=Paper=
{{Paper
|id=Vol-3711/paper15
|storemode=property
|title=Enhancing solar panel efficiency with LSTM-based MPPT controllers
|pdfUrl=https://ceur-ws.org/Vol-3711/paper15.pdf
|volume=Vol-3711
|authors=Alexander Gozhyj,Vladyslav Nechakhin,Irina Kalinina
|dblpUrl=https://dblp.org/rec/conf/momlet/GozhyjNK24
}}
==Enhancing solar panel efficiency with LSTM-based MPPT controllers==
https://ceur-ws.org/Vol-3711/paper15.pdf
Enhancing solar panel efficiency with LSTM-based MPPT
controllers
Alexander Gozhyj1,†, Vladyslav Nechakhin1,*,† and Irina Kalinina1,†
1Petro Mohyla Black Sea National University, 68 Desantnykiv St, Mykolaiv, 54056, Ukraine
Abstract
This paper investigates the application of Long Short-Term Memory (LSTM) neural networks as
Maximum Power Point Tracking (MPPT) controllers for solar panels. Traditional MPPT
algorithms, including Perturb and Observe (P&O), Incremental Conductance (IncCond), and Hill
Climbing (HC), are compared with LSTM-based approaches in terms of accuracy, efficiency, and
adaptability. Minute-level data on voltage, current, power output, temperature, and solar
irradiance from diverse locations are used to train and evaluate the LSTM model. Results
demonstrate that LSTM-based MPPT controllers outperform traditional algorithms, offering
superior tracking accuracy and adaptability to dynamic environmental conditions. The study
highlights the significance of LSTM-based controllers in enhancing solar panel efficiency and
maximizing energy harvesting. This research contributes to the advancement of renewable
energy technologies and underscores the potential of artificial intelligence in optimizing solar
energy systems.
Keywords
Solar panel optimization, Maximum Power Point Tracking, Long Short-Term Memory neural
networks
1. Introduction
The utilization of renewable energy sources has become increasingly imperative in light of
global efforts to mitigate climate change and reduce dependency on fossil fuels. Among
these sources, solar energy holds particular promise due to its abundance and
sustainability. Maximizing the efficiency of solar panels plays a crucial role in harnessing
this energy resource effectively for sustainable power generation.
The transition towards renewable energy sources is driven by the need to mitigate
environmental degradation caused by traditional energy production methods. Solar energy,
in particular, offers a clean and abundant alternative to fossil fuels. However, the efficiency
of solar panels, which directly impacts energy output, is paramount for ensuring the
viability of solar power as a sustainable energy solution [1, 2, 3]. Thus, efforts to enhance
solar panel efficiency are essential for advancing renewable energy utilization and reducing
carbon emissions.
MoMLeT-2024: 6th International Workshop on Modern Machine Learning Technologies, May, 31 - June, 1, 2024,
Lviv-Shatsk, Ukraine
∗ Corresponding author.
† These authors contributed equally.
alex.gozhyj@gmail.com (A. Gozhyj); vladyslav.nechakhin@l3s.de (V. Nechakhin);
irina.kalinina@chmnu.edu.ua (I. Kalinina)
0000-0002-3517-580X (A. Gozhyj); 0000-0003-0146-1207 (V. Nechakhin); 0000-0001-8359-2045 (I.
Kalinina)
© 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
Maximizing the power output of solar panels is inherently challenging due to variations
in environmental conditions such as sunlight intensity and temperature. Maximum Power
Point Tracking (MPPT) algorithms are instrumental in optimizing solar panel performance
by continuously adjusting the operating point to extract maximum power from the solar
array. Traditional MPPT techniques, including Perturb and Observe (P&O) [4], Incremental
Conductance (IncCond) [5], and Hill Climbing (HC) [6], have been widely employed for this
purpose. However, these methods have inherent limitations in terms of accuracy, efficiency,
and adaptability to changing environmental conditions [7].
