Sustainable Energy Technologies and Assessments 57 (2023) 103243

Contents lists available at ScienceDirect

Sustainable Energy Technologies and Assessments

journal homepage: www.elsevier.com/locate/seta

Power grid based renewable energy analysis by photovoltaic cell machine

learning architecture in wind energy hybridization
U. Sakthi a, T. Anil Kumar b, Kuraluka Vimala Kumar c, Shamimul Qamar d, *,
Girish Kumar Sharma e, Abdul Azeem f
a
Department of Computational Intelligence, School of Computing, SRM Institute of Science and Technology, Kattankulathur, 603203, Chengalpattu, Chennai, TN, India
b
Department of Electrical and Electronics Engineering, Anurag University, Hyderabad, Telangana, India
c
Department of Electrical and Electronics Engineering, JNTUK University College of Engineering Narasaraopet, Andhra Pradesh, India
d
Computer Science & Engineering Department, College of Sciences & Arts, Dhahran Al Janoub Campus 64261 King Khalid University, Abha, Saudi Arabia
e
Department of Computer Applications, Bhai Parmanand DSEU Shakarpur Campus-II (A Constituent Campus of Delhi Skill and Entrepreneurship University) GNCT of
Delhi, Delhi 110092, India
f
Department of Electrical Engineering, Jamia Millia Islamia, Delhi 110025, India

A R T I C L E I N F O A B S T R A C T

Keywords: Use of solar energy systems and related green energy technology has spread around the world. When compared
Energy systems to conventional energy sources, solar energy is still not a frequently used energy source due to the comparatively
Power consumption high installation prices, low conversion rates, and battery capacity concerns. Despite the difficulties, there are
Network efficiency
numerous creative studies of new substances and new techniques for enhancing the efficiency of solar energy
Photovoltaic cell
QoS (quality of service)
transformation to increase competitiveness of solar energy in market. This research proposes novel method in
Solar energy renewable energy analysis based on photovoltaic cell and machine learning technique for wind energy hybrid­
ization. The renewable analysis has been analysed using photovoltaic (PV) cell. Wind energy hybridization is
carried out using convolutional kernel support regression vector machine. Experimental analysis has been carried
out in terms of scalability, QoS (quality of service), power consumption, network efficiency, training accuracy.
Financial advantages of using new cooling methods for photovoltaic panels are also assessed through a cost
analysis.

Introduction centres [3]. Increased user knowledge of energy conservation and

environmental protection, along with the direct reduction of carbon
Recently, population growth and global economic expansion have dioxide caused by renewable energy producing systems, will result in
increased the demand for energy, which is crucial for both the socio­ significant energy savings and environmental contributions. The global
economic development of developing and industrialised nations [1]. As extent of environmental problems like global warming caused by carbon
a result, energy usage is very significant. There is widespread agreement dioxide emissions has been growing. Electric power generating is often
that significant energy efficiency policy necessitates a combination of the economic sector that produces the most CO2 emissions. This has a
actions, including financial incentives, regulatory tools, and data substantial impact on climate change, which is currently a major topic of
dissemination, and that mix of actions must be customised to circum­ discussion and controversy across the globe. The excessive release of
stances of every individual country. A sustainable as well as pure source global greenhouse gases (GHGs) into the atmosphere as a result of
of energy, renewable energy comes from the natural world [2]. One of human activity, such as carbon dioxide and methane, is one of the main
the solutions is renewable energy technology, which generates energy causes of climate change.
by converting natural occurrences (or natural resources) into usable Applications like high power point tracking, power prediction, fault
energy forms. Recently, there has been a noticeable rise in concern for diagnostics, and efficiency analysis are frequently used in solar PV
the research and teaching of renewable energy applications. Complete power producing systems. As a theoretical foundation for these appli­
courses on renewable energies have been established at some university cations, matching mathematical design of PV cells is needed [4]. As a

* Corresponding author.
E-mail address: sqamar@kku.edu.sa (S. Qamar).

