Key Laboratory of Modern Power System Simulation and Control & Renewable Energy Technology, Ministry of Education(Northeast Electric Power University), Jilin 132012, China

High quality of photovoltaic(PV)power data is the basis for conducting PV research, but the data collected from PV power plants contains a large proportion of abnormal data, which need to be identified. The PV power data with different illumination characteristics is modeled separately, and the probability distribution function of irradiance and power is constructed by the Copula joint distribution function. The BFGS parameter estimation method based on empirical function is used to estimate the parameters of various Copula functions. Refer to various Copula functions, Euclidean distance of empirical joint distribution function and K-S values, the Copula function is selected. The probability power curve under 90% confidence of the probability distribution of the PV power condition is obtained by combining the narrowest principle of the estimation interval. According to engineering experience and considering the time series characteristics of abnormal data, an abnormal data recognition model is established based on the criteria of four types of abnormal data. The simulation analysis of raw data and synthetic abnormal data of PV power station shows that the proposed method can identify various types of abnormal data effectively and accurately.