In recent years, artificial intelligence (AI) techniques have emerged as promising tools for
improving the efficiency of renewable energy systems. Long Short-Term Memory (LSTM)
neural networks, a type of recurrent neural network (RNN), offer significant potential for
enhancing MPPT precision and adaptability in solar panel systems. Unlike traditional MPPT
algorithms, LSTM networks can effectively capture temporal dependencies in the input data
and learn complex patterns, enabling more accurate and dynamic control of solar panel
operation [8, 9]. The application of LSTM neural networks represents a novel approach to
addressing the challenges associated with traditional MPPT techniques, thereby unlocking
new opportunities for optimizing solar energy harvesting.
This research paper aims to investigate the efficacy of LSTM-based MPPT controllers for
enhancing solar panel efficiency. The paper is structured as follows: after this introduction,
Section 2 provides a comprehensive review of related works, including traditional MPPT
techniques and recent advancements in AI-based approaches. Section 3 outlines the
methods and materials employed in the research, including data collection, LSTM
architecture, and training procedures. Following this, Section 4 presents the experimental
results, comparing the performance of LSTMbased MPPT controllers with conventional
algorithms. Section 5 discusses the implications of the findings, including limitations,
advantages, and potential applications of LSTM-based MPPT controllers. Finally, Section 6
concludes the paper by summarizing key findings and emphasizing the significance of
LSTM-based approaches for enhancing solar panel efficiency.
2. Related Works
Maximizing the power output of solar panels has been a subject of extensive research,
leading to the development of various Maximum Power Point Tracking (MPPT) techniques.
This section provides a thorough survey of both traditional MPPT algorithms and recent
advancements in AI-based approaches.
2.1. Survey of MPPT Techniques
Traditional MPPT algorithms play a fundamental role in optimizing solar panel
performance under varying environmental conditions. Perturb and Observe (P&O),
Incremental Conductance (IncCond), and Hill Climbing (HC) are among the most commonly
used techniques in this regard.
Perturb and Observe (P&O) [4] is a simple yet widely employed MPPT method that
perturbs the operating point of the solar panel and observes the resulting change in power
output to determine the direction of adjustment (Figure 1). While P&O is straightforward
to implement, it may suffer from oscillations around the maximum power point and slow
convergence under rapidly changing conditions.
Incremental Conductance (IncCond) [5] algorithm utilizes the derivative of the power-
voltage characteristic curve to dynamically adjust the operating point towards the
maximum power point (Figure 2). Compared to P&O, IncCond offers improved tracking
accuracy and faster convergence, especially under varying irradiance levels. However, it
may exhibit instability issues in certain scenarios, particularly when the system operates
near the maximum power point.
Hill Climbing (HC) algorithm [6] iteratively adjusts the operating point in the direction of
increasing power output until the maximum power point is reached (Figure 3). HC is known
for its simplicity and robustness in various environmental conditions. However, it may
suffer from slow convergence and susceptibility to local maxima, leading to suboptimal
performance, especially under rapidly changing conditions.
Figure 1: Flowchart of Perturb and Observe (P&O) method [10].
2.2. Advancements in AI for MPPT
Recent advancements in artificial intelligence (AI) have revolutionized maximum power
point tracking (MPPT) techniques for photovoltaic (PV) systems. Several studies have
explored the application of AI algorithms, particularly Long Short-Term Memory (LSTM)
neural networks, to enhance MPPT accuracy and efficiency, especially in challenging
scenarios such as partial shading conditions (PSC) and dynamic environmental changes.
In a comprehensive review, (Seyedmahmoudian et al., 2016) [13] discussed various AI-
based MPPT techniques, highlighting their robustness and reliability under diverse
conditions. The review categorized AI methods based on their performance and
applicability, providing valuable insights for researchers and engineers working with PV-
based power systems. Another survey by (Raj et al., 2022) [14] provided a thorough
examination of AI-based MPPT algorithms in PV systems, categorizing them based on their
application strategies and analyzing their merits and demerits. The study aimed to assist
users in selecting the most suitable AI-based MPPT technique according to their specific
project requirements and system constraints.