https://doi.org/10.1016/j.seta.2023.103243
Received 18 October 2022; Received in revised form 23 April 2023; Accepted 27 April 2023
Available online 14 May 2023
2213-1388/© 2023 Published by Elsevier Ltd.
U. Sakthi et al. Sustainable Energy Technologies and Assessments 57 (2023) 103243

well-constructed mathematical model, it ought to be able to faithfully Due to the requirement to list and evaluate every current parameter that
capture operation of PV cells. While this is going on, the created model influences effectiveness and performance of using PVPs, this lengthens
can as accurately represent the properties of real PV cells under various feasibility study for introduction of PVPPs. Information provided in
temperatures as well as sun irradiances. In other words, as far as manufacturer’s specs is adequate for the 5-parameter (single-diode)
possible, the output characteristics should match the data from the equivalent circuit that serves as the foundation for the majority of PVP
datasheet or the real characteristic data. models. The techniques and algorithms for figuring out PVP’s electrical
Artificial intelligence (AI) methods have a huge potential to advance properties have been extensively researched [17]. The error in deter­
renewable energy sources in energy sector, to point where many busi­ mining parameters is rather high because the majority of these analyt­
nesses in sector are experimenting as well as testing cutting-edge solu­ ical approaches simply use data under STC in datasheet material and are
tions to increase efficiency of various methods in order to meet set based on certain simplifying assumptions. Lambert W function, intro­
climate targets. As a result, technical as well as political and economic duced by Work [18] in the analytical method, simplifies and clarifies
issues face renewable energy sources [5]. The unpredictable nature of parameter statements, but it necessitates the use of mathematical soft­
power production is one of them. Supply as well as demand do not al­ ware to apply numerical solution [19]. It is inconvenient in practise for
ways match up, but when wind blows as well as sun shines, photovoltaic the sequential technique to generate I-V curves for PV cells under a
(PV) panels and wind turbines create a lot of electricity to meet rising variety of sun irradiance and temperature settings through trials. The
need for energy. However, production of electricity from renewable five model parameters are derived in reverse by Work [20] after
sources varies depending on a number of factors, most notably the obtaining the key parameters using translational formulas under vari­
weather. To address this problem and create methods that can forecast able climatic conditions. However, this method will result in significant
fluctuations in energy supply as well as alter electrical supply to match parameter errors because of error accumulation.
demand, scientists are now interested in AI [6].
Research gap
Contribution of this research is as follows:
This brief state of the art demonstrates that very few wind energy-
1. To propose novel technique in renewable energy analysis based on related works make use of deep learning. In any case, this procedure
photovoltaic cell and machine learning technique for wind energy has been principally used to estimate wind and to distinguish disap­
hybridization. pointments in WTs. Additionally, previous research that employs fuzzy
2. The renewable analysis has been analysed using photovoltaic cell. pitch control relies on the measured wind without taking into account
Wind energy hybridization is carried out using convolutional kernel the possibility that this wind will not reach the rotor. The combination of
support regression vector machine. a PV method, an electric vehicle, a battery, and thermal power storage
devices can result in an integrated energy management system (EMS).
Background study The EMS is based on complex forecasting as well as considers both
financial effects of availability of electric car as well as driving habits
HRES (hybrid renewable energy system) system architectures and and priorities.They used mathematical modelling to evaluate the pro­
classification were reported in [7]. The optimisation techniques for posed EMS’s (energy management system) efficiency and contrasted it
HRES component sizing were described in [8]. The systems for wind, PV, with a more straightforward management system built on the smart grid
super capacitors, and batteries were explored in [9]. The authors IoT architecture, which is designed to generate all the electricity
investigated HRES algorithms in stand-alone mode with DC (direct required for the residence. The findings demonstrate that in a Fog-based
current) loads in [10] and AC (alternate current) loads in [11]. Several energy management system, the proposed technique enables demand
papers [12] solely discuss HRES system’s operation in stand-alone and residential generation to adapt to electricity supply conditions.
mode, leaving out the topic of energy management measures. Most ar­
ticles studied a machine side converter coupled to a permanent magnet
Proposed methodology
synchronised generator (PMSG) using a simple topology [13].Simple
hysteresis control of converter was utilized in systems with grid mode
This section discussnovel technique in renewable energy analysis
operation for grid side converter [14].
based on photovoltaic cell and machine learning technique for wind
Grid converter’s hysteresis control is imprecise and may negatively
energy hybridization. The renewable analysis has been analysed using
affect the supply network. The functioning of HRES with sophisticated
photovoltaic cell. The wind energy hybridization is carried out using
energy storage technologies including fuel cells, super capacitors,
convolutional kernel support regression vector machine. The wind hy­
hydrogen techniques, and bioreactors was examined by several writers
bridization based renewable energy analysis using photovoltaic cell and
[15]. The calculation process, precision, and amount of factors used in
machine learning is shown in Fig. 1.
the calculations vary amongst the extant mathematical models defining
the action of PVP. The primary models, for instance, are as follows:
Equivalent circuit of PV cells
Power models, two-diode PV array models, and the ideal one-diode
model. At the same time, rating parameters are used as starting
Due to the diode-like properties of PV cells, their output current is a
(input) ones in all PVP (photovoltaic power) models [15]. MATLAB
superposition of dark current that flows opposite way of the photocur­
Simulink is one of the most widely used computer software programmes
rent. I-V characteristic expression for a PV cell is shown in Eq. (1) in
for simulating PVP as well as PVPPs ((photovoltaic power plants). When
accordance with a general analogous circuit.
modelling PVPP operating modes using a computer, the manufacturer-
{ [ ] }
specified parameters are used to replicate a PVP. These models are q V + IRs
effective at identifying PVP features [16]. However, MATLAB Simulink I = Iph − Io exp (V + IRs ) − 1 − (1)
nNs kT Rsh
models also work well for creating controller algorithms and examining
transitory processes. where I and V are photovoltaic cells’ output current and voltage, Iph is
photocurrent generated when the cells are lighted, Io is reverse satura­
PV cell analysis in wind energy hybridization tion current of a diode, and q is charge on a single electron. Rsh describes
the leakage current at PN (p-type and n-type) junctions. Because current
The procedure of modelling the operating modes of PVPPs is made I exists on both sides of the equation, Eq. (1) is a classic implicit tran­
more difficult by absence of median values for PVPs’ parameter values. scendental equation. When PV cells are at open circuit point (Voc, 0),

