Short-Term Photovoltaic Power Prediction Based on SVR-PSO-BP

The power generation of Photovoltaic (PV) fully influenced by the meteorological factors, such as intermittent, unstable and low energy density, etc., causing power grid disequilibrium or light abandonment problems. PV generation rate prediction is an important method to overcome these shortages of PV generation system. The SVR-PSO-BP algorithm is proposed to predict photovoltaic power generation based on different meteorological characteristics and mechanisms that affect photovoltaic power generation. The research object is a ground PV power station for 1.5 MWp. 2 years' meteorological data and power generation data of this station are taken as input data to study its solar irradiance, PV power generation, environmental temperature, and other key characteristics. Using path analysis and grey correlation analysis for data preprocessing, combined with adaptive OPS-GA-PB based short-term photovoltaic power generation prediction algorithm. Using Matlab, the algorithm is implemented. And the algorithm training is applied to get a higher fitting forecasting model. The results show that both solar irradiance and temperature have an effect on PV generation rate, and solar irradiation is the main effect factor. The prediction accuracy of the training set is about 93%, and the prediction accuracy of the actual test set is over 97%. The weather forecast model can be used to predict the before-oneday and within-one-day PV generation in sunny weather. Accurate and timely prediction of photovoltaic power generation is helpful for the advanced dispatch of the power grid.