Figure 2: Flowchart of Incremental Conductance (IncCond) method [11].
The study by (Aouchiche et al., 2018) [15] proposed a novel approach using the Moth-
Flame Optimization algorithm (MFO) for global MPPT in PV plants under partial shading.
By combining Global MPPT (GMPPT) and Distributed MPPT (DMPPT) techniques, the
proposed method effectively mitigated the drawbacks of PSC and improved PV system
performance. Furthermore, (Amrouche et al., 2007) [16] proposed an AI-based Perturb and
Observe (P&O) MPPT method for PV systems, aiming to overcome the drawbacks of
traditional algorithms such as slow response speed and oscillations around the maximum
power point. By utilizing Artificial Neural Networks (ANN) to approximate the perturbation
step, the proposed method achieved improved performance and stability.
Additionally, (Kumar et al., 2022) [8] introduced a novel MPPT controller based on a Rain
Optimization Algorithm (ROA) and Bidirectional LSTM (Bi-LSTM) neural network for
gridconnected hybrid solar-wind systems. The proposed controller effectively tracked the
maximum power from solar PV and wind sources under varying climatic conditions,
contributing to stable power flow and grid integration. Lastly, (Pengcheng et al., 2021) [9]
conducted simulation experiments using LSTM neural networks and attention mechanisms
for MPPT in PV systems, demonstrating the potential of AI-based approaches in improving
power generation efficiency.
Figure 3: Flowchart of Hill Climbing (HC) method [12].
The study highlighted the importance of addressing data nonlinearity and feature
sparseness in real-world scenarios to enhance MPPT performance.
Studies have demonstrated the effectiveness of LSTM-based MPPT controllers in
improving tracking accuracy and adaptability compared to traditional methods. By
leveraging historical data and learning from past experiences, LSTM networks can predict
the optimal operating point of solar panels more accurately, even under rapidly changing
environmental conditions. Additionally, LSTM-based approaches offer potential for real-
time optimization and can adapt to fluctuations in solar irradiance and temperature more
effectively than conventional algorithms.
3. Methods and Materials
This section outlines the methodology employed for data collection, LSTM architecture, and
training procedures for LSTM-based MPPT controllers in solar panel systems.
3.1. Data Collection
Data collection is a crucial step in training and validating LSTM-based MPPT controllers. To
ensure comprehensive coverage of environmental conditions, data is collected from various
solar power plants located across different regions, such as Europe, North America and
Oceania. The collected dataset includes measurements of voltage, current, and power
output of solar panels, recorded at regular intervals. Additionally, external factors such as
temperature and solar irradiance are incorporated into the dataset to capture their
influence on solar panel performance.
By gathering data from diverse geographical locations and environmental conditions, the
dataset provides a comprehensive basis for training and testing LSTM models for MPPT
control. A sample of the collected data is shown in the Table 1.
Table 1
Sample solar power plant data with 1-minute intervals.
Date Voltage Current Power Temperature Irradiance
2023-06-01 112.76 2.49 281.58 25 563.67
10:00:00
2023-06-01 113.09 2.48 280.88 25 565.55
10:01:00
2023-06-01 113.53 2.51 285.57 25 567.43
10:02:00
2023-06-01 113.85 2.50 284.80 25 569.31
10:03:00
2023-06-01 114.21 2.50 286.11 25 571.18
10:04:00
2023-06-01 114.56 2.51 288.48 25 573.06
10:05:00
2023-06-01 114.95 2.50 287.68 25 574.94
10:06:00
2023-06-01 115.31 2.50 289.39 25 576.82
10:07:00
2023-06-01 115.75 2.49 289.33 25 578.70
10:08:00
2023-06-01 116.13 2.48 288.35 25 580.58
10:09:00
2023-06-01 116.46 2.51 293.33 25 582.46
10:10:00
3.2. LSTM Architecture and Functioning
Long Short-Term Memory (LSTM) neural networks are a type of recurrent neural network
(RNN) designed to capture long-term dependencies in sequential data. LSTM networks are
particularly well-suited for processing time-series data, making them suitable for dynamic
MPPT control in solar panel systems. The architecture of an LSTM network consists of
multiple memory cells and gating mechanisms that regulate the flow of information through
the network. This allows LSTM models to retain information over extended time periods
and learn complex patterns in the input data [17].
