions. Renew Sustain Energy Rev 41:436–453

46. D. Verma, S. Nema, A. M. Shandilya, and S. K. Dash, “Maximum power point tracking
(MPPT) techniques: Recapitulation in solar photovoltaic systems,” Renewable and
Sustainable Energy Reviews, vol.54, pp. 1018–1034, Feb. 2016.
47. M. A. S. Masoum, H. Dehbonei, and E. F. Fuchs, “Theoretical and experimental analyses
of photovoltaic systems with voltage and current-based maximum power-point tracking,”
IEEE Transactions on Energy Conversion, vol. 17, no. 4, pp. 514–522, Dec. 2002.

You might also like