Forecast of Solar Photovoltaic Power Output Based on Polycrystalline Panel-based Employing Various Ensemble Machine Learning Methods

Abstract

This paper presents solar photovoltaic (PV) energy prediction based on Polycrystalline technology utilizing various ensemble machine learning (ML) models. Several ML models like Extra Tree Regressor (ETR), Decision Tree Regression (DTR), Random Forest Regressor (RFR), Adaptive Boosting (AdaBoost), and Gradient Boosting Regressor (GBR) were utilized to forecast PV power output and the performance of all models is evaluated according to performance metrics. The selected ensemble models are based on bagging and boosting approaches. The primary input parameters such as solar radiation, wind speed, time, and the actual power generated by the Polycrystalline PV panel based on the 2019 data set were considered for forecasting solar PV output power. The results showed that AdaBoost outperformed the other ensemble ML algorithms, whereas DTR performed the poorest. The AdaBoost model had the best performance, with Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) values of 15.36and 25.05, respectively. On the other hand, the DTR model performed poorly, with an RMSE of 35.72 and an MAE of 23.47, respectively.