Energy Harvesting from various renewable energy sources has experienced rapid growth due to the detrimental environmental effects caused by the use of fossil fuels. Solar energy is a prominent energy resource widely employed for power generation in wide range of applications. The output power of a Photovoltaic (PV) panel depends significantly on environmental parameters, which results in irregular and variable output power. This unstable nature of output power has evolved as a notable issue. A precise framework for forecasting the output power of PV panel is essential. This manuscript proposes a hybrid ensemble Machine Learning (ML) model combining Support Vector Regression (SVR), Decision Tree Regression (DTR), and K-nearest Neighbor (KNN) as base learner with Linear Regressor (LR) as meta meta model to forecast short term PV panel output power. The proposed model is trained on selected features using Pearson's correlation coefficient and Principal Component Analysis (PCA) as feature selection techniques. In addition, results are obtained by computing various ML models to compare and evaluate the proposed work. The proposed model outperforms conventional ML models by achieving maximum coefficient of determination (R2) of 0.82 and minimum Mean Square Error (MSE) of 0.19.
Machine Learning Based Framework for Prediction of Photovoltaic Output Power
09.08.2023
531927 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Photovoltaic Power Prediction Based on Improved Sparse Bayesian Regression
Online Contents | 2016
|