Ultra Short-term Output Forecasting of Distributed Photovoltaic Power Station Based on Feature Extraction

The power output feature of Distributed photovoltaic power stations (DPV) varies dramatically compared to the traditional photovoltaic power station due to various local weather conditions. Thus it is of great importance to analyze the output features to improve the power forecasting performance. In this paper, weather conditions of DPV are studied via statistical approach to obtain the distinguish forecasting models, which are suitable for sunny day, cloudy day, rainy day and foggy day. Then within each model, the clustering algorithm is applied to select most influencing elements among ambient temperature, radiation intensity, local weather, wind speed and direction. Case study shows that the precision of the proposed algorithm is improved by 0.07% (MAPE) and 31MW (RMSE) compared to the traditional neural network forecasting algorithm.