Research on modeling method describing wind power fluctuation probability density based on nonparametric kernel density estimation

Abstract

The research on the modeling method of wind power fluctuation probability density is of great importance for wind power integration and operation. In this paper a novel modeling method is proposed for the wind power fluctuation probability density based on nonparametric kernel density estimation. First, wavelet decomposition is used to extract the fluctuation components of wind power to build a model which is based on nonparametric kernel density estimation. This modelling process involves bandwidth optimization. Then we built a bandwidth optimization model constrained by goodness of fit test. Finally, constrained ordinal optimization is utilized to solve the model. Simulation results show that the model constructed by nonparametric kernel density estimation is determined by sample data, therefore this modelling method features with higher accuracy and more general applicability. In addition, an improved strategy, which is proposed in this paper for nonparametric kernel density estimation, also greatly improves the modeling accuracy and computational efficiency.