1 s2.0 S2352484723008065 Main

Energy Reports 9 (2023) 6063–6087
Contents lists available at ScienceDirect
Energy Reports
journal homepage: www.elsevier.com/locate/egyr
Review article
Renewables integration into power systems through intelligent

techniques: Implementation procedures, key features, and
performance evaluation
∗
Sayemul Islam, Naruttam Kumar Roy
Department of Electrical and Electronic Engineering, Khulna University of Engineering & Technology, Khulna 9203, Bangladesh
article info a b s t r a c t
Article history: Integrating renewable energy sources (RESs) such as solar photovoltaic (PV), wind, biogas, and
Received 27 November 2022 hydropower into the power system is a sustainable solution that can feasibly maintain the power
Received in revised form 22 March 2023 supply and demand response. The uncertainty in solar irradiance and wind speed impedes the supply
Accepted 18 May 2023
and demand response. The uncertainty problem can be solved by integrating an appropriate control
Available online 31 May 2023
technique that reasonably forecasts necessary information and maintains system operation. A critical
Keywords: analysis of different intelligent techniques with numerical data review, prediction accuracy, pros and
Renewable energy integration cons, and techno-economic feasibility is necessary for the reader’s perception. This paper analyzes the
Photovoltaic 89 research works of different intelligent techniques integrated into RESs and energy storage systems
Wind (ESSs). The intelligent techniques are classified according to the considered resources, such as PV, wind,
Energy storage system biogas, and hydropower to demonstrate a meaningful insight into the particular research field. The
Artificial intelligent analysis provides adequate information on each considered technique presenting the implementation
Machine learning
procedures, key features, and accuracy. The accuracy of each method is determined by integrating
different feasibility metrics such as root mean square error (RMSE), root mean absolute error (RMAE),
and root mean percentage error (RMPE). The intelligent integration into ESS emphasizes the possibility
of enhancing the storage backup for RESs connected power distribution systems. The review analysis
signifies the current view and potentiality of incorporating intelligent methods into power systems
and demonstrates a significant insight into the research field.
© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Contents
1. Introduction..................................................................................................................................................................................................................... 6064
1.1. Research motivation .......................................................................................................................................................................................... 6065
1.2. Framework & Key contributions ...................................................................................................................................................................... 6066
2. A critical review of existing intelligent techniques ................................................................................................................................................... 6067
2.1. Integration of intelligent techniques in Solar PV system ............................................................................................................................. 6067
2.1.1. Application of ANN in Solar PV system........................................................................................................................................... 6067
2.1.2. Application of CNN in Solar PV system........................................................................................................................................... 6068
2.1.3. Application of PSO in Solar PV system............................................................................................................................................ 6070
2.1.4. Application of other intelligent techniques in Solar PV system................................................................................................... 6071
2.2. Integration of intelligent techniques in wind power system....................................................................................................................... 6072
2.3. Integration of intelligent techniques in hybrid RESs..................................................................................................................................... 6074
2.4. Integration of intelligent techniques in ESS ................................................................................................................................................... 6078
3. Metrics of feasibility analysis........................................................................................................................................................................................ 6081
4. Findings and discussions ............................................................................................................................................................................................... 6083
5. Conclusions & future research ...................................................................................................................................................................................... 6084
Declaration of competing interest................................................................................................................................................................................ 6084
∗ Corresponding author.
E-mail addresses: sayemul100@gmail.com (S. Islam), nkroy@eee.kuet.ac.bd (N.K. Roy).
https://doi.org/10.1016/j.egyr.2023.05.063
2352-4847/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
S. Islam and N.K. Roy Energy Reports 9 (2023) 6063–6087
Data availability .............................................................................................................................................................................................................. 6084

References ....................................................................................................................................................................................................................... 6084
is proposed to achieve maximum point tracking (MPPT) of a

1. Introduction hybrid wind–PV system. The 2 MW hybrid system is simulated
in MATLAB/Simulink platform considering significant parame-
The global electricity demand is increasing with the rapid ters such as tip speed ratio, wind speed, and optimal torque.
growth of the world’s population and economy. Countries world- A case study for the hybrid PV–diesel–battery integration based
wide are constructing fossil fuel (oil, diesel, gas)-based power on the coyote optimization algorithm is demonstrated in Sari
generation systems to cope with the electricity demand. Through et al. (2022) that aims to maintain optimal component sizing,
these conventional power plants generate a large amount of minimum loss of energy minimum CO2 , and reduced cost of
electricity to meet the load demand, fossil fuel emits greenhouse operation. A concentrated solar (138 MW) and wind energy (146
gases (GHGs) such as CO2 , CO, SO2 , NOX that causes the break- MW) system is proposed in Sezer et al. (2019), where wind
down of environmental feasibility (Dong et al., 2015; Saidur et al., energy acts as a backup during insufficient solar irradiance. The
2011; Del Río and Burguillo, 2008). Emission causes acid rain energy efficiency for the combined operation of the two energy
and global warming, which is harmful to humankind. Integrating sources is 61.3%. In Soliman et al. (2021), a fuzzy logic and
renewable energy sources (RESs) such as wind, solar photovoltaic high sliding mode control-based intelligent control approach is
(PV), hydropower, and biogas into the power system can be an provided for a hybrid PV, wind, tidal, and ESS integrated system.
alternative to conventional power generation (Liu et al., 2019). The techno-economic and environmental feasibility is determined
The storage of fossil fuels is limited on the earth. However, RESs using HOMER Pro software for a solar PV, wind, and fuel cell-
is unlimited as solar PV requires sunlight, and wind power plant based energy system (Al-Badi et al., 2022). The simulation is
requires wind speed that can be achieved sustainably. RESs are carried out for different configurations of the energy resources,
an essential tool to maintain the electricity demand (Obama, where PV, wind, and fuel-cell-based system reduces CO2 emis-
2017). European Union nations have decided to integrate renew- sions and minimizes operation cost. In Al-Ghussain et al. (2023),
able resources into the power grid and supply 32% of the total an illustration for Estonia is provided to generate 100% of their
electricity by 2030 (Podder et al., 2020). A statistics of renewable energy from RESs (wind, PV, ESS, and thermal storage) within
energy generation for different regions in the world for years 2050. The study reveals that a large storage facility is required
(2011–2020) are illustrated in Fig. 1 (IRENA, 2021). The figure for optimal operation and the estimated energy efficiency of
emphasizes that the generation of power from RESs is improv- a hybrid PV–wind system is 14.33%. Another research work in
ing each year. There is a 52.5% increase in energy generation Mohammed et al. (2022) focuses on integrating several intelligent
from RESs between the years (2011–2020), which indicates a techniques to find the optimal sizing of components, system
considerable energy solution as an alternative to conventional efficiency, and minimum cost for a hybrid PV, wind, diesel, and
energy. The maximum increase in RESs-based power generation ESS-based system.
is in Asia (66.387%), and the lowest energy generation is in South The availability of solar radiation and wind speed varies de-
America (34.76%) for years (2011–2020). Sufficient energy gener- pending on the environment. Solar PV generates a dc power
ation in developing countries is a promising concern for economic output that needs to be converted to ac (Ferrero Bermejo et al.,
security and fast advancement in the nation’s development in- 2019). The inertia response and frequency stability are funda-
frastructure. Developing countries (Bangladesh, India, Pakistan, mental concerns of integrating solar PV and wind into the power
Afghanistan, Nigeria, etc.) focus on generating renewable energy grid. Hydropower has been reliably used for many years in dif-
from natural resources (hydropower, wind, solar PV, etc.) to sus- ferent countries that depend on the tide of water and emits
tain their non-renewable sources and maintain environmental no GHGs. Biogas is another vital source of RESs that incorpo-
reliability. Fig. 2 demonstrates different developing countries’ rates different biomass, such as poultry dumping and cow dung,
energy generation statistics from various RESs in 2020 (IRENA, to generate biogas (Perez-Navarro et al., 2010). The biogas can
2021). The statistics emphasize that hydropower is a great energy then be converted to electricity. It is expected that renewable
source highly utilized by developing countries, whereas wind and resources can replace conventional energy sources. Proper utiliza-
solar PV have also become promising factors. tion and appropriate techniques are necessary for every renew-
Sustainable energy solution is vital to ensure the quality of able source to obtain maximum energy to fulfill load demand.
life and energy security (Xiangchengzhen and Yilmaz, 2020; Shao Again, renewable-based power generation suffers intermittency
et al., 2021). Today’s world energy policy focuses on reducing due to the uncertainty in solar radiation, wind speed, etc. The
carbon emissions by eliminating the dependency on fossil fu- intermittency limits the renewable system to maintain appro-
els. Asian developing countries can boost their economic growth priate load demand to consumers, which hampers the system’s
by taking effective energy policies for utilizing renewable re- reliability. The limitation can be overcome by introducing ESS to
sources (Esquivias et al., 2022). However, Asian nations are ge- the RESs. The ESS’s maximum technical and economic feasibility
ographically in a great hub to make use of natural energy that can be achieved by applying appropriate control techniques. The
eliminates the nations to be less dependent on importing fossils optimal PV power generation from a solar PV system depends on
like oil, coal, and gas from other countries. The energy policy solar irradiance with two components: beam and diffuse solar
of these countries should follow socio-economic benefits that irradiance. Beam solar irradiation occurs when sunlight directly
signify poverty alleviation and income equality with minimum hits the solar panel. However, sunlight that hits the solar panel
environmental pollution (Khan et al., 2022; Vo and Vo, 2022; Gal- in another way is called diffuse solar irradiation. The combination
imova et al., 2022). The proper utilization of renewable resources of beam and diffuse solar irradiation is called global solar irradi-
ensures the feasibility of environments with rapid industrializa- ation (Chow et al., 2012). Again, the power generation of wind
tion, urbanization, and economic growth. farms depends on the quality of wind speed, weather conditions,
Several recently published research works emphasize signifi- etc. (Albadi and El-Saadany, 2010; Liu et al., 2014). Hence, the
cant aspects of wind, PV, and energy storage system (ESS) inte- accurate forecasting of solar irradiance and wind speed must be
gration in power systems. In Kumar (2022), a control approach evaluated to maintain optimal operations and fulfill load demand.
6064
Fig. 1. A statistics illustrating the renewable energy generation in MW for different regions in years (2011–2020) (IRENA, 2021).
Fig. 2. Statistics of renewable energy generation (MW) in different developing countries in the year 2020 (IRENA, 2021).
The forecasting values of solar irradiance and wind speed can 1.1. Research motivation
be evaluated by integrating different intelligent techniques such
as artificial neural network (ANN), particle swarm optimization The author’s motivation behind the implementation of the
(PSO), fuzzy logic control (FLC), recurrent neural network (RNN), research work is to demonstrate the current state of the power
convolutional neural network (CNN), genetic algorithm (GA), etc. system integrated with intelligent techniques, especially for re-
Intelligent techniques can also maintain frequency stability, en- newable resources. Several review works have already been pro-
ergy distribution, system sizing, and uncertainty reduction. This posed demonstrating the meaningful insight between the power
paper briefly analyzes the current state of energy generation in system and different control techniques (artificial intelligence,
different countries from RESs. The paper’s main objective is criti- machine learning, IoT, etc.). A review analysis based on different
cally analyzing different intelligent techniques applied to renew- islanding detection methods (IDMs) for RESs integrated micro-
able resources and storage systems. Relative analysis among these grid is demonstrated in Panigrahi et al. (2021) that classifies
intelligent techniques, including numerical data review, pros and IDMs in primarily two cases: conventional IDM and modern IDM.
Conventional IDM includes different active, passive, and hybrid
cons, and techno-economic feasibility, are demonstrated.
techniques, whereas modern IDMs incorporate signal processing-
The paper is organized into several sections considering much
based and intelligent-based techniques. Another review work
priority for readers’ perceptions on this topic. The introduction of
in Olatomiwa et al. (2016) presents a comprehensive study of
this paper in Section 1 contains the current state-of-art and future
the different standalone, grid, and smart grid-connected energy
aspects of the selected subject that covers the extensive insight of
control strategies integrated into the power system. Each con-
RESs, the research motivation, framework and key contribution. trol strategy is applied for a specific system configuration that
Section 2 demonstrates deep literature on different intelligent would be standalone, grid-connected, or smart grid-connected
techniques (ANN, CNN, RNN, PSO, FLC, GA, etc.) related to the with PV, wind, fuel cell, diesel, battery, etc., as energy resources.
selected topic. The section critically analyzes various existing The analysis aims to find out the strategies that are technically
methods considering prons and cons, system configuration, input and economically viable. The present state and the future aspect
variables, and system modeling. The section is subdivided into of integrating artificial intelligence (AI) into RESs-based power
several sections depending on the considered energy sources systems is illustrated in Jha et al. (2017), which includes the
(PV, wind, ESS, and combined sources). The metrics of feasibility system configuration, simulation platforms, considered resources,
analysis are introduced in Section 3. The significant findings and and outcomes for individual control schemes based on ANN,
discussion of the review work are included in Section 4. The con- CNN, FLC, PSO, GA, etc. A review analysis in Jaalam et al. (2016)
clusion and future directions of the proposed paper are presented demonstrates the different existing synchronization techniques
in Section 5. for grid-connected converters. Each approach aims to maintain
6065
Fig. 3. An overview of the working flow of the proposed review work.
the frequency variations and phase angle at a feasible range energy technologies in island power grids incorporating RESs
and sustain the immediate response to the grid. A literature (wind turbine, solar PV, hydropower, biomass, geothermal, and
review based on the past advancement and present research ocean energy) are analyzed in Kuang et al. (2016). The study also
activities on smart grid technology is presented in Alotaibi et al. represents the integration of the smart grid, demand-side man-
(2020) that perceives the effect of integrating RESs in the power agement (DSM), ESS, etc., for efficient energy management in the
system. The analysis dictates several significant features of the island power grid. The paper in Kanase-Patil et al. (2020) demon-
smart grid technology, such as infrastructure, pricing mecha- strates a review analysis of AI-integrated RESs for smart cities
nism, data management, components, cybersecurity, and control. that includes energy resources, optimization techniques, and the
Another review work in Krishna and Kumar (2015) presents smart grid application. A detailed analysis of different ML models
significant information on optimal sizing, control strategy, and for RESs forecasting is presented in Lai et al. (2020) including
energy management for the integration of RESs into the power data pre-processing and parameters selection techniques, perfor-
system. The study also describes different available controllers mance measuring methods, and accuracy-determining schemes.
for energy management, such as proportional–integral (PI) con- In Shafiullah (2016), a hybrid RESs model is developed to ob-
trollers, fuzzy logic controllers, microcontrollers, etc. A review tain techno-economic and environmental feasibility. The pro-
analysis presents different forecasting models for RESs connected
posed model predicts the generation of solar and wind power to
smart grid systems in Ahmad et al. (2020) that study three
maintain efficient load management on the consumer’s side. A
forecasting techniques: machine learning-based, ensemble-based,
review analysis of energy forecasting, energy management, and
and ANN-based techniques. The study is carried out consider-
demand response by DL models is proposed in Devaraj et al.
ing the forecasting interval in three categories: short interval,
(2021). The forecasting accuracy for wind and solar PV sources
medium interval, and long interval.
is analyzed and compared for different DL models.
A review of the energy modeling and planning based on fuzzy
logic for RESs-connected power systems is proposed in Sugan-
thi et al. (2015). Another review on deep learning (DL) based 1.2. Framework & Key contributions
power load forecasting techniques for RESs (wind and solar PV)
connected to smart microgrid is proposed in Aslam et al. (2021). Fig. 3 demonstrates the working flow of the proposed review
The review analysis considers hybrid DL schemes to evaluate activities to facilitate the reader’s perception of this research
the prediction accuracy and to perform the relative comparison. analysis. The introduction section incorporates two separate sec-
The paper in Ponnusamy et al. (2021) demonstrates a review of tions, where the research motivation section illustrates the recent
big data analysis for the smart grid technologies presenting the review activities on intelligent techniques integration into the
techno-economical and political impacts. The study also analyzes power system. The next section includes the major part of the
the importance of AI and machine learning (ML) technologies for review work, where 89 research works of different intelligent
real-time data management in smart grid systems. A comprehen- techniques for only solar PV, only wind, ESS, and hybrid RESs
sive review of different energy storage technology integrated to configurations are discussed. The significant findings from the se-
RESs-based systems is proposed in Tan et al. (2021). The study lected research works are also analyzed. The contributions of this
demonstrates the current state of different ESS, ESS integration to paper in comparison to existing research works is demonstrated
the power grid, and the impact and challenges of ESS integration in Table 1.
into RESs. A review of inertia response and frequency control The significant contributions from the proposed review work
techniques for RESs (wind turbine and solar PV) is presented in can be summarized as follows:
Dreidy et al. (2017), including the deloading techniques, inertia
emulation techniques, fast power reserve, and droop techniques. • Provides deep analysis of the 89 research works of different
The techno-economical feasibility for the integration of ESS into intelligent techniques integrated into RESs and ESS
RESs is also demonstrated. In Ellabban et al. (2014), the current • Analyzes intelligent techniques for different configurations
state and future aspects of RESs, power electronics technolo- such as standalone solar PV, wind, ESS, and also for hybrid
gies, and smart grid technologies are discussed briefly. The study RESs
aims to determine the potency of integrating these technolo- • Perceives the potency of ESS integration into RESs connected
gies to improve power generation and fulfill load demand. The power system for power backup and stable operation
6066
Table 1
The contributions of this paper in comparison to existing research works.
Contributions This Panigrahi Olatomiwa Jaalam et al. Alotaibi Krishna and Ahmad Suganthi Aslam et al. Devaraj
paper et al. (2021) et al. (2016) (2016) et al. (2020) Kumar (2015) et al. (2020) et al. (2015) (2021) et al. (2021)
Critical analysis of ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
different
intelligent
techniques
The potency of ✓ ✘ ✓ ✘ ✓ ✓ ✘ ✓ ✘ ✘
ESS integration
into RESs
System ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
construction,
features, and
limitations
Implementation ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
procedures and
simulation
platforms
Mathematical ✓ ✓ ✘ ✓ ✘ ✓ ✘ ✘ ✓ ✘
models and
equations
Performance ✓ ✓ ✘ ✘ ✘ ✘ ✓ ✘ ✘ ✓
evaluation metrics
and accuracy
Fig. 4. An overview of intelligent techniques that are considered in different energy systems.
• Analyzes the system configuration, simulation platform, key 2.1. Integration of intelligent techniques in Solar PV system
features, and accuracy for each of the intelligent techniques
and also considers different feasibility metrics (RMAE, RMSE, This sub-section demonstrates different intelligent techniques
MAPE, etc.) for relative comparison integrated into solar PV systems. Several techniques are pre-
sented in three separate sections depending on the intelligent
concept: ANN, CNN, and PSO. The other techniques are included
2. A critical review of existing intelligent techniques in another sub-section.
2.1.1. Application of ANN in Solar PV system

