Improved Importance Sampling for Reliability Evaluation of Composite Power Systems

Abstract

This paper presents an improved way of applying Monte Carlo simulation using the crossentropy method to calculate the risk of capacity deficit of a composite power system. By applying importance sampling for load states in addition to the generation and transmission states in a systematic manner, the proposed method is many orders of magnitude more efficient than the crude Monte Carlo simulation and considerably more efficient than other crossentropy-based algorithms that apply other ways of estimating the importance sampling distributions. An effective performance metric of system states is applied in order to find optimal importance sampling distributions during presimulation that significantly reduces the required computational effort. Simulations, using well-known IEEE reliability test systems, show that even problems that are nearly intractable using crude Monte Carlo simulation become very manageable using the proposed method.