2
U. Sakthi et al. Sustainable Energy Technologies and Assessments 57 (2023) 103243

Fig. 1. Wind hybridization based renewable energy analysis using photovoltaic cell.

Eq. (2) becomes

( )
qVo Voc
Iph + Io = Io exp − (2)
nNs kT Rsh
It is necessary to know the electrical current multiplied by the
voltage at greatest power point. The voltage and current at highest
power point are given by Eqs. (1)–(3).
( )
ΓNse kB T ΓNse kB T 1s
vmp = ln (3)
qe qel o Voc

where Boltzmann’s constant, electron charge’s absolute value, the

number of series configurations, short-circuit current, and open-circuit
voltage equations are kB, qe, Nse, Isc, and Voc (4).
qENC [ ]
Iph = Iph.Kef + μ1 (T − TRet ) (4)
qiNCNRet
Here, f1 = 50 or 60 Hz, and IRl, ISl and ITl _ are RMS values of current at
those frequencies, respectively (Rh,). The Eq. (5) can be used to get RMS
(root mean square) value of every phase voltage and line current.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
∫
1 T 2 Fig. 2. General energy flows and heat transfer system from PV panel.
vx = v (t)dt (5)
T 0 x

RMS value of voltage and current waveform utilizing DWT (discrete irradiation that PV panel receives as system’s energy input:
wavelet transform) may be extended to 3 phase approaches by ac­ QSolar = a⋅E⋅Ap (7)
counting for any of three phases (R, S, or T), as demonstrated by Eq. (6)
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
∫
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
∫ where an is the coefficient of absorption. Equation (8) can be used to
1 T 2 1 T 2 compute the total heat loss.
Vx = vx dt, Ix = i dt (6)
T 0 T 0 x
Qloss = QC + QR + QE (8)
The majority of solar radiation is utilized to boost internal energy of
Wind Turbine hybridization Model.
PV panels, known as DUpanel, while the remaining portion, known as
Qloss, is lost to the environment. Fig. 2 illustrates the general energy Wind turbine transforms mechanical rotational energy of the turbine
blades from wind’s kinetic energy. The following equation is utilized to
flows as well as heat transfer methods in PV panel.
The three main sources of heat loss—radiation, evaporation, and represent mechanical power Pt that wind turbine produces (9)
convection—are represented in this graph. Eq. (7) defines solar

3
U. Sakthi et al. Sustainable Energy Technologies and Assessments 57 (2023) 103243

1 methods—polycrystalline silicon Kyocera KC200GT, monocrystalline

Pt = ρACp (λ, β)v3w (9)
2 silicon Shell SP70, and thin-film Shell ST40—are run in Matlab. The
Tip speed ratio λ and blade β pitch angle are complex dependencies modelling approach used in this article assumes that the short circuit
on which turbine power coefficient Cp is stated. The common dq- point (0, Isc) and open circuit point (Voc, 0) of PV cells are independent
reference frame, which rotates at speed of electrical angular rotor, is of Rs and Rsh sizes.Short circuit current and open circuit voltage that
used to explain mathematical equations of the PMSG (power manage­ were determined in this work using a solution technique are therefore
ment smart grid). Eq. (10) gives the stator flux vector’s dq components first validated. I-V characteristic curves of three particular PV cells are
in dq reference frame. then displayed, and using mathematical design described in this paper,
mistakes related to them are also evaluated.
ψ sd = Ld isd + ψ PM (10) Short circuit current as well as solar irradiance have a good linear
connection, and the calculated and experimental results for Isc can
where stator flux vector sd, sq—dq components, stator voltage vector perfectly coincide. In various situations, evaluated as well as experi­
vsd, and the stator current vector ψsd, ψsq—dq components; PM-flux mental data remain essentially same. The absolute error between
created by permanent magnets The PMSG’s electrical and mechanical calculated data as well as experimental data is less than 0.015 A, despite