The integration of RESs in the power system causes frequency
An ANN-based power output prediction technique for a large
instability and uncertainties that impede optimal energy manage-
solar system is proposed in Kalogirou et al. (2014) that considers
ment. ESS is required as a backup of energy in case of power
the output of daily operation and the temperature level of the
uncertainty of wind and PV systems. The forecasting of solar
storage tank as essential factors. The experiment continued for
irradiation, wind speed, and power output through intelligent almost one year to collect relevant data, and eventually, ANN
techniques facilitates optimal energy management at the con- predicted accurate data of the PV system. Results show that the
sumer’s level. This section includes the 89 research works of proposed approach’s correlation coefficient (CC) is approximately
different intelligent techniques, where each technique aims to 0.93 for the daily output power prediction, implying better fore-
reduce the techno-economic costs and improve environmental casting accuracy. Another PV power prediction technique based
benefits. The considered energy systems are grouped into four on ANN and an analog ensemble (AnEn) is proposed in Cervone
sub-sections based on energy sources: the only PV system, the et al. (2017) that analyzes the short-term (72-h) operation of
only wind systems, the combination of different RESs, and the ESS the three PV power plants situated in Italy. The study simu-
system, as shown in Fig. 4. lates the system for three cases: only ANN, only AnEn, and the
6067
combination of ANN and AnEn, where the combined operation A PV and thermal system in Malaysia based on ANN is proposed
offers the best performance to evaluate the deterministic and in Al-Waeli et al. (2018) that predicts the system output current
probabilistic forecasts. A short-term power prediction technique and efficiency for three different cooling systems: tank filled with
based on ANN is presented in Chow et al. (2012) that evaluates water, phase change material (PCM), and nanofluid and nano-
accurate predictions of the PV system to minimize operational PCM. The result illustrates the current output increased by 0.35 A
costs and maximize power output. Solar radiation, temperature, and efficiency by 5.25% for the nanofluid and nano-PCM system
azimuth angle, and elevation angle are considered in the pre- than others. A PV system is modeled to predict the output power
diction method. The major advantage of integrating ANN into considering temperature and irradiance (Saberian et al., 2014).
the PV system is that it can accurately predict the daily solar General regression NN (GRNN) and feedforward backpropagation
irradiance and the output power generation without having a (FFBP) are introduced in the scheme to continue the prediction
developed relationship between input and output parameters. operation. Simulation results show that FFBP provides better
Results show that the CC varies from 0.618 to 0.9305, and the performance than GRNN. A solar irradiation prediction technique
confidence limit for forecasting accuracy is 95%. An ANN-based for remote solar monitoring stations in Northwestern Alberta,
intelligent technique is proposed in Yan et al. (2015) to evaluate Canada, is presented in Vanderstar et al. (2018) that integrates
the power reserve of the PV system considering the power out- ANN for solar irradiation forecasting and GA to measure the opti-
put and load forecasting in a 24-h time range. The uncertainty mal sizing and positioning of the system. 2 hours forecasting for 5
due to the integration of PV into the power system is highly input ANN estimates nRMSE as 10.8% and correlation coefficient
emphasized. The net forecasted demand (NFD) is introduced to as 0.912. The main limitation is that the scheme can predict only
analyze the power reserve from the proposed system. The scheme non-zero solar irradiance. Table 2 demonstrates a comparative
compensates for the risk considering the fixed risk index. The analysis of different ANN-based techniques integrated into PV
limitation of the plan is that the forecasting errors increase with systems.
increased time scheduling. Another ANN-based solar radiation
prediction technique is demonstrated in Rocha et al. (2019) that 2.1.2. Application of CNN in Solar PV system
incorporates the Broyden–Fletcher–Goldfarb–Shanno (BFGS) al- A CNN and SSwA (Salp Swarm Algorithm) based PV power
gorithm for prediction purposes. The study includes different forecasting technique for a 500 kW PV power plant in south
input predictors: maximum temperature, minimum temperature, Taiwan is proposed in Aprillia et al. (2020). CNN develops the
wind speed, relative humidity, cloudiness, evaporation, and so on. forecasting model, and SSwA identifies the CNN parameters. The
The experimental outcomes show that the method can provide study considers five weather types: rainy, heavy cloudy, cloudy,
the most accurate prediction of solar radiation. light cloudy, and sunny. The method is simulated in the MAT-
A solar irradiation prediction scheme based on ANN for 10 LAB/Simulink software platform, and the outcome implies that
years of data from 5 cities in India is presented in Premalatha the CNN-SSwA scheme provides better accuracy than long short-
and Valan Arasu (2016) that introduces 4 algorithms: gradient de- term memory (LSTM)-SSwA and SVM-SSwA methods. The mean
scent (GD), Levenberg–Marquardt (LM), scaled conjugate gradient relative error (MRE) and MAPE for the proposed approach for the
(SCG), and resilient backpropagation (RP) to determine the best rainy model are 2.62% and 21.17%, respectively. A CNN and LSTM-
technique. The scheme predicts solar irradiation for two models based solar radiation forecasting model is proposed in Lee et al.
(model 1 and model 2) based on 9 input data: latitude, altitude, (2018) that incorporates solar irradiation level and temperature
longitude, year, month, air temperature, pressure, wind speed, as system parameters. The model is designed to estimate fore-
and humidity. Comparing predicted values with error implies that casting data for a specific location from the weather center even
the proposed ANN model 2 with the LM algorithm is the best- when there are insufficient input variables. A solar irradiation
suited technique. An ANN-based PV output prediction model in prediction study integrating the CNN and the chaotic GA/PSO
Qatar is illustrated in Khandakar et al. (2019) that considers solar hybrid algorithm (CHA) is presented in Dong et al. (2020). The
irradiation, wind speed, humidity, temperature, and accumulated study estimates solar radiation and power output in a PV plant
dust as input data. This work introduces correlation feature selec- based on available input variables. The model reduces the annual
tion (CFS) and relief feature selection (RFS) techniques to evaluate mean absolute error by 0.1463 MJ m−2 than a single CNN scheme.
predicted values from input data. Experimental outcomes show A solar irradiance prediction technique based on CNN and LSTM
that the ANN technique with all features provides the best pre- is proposed in Zang et al. (2020), where CNN manages the spatial
diction with the lowest RMSE without CFS and RFS as 2.1436%. features and LSTM collects the temporal features. The combined
An experiment is carried out in Ghimire et al. (2019) to find the operation of CNN and LSTM provides better performance than
best global solar irradiation prediction from 4 techniques: ANN, other techniques such as SVM, ANN, CNN, ANN-CNN, etc. The
SVM, GPML, and GP. Different errors such as RMAE and RMSE, and MAE for the proposed scheme is reduced at a range (27.78%–
forecasted error (FE) act as prediction metrics to compare among 39.43%) compared to SVM and (22.23%–28.86%) compared to
these techniques. Results show that RMSE for ANN is in the range LSTM. A CNN and LSTM solar radiance prediction technique is
of (1.715–2.27) and RMAE in (7.97–11.74), which is more feasible introduced in Gao et al. (2020) that integrates complete ensem-
than others. A solar irradiance prediction technique based on ANN ble empirical mode decomposition adaptive noise (CEEMDAN)
in Kathmandu, Nepal, is proposed in Panthee and Jha (2016) that to manage existing data from input features of the system. The
takes temperature, humidity, rainfall amount, sunshine hour, and hybrid CEEMDAN-CNN-LSTM approach facilitates separating ex-
solar irradiance as input data. Five different models are estab- isting data’s spatial and temporal features. The hybrid system
lished for various combinations of input data. Simulation results performs better than other systems in different performance
emphasize that model 1 has the best RMSE and CC as 0.2787 assessment metrics (MAE, RMSE, nRMSE). A solar irradiation fore-
and 0.9880, respectively. An ANN-based solar PV temperature casting method based on 3D-CNN is proposed in Zhao et al. (2019)
prediction technique in Turkey is demonstrated in Ceylan et al. that considers ground-based cloud (GBC) images to collect cloud
(2014) that incorporates solar irradiance and ambient temper- information and obtains irradiance forecasting from direct normal
atures (10◦ , 20◦ , 30◦ , and 40◦ ) as input data. The temperature, irradiance (DNI) data. The comparative analysis of outcomes from
irradiance, efficiency, and power are determined every month for the proposed models with different metrics shows a 17.06% im-
8 different locations in the Aegean region in Turkey, where the provement in solar irradiance prediction accuracy. A CNN-based
third and seventh months provide the maximum power output. short-term solar irradiance forecasting method is proposed in
6068
Table 2
A comparative analysis of different ANN-based techniques integrated into PV systems.
Ref. Considered Year Research System Key features Accuracy
approach location configuration
Kalogirou et al. ANN-based solar 2013 N/A Large Solar System • Predicts two factors: daily • 95% confidence limit
(2014) power output + ANN power output and temperature • CC for the daily power
prediction level in the storage tank output is 0.93, which is
• Simple method feasible
• Accurate in case of less
number of inputs
Cervone et al. ANN and 2016 Italy Three PV Power • Short-term (72 hours) solar • RMSE is 8.66%, which
(2017) AnEn-based solar Plants + ANN + power prediction is less than other
power prediction AnEn • NCAR supercomputer to techniques
verify forecasting accuracy
Chow et al. ANN-based PV 2012 Hong Kong PV + ANN • Short-term power prediction • CC varies from 0.618
(2012) power prediction • Solar irradiation, to 0.9305
temperature, azimuth angle, • The confidence limit of
and elevation angle accuracy is 95%
• MLP compares the testing
and training data for
performance evaluation
Yan et al. ANN-based PV 2010 France 1 MW PV • Estimates power reserve • RMSE and MAE of the
(2015) system’s Generator + three from PV system’s uncertainty validation set for power
uncertainty 300 kW micro-gas • Predicts the output of the PV forecasting is 5.28% and
analysis turbine + ANN + system for the next 24 hours 2.93%
BP + NFD • NFD analyzes the power
reserve of the system
Rocha et al. ANN-based solar 2018 Fortaleza, Brazil PV + ANN + BFGS • Prediction of daily, weekly, • MAPE and nRMSE
(2019) irradiation and monthly solar irradiation prove that the monthly
forecasting • BFGS algorithm uses R prediction of more
language accurate
Premalatha and ANN-based solar 2016 India PV + ANN + • Analyzes 10 years of • ANN model-2 with LM
Valan Arasu irradiation LMBP irradiation data for monthly has less error, where
(2016) prediction irradiation prediction MAE and RMSE are
• GD, LM, RP, and SCG for 3.0281% and 3.6461%,
comparison purposes respectively.
Khandakar ANN-based PV 2019 Qatar PV + ANN + CFS • Calibration of sensors to • The lowest RMSE
et al. (2019) power prediction University +RFS accurately measure weather without CFS and RFS is
data 2.1436
• Focuses on minimizing RMSE
• MATLAB for simulation
Ghimire et al. ANN-based solar 2018 Queensland, PV + ANN + SVM • ML for estimating daily solar • RMSE for ANN is
(2019) irradiation Australia + GPML + GP irradiation (1.715–2.27)%
prediction • ARIMA, temperature model • RMAE for ANN is
(TM), and time series and (7.97–11.74)%
Fourier series (TSFS) are
introduced for performance
evaluation
Panthee and ANN-based solar 2016 Kathmandu, PV + ANN + • Considers 5 models to find • Model 1 has the
Jha (2016) irradiation Nepal LMBP the best solutions lowest RMSE of 0.2781
forecasting • Temperature, humidity, and the CC of 0.9880
rainfall, sunshine, and
irradiance, as input data
Ceylan et al. ANN-based solar 2013 Turkey PV + ANN + • Solar irradiance and ambient • Maximum power at
(2014) PV temperature LMBP temperature as input data Aegean region for third
prediction • LMBP optimizes the input and seventh months
data
Al-Waeli et al. ANN-based 2017 Malaysia PV + PCM + ANN • Considers 3 cooling systems: • For nanofluid and
(2018) thermal and PV + MLP tank filled with water, PCM, nano-PCM, current and
power prediction and nanofluid and nano-PCM efficiency increased by
• Estimates the PV output 0.35A and 5.25%,
current and efficiency respectively
Saberian et al. ANN-based PV 2014 Malaysia PV + ANN • Considers temperature and • FFBP is better than
(2014) output power solar irradiance GRNN
prediction • GRNN and FFBP for
comparative analysis
Vanderstar ANN-based solar 2018 Northwestern PV + ANN + GA • Solar irradiation prediction • nRMSE = 10.8%
et al. (2018) irradiation Alberta, Canada for the next 2 hours • CC = 0.912
prediction • Remote solar monitoring
platform
• GA optimizes the sizing and
monitoring of PV system
6069
Table 3
A comparative analysis of different CNN-based techniques integrated into PV systems.
Aprillia et al. CNN-SSwA-based 2020 South Taiwan PV + CNN + • SSwA extracts the best input • MRE = 3.14% and
(2020) PV power SSwA parameters MAPE = 15.27% for the
forecasting • Compared to SVM-SSwA and heavy-cloud model
LSTM-SSwA
Lee et al. CNN and 2018 South Korea PV + CNN + • Accurately measures solar • CNN + LSTM reveals
(2018) LSTM-based solar LSTM output with insufficient input better forecasting
irradiation variables accuracy than AE +
forecasting • Solar irradiation and LSTM
temperature as input
• LSTM can effectively handle
long-term data
Dong et al. CNN-based solar 2018 China PV + CNN + CHA • CHA enhances the prediction • MAE is 0.1463, which
(2020) irradiation + GBRT accuracy by introducing is less than the
forecasting hyperparameters and solves individual CNN
the MILP technique
• GBRT is introduced in
simulation activities
Zang et al. CNN-LSTM -based 2019 Texas, USA PV + CNN + • Short-term prediction using • 27.78%–39.43% MAE
(2020) solar irradiation LSTM spatiotemporal correlation compared to SVM
forecasting • CNN collects spatial features, • 22.23%–28.86% MAE
and LSTM collects temporal compared to LSTM
features
• Compared to ANN and SVM
Gao et al. CNN-LSTM -based 2020 Los Angles, PV + CNN + • CEEMDAN analyzes the • Better performance
(2020) solar irradiation Denver, Hawaii, LSTM + CEEMDAN historical data than CEEMDAN-BPNN,
forecasting Tamanrasset • Predicts solar irradiation of CEEMDAN-SVM, and
four different regions for the CEEMDAN-LSTM
next hour
• Compared with LSTM, BPNN,
and SVM
Zhao et al. CNN-3D -based 2018 China PV + CNN + MLP • GBC collects the cloud • nRME is 30.82%
(2019) solar irradiation information • MAE is 17.91%
forecasting • DNI connects the irradiation • Forecasting accuracy is
data 17.06% higher than
• MLP enhances the others
forecasting accuracy by
extracting input information
Ryu et al. CNN - based solar 2019 Waseda PV + CNN + TSI • Estimates short-term (5–20 • RMSE is (49–177)%,
(2019) irradiation University, min) forecasting (93–146)%, and
forecasting Japan • TSI collects the cloud (71–118)% for sunny,
information and solar partly cloudy, and
irradiance data cloudy days, respectively
Ryu et al. (2019) that integrates a total-sky imager (TSI) device deviation of the PV system can be expressed as
to measure the real-time solar irradiance. The system aims to 3 ∫ T 3 ∫ T
∑ ∑
predict irradiance for the short-term (5–20 min). Two models fmin = |∆fi | dt + |Pcon − Pint | dt , (1)
are developed to test the system’s efficiency for three weather i=1 0 i=1 0
conditions: sunny, partly cloudy, and overcast. Table 3 demon-
where ∆fi is the frequency deviation, Pcon is the power output
strates a comparative analysis of different CNN-based techniques