In the context of MPPT control, LSTM models function by processing historical data of
solar panel performance and environmental conditions to predict the optimal operating
point for maximizing power output. By leveraging past observations, LSTM networks can
adaptively adjust the operating point in response to changing environmental conditions,
thus improving the efficiency of solar panel systems.
3.3. Training and Implementation
The training of LSTM networks for MPPT control involves several steps. Firstly, the
collected dataset is preprocessed to remove noise and outliers and normalize the input
features. Next, the dataset is partitioned into training, validation, and testing sets to
facilitate model evaluation. The LSTM network is then trained using the training data, with
the objective of minimizing the prediction error between the actual and predicted
maximum power points.
After training, the LSTM model is integrated into the control system of solar panel arrays.
Real-time data from sensors in a simulated environment measuring voltage, current,
temperature, and solar irradiance are fed into the LSTM model, which generates predictions
of the optimal operating point. These predictions are then used to dynamically adjust the
operating parameters of the solar panels, thereby maximizing power output.
4. Experiment
This section elucidates the experimental procedures employed to evaluate the performance
of LSTM-based MPPT controllers in solar panel systems.
4.1. Parameter Selection and Control Mechanism
The selection of input parameters and the design of the output control mechanism are
crucial aspects of developing an effective LSTM-based MPPT controller. In our case the input
parameters include measurements of solar irradiance, temperature, voltage, and current,
which are essential for accurately predicting the maximum power point of the solar panel.
Additionally, historical data of power output and environmental conditions are used to
capture temporal dependencies and improve prediction accuracy.
The output control mechanism of the LSTM model involves determining the optimal
operating point of the solar panel based on the predicted maximum power point. This
involves adjusting the duty cycle of a DC-DC converter or controlling the voltage and current
levels to maximize power output. The design of the output control mechanism aims to
dynamically adapt the operating parameters of the solar panel system in response to
changing environmental conditions, thereby optimizing energy harvesting efficiency.
4.2. Training Procedure
The training procedure for the LSTM model involves several steps to ensure optimal
performance and generalization ability. Firstly, the collected dataset is divided into training,
validation, and testing sets using a suitable partitioning strategy. Between random sampling
and time-based splitting the latter was chosen to maintain the continuity among the
datapoints which gives the LSTM more opportunities to learn repeating patterns in the data.
The training set is used to update the parameters of the LSTM network through
backpropagation and gradient descent,
while the validation set is utilized to monitor model performance and prevent overfitting.
During training, hyperparameters such as learning rate, batch size, and number of epochs
are tuned to optimize model performance. Regularization techniques such as dropout and
early stopping are also employed to prevent overfitting and improve generalization ability.
Once training is complete, the trained LSTM model is evaluated on the testing set to assess
its performance in unseen data.
4.3. Fine-tuning of the LSTM Model
Fine-tuning the LSTM model is an iterative process aimed at improving its performance in
MPPT control [18]. This involves adjusting hyperparameters, retraining the model with
additional data, and fine-tuning the network architecture to better capture complex
patterns in the input data. Between fine-tuning techniques such as grid search and Bayesian
optimization, grid search was employed due to it’s simplicity to systematically explore the
hyperparameter space and identify the optimal configuration for the LSTM model. Fine-
tuning the LSTM model is essential for achieving high accuracy and robustness in real-world
applications of solar panel systems.