angular speeds ωe,ωm, the stator phase resistance, the direct and scale factor between Isc and T of Shell ST40 being considerably varied
quadrature inductances, and the number of pole pairs are all represented from experimental data for S = 1000 W/ m2. Accuracy is sufficient for
by the variables e and m convolutional kernel support regression vector the majority of engineering applications. The experimental data and the
machine based wind energy hybridization: calculated value of Voc for n = 1.9381 are nearly identical, with a
A nonlinear SVM (support vector machine) locates a linear boundary maximum absolute error of 0.16 V.
in feature space F by maximising the lowest distance of instances to this Projected value of Voc departs greatly from the experimental data if
boundary after mapping input data × = x1,…,xn into an HD feature n = 1.6 in accordance with material property of PV cells. As can be
space, F = R l, utilizing a nonlinear mapping function s: R p R l. This is shown, due to its smaller inaccuracy than the double-diode model,
technically equal to solving by Eq. (11). single-diode method appears to be extremely suitable to model the
∑ Mono-Si modules. On the other hand, the double-diode model offers a
Min w, b 1/2 WTw + c ai (11)
i∈SV crucial approach to modelling the PolySi PV panel, particularly at me­
dium and high temperatures. Additionally, the DDM (Digital Diagnostic
where C specifies trade-off between best combinatorial choice of margin Monitoring) is more appropriate for mid and high ranges of irradiance
and classification error, and ξ = (ξ1, …,ξn) is so-called soft margin whereas the proposed technique shows outstanding results for low
parameter. T is a slack variable vector with a non-negative value that irradiance. Both PV panel technologies are used to compare the selected
regulates misclassification. Consequently, sign(D(x)), with eq, (12) as­ equivalent-circuit models for the short-circuit current within the same
signs label of an instance x. range of climatic circumstances.
∑ Table 1 above compares and contrasts planned and current PV
D(x) = ̂k i yi K(xi , x) + b. (12) module-based techniques. Here, the PV modules made of polycrystalline
i∈SV
silicon (Kyocera KC200GT), monocrystalline silicon (Shell SP70), and
where the projected value of an is represented by a. The bias term bj is thin-film silicon (Shell ST40) have all been analysed. Scalability, QoS,
theoretically shown to be the same for every instance in the SV. Utilizing power consumption, network efficiency, and training accuracy are
j-th support vector (SV) xj by equation, the biassed term b is really among the parameters that were examined.
calculated as average of all estimated bj values at all support vectors (13) When the wind speed factor is larger, the hybrid renewable energy
∑ ( ) system operates more effectively. Wind subsystem will provide appro­
bj = yj − ȧi yi K xi , xj (13) priate amount of power from a wind turbine in the range of 4 to 8 m/s. A
i∈SV
beneficial effect will be seen when wind speed is increased from 8 to 12
The one-versus-all technique may often be used to break down a k- m/s and continues up to 15 m/s with a reduced slope on power. For
category classification problem into a series of binary classification is­ small-scale wind turbines (400 W), the optimal wind speed is greater
sues with class label yi taking value from 1,…, k. The training sample xi, than 4 m/s and between 8 and 10 m/s.Additionally, it is undesirable to
y (m) I is set up for the m-th binary classification, where y (m) I = I(yi = raise temperature of solar cells to more than 80 ◦ C, as this will signifi­
m) I(yi 6 = m) and I(m) is indicator function by eq. (14). cantly reduce the system’s power. Solar subsystem power generation
∑ (m) (m) will be influenced most by radiation, and wind speed will be influenced
Dm (x) = ai yi Km (xi , x) + b(m) (14) most by wind power generation. The joined effect of radiation and
sunlight based cell temperature is exceptionally near one another and
i∈SVm