of the DC/DC converter, and Pint is the power output of the
integrated into PV systems.
DC/AC converter. A wavelet transform (WT), PSO, and generalized
regression neural network (GRNN)-based PV system uncertainty
2.1.3. Application of PSO in Solar PV system quantification technique is proposed in AlHakeem et al. (2015)
An improved PSO technique to determine the global peak that integrates bootstrap confidence intervals (CI) to determine
(GP) for the partial shading conditions (PSC) in the PV system uncertainties. The efficiency of the proposed system is compared
is introduced in Eltamaly et al. (2020), which saves 650% of the with other hybrid systems to verify the system’s effectiveness.
Simulation results in MATLAB represent that the RMSE for day
time to find the GP as no reinitialization is required. The proposed
1 is 0.74, 4.12, 2.04, and 1.84 for winter, fall, summer, and
PSO scheme avoids premature convergence and captures GP more
spring, respectively, which is less than other techniques’ RMSE. A
efficiently than the conventional PSO techniques. A PSO-based
hybrid WT, PSO, and SVM-based PV power forecasting scheme is
optimal Sugeno FLC (SFLC) technique is integrated into a PV plant proposed in Eseye et al. (2018) that incorporates supervisory con-
to maintain robust frequency stability while supplying power trol and data acquisition (SCADA) data with numerical weather
in different areas (Sa-ngawong and Ngamroo, 2015). MATLAB is prediction (NWP) meteorological data. The MAPE and nMAPE are
utilized to perform simulations for 5 different cases considering determined as 4.22% and 0.4%, respectively, implying that the
frequency variations and solar insolation as input variables. The proposed model has better forecasting accuracy than the other
technique is capable of minimizing frequency fluctuations with 7 forecasting models. A hybrid PSO and adaptive neuro-fuzzy
load variations. The objective function, fmin to minimize frequency interference system (ANFIS) is presented in Priyadarshi et al.
6070
Table 4
A comparative analysis of different PSO-based techniques integrated into PV systems.
Eltamaly et al. An improved PSO 2020 Saudi Arabia PV + PSO + MPPT • Determines the GP for PSC • Convergence time
(2020) technique for PV • Avoids premature reduced by 650% for
convergence random reinitialization
• Eliminates reinitialization • Improves stability and
and saves operation time power efficiency
• MATLAB/Simulink for
simulations
Sa-ngawong PSO-based 2014 Thiland PV + PSO + SFLC • Minimizes the frequency and • Better frequency
and Ngamroo frequency + MPPT power fluctuations stability than individual
(2015) stabilization • Frequency variations and PSO and SFLC technique
solar insolation as input data
AlHakeem et al. PSO-based 2015 USA PV + PSO + WT • Determines the uncertainty • RMSE for winter day 1
(2015) uncertainty + GRNN for PV integrated system using is 0.74, for fall is 4.12,
determination Bootstrap CI for summer is 2.04, for
• WT decomposes the original spring 1.84, which is less
input historical data than other techniques
Eseye et al. PSO-based PV 2017 Beijing, China PV + WT + PSO • WT decomposes input • MAPE is 4.22%
(2018) power forecasting + SVM + SCADA information from SCADA and • MAE is 0.4%
NWP
• SVM optimizes the data
• PSO estimates the forecasting
output
Priyadarshi PSO-based MPPT 2018 Denmark PV + PSO + • Tracks maximum power • The least execution
et al. (2019) ANFIS + MPPT point with zero oscillations time is 0.3 s
• Extra sensors are not • Lowest RMSE 0.327
required for solar irradiation
and temperature
• Compares with PSO, ACO,
and ABC algorithm
Semero et al. PSO-GA-ANFIS- 2018 Beijing PV + PSO + GA • Short term power prediction • MAE for PV1, PV2, PV3
(2018) based PV power + ANFIS technique is 23.36%, 5.16%, and
forecasting • GA eliminates unnecessary 12.66%, respectively,
information which is less than ANN
• ANFIS optimizes the forecast
output from PSO-GA
• Compares with ANN, LR, and
persistence model
(2019) to track maximum power points with zero oscillations NN-based solar radiation prediction technique is proposed in
for a PV system. A space vector modulation hysteresis current Boussaada et al. (2018) that considers the cloud and solar char-
controller is integrated into the system to achieve sinusoidal acteristics and sailboat mobility as input parameters for the pre-
reference current from the inverter. The execution time and RMSE diction proceeding. The daily mean of the power error (DMPE)
for the proposed approach are 0.3 s and 0.327, respectively, which is considered the main factor in verifying the feasibility of the
is more feasible than the ant colony optimization (ACO) and proposed approach and can be expressed as
artificial bee colony (ABC) method. A hybrid GA, PSO, and ANFIS ∑N
scheme is demonstrated in Semero et al. (2018), where GA deter- 1 (yim − yip )
mines the significant inputs that influence the power output. The
DMPE = , (2)
N
combination of GA and PSO emphasizes the optimal operation of
where N is the number of pattern pairs, yim and yip are measured
ANFIS. The system verification is continued by integrating some
and predicted solar irradiation, respectively. The simulation out-
other configurations, and the comparison is made considering
comes show that the MSE and DMPE of the approach are 0.00279
the 3 metrics assessment: RMSE, MAE, and nMAE. A comparative
and 24.0584, respectively. A nonlinear intelligent daily PV power
analysis of different PSO-based techniques integrated into PV
forecasting technique based on multivariate adaptive regression
systems is illustrated in Table 4.
splines (MARS) is proposed in Li et al. (2016). The prediction
2.1.4. Application of other intelligent techniques in Solar PV system accuracy of the technique is compared with SVM, ANN, KNN,
A DL based hybrid mapping model to forecast solar irradiance and classification and regression tree (CART). Simulation results
is proposed in Zhen et al. (2020) that collects the information show that the MARS technique performs better with nonlinear
of sky-cloud based on convolution autoencoder (CAE) and k- values than other techniques. An intelligent PV power forecasting
means clustering (KMC) algorithm. The other three techniques: technique based on wavelet transform (WT) and deep CNN is
CNN, LSTM, and ANN, validate the proposed hybrid model’s fore- proposed in Wang et al. (2017). The WT decomposes the origi-
casting accuracy, implying that the hybrid method causes fewer nal input data, and deep CNN detects the nonlinear features of
errors than others. The technique has limitations to predict solar the data to estimate accurate forecasting data. The numerical
irradiance for a distributed PV system and maintaining demand- simulation illustrates that the average MAPE of the proposed
side management. A nonlinear autoregressive exogenous (NARX) technique is 3.85%, where the MAPE for BPNN is 9.33%, and for
6071
Table 5
A comparative analysis of other intelligent techniques integrated into PV systems.
Zhen et al. DL-based solar 2020 China PV + DL + CAE + • Sky image and radiation data • Better accuracy and
(2020) irradiation KMC are extracted through CAE and reliability in CC, MAE,
forecasting KMC and RMSE
• Compared with CNN, LSTM,
and ANN
Boussaada NARX NN based 2018 France PV + NARX • Considers sailboat mobility, • MSE is 0.00279
et al. (2018) solar radiation cloud, and solar characteristics • DMPE is 24.0584
prediction as input
• Integrates the sailboat
operation optimally
Li et al. (2016) MARS-based PV 2015 The Coloane PV + MARS + • MARS estimates accurate • RMSE is 103.2 and
power forecasting island, Macau MLR forecasting from nonlinear MAPE is 28.8%, which is
input variables less than others.
• Compared with ANN, KNN,
CART, and SVM
Wang et al. WT and 2017 Belgium PV + DCNN + WT • A deterministic and • MAPE is 3.85%, where
(2017) DCNN-based PV + Quantile probabilistic forecasting model the MAPE for BPNN is
power forecasting regression (QR) • WT decomposes the input 9.33% and for SVM is
variables (4.93%–10.60%)
• Quantile regression (QR)
handles the uncertainties of
the PV system
• Compared with BPNN and
SVM
Khodayar et al. CGAE-based solar 2019 USA PV + CGAE • A spatio-temporal forecasting • 4% better reliability
(2019) irradiance model than other techniques
forecasting • Considers SGC, DL, and VBI
to design an accurate
forecasting model
SVM, MAPE varies from 4.93% to 10.60%. A convolution graph performance of the proposed VMD-MKRR-MOCWCA technique
autoencoder (CGAE) based solar irradiation forecasting technique is more reliable than others, where the prediction result with
is proposed in Khodayar et al. (2019) that considers spectral graph PINC implies 95% forecasting accuracy. A wind power forecast-
convolutions (SGC), DL, and variational Bayesian interface (VBI). ing technique for a multi-scale wind firm based on the multi
The collected historical solar irradiation data is analyzed and then to multi (M2M) mapping network and stacked denoising auto-
sent to the Spatio-temporal probabilistic forecasting framework encoder (SDAE) is proposed in Yan et al. (2017). The method
to estimate the forecasted output. Results show that the proposed can restore different essential features of multiple wind firms to
CGAE provides better reliability (approximately 4% better) than improve prediction accuracy. Still, there is a lacking of handling
other techniques. Table 5 demonstrates a comparative analysis of forecasting uncertainties for multiple temporal–spatial scales. A
intelligent techniques integrated into PV systems. multi-objective optimization technique incorporated with a hy-
brid flower pollination algorithm (FPA) and a bat search algorithm
2.2. Integration of intelligent techniques in wind power system (BSA) is demonstrated in Qu et al. (2017) that aims to maintain
system stability with accurate forecasting. The forecasting is con-
A hybrid intelligent wind power forecasting technique is pro- tinued by collecting data from 12 wind-speed datasets from 2
posed in Osório et al. (2015) that integrates the adaptive neuro-
wind farms. The simulation results for 6 cases imply that the
fuzzy inference system (ANFIS), evolutionary PSO (EPSO), wavelet
proposed method provides better forecasting accuracy with lower
transform (WT), and mutual information (MI). The hybrid system
SSE, MAE, MAPE, and RMSE. A hybrid wind speed forecasting
improves the forecasting operation with higher accuracy and
technique is proposed in Meng et al. (2016) that considers two
an average MAPE of 3.75%. An intelligent technique to predict
essential parameters: the vertical crossover probability and the
wind speed is demonstrated in Liu et al. (2018) that comprises 4
number of layers in wavelet packet decomposition (WPD). The
different models: variation mode decomposition (VMD), singular
proposed hybrid scheme is comprised of 3 models: WPD, ANN,
spectrum analysis (SSA), LSTM, and extreme learning machine
(ELM). The proposed hybrid technique is compared with dif- and crisscross optimization algorithm (COA). The scheme is vali-
ferent models: ARIMA, LSTM, ELM, VMD-ELM, VMD-LSTM-ELM, dated with the back-propagation algorithm and PSO for different
etc., that illustrates lower MAPE, MAE, and RMSE. A statistical errors: MAE, MAPE, and RMSE.
hybrid wind power forecasting technique (SHWIP) is proposed A wavelet neural network (WNN)-based wind speed forecast-
in Ozkan and Karagoz (2015) that predicts power output based ing technique is presented in Meyyappan (2018) that considers
on historical wind data. The main advantage of the proposed temperature, humidity, wind speed, wind direction, dew point,
method over ANN and SVM is that it requires less amount of and pressure as input variables of the forecasting technique. The
recorded wind data (only 3 months). The limitation arises from study analyzes the economic dispatch (ED) by introducing a shuf-
the proposed scheme is that it cannot accurately determine the fled frog leap algorithm (SFLA) for 5 units, 6 units, and 15 units
individual wind turbine power outcome. A wind speed and wind power system and prime-dual interior point (PDIP). The result
power prediction technique in Spain are presented in Naik et al. shows that the scheme can reduce the cost of operation by (2–
(2019) that consists of 3 models: variational mode decomposition 3)%. A combined wind power forecasting scheme is demonstrated
(VMD), multi-kernel robust ridge regression (MKRR), and multi- in Li et al. (2018), where the first step introduces the empirical
objective chaotic water cycle algorithm (MOCWCA). The overall mode decomposition (EMD) to decompose the input data. The
6072
results show that the NRMSE and MAPE for 1-hour forecasting
are 11% and 14%, respectively, lower than other techniques. A
relevance vector machine (RVM) for wind speed prediction based
on wavelet decomposition (WD) and artificial bee colony (ABC)
algorithm is demonstrated in Fei and He (2015). The feasibility
of the proposed hybrid technique is compared with 9 other RVM
techniques. The result indicates that experiments 1, 2, and 3 have
21.61%, 22.93%, and 19.83% prediction error, resectively which are
less than the other 9 RVM techniques. A stacked autoencoder
(SAE) and backpropagation (BP) algorithm-based wind power
prediction technique is proposed in Jiao et al. (2018), where SAE
Fig. 5. A schematic representation of DCOM and DcapsNet-based multi-objective
optimization model (Al-Janabi et al., 2020).
optimally extracts data from data sequence and BA algorithm can
tune weights for the system network. PSO algorithm is adopted
to determine the best optimal solutions for the extracted data.
The predicted wind power for the proposed hybrid technique has
second step creates 3 hybrid models to improve the forecast-
15.96% MAPE, whereas only the BP algorithm and only the SVM
ing accuracy: EMD-ARIMA, EMD- Elman neural network (ENN)-
technique have MAPE of 47.33% and 27.88%, respectively. The
bat optimization algorithm (BOA), and EMD-BOA-BPNN that are
research does not consider the effect of wind speed uncertainty
implying the MAPE is reduced by 23.21%, 61.56%, and 13.16%,
on wind power prediction. A wavelet neural network (WNN)
respectively. A wavelet transform (WT) and Fuzzy ARTMAP (FA)
based wind power forecasting technique is proposed in Chitsaz
based intelligent technique is presented in Haque et al. (2013)
et al. (2015) that integrates the clonal selection algorithm (CSA) to
that predicts the daily and weekly wind power considering the
measure the optimal solutions from the input data. The maximum
temperature, wind speed, and wind directions. The research work
correntropy criterion (MCC) is introduced in the research work to
is implemented in the Kent Hill wind farm in Canada. The simu-
find errors in training data. The MCC can be expressed as
lation results show that the proposed scheme’s daily and weekly
prediction error reduction are 50% and 30%, respectively com-
{ N
}
1 ∑
pared to other techniques such as BPNN, RBFNN, FA, WT + BPNN, MCC = max G(ϵi , σ ) ,
2
(4)
etc. The research is not capable of determining the uncertainty for N
i=1
predicted power. A CART-Bagging algorithm-based intelligent ap-
where N is the number of training samples, G is the Gaussian
proach is presented in Fischer et al. (2017) that collects observed
Kernel, ϵi is the error of ith samples, and σ 2 is the variance. The
data (wind speed, power output) from a wind farm for 10 min
forecasting error of the proposed approach for a 1-h and 6-h
intervals. The predicted power output for only wind speed can
ahead forecast is 5.04% and 15.94%, respectively.
be expressed as
A neural network (NN)-based integral sliding mode control
C (ISMC) is proposed in Qian et al. (2016) that maintains load
Pt = , (3)
1 + exp(a0 + a1 Wt + a2 Wt2 + a3 Wt3 ) frequency control (LFC) for a wind integrated power system. The
Lyapunov direct method (LDM) is employed to handle the weight
where Pt is the predicted power output, Wt is the wind speed at
update law for the NN. The radial basis function (RBF) NN is inte-
time t. C and ai are estimated parameters, where i = 0,1,2. . . N.
grated into the research to eliminate generation rate constraints
The simulation result shows that the estimated power output
(GRC) and system uncertainty. The simulation result implies that
error for the proposed algorithm for only the wind speed vari-
the proposed approach can maintain wind-based power system
able is 2.04% and for all variables is 1.65%, compared to other