5. Results
This section presents the experimental findings obtained from comparing the performance
of LSTM-based MPPT controllers with conventional algorithms, along with an evaluation of
the accuracy, efficiency, and adaptability of the LSTM model under various conditions.
5.1. Experimental Findings
The averaged output of a photovoltaic system with different MPPT controllers across
different times of day is shown in the Table 2. The experimental results demonstrate the
efficacy of LSTMbased MPPT controllers in optimizing solar panel performance compared
to traditional algorithms such as Perturb and Observe (P&O), Incremental Conductance
(IncCond), and Hill Climbing (HC). The LSTM-based controller exhibits superior tracking
accuracy and adaptability to changing environmental conditions, resulting in higher energy
yields across different scenarios.
Table 2
Solar photovoltaic generation (W) with different MPPT controllers
Time of Day Perturb and Observe Incremental Conductance Hill Climbing LSTM-based
approach
00:00 0.00 0.00 0.00 0.00
01:00 0.00 0.00 0.00 0.00
02:00 0.00 0.00 0.00 0.00
03:00 0.00 0.00 0.00 0.00
04:00 0.00 0.00 0.00 0.00
05:00 0.00 0.00 0.00 0.00
06:00 15.83 17.55 16.69 19.55
07:00 93.74 103.94 98.83 115.75
08:00 175.17 194.23 184.67 216.30
09:00 211.86 234.92 223.36 261.60
10:00 234.98 260.56 247.73 290.15
11:00 247.04 273.94 260.45 305.05
12:00 251.32 278.68 264.96 310.33
13:00 246.04 272.82 259.39 303.81
14:00 240.01 266.13 253.03 296.36
15:00 228.95 253.87 241.37 282.71
16:00 211.86 234.92 223.36 261.60
17:00 170.64 189.22 179.90 210.71
18:00 94.74 105.06 99.88 116.99
19:00 42.97 47.65 45.30 53.06
20:00 0.00 0.00 0.00 0.00
21:00 0.00 0.00 0.00 0.00
22:00 0.00 0.00 0.00 0.00
23:00 0.00 0.00 0.00 0.00
Under varying solar irradiance levels and temperature gradients, the LSTM-based MPPT
controller consistently outperforms traditional algorithms in maintaining the solar panel
operating point at or near the maximum power point. This is evidenced by the higher power
output achieved by the LSTM-based controller compared to conventional methods,
particularly during transient conditions and partial shading events.
5.2. Evaluation of Accuracy, Efficiency, and Adaptability
The comparison of MPPT controllers’ performance on a cloudless day is shown on Figure 4.
However, the accuracy of the LSTM model is assessed based on its ability to predict the
maximum power point of the solar panel accurately under diverse environmental
conditions. Comparative analysis with traditional algorithms reveals that the LSTM-based
MPPT controller achieves higher accuracy in tracking the optimal operating point, resulting
in increased energy harvesting efficiency. For instance, during the hours with variable solar
irradiance, the LSTM-based MPPT controller managed to produce 19.01% more power than
P&O controller, 10.19% more power than IncCond controller and 14.62% more power than
HC controller. The LSTM-based controller exhibits robustness in adapting to rapid
fluctuations in environmental conditions, effectively optimizing power output and
mitigating the effects of partial shading and other transient phenomena. During the ideal
conditions all controllers operate within the margin of error from each other.
Figure 4: Comparison of solar panel output with different MPPT controllers.
Furthermore, the efficiency of the LSTM model is evaluated in terms of its computational
complexity and real-time performance. Despite the additional computational overhead
associated with training and implementing LSTM networks, the experimental results
demonstrate that the LSTM-based MPPT controller maintains high efficiency in predicting
the maximum power point and dynamically adjusting the operating parameters of the solar
panel system.
Overall, the experimental findings underscore the effectiveness of LSTM-based MPPT
controllers in enhancing solar panel efficiency and energy harvesting capabilities, thereby
contributing to the advancement of renewable energy technologies.