A majority voting approach can be used to determine final class label positive in power age. Interaction between wind speed and radiation
for an instance utilizing estimated decision functions from all m-th bi­ intensity is most significant effect on system power generation.
nary classifications. There are many common kernels available for SVM The above Fig. 3 shows analysis based on Polycrystalline silicon
process. From a geometrical perspective, Riemannian metric is induced Kyocera KC200GT. Proposed technique attained scalability of 82%, QoS
in input space I when feature space F is Euclidean space. For example, in of 51%, power consumption of 81%, network efficiency of 85%, training
a two-dimensional situation, a tiny change in input space, d(x), will be accuracy of 92%; while existing HRES attained scalability of 77%, QoS
represented as ds(x) in feature space, Eq. (15) of 45%, power consumption of 77%, network efficiency of 81%, training
accuracy of 83%, PVP attained scalability of 79%, QoS of 48%, power
ds(x) = ∇s⋅dx (15)
consumption of 79%, network efficiency of 83%, training accuracy of
Although Lp-norms (p > 0) can be used generally in practical situ­ 88%.
ations, such as the L1-norm, we discovered that the L2-norm would The above Fig. 4 shows analysis based on monocrystalline silicon
perform the best empirically. We additionally propose the notion of Shell SP70. Proposed technique attained scalability of 88%, QoS of 55%,
building c(x) in binary case to include data from × since wms do not power consumption of 85%, network efficiency of 91%, training accu­
involve knowledge of x.Experimental analysis racy of 93%; while existing HRES attained scalability of 83%, QoS of
To confirm the mathematical approach used for PV cells in this 52%, power consumption of 79%, network efficiency of 86%, training
research, simulations for 3 different types of PV accuracy of 85%, PVP attained scalability of 85%, QoS of 53%, power
consumption of 82%, network efficiency of 88%, training accuracy of

4
U. Sakthi et al. Sustainable Energy Technologies and Assessments 57 (2023) 103243

Table 1
Analysis based on PV modules.
PV Modules Techniques Scalability QoS Power consumption Network efficiency Training accuracy