intelligent techniques such as linear regression, logistic regres- stability. An artificial wavelet neural network (AWNN) and multi-
sion, CART, SVM, etc. A multi-objective optimization model for resolution analysis (MRA)-based wind speed prediction technique
wind power generation is developed in Al-Janabi et al. (2020) is proposed in Doucoure et al. (2016), where the Hurst coefficient
based on two optimization models: develop cuckoo optimization (HC) is employed to eliminate less necessary input information.
model (DCOM) and develop capsule network model (DCapsNet), The simulation outcomes show that the prediction error is re-
as shown in Fig. 5. DCOM is best suited for a system having fewer duced by 29%. A fuzzy k-means clustering (FKC) algorithm and en-
input variables. Simulation results claim that DCapsNet is better hanced harmony search (EHS) algorithm-based short-term wind
than DCOM in determining an optimization model with minimum power forecasting technique is proposed in Huang et al. (2017)
cost and the highest efficiency. that collects the wind speed as input data. FKC algorithm analyzes
A multi-objective salp swarm optimization algorithm (MSSOA) the input data and detects the pattern of data. The EHS model
for forecasting wind power is proposed in Jiang et al. (2019) is developed using multiple support vector regression (SVR). The
that considers only wind speed. The noise of the wind speed feasibility of the proposed scheme is evaluated by comparing it
data is eliminated by introducing the variational mode decom- with ANN, RNN, SVR, and PSO-based SVR techniques. The simu-
position method. The MSSOA determines the optimal solutions, lation results show that the proposed technique has an MRE of
and then a fuzzy-based mechanism is employed to find the best the 2.2326%, which is less than other techniques. The schematic
optimal solution. Results show that the proposed scheme pro- representation of the proposed technique is demonstrated in
vides the highest possible prediction accuracy than other men- Fig. 6.
tioned techniques. An intelligent wind power forecasting tech- An RNN-based wind power forecasting technique is proposed
nique in Canada based on wavelet transform, hybrid neural net- in Dong et al. (2018) that introduces the LSTM model to collect
works (HNNs), and the imperialist competitive algorithm (ICA) historical wind speed data for prediction purposes. The predic-
is proposed in Aghajani et al. (2016) that analyzes the effect of tion accuracy for the LSTM model is compared with the linear
6 input parameters: temperature, pressure, wind power, wind regression (LR), random forecast (RF), and gradient boosting (GB)
speed, wind direction, and humidity. The input data is filtered method and proves the accuracy of the proposed scheme. The re-
using wavelet transform, and different learning algorithms such search cannot provide the idea for the series of input lengths and
as Levenberg–Marquardt (LM), Bayesian Regularization (BR), etc., forecasting uncertainty. An ANN-based wind speed prediction
are employed to measure optimal power prediction. Simulation technique is demonstrated in Filik and Filik (2017) that considers
6073
Fig. 6. A schematic representation of FKC and EHS algorithm based power forecasting technique (Huang et al., 2017).
Fig. 7. A schematic representation of the two-hybrid wind speed forecasting technique (Liu et al., 2015).
wind speed, wind direction, temperature, and pressure as input are 5.75% and 9.20% less than other models. A wavelet-based
variables. The wind speed prediction is carried out for three ensemble (WNE) model for wind power forecasting is demon-
different combinations of inputs: only wind speed, wind speed strated in Li et al. (2015) that integrates NN, WT, partial least
+ temperature, and wind speed + wind pressure + tempera- squares regression (PLSR), and feature selection to improve the
ture. The simulation results show that the maximum accuracy prediction accuracy. The feature selection process extracts the
for wind speed prediction is found for the third combination of essential input features from the existing features to develop
input variables. A GA, adaptive PSO (APSO), and WNN-based wind the WNE model. The simulation results in case-1 show that the
speed forecasting technique is illustrated in Wang et al. (2016) proposed scheme is 82.5% and 77.9% better than the PER and
that incorporates an ensemble empirical mode decomposition NR models, respectively, in MAPE. A nonlinear hybrid intelligent
(EEMD) to find valuable data from collected input data. The WNN technique for wind speed forecasting based on LSTM, hysteresis
model is developed by incorporating GA and APSO to optimize extreme learning machine (HELM), and differential evolution (DE)
input parameters and provide maximum forecasting accuracy. algorithm is proposed in Hu and Chen (2018). HELM and LSTM
The input–output mapping for WNN can be expressed as network generates hidden layers, and the DE algorithm optimizes
h the hidden layers. The developed forecasting model is validated
∑ x2
Yk = Wkj Cos(1.75x) exp(− ), (5) for two scenarios: utmost short-term and short-term wind speed
2 forecasting. The prediction outcomes are compared with ANN,
j=1
ARIMA, ELM, and SVM, emphasizing that the proposed approach
where Yk is the output, Wkj is the weight, and x is the input
is better than others. Table 6 demonstrates a comparative analysis
of the network. The MAPE and MSE for the proposed intelligent
of different intelligent techniques integrated into wind systems.
technique are 2.4907% and 0.0271, which is more accurate than
other techniques (WNN, FLC, BP, SVM, etc.).
2.3. Integration of intelligent techniques in hybrid RESs
Two-hybrid wind speed forecasting approaches such as fast
EEMD (FEEMD) + mind evolutionary algorithm (MEA) + multi
layer perception NN (MLPNN) and FEEMD + GA + MLP are A DL-based hybrid RESs energy forecasting technique in Jeju
proposed in Liu et al. (2015). The FEEMD decomposes the input Island, the Republic of South Korea, is proposed in Nam et al.
data, and then MEA or GA is introduced to develop the MLPNN (2020) that determines accurate forecasting considering techni-
model, which can optimize the input data to determine accurate cal, economic, and environmental feasibility. The approach of-
forecasting. Experimental results emphasize that the FEEMD + fers 4 scenarios based on available RESs and storage systems.
MEA + MLPNN-based approach is more promising than another EMD decomposes the collected input data from PV and wind
hybrid approach in MAE, MAPE, and RMSE. The schematic repre- farms, and then the DL technique evaluates the predicted values
sentation of the workflow of the proposed approach is illustrated for individual scenarios. Simulation results show that scenario
in Fig. 7. 2 decreases the combined economic and environmental cost by
A radial basis function neural network (RBFNN) model is 36.16%, implying that scenario 2 is the most suitable for the
established to predict wind power is proposed in Li and Liao island. The integrated gasification combined cycle (IGCC) provides
(2018) that introduces multiple SVM to understand wind power advantages for scenario 2 by reducing pollution and supply-
gradient events. The maximum overlap discrete wavelet trans- ing sustainable energy. An ANN-based hydropower optimization
form (MODWT) extracts the input data. Adaptive wavelet NN technique is proposed in Shaw et al. (2017) that considers the
(AWNN) is utilized in the experiment for prediction purposes up high-fidelity hydrodynamic and water quality model (CE-QUAL-
to 24 hours ahead. The prediction output of the AWNN model can W2) to predict the system power output. The approach takes
be expressed as meteorological, calibration, and inflow data as input variables.
m n The GA is integrated into the approach to maximize the power
output. The objective function to maximize system output can be
∑ ∑
y= ωj zj + v i ui + g , (6)
expressed as
j=1 i=1
n
where y is the prediction output, ωj is the jth wavelet and output ∑
Objmax = max c (i) × x (i) × r , (7)
node, vi is the ith input node and the output node, and g is the
i=1
arithmetic mean deviation of the input node. The effectiveness
of the model is evaluated by comparing it with the feedforward where n is the number of hours, c (i) is the power at the time i,
NN (FFNN), persistence model (PER), and new reference (NR) and r is the turbine power rating in MW. Experiment results show
model. The MAE and RMSE of the proposed prediction approach that the cost for achieving optimal power output is reduced from
6074
Table 6
A comparative analysis of different intelligent techniques integrated into wind systems.
Osório et al. Hybrid wind 2014 Portugal Wind + ANFIS + • Short-term power forecasting • Average MAPE is 3.75%
(2015) power forecasting EPSO + WT + MI • WT decomposes input less than other
variables techniques (persistence,
• MI reduces the randomness ARIMA, NN, etc.)
of input data
• EPSO and ANFIS enhance the
forecasting accuracy
Liu et al. DL-based wind 2017 China Wind + DL + • VMD decomposes the • Less RMSE, MAE, and
(2018) speed forecasting VMD + SSA + original input data MAPE than other
LSTM + ELM • LSTM and ELM forecasts for techniques (ARIMA, ELM,
low frequency and LSTM, and their different
high-frequency layers, combinations
respectively
Ozkan and SHWIP-based wind 2015 Turkey Wind + SHWIP + • Short-term power forecasting • The lowest MAE is
Karagoz (2015) power forecasting NWP (next 48 hours) 11.80%
• Accurate forecasting with a
fewer input information
• Compared with ANN and
SVM
Naik et al. Wind speed and 2018 Spain Wind + VMD + • VMD decomposes input data • 95% forecasting
(2019) power forecasting MKRR + MOCWCA • Compares to VMD-KRR accuracy
models
Yan et al. M2M mapping 2017 China Wind + M2M • Forecasts power output for • nRMSE for the
(2017) and SDAE wind mapping + SDAE the next 24–72 hours ensemble SDAE and only
power forecasting • Compared with only SDAE, SDAE is 9.6% and 10.5%,
NN, and SVM respectively
Qu et al. (2017) FPA-BSA-based 2017 Shandong, Wind + FPA + • Improving multi-objective • Reduces MAPE by
wind power China BSA optimization using the 9.1549%, 6.1139%, and
forecasting ensemble model 18.9854% than
• Compared with PSO, FPA, SSA-BPNN, SSA-RNN, and
and BA SSA-GRNN, respectively
Meng et al. Hybrid short-term 2015 Netherlands Wind + WPD + • WPD decomposes the • MAE, MAPE, and RMSE
(2016) wind speed NN + COA original data, and NN-COA for one-step are 0.1365,
forecasting estimates the forecasted output 1.214%, and 0.2216,
• Compared with PSO and respectively, which is
BPNN less than other hybrid
• MATLAB for simulation techniques
Meyyappan WNN-based wind 2017 India Wind + WNN + • Aims to minimize the power • Manages ED problems
(2018) speed forecasting ANN + ED + SFLA generation cost by reducing the
+ PDIP • Forecasts wind speed for the operation cost by 2%–3%
next 24 hours
• SFLA and PDIP enhance ED
Li et al. (2018) Short-term wind 2017 Penglai, China Wind + EMD + • EMD decomposes the • MAPE is reduced by
power forecasting BPNN + ARIMA + random and nonlinear input 23.21%, 61.56%, and
ENN + SVR data 13.16% for EMD-ARIMA,
• Forecasting model using the EMD-ENN-BOA, and
different combinations of EMD-BOA-BPNN,
BPNN, ARIMA, and ENN respectively
Haque et al. WT- FA-based 2013 Brunswick, Wind + WT + FA • WT extracts input • MAPE is reduced by
(2013) wind power Canada + information 50% and 30% for daily
prediction • FA observes the non-linearity and weekly power
and fluctuations of information prediction, respectively
• Compared with the different
combinations of BPNN, ANFIS,
WT, and RBFNN
Fischer et al. CART-Bagging 2016 France Wind + • Observes wind speed and • Reduces prediction
(2017) based wind power CART-Bagging power output error by 2.04% for only
prediction • Compared with linear wind speed as input and
regression, logistic regression, 1.65% for all input
CART, SVM, etc. variables
Al-Janabi et al. DL-based model 2020 Iraq Wind + DCOM + • Considers environmental and • DCapsNet is better
(2020) for power DCapsNet economic benefits than DCOM in costing
generation • DCOM efficiently performs and efficiency
with fewer features
• A comparative study
between DCOM and DCapsNet
(continued on next page)
6075
Table 6 (continued).
Jiang et al. MSSOA-based 2018 Shandong, Wind + MSSOA + • Only considers wind speed as the • The highest possible
(2019) wind power China Fuzzy input prediction accuracy
forecasting • A fuzzy mechanism estimates the
best solution
Aghajani et al. Intelligent wind 2015 Alberta, Canada Wind + WT + • WT decomposes input information • nRMSE and MAPE is
(2016) power prediction HNN + ICA + LM • LM, ICA, and BR optimizes reduced by 11% and 14%,
+ BR forecasted output respectively
Fei and He RVM for wind 2014 China Wind + RVM + • WD decomposes input information • The prediction error is
(2015) speed prediction WD + ABC • ABC detects the optimal kernel reduced by 21.61%,
parameters 22.93%, and 19.83% for
experiments 1, 2, and 3,
respectively
Jiao et al. SAE-BP based 2018 China Wind + SAE + BP • Short-term power prediction • Improves accuracy by
(2018) wind power + PSO • PSO is incorporated to estimate 12%
prediction optimal forecasting
• Compares to BPNN and SVM
Chitsaz et al. WNN-based wind 2014 Alberta, Canada Wind + WNN + • CSA optimizes the prediction • Improves accuracy by
(2015) power forecasting CSA + MCC output 30.29% for 1-hour ahead
• MCC is used as feasibility metrics forecasting
Qian et al. ISMC for load 2015 China Wind + ISMC + • LDM manages the weight update • Effective method to
(2016) frequency control NN + RBF law control load frequency
• RBF eliminates system uncertainty
and nonlinearity
Doucoure et al. AWNN-MRA based 2014 Canada Wind + AWNN + • Considers only wind speed as input • The prediction error is
(2016) wind speed MRA + HC information reduced by 29%
forecasting • HC and MRA collect the essential
input information
Huang et al. KFC-EHS based 2016 Taiwan Wind + KFC + • Short-term (next 3-hours) power • MRE is reduced by
(2017) wind power SVR prediction 2.2326%
forecasting • EHS estimates parameters for the
SVR model
• Compared with ANN, RNN, SVR,
and PSO-SVR
Dong et al. RNN-based wind 2018 Canada Wind + RNN + • Wind speed data as input • Better accuracy than
(2018) power forecasting LSTM information others
• Compared with LR, RF, and GB
Filik and Filik ANN-based wind 2016 Eskisehir, Wind + ANN • Wind speed, wind direction, • The most accurate
(2017) speed forecasting Turkey temperature, and pressure as input forecasting for wind
• Forecasting for different speed + pressure +
combinations of input variables temperature
Wang et al. Hybrid intelligent 2015 China Wind + GA + • GA and APSO optimize the input • MAPE and MSE are
(2016) wind speed APSO + WNN parameters reduced by 2.4907% and
forecasting • WNN model estimates forecasted 0.0271, respectively
output
• Compared with WNN, FLC, BP, and
SVM
Liu et al. Hybrid wind 2015 China Wind + FEEMD + • FEEMD decomposes the input • FEEMD + MEA +
(2015) speed forecasting MEA + MLPNN + information MLPNN provides better
GA • Feasibility analysis for different accuracy than others
combinations of algorithms
Li and Liao RBFNN-based wind 2018 Hunan Wind + RBFNN + • MODWT extracts input information • MAE and RMSE are
(2018) power prediction Province, China SVM + AWNN • MATLAB for simulation reduced by 5.75% and
• Compared with FFNN, PR, and NR 9.20%, respectively
Li et al. (2015) WNE for wind 2015 N/A Wind + WNE + • A feature selection technique • 82.5% and 77.9%
power forecasting NN + WT + PLSR • MATLAB for simulation improvement of
forecasting accuracy
compared to PER and
NR, respectively
Hu and Chen Hybrid intelligent 2018 Inner Mongolia, Wind + LSTM + • Short-term and long-term wind • Better performance
(2018) wind speed China HELM + DE speed prediction than other techniques
forecasting • Manages the nonlinearity of input
information
• Compared with ANN, ARIMA, ELM,
and SVM
$812,750 to $316,000, demonstrating the system’s feasibility. The are demonstrated in Chatziagorakis et al. (2016) to achieve an
operational flow of the proposed approach is illustrated in Fig. 8. optimal power management system (PMS). The experiment of
An RNN and nonlinear autoregressive (NAR) architecture- the proposed method is carried out in Olvio, Greece. Simulation
based prediction approach for solar PV and wind hybrid systems outcomes show that the estimated regression coefficient is 0.93
6076
is included in the system to assist PV to handle frequency fluc-

