A Novel Hybrid CNN-XGBoost Model for Photovoltaic System Power Forecasting
Safia Babikir Bashir, Mena Maurice Farag, Abdul Hamid, Ali Ahmed Adam, Ahmed G. Abo‐Khalil, Ramesh C. Bansal
Abstract
Photovoltaic (PV) systems have been at the forefront of renewable energy resources, being a primary alternate source of energy within the electricity supply. However, PV systems are vulnerable to abrupt changes in weather conditions, which affects the stability of DC output generation. Therefore, forecasting the DC power with respect to variable weather conditions is a necessity. This paper proposes a hybrid forecasting model composing of convolution neural network (CNN) and XGBoost models, to enhance the forecasting accuracy during dynamic weather conditions. The model is validated based on a 2.88 kW grid-connected PV system located in Sharjah, UAE. The proposed model has shown superior performance with respect to contemporary models, depicting an RMSE of 44.18 and a R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.9962, which provides accurate forecasting of DC power under dynamic weather conditions. The infusion of CNN model has significantly enhanced the forecasting accuracy, as compared to the standalone XGBoost.