Solar photovoltaic power prediction using different machine learning methods

Abstract

The main aim of the present study is to explore the relationship between numerous input parameters and the solar photovoltaic (PV) power using machine learning (ML) models. Two different ML approaches such as support vector machine (SVM) and Gaussian process regression (GPR) were considered and compared. The basic input parameters including solar PV panel temperature, ambient temperature, solar flux, time of the day and relative humidity were considered for predicting the solar PV power. The results showed that among the proposed ML approaches, Matern 5/2 GPR algorithm provided the optimal performance; whereas cubic SVM had the worst performance. Furthermore, the predicted output results are in good agreement with the experimental values, indicating that the proposed ML approaches are appropriate for use in predicting the power of different solar PV panel. Additionally, to showcase the effectiveness and the accuracy of SVM and GPR models in forecasting solar PV power, the results of these models are compared using root mean squared error (RMSE) and mean absolute error (MAE) criteria.