tuations and reduce power loss. A PSO-based techno-economic
optimization algorithm is proposed in Sanchez et al. (2014) to
ensure optimal sizing for wind–PV–fuel cell-based hybrid power
systems. Solar insolation, wind speed, and load demand are con-
sidered the input data of the system. The scheme aims to reduce
investment, replacement, operation, and replacement costs. The
objective function, fobj to minimize costs can be expressed as
∑ ( ) ∑
fobj = min[ CACm + COMm + CCREMP n }, (8)
m=PV ,elec ,Wind,FC n=elec ,FC
where CACm is the annualized investment cost, COMm is the op-

eration and management cost, and CRAn is the replacement an-
nualized cost. The simulation result shows that the PSO-based
Fig. 8. A schematic representation of ANN-based hydropower optimization operation is faster than DE and saves 10% of the cost.
technique (Shaw et al., 2017). An evolutionary algorithm-based approach to evaluate opti-
mal sizing for hybrid PV, wind, and battery power systems is
proposed in Maleki and Pourfayaz (2015). The approach aims to
to 0.99, which is significant to maintain green energy efficiently. provide optimal sizing of the hybrid system to minimize system
The overview of the proposed method is demonstrated in Fig. 9. cost and achieve low power supply probability (LPSP). Differ-
An intelligent approach for predicting the hydropower in the ent evolutionary algorithms: PSO, tabu search (TS), simulated
Amazon region is proposed in Lopes et al. (2019) that evaluates annealing (SA), improved harmony search-based simulated an-
two techniques, such as the group method of data handling nealing (IHSBSA), artificial bee search optimization (ABSO), and
(GMDH) and conventional ANN, to find the best prediction perfor- improved harmony search (HIS) are employed in the scheme to
mances. The GHDH method includes the DL concept and polyno- verify the performance specifications. Simulation results show
mial NN to continue the prediction operation, whereas the other that ABSO provides better performance than others, and the hy-
technique integrates the multilayer perception (MLP) and feed- brid PV–battery system is best suited for cost efficacy for 5% LPSP.
forward BP algorithm. The ANN method is implemented for ANN An automatic generation control (AGC) approach for renewables
+ Levenberg Bayesian (LB) and ANN + Bayesian Regulation (BR). integrated power systems is proposed in Arya (2019) that incor-
Simulation results imply that GMDH has a minimum prediction porates capacitive energy storage (CES) and multi-stage muzzy
error (7.92%) than others and emphasize that the GMDH is the assisted PID with filter (FPIDF-(1 + PI)) controller. The approach
best suited for hydropower plants. Another intelligent technique aims to balance supply and demand response in frequency fluc-
for predicting hydropower is demonstrated in Dehghani et al. tuations and load disturbances. Simulation outcomes show that
(2019) that incorporates the Grey wolf optimization (GWO) and the proposed approach performs better than other considered
adaptive neuro-fuzzy inference system (ANFIS). 20 input–output approaches, such as PI, PID, FPID, and FPIDF. An intelligent tech-
combinations are considered to verify the prediction accuracy nique based on ANN for hybrid RESs is illustrated in Lujano-Rojas
between ANFIS and GWO + ANFIS configurations. Simulation et al. (2013) that considers the uncertainty for solar irradiation,
outcomes demonstrate that the ANFIS cannot accurately predict wind speed, battery lifetime, and fuel prices. A GA optimization
9 combinations out of 20, and the GWO + ANFIS approach has model is developed by the combined approach of ANN and MCS
simulation. The numerical simulation outcomes that the RMSE,
better performance than ANFIS. A DL-based supply–demand fore-
MAE, and R-squared are 36.67%, 16.09%, and 0.9836, respec-
casting model for a hydrogen-based self-sustaining integrated
tively, which proves system‘s feasibility and reliability. A firefly
renewable energy network (HySIREN) is proposed in Hwangbo
algorithm (FA) for optimal sizing of wind and PV-based hybrid
et al. (2019), which is implemented in Jeju Island in the Republic
systems is proposed in Kaabeche et al. (2017) that considers the
of Korea. The research mainly focuses on improving economic
load dissatisfaction rate (LDR) and energy cost (EC) to maintain
feasibility. The EMD is employed to decompose the historical data
system reliability and minimum costing. The objective function
for DL that evaluates the forecasting value of the supply–demand
for minimizing costing, ECmin can be expressed as
model by the deep neural network (DNN), gated recurrent unit
(GRU), and LSTM. The approach also focuses on balancing the LCC × CRF
ECmin = ∑8760 , (9)
hydrogen in the environment. Results show that 64.5% of the
t =1 Eload (t)
cost for environmental feasibility is reduced in the experiment. A
stochastic framework for hybrid RESs is proposed in Arabali et al. where LCC is the life cycle cost, CRF is the capital recovery
(2013) to achieve cost efficiency and reliability. The sequential factor, and Eload (t) is the required energy at time t. The proposed
Monte Carlo simulation (SMCS) is introduced in the approach approach is simulated in MATLAB and compared with different
to reduce system cost. The uncertainty of hybrid RESs such as intelligent techniques such as accelerated PSO (APSO), general-
wind and solar PV are handled by autoregressive moving aver- ized evolutionary walk algorithm (GEWA), and bat algorithm (BA)
that prove the economic feasibility of the proposed FA technique.
age (ARMA). Simulations are carried out for two configurations:
An ANN-based hydropower prediction technique is proposed in
hybrid systems with reliability constraints and without reliability
Hammid et al. (2018) that considers the rate at which water
constraints. A fuzzy-based frequency control approach for a PV
flows through a hydropower plant. ANN includes BP to estimate
and diesel-based power system is demonstrated in Datta et al.
power production based on 10 years of historical data. The predic-
(2010) that aims to eliminate frequency fluctuation of the PV
tion accuracy is checked using MSE, MAE, RMSE, and correlation
system and ensure maximum energy utilities. Solar PV cannot
coefficient (CC). The CC can be expressed as
supply power at night, and the power requirements at night √∑
depend on diesel generators. At daylight, the PV system and diesel n
| (hn − hm ) |2
generators work to supply load demand, but power loss occurs CC = 1 − 1
, (10)
to the PV system to maintain frequency fluctuations. Hence ESS hn
6077
Fig. 9. A schematic representation of RNN and NAR based PV and wind power prediction technique (Chatziagorakis et al., 2016).
where hn is the measured output values and hm is the fore- where f1 is the objective function for optimal BESS sizing, f2 is
casted output values. Simulation results demonstrate that the for minimum capital cost, and f3 is for minimum operation and
MASE and CC of the proposed approach are 0.0032734 and 0.96, management costs. Simulation results in DIgSilENT PowerFactory
respectively, which emphasizes the best prediction performance. claim that the proposed approach can control system frequency
The integration of RESs into power systems creates uncertain- through DR, enabling the BESS to provide optimal power flow
ties for grid utility. A protective approach for RESs integrated during regular and emergencies. A modified model predictive
systems is demonstrated in Ram Ola et al. (2020) that incorpo- control (MMPC) technique is applied to a bidirectional ac/dc
rates the current-based Wigner distribution index (WD-index) converter connected to ESS (Akter et al., 2015). The method in-
and the voltage-based alienation index (ALN-index) to estimate corporates the Lyapunov function (LF) with the MMPC to improve
the fault index (FI). The comparative analysis of the proposed the charging capacity, dynamic response, and nonlinear stability.
scheme with the discrete wavelet transform (DWT) and Stockwell The quantization error of the converter is reduced to provide fast
transform (ST) emphasizes that the proposed technique is much operation with a low execution time. The simulation results in
faster than others in detecting faults and is not affected by low MATLAB/Simulink show that the execution time for the proposed
noise. The equation of the WD-index can be expressed as approach is reduced by 18% compared to conventional MPC tech-
niques. An optimal control strategy for battery-supercapacitor
∞
τ) τ
∫ (
WD_index = I t+ I ∗ (t + )e−jωτ dτ , (11) (SC)-based hybrid ESS is demonstrated in Chong et al. (2016)
−∞ 2 2 that integrates low pass filter (LPF) and FLC to enhance battery
where I(t) is the current signal, ω is the angular signal frequency, life. The main objective of the proposed approach is to reduce
and τ is the time domain-based signal function. Again the ALN- dynamic response by LPF and the battery peak current by FLC.
index in terms of the correlation coefficient (R2 ) can be expressed The study considers 4 models to present a comparative anal-
as ysis: standalone PV with battery, standalone PV with battery
and SC, standalone PV with battery, and SC integrated to high
ALN − index = 1 − R2 , (12) pass filter and filtration based controller (FBC), and the proposed
∑ ∑ ∑ system configuration. Simulation results in Simulink show that
Ns xy − ( x)( y) the proposed configuration can improve the dynamic stress and
and R = √
∑ (∑ )2 ∑ (∑ )2 , (13)
peak current demand. A hybrid differential evolution optimiza-
[Ns x2 − x ][Ns y2 − y ] tion (DEO) and ANN-based intelligent PI controller for BESS is
where x and y is the voltage variable and Ns is the numbers in a proposed in Alshehri et al. (2019) to enhance the power quality
cycle. and minimize the sudden disturbance by maintaining voltage
A reliable and sustainable solution for RESs integrated reverse and frequency at the steady-state level. The objective function to
osmosis (RO) desalination technique is proposed in Li et al. (2019) improve power quality can be expressed as
that aims to forecast power output for responding to water de-
∫ t ∫ t
mand through the RO process. The forecasting is implemented by min (∆ω)2 dt = (ω0 − ω)2 dt , (17)
0 0
RNN considering ambient temperature, wind speed, solar radia-
tion, and water demand. Extended mathematical programming where ω0 is the reference frequency and ω is the system fre-
(EMP) is introduced to maintain multi-objective optimization quency. Simulation results in MATLAB emphasize that the con-
that provides economic and environmental feasibility. Simulation troller can effectively flow the active and reactive current in
results show that case B-1 has the lowest cost and GHGs emission. steady-state conditions to improve power quality. An FLC-based
Again the case P-1 has better performance than others in terms of intelligent control technique for the grid, RESs, and ESS connected
economic, environmental, and energy reliability. Table 7 demon- microgrid is proposed in Teo et al. (2018) that aims to reduce
strates a comparative analysis of different intelligent techniques power flow between sources and loads and minimize costs. The
integrated into combined RESs systems. operating cost, Coc of the proposed approach can be expressed as
T
∑
2.4. Integration of intelligent techniques in ESS Coc = [Pbal + Pc (t ) − Pd (t )] Cp (t), (18)
t =1
A PSO-based optimal battery ESS (BESS) system is proposed in
where Pbal is the difference between actual load and renewable
Kerdphol et al. (2016) that considers the demand response (DR)
energy, Pc (t) and Pd (t ) are the charging and discharging power
to provide a fast, stable, and secure BESS system. It is a multi-
output at time t, and Cp is the electricity price. The proposed ap-
objective optimization technique whose objective functions can
proach is evaluated in MATLAB/Simulink platform and compared
be expressed as
with self-adaptive ESS controller (SAEC) and mixed-integer linear
f1 = min(PBESS ), (14) programming (MILP). Results show that the proposed scheme
has 1.87% less operating cost than SAEC. The overview of the
f2 = min(CCapital ), (15) proposed scheme is illustrated in Fig. 10.
An improved bat algorithm (IBA) for optimal sizing of BESS
is demonstrated in Bahmani-Firouzi and Azizipanah-Abarghooee
f3 = min(CO&M ), (16) (2014) to obtain the optimal operation with minimum cost for
6078
Table 7
A comparative analysis of different intelligent techniques integrated into combined RESs systems.
Nam et al. DL-based RESs 2019 Jeju Island, Wind + PV + FC • Maintains technical, • The environmental and
(2020) energy forecasting South Korea + EMD + IGCC economic, and environmental economic cost is reduced
feasibility by 36.16% for scenario-1
• EMD decomposes input
information of wind and PV
plant
Shaw et al. ANN-based 2017 Nashville, Hydropower + • ANN creates CE-QUAL-W2 • Reduces power
(2017) hydropower Tennessee, USA ANN + GA + model for power prediction generation cost by 61.2%
optimization CE-QUAL-W2 • GA maximizes the • Increases production
hydropower generation by 6.8%
Chatziagorakis RNN-NAR based 2015 Olvio, Greece PV + Wind + • Daily and hourly energy • RC enhanced from
et al. (2016) RESs power RNN + NAR prediction 0.93 to 0.99
prediction • Regression coefficient (RC) is
the efficacy of generating
green energy
Lopes et al. Hydropower 2019 Amazon region Hydropower + • Two different ANN approach: • GMDH is better than
(2019) prediction ANN + MLP + GMDH and MLP-FFBP others in that reduces
approach GMDH + FFBP • Compared with ANN + LB error by 7.92%
and ANN + BR
Dehghani et al. GWO-ANFIS based 2018 Iran Hydropower + • Comparative analysis • ANFIS + GWO shows
(2019) hydropower GWO + ANFIS + between ANFIS and ANFIS + better forecasting
prediction GWO accuracy than ANFIS
• Integration of ANFIS with
GWO reduces the uncertainty
of input information
Hwangbo et al. DL-based 2018 Jeju Island, RESs + DL + • HySIREN provides an optimal • Reduces 64.5% of
(2019) supply-demand South Korea HySIREN + EMD supply and demand forecasting environmental cost
forecasting model
• EMD decomposed the
original input information of
RESs
Arabali et al. A stochastic 2013 USA Wind + PV + ESS • ARMA eliminates the • Reduces system cost
(2013) framework for + SMCS + ARMA uncertainty in RESs and enhances reliability
hybrid RESs • Aims to reduce investment
cost
• Optimal energy distribution
among sources and loads
Datta et al. Fuzzy based 2010 Okinawa, Japan PV + Diesel + • Considers frequency • Minimizes frequency
(2010) frequency control Fuzzy Logic fluctuations, average insolation, fluctuations and
and change in insolation as operation cost
input
• Compared with conventional
MPPT and ESS
Sanchez et al. PSO based 2014 Mexico Wind + PV + • Aims to maintain optimal • Reduces cost by 10%
(2014) techno-economic Hydrogen + PSO sizing of power plant with
optimization + FC minimum cost by PSO
• Compared with DE
