1. State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources(North China Electric Power University), Baoding 071003, China;2. China Electric Power Research Institute, Beijing 100192, China;3. State Grid Qinghai Electric Power Company, Xining 810000, China

The forecasting error of photovoltaic(PV)output is difficult to avoid and cannot be ignored. The accurate description of the forecasting error distribution is beneficial to the optimal dispatch and stable operation of the power system. The correlation between the forecasting error distribution and influencing factors is analyzed, and a probabilistic model of the PV output forecasting error based on numerical characteristic clustering is proposed. The whole level of forecasting errors will be classified through the fuzzy C-means clustering and the classification on the output forecasting will be conducted according to its numerical characteristics. A general Gauss mixed model available to estimate error distribution is established. Considering the influences of meteorological factors and numerical characteristics on forecasting errors, the model could estimate forecasting errors at different times and produce the confidence interval of error distribution. At the same time, the proposed model is free from the influences brought by different forecasting algorithms and the geographic information of PV power station. Based on the historical data of PV systems in Belgium and Northwest China, analysis results show that the skewness, multi-peak and kurtosis diversity of PV power forecasting error distribution can be described by the proposed model more accurately compared with other distribution models, which can be used to describe the forecasting error distribution of day-ahead PV power in different situations.