Research of Photovoltaic Power Forecasting Based on Big Data and mRMR Feature Reduction

Due to the multiple influential factors on solar photovoltaic (PV) power outputs, it is rather difficult to predict the short-term PV power generation accurately. Conventional models have taken into consideration the temperature, humidity and wind speed data for forecasting, but these models might not be accurate enough under extreme weather conditions. A novel support vector regression (SVR) -based PV power forecasting model is proposed in this paper, based on big data from multiple photovoltaic, meteorological and weather data sources. The mRMR feature reduction technique is utilized to optimize the feature extraction and decrease the computational burden. Case studies on a real solar PV power station in Salem, USA demonstrate that the forecasting results coincide well with measurement data. And the proposed model has also shown the ability of improving the forecasting accuracy, while reducing the computational time considerably.