Polycrystalline silicon Kyocera KC200GT HRES 77 45 77 81 83

PVP 79 48 79 83 88
PGREA_PCMLA_WEH 82 51 81 85 92

Monocrystalline silicon Shell SP70 HRES 83 52 79 86 85

PVP 85 53 82 88 91
PGREA_PCMLA_WEH 88 55 85 91 93

Thin-film Shell ST40 HRES 85 56 81 89 87

PVP 86 58 83 93 92
PGREA_PCMLA_WEH 89 61 86 95 96

Fig. 5. Analysis based on Thin-film Shell ST40 PV module.

accuracy of 87%, PVP attained scalability of 86%, QoS of 58%, power

consumption of 83%, network efficiency of 93%, training accuracy of
92%.
Fig. 3. Analysis based on Polycrystalline silicon Kyocera KC200GT PV module.

Discussion and Limitation:

The creation of a cutting-edge power grid that makes it possible for

energy providers and consumers to communicate in both directions for
the purpose of fine-grained calculation, control, and feedback is
becoming increasingly significant all over the world. The Smart Grid’s
key features include resource management and increased energy effi­
ciency. The smart grid is a cutting-edge power infrastructure that pro­
vides two-way communication between energy providers and customers
for advanced metering, control, and feedback. Its development is
expanding on a global scale. The smart grid’s better energy efficiency
and use of current resources are key components. Every appliance has
been initialised to start at the best time and run in the best modes before
the optimisation procedure starts. When user cost variations between
iterations for all users fall below a certain threshold, the distributed
algorithm’s iteration is finished. The cost cannot be penalised in this
circumstance due to the contemporaneous changes, and the penalty
ratio declines with the generation cost index of the prediction ratio-
Fig. 4. Analysis based on monocrystalline silicon Shell SP70 PV module.
based organisation. Majority of utilities have adopted a load shedding
strategy as a result of the sharp rise in electricity demand, which has
presented a significant challenge to the electricity distribution networks;
91%.
In order to avoid system failure or major collapse, it represents a strategy
The above Fig. 5 shows analysis based on Thin-film Shell ST40.
for handling load requirements by shedding them in critical situations
Proposed technique attained scalability of 89%, QoS of 61%, power
when demand is greater than the total generation. A demand-side
consumption of 86%, network efficiency of 95%, training accuracy of
management system known as load management is operated by an en­
96%; while existing HRES attained scalability of 85%, QoS of 56%,
ergy management system or consumer power supply.
power consumption of 81%, network efficiency of 89%, training