6. Discussions
This section engages in discussions surrounding the limitations and challenges encountered
during experimentation, interpretation of findings, analysis of advantages and
disadvantages of using LSTM neural networks for MPPT control, and proposes future
research directions in the field.
6.1. Limitations and Challenges
Despite the promising results obtained, several limitations and challenges were
encountered during the experimentation phase. One notable challenge is the complexity of
training LSTM networks, which requires large amounts of data and computational
resources. Additionally, the generalization ability of the LSTM model may be limited by the
specific conditions and datasets used for training, leading to potential performance
degradation in real-world applications.
6.2. Interpretation of Findings
The experimental findings underscore the potential of LSTM-based MPPT controllers in
significantly improving solar panel efficiency and energy harvesting capabilities. By
leveraging the temporal dependencies in input data, LSTM networks demonstrate superior
accuracy and adaptability compared to traditional MPPT algorithms. The interpretation of
findings suggests that LSTM-based approaches hold promise for enhancing the
performance of solar panel systems under diverse environmental conditions, contributing
to the advancement of renewable energy technologies and sustainability efforts.
6.3. Advantages and Disadvantages
The analysis of advantages of using LSTM neural networks for MPPT control highlights their
ability to capture long-term dependencies and complex patterns in time-series data,
enabling more accurate and dynamic control of solar panel systems. Additionally, LSTM-
based approaches offer potential for real-time optimization and adaptability to changing
environmental conditions. However, disadvantages such as computational complexity,
requirement for extensive training data, and potential overfitting remain challenges to be
addressed.
6.4. Future Research Directions
Future research in the field of LSTM-based MPPT controllers may explore potential
applications, scalability, and hybrid approaches to further enhance performance and
applicability. One direction is the investigation of hybrid models that combine traditional
MPPT algorithms with LSTM-based approaches to leverage the strengths of both methods.
Additionally, research on optimizing the computational efficiency of LSTM networks and
improving generalization ability in realworld scenarios is warranted. Furthermore,
scalability and deployment of LSTM-based MPPT controllers in large-scale solar panel
systems merit exploration to facilitate widespread adoption and maximize energy
harvesting efficiency.
7. Conclusions
This section provides a summary of the key findings and contributions of the study,
emphasizing the significance of LSTM-based MPPT controllers in enhancing solar panel
efficiency and contributing to sustainable energy production.
7.1. Key Findings and Contributions
In summary, the study investigated the efficacy of LSTM-based MPPT controllers for
optimizing solar panel performance. Experimental results demonstrated that LSTM-based
controllers outperform traditional MPPT algorithms in terms of accuracy, efficiency, and
adaptability. By leveraging the temporal dependencies in input data, LSTM networks
accurately predict the maximum power point of solar panels under diverse environmental
conditions, thereby maximizing energy harvesting efficiency. The study contributes to
advancing the field of renewable energy technologies by showcasing the potential of AI-
based approaches in enhancing solar panel efficiency.
7.2. Significance of LSTM-based MPPT Controllers
The significance of LSTM-based MPPT controllers lies in their ability to address the inherent
challenges of traditional algorithms and improve the efficiency of solar panel systems. By
leveraging the capabilities of LSTM neural networks, these controllers offer enhanced
accuracy and adaptability, enabling more effective utilization of solar energy resources.
LSTM-based MPPT controllers play a crucial role in advancing sustainable energy
production by maximizing the power output of solar panels and reducing dependency on
fossil fuels. As such, they represent a promising avenue for realizing the transition towards
a more sustainable and environmentally friendly energy future.
References
[1] V. Fthenakis, J. E. Mason, K. Zweibel, The technical, geographical, and economic
feasibility for solar energy to supply the energy needs of the us, Energy policy 37
(2009) 387–399.
[2] M. A. H. Mondal, Economic viability of solar home systems: Case study of bangladesh,
Renewable Energy 35 (2010) 1125–1129.
