ABC-SVM and PSO-RF Model for Photovoltaic Forecasting Based on Big Data

摘要

Prediction of photovoltaic output is of great significance to the stable operation of microgrid system. Firstly, the artificial bee colony based support mechine (ABC-SVM) method is used to train historical meteorological data and photovoltaic output data, which can divide the weather condition into four categories. Secondly, tens of thousands of data are selected under four types of meteorological conditions, and each group of data is trained by particle swarm optimization based random forest (PSO-RF) model. After training, the four different PSO-RF model with different parameters can be obtained for the photovoltaic forecasting individually. Finally, we collect weather information and photovoltaic data from a microgrid station in Yangjiang Guangdong province to test our combined model. Numerical results show that the proposed approach achieves better prediction accuracy than the simple SVR and traditional RF methods.