Maleki and An evolutionary 2014 Rafsanjan, Wind + PV + • Optimal sizing and power • Reduces system cost
Pourfayaz algorithm for South of Iran Battery + LPSP distribution by 5%
(2015) optimal sizing of • MATLAB for simulation
RESs • Compared with PSO, SA, TS,
IHSBSA, ABSO, HIS
Arya (2019) AGC approach for 2018 India PV-Thermal + • Eliminates frequency • Achieves optimal
supply–demand Hydro-Thermal + fluctuation due to load performance than other
response CES + disturbance techniques
FPIDF-(1+PI) • MATLAB for Simulation
• Compared with PI, PID, FPID,
and FPIDF
Lujano-Rojas ANN-based hybrid 2012 Zaragoza, Spain Wind + PV + FC • Analyzes the uncertainty for • RMSE, MAE, and
et al. (2013) RESs + ANN + GA + solar irradiation, wind speed, R-squared are 36.67%,
Battery FC, and battery prices 16.09%, and 0.9836,
• A hybrid ANN and MCS respectively
approach
Kaabeche et al. FA for optimal 2017 Algeria Wind + PV + FA • FA considers LDR and EC to • Reduces system cost
(2017) sizing maintain cost efficacy by 36%
• Compared with APSO, GEWA,
and BA
6079
Hammid et al. ANN-based 2016 Himreen Lake Hydropower + • Considers the rate of water • MASE and CC are
(2018) hydropower dam, Iraq ANN + BP flow 0.0032734 and 0.96,
prediction • Collects 10 years of historical respectfully
data on water flow rate and
power output
Ram Ola et al. WD-index and 2020 India Wind + PV + • Estimates FI from WD-index • Better performance
(2020) ALN-index-based WD-index + and ALN-index than DWT, ST, and
uncertainty ALN-index + FI • IEEE-13 node network for standalone ALN-index
analysis system modeling
• Compared with DWT and ST
Li et al. (2019) RO desalination 2018 UK Wind + PV + RO • Considers economic and • Reduces power supply
technique + RNN + EMP + environmental feasibility loss by 18.3%
BSS • EMP minimizes the operation
cost and GHGs emissions
• RNN for wind speed and PV
power forecasting
Fig. 10. A schematic representation of FLC-based intelligent control technique (Teo et al., 2018).
RESs connected microgrid (MG). The algorithm considers the García-Santacruz et al. (2021). The proposed system is a multi-
charging and discharging efficiency, demand response, power optimization approach that maintains optimal sizing and power
reserve, and charge storage capability to formulate the model for flow of ESS, considering economic feasibility. EBA estimates the
optimal sizing. The objective function for minimum cost, fcost can priority for optimal management of power flow between sup-
be expressed as pliers and consumers. The multi-objective optimization can be
NT
accomplished by introducing the Pareto frontier. The net present
value (NPV) and the internal rate of return (IRR) are significant
∑
fcost = min ft + OMDG + TCDBES , (19)
for obtaining optimal sizing and can be expressed as
t =1
n
∑ CF
and ft = Costgrid,t + CostDG,t + CostBES ,t + SUCFC ,t + SUCMC ,t NPV = −I0 + , (21)
(1 + k)t
t =1
+ SDCFC ,t + SDCMC ,t , (20)
n
where OMDG is the operation and maintenance cost of the dis- ∑ CF
tributed generator (DG), TCDBES is the total cost per day of BES, I0 + = 0, (22)
(1 + IRR)t
Costgrid,t and CostDG,t are the cost of trade in the grid and the DG t =1
at time t, SUCFC ,t and SUCMT ,t are the shut-up cost of fixed cost where I0 is the total investment, k is the discount rate, and CF is
(FC) and maintenance cost (MC). SDCFC ,t and SDCMC ,t are the shut- the annual cash flow from the investment. Results show that the
down cost of FC and MC. The approach considers three cases: case proposed method is profitable for BESS connected power grid.
A considers only MG, case B considers MG with BESS without An adaptive wavelet transform (AWT) and FLC-based EMS for
initial charge, and case C considers BESS connected MG with battery-SC-based ESS is proposed in Hu et al. (2020) that incor-
initial charge. Simulation results emphasize that the cost for each porates the driving pattern recognition (DPR) to determine elec-
case is more feasible than other techniques such as PSO and GA. A tric vehicle’s real-time driving pattern and power flow manage-
multi-agent system (MAS) based autonomous EMS is proposed in ment. SC is included in the ESS to provide peak power demand,
Yoo et al. (2013) for BESS integrated MG, as shown in Fig. 11. The whereas the battery provides the steady-state power demand.
system proposed two layers of decision-making proceedings: one The FLC controls the SOC level of SC. Simulation results in MAT-
layer is in the MG central controller (MGCC), and another is in the LAB/Simulink show that the approach improves the battery life-
agents. The agents estimate the optimal solutions by introducing time by 6.16% and decreases the maximum charging/discharging
intelligent techniques such as FLC and considering load demand, current by 58.2%. An online energy management control scheme
current capacity, operation cost, and incentives at the consumer’s for ultracapacitor (UC) and battery-based ESS is presented in Shen
level. The battery agents aim to sustain battery SOC level at a and Khaligh (2015) that aims to reduce battery life and ensure
feasible range and reduce peak current. optimal power distribution to minimize power losses. Dynamic
An E-Broker algorithm (EBA) and the PSO-based price control programming (DP) is integrated to estimate numerical solutions
approach for the ESS integrated large power plant is proposed in from different driven cycles, and a rule-based control scheme is
6080
Fig. 11. A schematic representation of a multi-agent system (MAS) based autonomous EMS (Yoo et al., 2013).
introduced to provide a comparative analysis with the proposed An intelligent generation control strategy for ESS is proposed
approach. The two objective functions f1 and f2 of the proposed in Mahmoud et al. (2019) that focuses on enhancing battery
approach can be expressed as lifetime, optimal power distribution, and minimizing costs. The
Ploss (k) charging/discharging time for batteries is analyzed by the FLC
f1 (k) = , k = 1, 2, . . . N , (23) scheme. The equation for optimal sizing of battery, SizeBESS for the
Ploss_max proposed scheme can be illustrated as
MaxSize
(Ib (k) − Ib (k − 1))2 All
⎧
, k = 2...N
⎪ ∑ ∑
SizeBESS = min( CostCap + CostAnnMG ) (27)
⎪
⎪ )2
∆I
⎪ ( Gen
b,max
⎨
f2 (k) = , (24) GnTp=1 Dispatch
(Ib (k))2 where CostCap is the capital cost and CostAnnMG is the annual MG
)2 , k = 1
⎪
⎪
Gen
⎪
⎪
∆Ib,max
⎩ (
generation cost in $/kWh. Simulation results show that the annual
where Ploss is the power loss and Ib is the battery current. Simula- cost is reduced by 6.5% for islanded mode, whereas the grid-
tion results demonstrate that the battery peak current is reduced connected non-export mode reduces the annual cost by 7.6%. The
by 60%, whereas the rule-based strategy reduces it by only 20%. grid-connected with export mode reduces annual cost by 6.4%.
The proposed strategy increases battery lifetime by 64.8% and An iterative bat optimization algorithm (BOA) based intelligent
the SOC level by 15%. An artificial bee colony (ABC) algorithm- technique to maintain optimal sizing and placement of BESS is
based intelligent technique is proposed in Das et al. (2018) for demonstrated in Ramírez et al. (2018). The study aims to maintain
the optimal placement of ESS. The technique aims to minimize the frequency stability under the RESs connected power systems.
system costs and reduce peak current demand and power losses. The cost function, J for the sizing of the proposed BESS system
The objective function CF to minimize costs can be expressed as can be expressed as
t
CF = min{ γVD ∗ CVD
n
+ γPL∗ CPL
l
+ γLL ∗ CLLl + (γESS ∗ CESS
UT ∆tk
( )
)}, (25)
∑
α ∗ (∆flk + flk−1 ) + SBk + SBk −1 × ( ),
[ ( )]
J = (28)
2
where γVD is the weighting factor (WF) for voltage deviation (VD) k=1
n
cost, CVD is the cost for CV, γPL is the WF for power losses (PL) where t is the simulation time, α is a constant, ∆fl is the absolute
l
cost, CPL is the cost for PL, γLL is the WF for line loading (LL) value of the measured frequency, and SB is the BESS output. The
simulation results in MATLAB emphasize that the proposed BOA
l
cost, CLL is the cost of LL, γESS is the WF of ESS cost, and CESS
UT
is technique minimizes costing and frequency fluctuations com-
the cost for ESS. The simulation is performed in DIgSILENT Pow- pared to GA. An adaptive dynamic programming (ADP)-based
erFactory, considering an IEEE-33 bus distribution system with intelligent frequency control strategy is proposed in Mu et al.
three cases and compared with a PSO model. The comparative (2019) that eliminates the intermittency due to the integration
analysis emphasizes the system efficacy in improving voltage of RESs to power systems. The ADP reduces the frequency fluc-
profile, minimizing line loading, and power loss compared to PSO. tuations and enhances the optimal power distribution between
A convex MPC strategy for BESS is proposed in Morstyn et al. ESS and loads. The simulation result is compared with PID, FLC,
(2017) that maintains an optimal power flow between BESS and and linear quadratic regulator (LQR) to verify the scheme’s effec-
MG considering a d-q reference frame voltage-current model. The tiveness. Table 8 demonstrates a comparative analysis of different
objective function, Jobj to minimize power loss can be expressed intelligent techniques integrated into ESS.
as
∑ 3. Metrics of feasibility analysis
Jobj = JLCL (k) + JLine (k) + JBat (k) , (26)
k∈τ
The study presents a deep analysis of different intelligent
where JLCL is the losses in the LCL filters, JLine is the losses in techniques integrated into RESs based systems. Feasibility anal-
line resistance, and JBat is the losses in the battery. The real-time ysis with appropriate metrics is necessary for the forecasting
digital simulator is introduced to continue the simulation process. outcomes such as solar irradiance, solar power, wind speed, wind
The results show that the battery SOC level ranges from 19.98% power, hydropower, etc. Feasibility metrics determine the rela-
to 100.05%, and the average power loss is 12.638 kW. tive accuracy of the proposed forecasting model, which facilitates
6081
Table 8
A comparative analysis of different intelligent techniques integrated into ESS.
Kerdphol et al. PSO-based BESS 2015 Japan PSO + BESS • Focuses on optimal sizing of • 5.5% less capital cost
(2016) BESS with fast DR and 5.5% less O&M cost
• Compared with BESS + SA than BESS + SA
• DIgSILENT PowerFactory as
simulation Platform
Akter et al. MMPC for 2015 Malaysia ESS + MMPC + LF • A voltage control approach • Reduces the execution
(2015) nonlinear system • LF is used for nonlinear time by 18% than
stability system stability conventional MPC
Chong et al. An optimal 2016 Malaysia Battery + SC + • Focuses on minimizing • Reduces the battery
(2016) approach for ESS LPF + FLC + PSO dynamic stress and peak peak current and peak
+ FBC current through FLC power by 16.05% and
• SC provides peak current 15.19%, respectively
demand during operation
• Simulink for simulation
Alshehri et al. An intelligent 2019 Saudi BESS + ANN + • Focuses on increasing power • Better performance
(2019) controller for BESS Arabia DEO + PI quality and minimize load than the benchmark
fluctuations controller
• Compared with the
benchmark PID controller
Teo et al. FLC-based RESs 2018 Singapore RESs + ESS + FLC • Focuses on minimizing • Reduces operating cost
(2018) and ESS control operating cost and maintaining by 1.87%
the optimal power distribution
• Considers the SOC of ESS,
load demand, and electricity
price
• Compared with SAEC and
MILP
Bahmani- IBA for optimal 2013 Iran BESS + IBA + • BA ios employed to • Cost reduced by 15%
Firouzi and sizing of BESS RESs minimize operation cost for optimal sizing of 250
Azizipanah- • MATLAB for simulation kWh than 150 kWh
Abarghooee • Compared with PSO and GA
(2014)
Yoo et al. MAS-based 2013 South Korea BESS + MGCC + • Agents estimate the load • Efficient in
(2013) autonomous EMS FLC demand, current capacity, and maintaining SOC level
consumers’s behavior through and peak current in a
FLC feasible range
• MGCC controls all the agents
García- PSO-based price 2021 Mexico ESS + EBA + PSO • EBA maintains priority for • A profitable BESS than
Santacruz et al. control optimal power distribution others
(2021) • NPV and IRR for feasibility
analysis
• Optimal sizing of ESS
Hu et al. (2020) AWC-FLC based 2019 China Battery + SC + • DPR detects driving patterns • Improves battery
EMS AWC + FLC from the input information lifetime by 6.16%
• SC responds to the peak • Decreases peak current
current demand by 58.2%
• FLC maintains SC SOC level
in a feasible range
Shen and Battery-UC-based 2015 USA Battery + UC + • DP estimates the • Improves battery
Khaligh (2015) ESS DP + optimization solution from the lifetime by 64.8% and
input information SOC level by 15%
• Maintains optimal sizing • Reduces battery peak
• Compared with rule-based current by 60%
strategy
Das et al. ABC algorithm for 2018 Australia ESS + ABC + PV • Focuses on optimal • Better performance
(2018) ESS + Wind placement of ESS • Includes than PSO
IEEE-33 bus distribution
system
• Compared with PSO
• DIgSILENT PowerFactory for
simulation
6082
Morstyn et al. Convex MPC for 2017 Australia Battery + Convex • The d-q reference frame • The range of battery
(2017) BESS MPC voltage-current model SOC is 19.98%–100.05%
maintains an optimal power • The average power
distribution loss is 12.638 kW
• IEEE 13 bus distribution
system for simulation
• Analyzes the nonlinear
charging–discharging
characteristics of Li-ion battery
Mahmoud et al. FLC-based ESS 2019 Australia ESS + FLC • Focuses on maintaining • Reduces the annual
(2019) optimal sizing with the cost by 6.5% and 6.4%
minimum cost for islanded mode and
• FLC analyzes battery quality grid-connected export
and charging–discharging time mode, respectively
• Compared with PSO, GA, and
FPA
Ramírez et al. BOA for BESS 2017 Mexico BESS + BOA + • Stability for RESs connected • Better performance for
(2018) RESs MG optimal sizing and
• BOA optimizes the input costing than GA
parameters
• Compared to GA
Mu et al. ADP strategy for 2019 China ESS + RESs + ADP • ADP for frequency stability • Better performance
(2019) ESS • Considers uncertainty for than others
RESs
• Compared with PID, LQR,
and FLC
researchers in finding the best solution. Different feasibility met- the 89 research work of different intelligent techniques facilitates
rics are addressed in different described control approaches, such realizing the current state and future acceptance of different
as root mean absolute error (RMAE), root mean square error techniques in practical implementations. The study presents each
(RMSE), mean absolute percentage error (MAPE), the sum of control technique’s system configuration, simulation platform,
squares error (SSE), and correlation coefficient (CC) (Qu et al., key features, and system accuracy. The significant findings of the
2017). The formula for RMAE can be expressed as research work can be expressed below:

