Gaussian Processes for Efficient Photovoltaic Power Prediction

Photovoltaic (PV) systems play a pivotal role in the transition towards sustainable energy sources. Grid operators rely on dependable forecasts to balance supply and demand dynamics, ensure grid stability, and optimize energy dispatch strategies. Accurate prediction of PV power enables effective integration of solar energy into the electrical grid. Grid operators need reliable forecasts to balance supply and demand, manage grid stability, and optimize energy dispatch. This study explores the use of Gaussian Process Regression (GPR) models for accurate PV power prediction, specifically comparing GPR with four different kernel functions: Rational Quadratic (QR), Squared Exponential (SE), Matern 5/2 (M52), and Exponential (Exp). Real-world data from a grid-connected PV system in Algeria is employed for evaluation. Results demonstrate that the GPR with SE kernel outperforms the other kernels, exhibiting the highest prediction accuracy based on statistical scores such as root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R2).