基于离散随机变量相关性重要抽样的电网可靠性评估交叉熵法OA

The classical cross-entropy method(CEM)typically employs independent importance sampling for discrete random variables(RVs)that represent operational states,such as the"normal"and"failure"states of power devices.While this simplifies CEM,it may limit its performance.This paper introduces an enhanced version with dependent importance sampling for discrete RVs to fully harness CEM's potential.A discrete mixture model(DMM)is proposed to construct the importance sampling probability density function(IS-PDF)for these RVs,with analytical expressions derived for the DMM parameters.The Expectation-Maximization(E-M)algorithm is used to estimate these parameters.The DMM effectively captures discrete RVs' critical joint variation patterns that significantly impact system reliability,leading to more efficient dependent importance sampling.Compared to independent importance sampling,the proposed method improves sampling efficiency for high-risk loss-of-load events caused by the joint failure of multiple devices,thereby accelerating simulation convergence.Reliability evaluation results for the IEEE-RTS79,IEEE-RTS96,and their modified versions demonstrate the superiority of the proposed method over the classical CEM.