Nonparametric Modeling Method for Correlation Analysis of Performance Indexes against the Load Loss Caused by Power System Cascading Failures

This paper study and analyze the correlation relationship between performance indexes of power system and the load loss caused by cascading failures by nonparametric correlation model. A simulation platform is firstly established based on a multi-layer time-series operation evolution model to generate a large amount of representative cascading failures samples. In general, electric state indexes, network structural indexes and distribution entropy indexes, as three kinds of most commonly utilized power system performance indexes, are investigated by this paper to extract their inner correlation relationship with power loss. The weakly coupled indexes are deleted by nonparametric independence screening method, and then the strongly coupled indexes are further studied by Group Lasso nonparametric regression analysis to extract the correlation relationship with load loss. Simulation results demonstrate that the proposed nonparametric modeling method perform well to discover the performance indexes that have strong correlation relationship with load loss caused by cascading failures. Moreover, these correlation rules can further serve as alarms for cascading failures of power systems.