Composite Reliability Evaluation Using Monte Carlo Simulation and Least Squares Support Vector Classifier

Abstract

This paper presents a fast and efficient method which combines the Monte Carlo simulation (MCS) and the least squares support vector machine (LSSVM) classifier, for reliability evaluation of composite power system. LSSVM is used to accurately pre-classify the power system operating states as either success or failure states during the Monte Carlo sampling. These pre-classified failure states are then evaluated for adequacy analysis using DC power flow to calculate reliability indices. As a result, the computing time to perform power flow analysis of the system success states is eliminated. The proposed hybrid method is applied to the IEEE Reliability Test System (IEEE-RTS-79) and simulation results obtained using LSSVM with linear and nonlinear kernels are compared with that of nonsequential MCS. These promising results demonstrate the efficacy of the proposed MCS-LSSVM based hybrid method in terms of both classification accuracy and computational time in evaluating the composite power system reliability.