A Regression Model-Based Short-Term PV Power Generation Forecasting

Solar photovoltaic (PV) modules have been increasingly integrated into power systems. However, their intermittency and variability have considerable impacts on power grids and could jeopardize the grid's stability when the penetration is high. Developing accurate PV power generation forecasting methods is key to enhancing reliable and secure grid operation. In this paper, a data-driven regression model-based short-term PV power generation forecasting is proposed, where nineteen regression models (including both deterministic and probabilistic predictors) from five regression families are evaluated, and performance assessment indices, such as RMSE and R-squared, are adopted to find the best models. To further improve the performance of forecasting models, hyperparameter optimization and tuning are conducted using MATLAB Regression Learner App. A real-world historical dataset of PV power generation is used to train and further test the models. It is found that the interactions linear, medium Gaussian support vector machine (SVM), and the ensemble of bagged trees outperform other regression models in this study. The proposed method can be utilized by the system operator for effective scheduling future power systems.