[3] Z. Li, F. Boyle, A. Reynolds, Domestic application of solar pv systems in ireland: The
reality of their economic viability, Energy 36 (2011) 5865–5876.
[4] J. J. Nedumgatt, K. Jayakrishnan, S. Umashankar, D. Vijayakumar, D. Kothari, Perturb
and observe mppt algorithm for solar pv systems-modeling and simulation, in: 2011
Annual IEEE India Conference, IEEE, 2011, pp. 1–6.
[5] A. Safari, S. Mekhilef, Incremental conductance mppt method for pv systems, in: 2011
24th Canadian Conference on Electrical and Computer Engineering (CCECE), IEEE,
2011, pp. 000345–000347.
[6] W. Xiao, W. G. Dunford, A modified adaptive hill climbing mppt method for photovoltaic
power systems, in: 2004 IEEE 35th annual power electronics specialists conference
(IEEE Cat. No. 04CH37551), volume 3, Ieee, 2004, pp. 1957–1963.
[7] N. Karami, N. Moubayed, R. Outbib, General review and classification of different mppt
techniques, Renewable and Sustainable Energy Reviews 68 (2017) 1–18.
[8] C. Kumar, G. Subramaniam, J. Jasper, et al., A novel roa optimized bi-lstm based mppt
controller for grid connected hybrid solar-wind system, COMPEL-The international
journal for computation and mathematics in electrical and electronic engineering 42
(2022) 378–401.
[9] S. Pengcheng, Q. Jiawei, Research on maximum power point tracking based on lstm
neural network, in: 2021 4th International Conference on Advanced Electronic
Materials, Computers and Software Engineering (AEMCSE), IEEE, 2021, pp. 866–869.
[10] M. M. Rana, M. R. Ali, A. K. Ajad, M. Moznuzzaman, Analysis of p&o and inc mppt
techniques for pv array using matlab, IOSR journal of electrical and electronics
engineering 11 (2016) 80–86.
[11] M. Kaouane, A. Boukhelifa, A. Cheriti, Implementation of incremental-conductance
mppt algorithm in a photovoltaic conversion system based on dc-dc zeta converter, in:
2016 8th International Conference on Modelling, Identification and Control (ICMIC),
IEEE, 2016, pp. 612–617.
[12] B. El Hassouni, M. Ourahou, W. Ayrir, A. Haddi, A. Amrani, A study of efficient mppt
techniques for photovoltaic system using boost converter, International Journal of
Emerging Electric Power Systems 19 (2018) 20170180.
[13] M. Seyedmahmoudian, B. Horan, T. K. Soon, R. Rahmani, A. M. T. Oo, S. Mekhilef, A.
Stojcevski, State of the art artificial intelligence-based mppt techniques for mitigating
partial shading effects on pv systems–a review, Renewable and Sustainable Energy
Reviews 64 (2016) 435–455.
[14] S. A. Raj, G. G. Samuel, Survey of ai based mppt algorithms in pv systems, in: 2022 4th
international conference on smart systems and inventive technology (ICSSIT), IEEE,
2022, pp. 597–604.
[15] N. Aouchiche, M. Aitcheikh, M. Becherif, M. A. Ebrahim, Ai-based global mppt for partial
shaded grid connected pv plant via mfo approach, Solar Energy 171 (2018) 593–603.
[16] B. Amrouche, M. Belhamel, A. Guessoum, Artificial intelligence based p&o mppt
method for photovoltaic systems, Revue des Energies Renouvelables ICRESD-07
Tlemcen (2007) 11–16.
[17] A. Sherstinsky, Fundamentals of recurrent neural network (rnn) and long short-term
memory (lstm) network, Physica D: Nonlinear Phenomena 404 (2020) 132306.
[18] V. Nechakhin, I. Kalinina, A. Gozhyj, Hyperparameter optimization of lstm mppt
controller for solar power plants, in: 2023 IEEE 18th International Conference on
Computer Science and Information Technologies (CSIT), IEEE, 2023, pp. 1–4.