Data-Driven Reliability Assessment of Wind Power System Operation State

Abstract

The penetration of wind energy into the power grid brings uncertainty to the health of power system. Most of the frontier research on wind power is probabilistic modeling, ignoring the time correlation of reliability evaluation, which has poor mathematical characteristics. This paper studies a data-driven reliability evaluation method for wind power system operation state. In the first part of this paper, the data collected by Supervisory Control And Data Acquisition system (SCADA) is preprocessed. Based on the feature extraction of rough set information entropy, the index vector is used to train the Support Vector Regression (SVR) model and predict the future change of indexes. On this basis, the classification of wind turbine state level is completed. In the second part, through the establishment of Gaussian Mixture Model (GMM), the accurate modeling of practical situation is completed, and the reliability evaluation of wind power output is completed by Bayesian risk indices. The method is verified on the modified IEEE 24-bus reliability test system (RTS), the superiority of the method mentioned in this paper is shown by comparing with the ordinary Unimodal Gaussian model.