N

1 ∑ • ANN effectively predicts solar irradiation, wind speed, and
RMAE = √ |Rn − Pn |, (29) power output for less number of input variables. The range
N
n=1 of RMSE for solar irradiation and power forecasting re-
where Rn is the actual value and Pn is the predicted value. The mains at 2%–10% which is feasible. The combined operation
formula of RMSE can be expressed as of ANN and analog ensemble (AnEn) provides significant
 improvement of the deterministic and probabilistic forecast-
N

 1∑ ing procedures. The integration of ANN in PV and thermal
RMSE = √( (Rn − Pn )2 ). (30) system improves efficiency by 5.25% than others.
N
n=1 • CNN provides short-term and long-term solar and wind
The formula for MAPE can be illustrated as forecasting with higher prediction accuracy. CNN-LSTM
N ⏐ ⏐ based prediction technique provide 27.78%–39.43% MAE
1 ∑ ⏐ Rn − Pn ⏐ compared to SVM and 22.23%–28.86% MAE compared to
MAPE = ⏐ ⏐ × 100%. (31)
N ⏐ Rn ⏐ LSTM. A hybrid CEEMDAN-CNN-LSTM approach improves
n=1
performance accuracy maintaining MAE, RMSE, and nRMSE
The formula of SSE can be expressed as
in a feasible range.
N • PSO is integrated into the PV system for several purposes:
∑
SSE = (Rn − Pn )2 (32) to analyze the frequency stability, to track maximum power
n=1 point, to eliminate uncertainty, and to maximize power out-
The formula of CC can be expressed as put. PSO-based MPPT in solar PV system provides the lowest
∑N RMSE (0.327%). A PSO-GA-ANFIS-based PV power prediction
{ti − tm ) (pi − pm ) technique provides MAE ranges of 5.16%–23.36%.
CC = ∑ i=1 1
, (33)
N
( i=1 (ti − tm )2 (pi − pm )2 ) 2 • MARS-based PV forecasting takes nonlinear input to obtain
better accuracy than ANN, KNN, CART, and SVM. NARX NN-
where ti and pi are the target and predicted values, respectively.
based solar radiation prediction technique outcomes the
tm and pm are the mean value of the target and predicted values,
least MAE (0.279%). CGAE-based solar irradiance forecasting
respectively.
technique provides 4% better accuracy than other consid-
4. Findings and discussions ered techniques.
• CART-Bagging-based wind power prediction method redu-
This paper aims to provide an in-depth view of intelligent ces prediction error by 2.04%. The SHWIP-based technique
techniques to sustain the stability and techno-economic feasi- provides the least MAE (11.80%) for wind power forecasting,
bility of RESs connected power systems. The critical review of whereas the RBFNN-based approach reduces MAE by 5.75%.
6083
• A combined wind, PV, FC, and IGCC reduces environmental management system. Additionally, the integration of intelligent
and economic costs by 36.16%. The integration of ANN in algorithms in controller design may improve the dynamic per-
hydropower systems reduces power generation costs by formance of power systems. Demand side management (DSM) for
61.2% and increases production by 6.8%. RNN-NAR-based smart grid technology is becoming essential for load management
RESs enhance RC from 0.93 to 0.99. HySIREN and EMD and cost reduction. Future research could explore integrating
based supply–demand forecasting model decreases 64.5% intelligent techniques into DSM to improve the adaptability of
of environmental cost. The evolutionary algorithm for the smart grid technologies.
optimal sizing of RESs reduces the overall cost by 5%.
• Integrating PSO in the wind, PV, and hydrogen-based system Declaration of competing interest
reduces techno-economic costs by 10%, whereas the RO
desalination technique reduces power supply loss by 18.3%. The authors declare that they have no known competing finan-
The FA-based wind and PV system lowers system costs by cial interests or personal relationships that could have appeared
36%. to influence the work reported in this paper.
• MMPC is integrated into ESS for nonlinear system stabil-
ity that reduces the execution time by 18%. FLC-based ESS Data availability
reduces the annual cost by 6.5% and 6.4% for islanded and
grid-connected export modes, respectively. FLC-based ESS No data was used for the research described in the article.
reduces the annual cost by 6.5% for islanded mode of op-
eration. The iterative bat optimization algorithm (BOA) for
References
ESS provides optimal sizing, stability, and performance than
genetic algorithm (GA). Aghajani, A., Kazemzadeh, R., Ebrahimi, A., 2016. A novel hybrid approach for
predicting wind farm power production based on wavelet transform, hybrid
5. Conclusions & future research neural networks and imperialist competitive algorithm. Energy Convers.
Manage. 121, 232–240.
Ahmad, T., Zhang, H., Yan, B., 2020. A review on renewable energy and electricity
This article emphasizes the importance of utilizing advanced requirement forecasting models for smart grid and buildings. Sustainable
intelligent techniques, specifically AI and ML-based technologies, Cities Soc. 55, 102052.
to feasibly maintain renewable-integrated power systems oper- Akter, M.P., Mekhilef, S., Tan, N.M.L., Akagi, H., 2015. Modified model predictive
ations under normal conditions. The proposed review considers control of a bidirectional AC–DC converter based on Lyapunov function for
intelligent techniques for various system configurations, such as energy storage systems. IEEE Trans. Ind. Electron. 63 (2), 704–715.
Al-Badi, A., Al Wahaibi, A., Ahshan, R., Malik, A., 2022. Techno-economic
power systems with solar PV, wind, hybrid RESs, and ESS. This feasibility of a solar-wind-fuel cell energy system in duqm, oman. Energies
study highlights that intelligent techniques are essential in man- 15 (15), 5379.
aging the uncertainties and intermittency associated with RESs Al-Ghussain, L., Ahmad, A.D., Abubaker, A.M., Hovi, K., Hassan, M.A., Annuk, A.,
integration by enabling real-time data prediction and analysis. 2023. Techno-economic feasibility of hybrid PV/wind/battery/thermal stor-
age trigeneration system: Toward 100% energy independency and green
It presents a comprehensive overview of various algorithms that
hydrogen production. Energy Rep. 9, 752–772.
have been employed for prediction, uncertainty analysis, forecast- Al-Janabi, S., Alkaim, A.F., Adel, Z., 2020. An innovative synthesis of deep learning
ing, and optimal sizing of systems, including ANN, CNN, LSTM, techniques (DCapsNet & DCOM) for generation electrical renewable energy
PSO, NARX-NN, MARS, DCNN, CGAE, WPD, and NN-COA. from wind energy. Soft Comput. 24 (14), 10943–10962.
This study highlights the significant improvements these tech- Al-Waeli, A.H., et al., 2018. Comparison of prediction methods of PV/T nanofluid
and nano-PCM system using a measured dataset and artificial neural
niques have brought to the accuracy, stability, optimization, con-
network. Sol. Energy 162, 378–396.
trol procedures, sizing, lifetime, operating costs, and storage ca- Albadi, M., El-Saadany, E., 2010. Overview of wind power intermittency impacts
pabilities of power systems with RESs integration. For instance, on power systems. Electr. Power Syst. Res. 80 (6), 627–632.
ANN algorithms have been integrated into different PV systems AlHakeem, D., Mandal, P., Haque, A.U., Yona, A., Senjyu, T., Tseng, T.-L., 2015. A
for prediction purposes and uncertainty analysis from solar irradi- new strategy to quantify uncertainties of wavelet-GRNN-PSO based solar PV
power forecasts using bootstrap confidence intervals. In: 2015 IEEE Power
ance, achieving maximum accuracy of approximately 95%. Addi-
& Energy Society General Meeting. IEEE, pp. 1–5.
tionally, CNN algorithms with LSTM for solar PV forecasting have Alotaibi, I., Abido, M.A., Khalid, M., Savkin, A.V., 2020. A comprehensive review of
been found to provide better accuracy than standalone CNN algo- recent advances in smart grids: A sustainable future with renewable energy
rithms or AE-LSTM-based applications. PSO algorithms have been resources. Energies 13 (23), 6269.
employed for MPPT control and forecasting purposes, demon- Alshehri, J., Khalid, M., Alzahrani, A., 2019. An intelligent battery energy storage-
based controller for power quality improvement in microgrids. Energies 12
strating the RMSE. Hybrid short-term wind speed forecasting
(11), 2112.
approaches that incorporate WPD and NN-COA for data decom- Aprillia, H., Yang, H.-T., Huang, C.-M., 2020. Short-term photovoltaic power
position and output estimation, respectively, have also achieved forecasting using a convolutional neural network–salp swarm algorithm.
the lowest MAPE of 1.214%. Furthermore, intelligent techniques Energies 13 (8), 1879.
integrated into combined RESs systems facilitate cost analysis Arabali, A., Ghofrani, M., Etezadi-Amoli, M., Fadali, M.S., 2013. Stochas-
tic performance assessment and sizing for a hybrid power system of
and optimal sizing of systems, while those integrated into ESS solar/wind/energy storage. IEEE Trans. Sustain. Energy 5 (2), 363–371.
enhance the efficient control procedure, optimal sizing, lifetime Arya, Y., 2019. AGC of PV-thermal and hydro-thermal power systems using CES
enhancement, operating cost, and storage capability. Overall, this and a new multi-stage FPIDF-(1+ PI) controller. Renew. Energy 134, 796–806.
study shows that intelligent techniques are crucial for power Aslam, S., Herodotou, H., Mohsin, S.M., Javaid, N., Ashraf, N., Aslam, S., 2021.
system research and offer a promising avenue for maintaining the A survey on deep learning methods for power load and renewable energy
forecasting in smart microgrids. Renew. Sustain. Energy Rev. 144, 110992.
feasibility of power system in the future. Bahmani-Firouzi, B., Azizipanah-Abarghooee, R., 2014. Optimal sizing of battery
Future research can be expanded by utilizing intelligent tech- energy storage for micro-grid operation management using a new improved
niques into RESs for optimal energy generation and distribu- bat algorithm. Int. J. Electr. Power Energy Syst. 56, 42–54.
tion. This task might involve analyzing weather forecast (solar Boussaada, Z., Curea, O., Remaci, A., Camblong, H., Mrabet Bellaaj, N., 2018. A
irradiance, wind speed), frequency fluctuations, intermittency, nonlinear autoregressive exogenous (NARX) neural network model for the
prediction of the daily direct solar radiation. Energies 11 (3), 620.
and power outcomes under variable load conditions. This would Cervone, G., Clemente-Harding, L., Alessandrini, S., Delle Monache, L., 2017.
involve developing advanced intelligent techniques that can pre- Short-term photovoltaic power forecasting using Artificial Neural Networks
dict and analyze real-time data to ensure an efficient energy and an Analog Ensemble. Renew. Energy 108, 274–286.
6084
Ceylan, İ., Erkaymaz, O., Gedik, E., Gürel, A.E., 2014. The prediction of pho- Ghimire, S., Deo, R.C., Downs, N.J., Raj, N., 2019. Global solar radiation prediction
tovoltaic module temperature with artificial neural networks. Case Stud. by ANN integrated with European Centre for medium range weather forecast
Therm. Eng. 3, 11–20. fields in solar rich cities of Queensland Australia. J. Clean. Prod. 216, 288–310.
Chatziagorakis, P., et al., 2016. Enhancement of hybrid renewable energy systems Hammid, A.T., Sulaiman, M.H.B., Abdalla, A.N., 2018. Prediction of small hy-
control with neural networks applied to weather forecasting: the case of dropower plant power production in Himreen Lake dam (HLD) using artificial
olvio. Neural Comput. Appl. 27 (5), 1093–1118. neural network. Alex. Eng. J. 57 (1), 211–221.
Chitsaz, H., Amjady, N., Zareipour, H., 2015. Wind power forecast using wavelet Haque, A.U., Mandal, P., Meng, J., Srivastava, A.K., Tseng, T.-L., Senjyu, T., 2013.
neural network trained by improved clonal selection algorithm. Energy A novel hybrid approach based on wavelet transform and fuzzy ARTMAP
Convers. Manage. 89, 588–598. networks for predicting wind farm power production. IEEE Trans. Ind. Appl.
Chong, L.W., Wong, Y.W., Rajkumar, R.K., Isa, D., 2016. An optimal control strat- 49 (5), 2253–2261.
egy for standalone PV system with Battery-Supercapacitor Hybrid Energy Hu, Y.-L., Chen, L., 2018. A nonlinear hybrid wind speed forecasting model using
Storage System. J. Power Sources 331, 553–565. LSTM network, hysteretic ELM and Differential Evolution algorithm. Energy
Chow, S.K., Lee, E.W., Li, D.H., 2012. Short-term prediction of photovoltaic energy Convers. Manage. 173, 123–142.
generation by intelligent approach. Energy Build. 55, 660–667. Hu, J., Liu, D., Du, C., Yan, F., Lv, C., 2020. Intelligent energy management strategy
Das, C.K., Bass, O., Kothapalli, G., Mahmoud, T.S., Habibi, D., 2018. Optimal of hybrid energy storage system for electric vehicle based on driving pattern
placement of distributed energy storage systems in distribution networks recognition. Energy 198, 117298.
using artificial bee colony algorithm. Appl. Energy 232, 212–228. Huang, C.-M., Kuo, C.-J., Huang, Y.-C., 2017. Short-term wind power forecasting
Datta, M., Senjyu, T., Yona, A., Funabashi, T., Kim, C.-H., 2010. A frequency-control and uncertainty analysis using a hybrid intelligent method. IET Renew. Power
approach by photovoltaic generator in a PV–diesel hybrid power system. IEEE Gener. 11 (5), 678–687.
Trans. Energy Convers. 26 (2), 559–571. Hwangbo, S., Nam, K., Heo, S., Yoo, C., 2019. Hydrogen-based self-sustaining
Dehghani, M., et al., 2019. Prediction of hydropower generation using grey wolf integrated renewable electricity network (HySIREN) using a supply–demand
optimization adaptive neuro-fuzzy inference system. Energies 12 (2), 289. forecasting model and deep-learning algorithms. Energy Convers. Manage.
Del Río, P., Burguillo, M., 2008. Assessing the impact of renewable energy 185, 353–367.
deployment on local sustainability: Towards a theoretical framework. Renew. 2021. Renewable Energy Statistics 2021. International Renewable Energy Agency
Sustain. Energy Rev. 12 (5), 1325–1344. (IRENA), Available: https://www.irena.org/.
Devaraj, J., Madurai Elavarasan, R., Shafiullah, G., Jamal, T., Khan, I., 2021. A Jaalam, N., Rahim, N., Bakar, A., Tan, C., Haidar, A.M., 2016. A comprehensive re-
holistic review on energy forecasting using big data and deep learning view of synchronization methods for grid-connected converters of renewable
models. Int. J. Energy Res. 45 (9), 13489–13530. energy source. Renew. Sustain. Energy Rev. 59, 1471–1481.
Jha, S.K., Bilalovic, J., Jha, A., Patel, N., Zhang, H., 2017. Renewable energy: Present
Dong, N., Chang, J.-F., Wu, A.-G., Gao, Z.-K., 2020. A novel convolutional neural
research and future scope of artificial intelligence. Renew. Sustain. Energy
network framework based solar irradiance prediction method. Int. J. Electr.
Rev. 77, 297–317.
Power Energy Syst. 114, 105411.
Jiang, P., Li, R., Li, H., 2019. Multi-objective algorithm for the design of prediction
Dong, D., Sheng, Z., Yang, T., 2018. Wind power prediction based on recurrent
intervals for wind power forecasting model. Appl. Math. Model. 67, 101–122.
neural network with long short-term memory units. In: 2018 International
Jiao, R., Huang, X., Ma, X., Han, L., Tian, W., 2018. A model combining stacked
Conference on Renewable Energy and Power Engineering. REPE, IEEE, pp.
auto encoder and back propagation algorithm for short-term wind power
34–38.
forecasting. Ieee Access 6, 17851–17858.
Dong, J., Xue, G., Dong, M., Xu, X., 2015. Energy-saving power gen-
Kaabeche, A., Diaf, S., Ibtiouen, R., 2017. Firefly-inspired algorithm for optimal
eration dispatching in China: Regulations, pilot projects and policy
sizing of renewable hybrid system considering reliability criteria. Sol. Energy
recommendations—A review. Renew. Sustain. Energy Rev. 43, 1285–1300.
155, 727–738.
Doucoure, B., Agbossou, K., Cardenas, A., 2016. Time series prediction using
Kalogirou, S.A., Mathioulakis, E., Belessiotis, V., 2014. Artificial neural networks
artificial wavelet neural network and multi-resolution analysis: Application
for the performance prediction of large solar systems. Renew. Energy 63,
to wind speed data. Renew. Energy 92, 202–211.
90–97.
Dreidy, M., Mokhlis, H., Mekhilef, S., 2017. Inertia response and frequency control
Kanase-Patil, A.B., Kaldate, A.P., Lokhande, S.D., Panchal, H., Suresh, M., Priya, V.,
techniques for renewable energy sources: A review. Renew. Sustain. Energy
2020. A review of artificial intelligence-based optimization techniques for
Rev. 69, 144–155.
the sizing of integrated renewable energy systems in smart cities. Environ.
Ellabban, O., Abu-Rub, H., Blaabjerg, F., 2014. Renewable energy resources: Technol. Rev. 9 (1), 111–136.
Current status, future prospects and their enabling technology. Renew.
Kerdphol, T., Qudaih, Y., Mitani, Y., 2016. Optimum battery energy storage system
Sustain. Energy Rev. 39, 748–764.
using PSO considering dynamic demand response for microgrids. Int. J. Electr.
Eltamaly, A.M., Al-Saud, M., Abo-Khalil, A., 2020. Performance improvement of Power Energy Syst. 83, 58–66.