5
U. Sakthi et al. Sustainable Energy Technologies and Assessments 57 (2023) 103243

Conclusion [2] Gajewski P, Pieńkowski K. Control of the hybrid renewable energy system with
wind turbine, photovoltaic panels and battery energy storage. Energies 2021;14(6):
1595.
In this research the novel technique for renewable energy based [3] Shinong W, Qianlong M, Jie X, Yuan G, Shilin L. An improved mathematical model
photovoltaic cell analysis with machine learning in hybridization of of photovoltaic cells based on datasheet information. Sol Energy 2020;199:437–46.
wind energy is carried out. convolutional kernel support regression [4] Hadipour A, Zargarabadi MR, Rashidi S. An efficient pulsed-spray water cooling
system for photovoltaic panels: experimental study and cost analysis. Renew
vector machine with photovoltaic cell has been used for wind energy Energy 2021;164:867–75.
hybridization with reduction of reliance on grid as well as electricity [5] Shirzadi N, Nasiri F, El-Bayeh C, Eicker U. Optimal dispatching of renewable
costs, particularly during peak hours, techniquesenables monitoring as energy-based urban microgrids using a deep learning approach for electrical load
and wind power forecasting. Int J Energy Res 2022;46(3):3173–88.
well as controlling renewable energy generation, percentage of energy [6] Jani HK, Kachhwaha SS, Nagababu G, Das A. Temporal and spatial simultaneity
imports from grid.Outcomes show that suggested approach can predict assessment of wind-solar energy resources in India by statistical analysis and
building energy demand as well as supply. Experimental analysis has machine learning clustering approach. Energy 2022;248:123586.
[7] Mayer MJ. Benefits of physical and machine learning hybridization for
been carried out in terms of scalability, QoS, power consumption, photovoltaic power forecasting. Renew Sustain Energy Rev 2022;168:112772.
network efficiency, training accuracy.Proposed technique attained [8] Nabavi SA, Motlagh NH, Zaidan MA, Aslani A, Zakeri B. Deep learning in energy
scalability of 89%, QoS of 61%, power consumption of 86%, network modeling: Application in smart buildings with distributed energy generation. IEEE
Access 2021;9:125439–61.
efficiency of 95%, training accuracy of 96%.Intend to examine the [9] Nwokolo SC, Obiwulu AU, Ogbulezie JC, Amadi SO. Hybridization of statistical
interaction between DL (deep learning) techniques as well as optimi­ machine learning and numerical models for improving beam, diffuse and global
zation methods like Sine Cosine method, GA (genetic algorithm), and solar radiation prediction. Clean Eng Technol 2022;9:100529.
[10] Jani HK, Kantipudi MVVP, Nagababu G, Prajapati D, Kachhwaha SS. Simultaneity
Wolf Pack method in our future work on net-zero smart buildings with
of wind and solar energy: a spatio-temporal analysis to delineate the plausible
renewable energy sources. regions to harness. Sustain Energy Technol Assess 2022;53:102665.
[11] Soundarya G, Sitharthan R, Sundarabalan CK, Balasundar C, Karthikaikannan D,
Sharma J. Design and modeling of hybrid DC/AC microgrid with manifold
Declaration of Competing Interest renewable energy sources. IEEE Can J Electr Comput Eng 2021;44(2):130–5.
[12] Wehbi Z, Taher R, Faraj J, Castelain C, Khaled M. Hybrid thermoelectric
generators-renewable energy systems: A short review on recent developments.
The authors declare that they have no known competing financial Energy Rep 2022;8:1361–70.
interests or personal relationships that could have appeared to influence [13] May EC, Bassam A, Ricalde LJ, Soberanis ME, Oubram O, Tzuc OM, et al. Global
the work reported in this paper. sensitivity analysis for a real-time electricity market forecast by a machine learning
approach: a case study of Mexico. Int J Electr Power Energy Syst 2022;135:107505.
[14] Wang Z, Liu Xu, Yue M, Yao H, Tian H, Sun X, et al. Hybridized energy harvesting
Data availability device based on high-performance triboelectric nanogenerator for smart
agriculture applications. Nano Energy 2022;102:107681.
[15] Kumar GBA, Shivashankar. Optimal power point tracking of solar and wind energy
The data that has been used is confidential. in a hybrid wind solar energy system. Int J Energy Environ Eng 2022;13(1):
77–103.
Acknowledgements [16] Cheng L, Zang H, Wei Z, Zhang F, Sun G. Evaluation of opaque deep-learning solar
power forecast models towards power-grid applications. Renew Energy 2022;198:
960–72.
The authors extend their appreciation to the Deanship of Scientific [17] Garg S, Krishnamurthi R. Powernet: A novel method for wind power predictive
Research at King Khalid University, Abha, Kingdom of Saudi Arabia for analytics using Powernet deep learning model. J Renew Sustain Energy 2022;14
(4):043301.
funding this work through Large Groups RGP.2/170/1444 [18] Shahiduzzaman KM, Jamal MN, Nawab MRI. Renewable energy production
forecasting: a comparative machine learning analysis. Int J Eng Adv Technol 2021;
References 10(6):11–8.
[19] Singh U, Rizwan M. SCADA system dataset exploration and machine learning based
forecast for wind turbines. Results Eng 2022;16:100640.
[1] Wang H, Li W, Liu T, Liu X, Hu X. Thermodynamic analysis and optimization of [20] Rodríguez F, Azcárate I, Vadillo J, Galarza A. Forecasting intra-hour solar
photovoltaic/thermal hybrid hydrogen generation system based on complementary photovoltaic energy by assembling wavelet based time-frequency analysis with
combination of photovoltaic cells and proton exchange membrane electrolyzer. deep learning neural networks. Int J Electr Power Energy Syst 2022;137:107777.
Energ Conver Manage 2019;183:97–108.