Short-Term Photovoltaic Power Forecasting Based on Stacking-SVM

Short-term photovoltaic(PV) power forecast is of great significance for maintaining the safety of the power grid and coordinating the use of resources. This paper introduces the idea and method of ensemble learning, and proposes a short-term photovoltaic power forecast model based on Stacking-SVM. This model uses SVM as base learner and meta learner of the Stacking algorithm, the K-means algorithm is used to cluster the training set, and the meta learner is trained by training samples of the same category as the forecast sample. In order to investigate the forecast performance, the proposed model is used for a 20kW PV power plant, the traditional forecast model is compared with the proposed model in stable meteorological conditions and abrupt meteorological conditions. The results show that the performance of proposed model has been significantly improved.