Modeling the temporal correlation of ultra-short term wind power forecast error based on copula theory

Abstract

With the increasing wind power penetration, the randomness and uncertainty of wind power forecast errors effect the operation of power system deeply. Therefore, a high-dimensional model of ultra-short wind power forecast errors and probability distributions of different time periods forecast error is proposed based on copula theory. Firstly, the parametric methods and nonparametric methods are introduced and used to fit the probability distribution of wind power forecast error. By comparing fitting accuracy of different fitting methods, KDE-based method with highest fitting accuracy is chose to fit marginal distribution of forecast error. Then, this paper models the temporal correlation of ultra-short term wind power forecast using copula function and obtains joint cumulative distribution function (JCDF) of forecast errors. Finally, the actual forecast error data of a wind farm located on Hebei province is used to verify the model. Comparing with the actual dependence structure, the method based on copula theory can effectively model the temporal correlation and detect zero dependence of ultra-short term wind power forecast errors. Thus the effectiveness of proposed method is proved by simulated results.