PV systems’ maximum power point tracker based on a scanning PSO particle Khan, S., Yahong, W., Zeeshan, A., 2022. Impact of poverty and income inequality
strategy. Sustainability 12 (3), 1185. on the ecological footprint in Asian developing economies: Assessment of
Eseye, A.T., Zhang, J., Zheng, D., 2018. Short-term photovoltaic solar power Sustainable Development Goals. Energy Rep. 8, 670–679.
forecasting using a hybrid Wavelet-PSO-SVM model based on SCADA and Khandakar, A., et al., 2019. Machine learning based photovoltaics (PV) power
Meteorological information. Renew. Energy 118, 357–367. prediction using different environmental parameters of Qatar. Energies 12
Esquivias, M.A., Sugiharti, L., Rohmawati, H., Rojas, O., Sethi, N., 2022. Nexus (14), 2782.
between technological innovation, renewable energy, and human capital Khodayar, M., Mohammadi, S., Khodayar, M.E., Wang, J., Liu, G., 2019. Convolu-
on the environmental sustainability in emerging Asian economies: a panel tional graph autoencoder: A generative deep neural network for probabilistic
quantile regression approach. Energies 15 (7), 2451. spatio-temporal solar irradiance forecasting. IEEE Trans. Sustain. Energy 11
Fei, S.-w., He, Y., 2015. Wind speed prediction using the hybrid model of wavelet (2), 571–583.
decomposition and artificial bee colony algorithm-based relevance vector Krishna, K.S., Kumar, K.S., 2015. A review on hybrid renewable energy systems.
machine. Int. J. Electr. Power Energy Syst. 73, 625–631. Renew. Sustain. Energy Rev. 52, 907–916.
Ferrero Bermejo, J., Gomez Fernandez, J.F., Olivencia Polo, F., Crespo Marquez, A., Kuang, Y., et al., 2016. A review of renewable energy utilization in islands.
2019. A review of the use of artificial neural network models for energy and Renew. Sustain. Energy Rev. 59, 504–513.
reliability prediction. a study of the solar PV, hydraulic and wind energy Kumar, G.A., 2022. Optimal power point tracking of solar and wind energy in a
sources. Appl. Sci. 9 (9), 1844. hybrid wind solar energy system. Int. J. Energy Environ. Eng. 13 (1), 77–103.
Filik, Ü.B., Filik, T., 2017. Wind speed prediction using artificial neural networks Lai, J.-P., Chang, Y.-M., Chen, C.-H., Pai, P.-F., 2020. A survey of machine learning
based on multiple local measurements in Eskisehir. Energy Procedia 107, models in renewable energy predictions. Appl. Sci. 10 (17), 5975.
264–269. Lee, W., Kim, K., Park, J., Kim, J., Kim, Y., 2018. Forecasting solar power using
Fischer, A., Montuelle, L., Mougeot, M., Picard, D., 2017. Statistical learning for long-short term memory and convolutional neural networks. IEEE Access 6,
wind power: A modeling and stability study towards forecasting. Wind 73068–73080.
Energy 20 (12), 2037–2047. Li, Y., He, Y., Su, Y., Shu, L., 2016. Forecasting the daily power output of a grid-
Galimova, T., Ram, M., Breyer, C., 2022. Mitigation of air pollution and corre- connected photovoltaic system based on multivariate adaptive regression
sponding impacts during a global energy transition towards 100% renewable splines. Appl. Energy 180, 392–401.
energy system by 2050. Energy Rep. 8, 14124–14143. Li, F., Liao, H.y., 2018. An intelligent method for wind power forecasting based
Gao, B., Huang, X., Shi, J., Tai, Y., Zhang, J., 2020. Hourly forecasting of on integrated power slope events prediction and wind speed forecasting.
solar irradiance based on CEEMDAN and multi-strategy CNN-LSTM neural IEEJ Trans. Electr. Electron. Eng. 13 (8), 1099–1105.
networks. Renew. Energy 162, 1665–1683. Li, Q., Loy-Benitez, J., Nam, K., Hwangbo, S., Rashidi, J., Yoo, C., 2019. Sustainable
García-Santacruz, C., Galván, L., Carrasco, J.M., Galván, E., 2021. Sizing and and reliable design of reverse osmosis desalination with hybrid renewable
management of energy storage systems in large-scale power plants using energy systems through supply chain forecasting using recurrent neural
price control and artificial intelligence. Energies 14 (11), 3296. networks. Energy 178, 277–292.
6085
Li, S., Wang, P., Goel, L., 2015. Wind power forecasting using neural net- Premalatha, N., Valan Arasu, A., 2016. Prediction of solar radiation for solar
work ensembles with feature selection. IEEE Trans. Sustain. Energy 6 (4), systems by using ANN models with different back propagation algorithms.
1447–1456. J. Appl. Res. Technol. 14 (3), 206–214.
Li, H., Wang, J., Lu, H., Guo, Z., 2018. Research and application of a combined Priyadarshi, N., Padmanaban, S., Holm-Nielsen, J.B., Blaabjerg, F., Bhaskar, M.S.,
model based on variable weight for short term wind speed forecasting. 2019. An experimental estimation of hybrid ANFIS–PSO-based MPPT for
Renew. Energy 116, 669–684. PV grid integration under fluctuating sun irradiance. IEEE Syst. J. 14 (1),
Liu, H., Mi, X., Li, Y., 2018. Smart multi-step deep learning model for wind speed 1218–1229.
forecasting based on variational mode decomposition, singular spectrum Qian, D., Tong, S., Liu, H., Liu, X., 2016. Load frequency control by neural-
analysis, LSTM network and ELM. Energy Convers. Manage. 159, 54–64. network-based integral sliding mode for nonlinear power systems with wind
Liu, D., Niu, D., Wang, H., Fan, L., 2014. Short-term wind speed forecasting turbines. Neurocomputing 173, 875–885.
using wavelet transform and support vector machines optimized by genetic Qu, Z., Zhang, K., Mao, W., Wang, J., Liu, C., Zhang, W., 2017. Research
algorithm. Renew. Energy 62, 592–597. and application of ensemble forecasting based on a novel multi-objective
Liu, H., Tian, H., Liang, X., Li, Y., 2015. New wind speed forecasting approaches optimization algorithm for wind-speed forecasting. Energy Convers. Manage.
using fast ensemble empirical model decomposition, genetic algorithm, Mind 154, 440–454.
Evolutionary Algorithm and Artificial Neural Networks. Renew. Energy 83, Ram Ola, S., et al., 2020. Alienation coefficient and wigner distribution function
1066–1075. based protection scheme for hybrid power system network with renewable
Liu, H., Zhai, R., Fu, J., Wang, Y., Yang, Y., 2019. Optimization study of thermal- energy penetration. Energies 13 (5), 1120.
storage PV-CSP integrated system based on GA-PSO algorithm. Sol. Energy Ramírez, M., Castellanos, R., Calderón, G., Malik, O., 2018. Placement and sizing
184, 391–409. of battery energy storage for primary frequency control in an isolated section
Lopes, M.N.G., da Rocha, B.R.P., Vieira, A.C., de Sá, J.A.S., Rolim, P.A.M., of the Mexican power system. Electr. Power Syst. Res. 160, 142–150.
da Silva, A.G., 2019. Artificial neural networks approaches for predicting Rocha, P.C., Fernandes, J., Modolo, A., Lima, R.P., da Silva, M.V., Bezerra, C.D.,
the potential for hydropower generation: a case study for amazon region. J. 2019. Estimation of daily, weekly and monthly global solar radiation using
Intell. Fuzzy Systems 36 (6), 5757–5772. ANNs and a long data set: a case study of Fortaleza, in Brazilian Northeast
Lujano-Rojas, J.M., Dufo-López, R., Bernal-Agustín, J.L., 2013. Probabilistic mod- region. Int. J. Energy Environ. Eng. 10 (3), 319–334.
elling and analysis of stand-alone hybrid power systems. Energy 63, Ryu, A., Ito, M., Ishii, H., Hayashi, Y., 2019. Preliminary analysis of short-term
19–27. solar irradiance forecasting by using total-sky imager and convolutional
Mahmoud, T.S., Ahmed, B.S., Hassan, M.Y., 2019. The role of intelligent generation neural network. In: 2019 IEEE PES GTD Grand International Conference and
control algorithms in optimizing battery energy storage systems size in Exposition Asia (GTD Asia). IEEE, pp. 627–631.
microgrids: A case study from Western Australia. Energy Convers. Manage. Sa-ngawong, N., Ngamroo, I., 2015. Intelligent photovoltaic farms for robust
196, 1335–1352. frequency stabilization in multi-area interconnected power system based on
Maleki, A., Pourfayaz, F., 2015. Optimal sizing of autonomous hybrid pho- PSO-based optimal sugeno fuzzy logic control. Renew. Energy 74, 555–567.
tovoltaic/wind/battery power system with LPSP technology by using Saberian, A., Hizam, H., Radzi, M., Ab Kadir, M., Mirzaei, M., 2014. Modelling and
evolutionary algorithms. Sol. Energy 115, 471–483. prediction of photovoltaic power output using artificial neural networks. Int.
Meng, A., Ge, J., Yin, H., Chen, S., 2016. Wind speed forecasting based on wavelet J. Photoenergy 2014.
packet decomposition and artificial neural networks trained by crisscross Saidur, R., Rahim, N., Islam, M., Solangi, K., 2011. Environmental impact of wind
optimization algorithm. Energy Convers. Manage. 114, 75–88. energy. Renew. Sustain. Energy Rev. 15 (5), 2423–2430.
Meyyappan, U., 2018. Wavelet neural network–based wind speed forecasting Sanchez, V.M., Chavez-Ramirez, A., Duron-Torres, S.M., Hernandez, J., Arriaga, L.,
and application of shuffled frog leap algorithm for economic dispatch with Ramirez, J.M., 2014. Techno-economical optimization based on swarm in-
prohibited zones incorporating wind power. Wind Eng. 42 (1), 3–15. telligence algorithm for a stand-alone wind-photovoltaic-hydrogen power
Mohammed, Y.S., Adetokun, B.B., Oghorada, O., Oshiga, O., 2022. Techno- system at south-east region of Mexico. Int. J. Hydrogen Energy 39 (29),
economic optimization of standalone hybrid power systems in context 16646–16655.
of intelligent computational multi-objective algorithms. Energy Rep. 8, Sari, A., et al., 2022. New optimized configuration for a hybrid PV/diesel/battery
11661–11674. system based on coyote optimization algorithm: A case study for hotan
Morstyn, T., Hredzak, B., Aguilera, R.P., Agelidis, V.G., 2017. Model predictive county. Energy Rep. 8, 15480–15492.
control for distributed microgrid battery energy storage systems. IEEE Trans. Semero, Y.K., Zhang, J., Zheng, D., 2018. PV power forecasting using an integrated
Control Syst. Technol. 26 (3), 1107–1114. GA-PSO-ANFIS approach and Gaussian process regression based feature
Mu, C., Zhang, Y., Jia, H., He, H., 2019. Energy-storage-based intelligent frequency selection strategy. CSEE J. Power Energy Syst. 4 (2), 210–218.
control of microgrid with stochastic model uncertainties. IEEE Trans. Smart Sezer, N., Biçer, Y., Koç, M., 2019. Design and analysis of an integrated concen-
Grid 11 (2), 1748–1758. trated solar and wind energy system with storage. Int. J. Energy Res. 43 (8),
Naik, J., Dash, P.K., Dhar, S., 2019. A multi-objective wind speed and wind power 3263–3283.
prediction interval forecasting using variational modes decomposition based Shafiullah, G., 2016. Hybrid renewable energy integration (HREI) system for
multi-kernel robust ridge regression. Renew. Energy 136, 701–731. subtropical climate in Central Queensland, Australia. Renew. Energy 96,
Nam, K., Hwangbo, S., Yoo, C., 2020. A deep learning-based forecasting model 1034–1053.
for renewable energy scenarios to guide sustainable energy policy: A case Shao, M., Tan-Mullins, M., Xie, L., 2021. Asian Infrastructure Investment Bank
study of Korea. Renew. Sustain. Energy Rev. 122, 109725. (AIIB)’s sustainable safeguard mechanism on energy projects. Energy Strategy
Obama, B., 2017. The irreversible momentum of clean energy. Science 355 Rev. 38, 100711.
(6321), 126–129. Shaw, A.R., Smith Sawyer, H., LeBoeuf, E.J., McDonald, M.P., Hadjerioua, B., 2017.
Olatomiwa, L., Mekhilef, S., Ismail, M.S., Moghavvemi, M., 2016. Energy man- Hydropower optimization using artificial neural network surrogate models
agement strategies in hybrid renewable energy systems: A review. Renew. of a high-fidelity hydrodynamics and water quality model. Water Resour.
Sustain. Energy Rev. 62, 821–835. Res. 53 (11), 9444–9461.
Osório, G., Matias, J., Catalão, J., 2015. Short-term wind power forecasting using Shen, J., Khaligh, A., 2015. A supervisory energy management control strategy
adaptive neuro-fuzzy inference system combined with evolutionary particle in a battery/ultracapacitor hybrid energy storage system. IEEE Trans. Transp.
swarm optimization, wavelet transform and mutual information. Renew. Electrif. 1 (3), 223–231.
Energy 75, 301–307. Soliman, M.S., et al., 2021. Supervisory energy management of a hybrid
Ozkan, M.B., Karagoz, P., 2015. A novel wind power forecast model: Statistical battery/PV/tidal/wind sources integrated in DC-microgrid energy storage
hybrid wind power forecast technique (SHWIP). IEEE Trans. Ind. Inform. 11 system. Energy Rep. 7, 7728–7740.
(2), 375–387. Suganthi, L., Iniyan, S., Samuel, A.A., 2015. Applications of fuzzy logic in
Panigrahi, B.K., Bhuyan, A., Shukla, J., Ray, P.K., Pati, S., 2021. A comprehensive renewable energy systems–a review. Renew. Sustain. Energy Rev. 48,
review on intelligent islanding detection techniques for renewable energy 585–607.
integrated power system. Int. J. Energy Res. 45 (10), 14085–14116. Tan, K.M., Babu, T.S., Ramachandaramurthy, V.K., Kasinathan, P., Solanki, S.G.,
Panthee, K., Jha, A.K., 2016. Estimation of global solar radiation using Artificial Raveendran, S.K., 2021. Empowering smart grid: A comprehensive re-
Neural Network in Kathmandu, Nepal. Int. J. Eng. Res. Sci. 2 (1), 62–68. view of energy storage technology and application with renewable energy
Perez-Navarro, A., Alfonso, D., Álvarez, C., Ibáñez, F., Sanchez, C., Segura, I., 2010. integration. J. Energy Storage 39, 102591.
Hybrid biomass-wind power plant for reliable energy generation. Renew. Teo, T.T., Logenthiran, T., Woo, W.L., Abidi, K., 2018. Intelligent controller for
Energy 35 (7), 1436–1443. energy storage system in grid-connected microgrid. IEEE Trans. Syst. Man
Podder, A.K., et al., 2020. Systematic categorization of optimization strategies for Cybern.: Syst..
virtual power plants. Energies 13 (23), 6251. Vanderstar, G., Musilek, P., Nassif, A., 2018. Solar forecasting using remote solar
Ponnusamy, V.K., et al., 2021. A comprehensive review on sustainable aspects monitoring stations and artificial neural networks. In: 2018 IEEE Canadian
of big data analytics for the smart grid. Sustainability 13 (23), 13322. Conference on Electrical & Computer Engineering. CCECE, IEEE, pp. 1–4.
6086
Vo, D.H., Vo, L.H., 2022. International volatility transmission among income, CO2 Yan, J., Zhang, H., Liu, Y., Han, S., Li, L., Lu, Z., 2017. Forecasting the high
emission, non-renewable and renewable energy consumption: Which causes penetration of wind power on multiple scales using multi-to-multi mapping.
which and when? Energy Rep. 8, 10061–10071. IEEE Trans. Power Syst. 33 (3), 3276–3284.
Wang, J., Zhang, F., Liu, F., Ma, J., 2016. Hybrid forecasting model-based data Yoo, C.-H., Chung, I.-Y., Lee, H.-J., Hong, S.-S., 2013. Intelligent control of
mining and genetic algorithm-adaptive particle swarm optimisation: a case battery energy storage for multi-agent based microgrid energy management.
study of wind speed time series. IET Renew. Power Gener. 10 (3), 287–298. Energies 6 (10), 4956–4979.
Wang, H., et al., 2017. Deterministic and probabilistic forecasting of photo- Zang, H., Liu, L., Sun, L., Cheng, L., Wei, Z., Sun, G., 2020. Short-term global
voltaic power based on deep convolutional neural network. Energy Convers. horizontal irradiance forecasting based on a hybrid CNN-LSTM model with
Manage. 153, 409–422. spatiotemporal correlations. Renew. Energy 160, 26–41.
Xiangchengzhen, M., Yilmaz, S., 2020. Renewable energy cooperation in North- Zhao, X., Wei, H., Wang, H., Zhu, T., Zhang, K., 2019. 3D-CNN-based fea-
east Asia: Incentives, mechanisms and challenges. Energy Strategy Rev. 29, ture extraction of ground-based cloud images for direct normal irradiance
100468. prediction. Sol. Energy 181, 510–518.
Yan, X., Francois, B., Abbes, D., 2015. Operating power reserve quantification Zhen, Z., et al., 2020. Deep learning based surface irradiance mapping model for
through PV generation uncertainty analysis of a microgrid. In: 2015 IEEE solar PV power forecasting using sky image. IEEE Trans. Ind. Appl. 56 (4),
Eindhoven PowerTech. IEEE, pp. 1–6. 3385–3396.
6087

1 s2.0 S2352484723008065 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S2352484723008065 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S2352484723008065 Main

Uploaded by

Copyright:

Available Formats

Energy Reports 9 (2023) 6063–6087

Contents lists available at ScienceDirect

Renewables integration into power systems through intelligent

Data availability .............................................................................................................................................................................................................. 6084

is proposed to achieve maximum point tracking (MPPT) of a

Fig. 3. An overview of the working flow of the proposed review work.

2.1.1. Application of ANN in Solar PV system

(continued on next page)

is included in the system to assist PV to handle frequency fluc-

where CACm is the annualized investment cost, COMm is the op-

(continued on next page)

(continued on next page)

You might also like