Applications of Machine Learning Algorithms in Renewable Energies
Pavan Gangwar; Aishvarya Narain
Abstract
This paper presents various maximum power point tracking (MPPT) techniques for solar photovoltaic (SPV) systems operating under partial shading conditions (PSCs). Traditional methods such as perturb and observe (P&O) used as a base model which face significant challenges in accurately identifying ...
Read More
This paper presents various maximum power point tracking (MPPT) techniques for solar photovoltaic (SPV) systems operating under partial shading conditions (PSCs). Traditional methods such as perturb and observe (P&O) used as a base model which face significant challenges in accurately identifying the maximum power point (MPP) in the power-voltage curve. To overcome these challenges other optimization techniques like particle swarm optimization (PSO), grey wolf optimization (GWO), and cuckoo search algorithm (CSA) and machine learning (ML) is used. A multilayer perceptron (MLP) based MPPT framework designed to predict duty cycles based on SPV voltage and current inputs. Simulation results shows that the MLP-based approach achieves faster convergence in power output, and improved voltage stability compared to P&O, PSO, GWO, and CSA methods. The result highlights the potential of integrating ML techniques into SPV systems to enhance efficiency in challenging PSC scenarios. This research contributes to the advancement of sustainable solar energy technologies by leveraging adaptive intelligence for optimal energy harvesting.
Applications of Machine Learning Algorithms in Renewable Energies
Marouan Ben El Haj; Said Ziani
Abstract
Reliable extraction of PV module electrical parameters constitutes a critical step toward enabling fault detection and performance assessment. The strong dependence of these parameters on environmental factors, particularly solar irradiance and cell operating temperature, introduces significant nonlinearity ...
Read More
Reliable extraction of PV module electrical parameters constitutes a critical step toward enabling fault detection and performance assessment. The strong dependence of these parameters on environmental factors, particularly solar irradiance and cell operating temperature, introduces significant nonlinearity that complicates their accurate identification. This paper proposes a hybrid metaheuristic framework, combining the Whale Optimization Algorithm (WOA) for global search with the Fishier Mantis Optimizer (FMO) for local refinement, for the parameter identification of single-, double-, and triple-diode PV models (SDM, DDM, TDM). A Bidirectional Long Short-Term Memory (BiLSTM) network is additionally investigated as a candidate residual corrector between the equivalent-circuit simulation and the measured current. Experimental validation uses current-voltage (I-V) data measured from the Solarex MSX60 module at three irradiance levels (600, 800, 1000 W/m², 25 °C). In a fair benchmark where PSO, GA, standard WOA, and two modern (post-2019) metaheuristics — the Arithmetic Optimization Algorithm (AOA) and the Slime Mould Algorithm (SMA) — are equipped with the same Latin-Hypercube initialization, opposition-based learning and reflection-based boundary handling as the proposed method, WOA-FMO attains a physical-parameter RMSE of 3.71×10⁻² A on all three topologies, significantly outperforming every baseline, classic and modern alike (Wilcoxon rank-sum, p<0.01 in all ten pairwise comparisons), at a comparable computational cost.
Applications of Machine Learning Algorithms in Renewable Energies
Kossoko Babatoundé Audace Didavi; Richard Gilles Agbokpanzo; Bienvenu Macaire Agbomahena
Abstract
In this work, the photovoltaic power forecast for the next 24 hours by combining a time series forecasting model (LSTM) and a regression model (XGBoost) from direct irradiation only is performed. Several meteorological parameters such as irradiance, ambient temperature, wind speed, relative humidity, ...
Read More
In this work, the photovoltaic power forecast for the next 24 hours by combining a time series forecasting model (LSTM) and a regression model (XGBoost) from direct irradiation only is performed. Several meteorological parameters such as irradiance, ambient temperature, wind speed, relative humidity, sun position, dew point were identified as influencing parameters of PV power variability. Thanks to the parameter extraction and selection techniques of the XGBoost model, only the direct irradiation could be kept as input parameters. The LSTM model was used to predict the direct irradiation for the next 24 hours and the XGBoost model to estimate the future power from the predicted irradiation. These models were developed under Python 3, the exploited data were downloaded in the PVGIS database for the city of Abomey-Calavi in Benin and the prediction was carried out on a panel of 1000W of peak power. An experimental validation was then performed by comparing the predicted irradiance values to the measured values on site. It was obtained for the LSTM model a root mean square error of 3.66 W/m2 and for the XGBoost model a root mean square error and a regression coefficient of 1.72 W and 0.992129 respectively. These results were compared to the LSTM-XGBoost performances with irradiation, temperature, sun position and wind speed as inputs. It was found that the use of irradiation alone as input did not as such impair the forecast performance. The proposed method was also found to be more efficient than LSTM and CNN models used alone.
