A COMPARATIVE STUDY ON SAMPLING BASED ALGORITHMS FOR NETWORK RELIABILITY ASSESSMENT

Abstract

Infrastructure networks, such as electrical power grids, transportation and water supply systems, support critical societal functions of society. Failures of such networks can have severe consequences, and quantification of the probability of failure of such systems is essential for understanding and managing their reliability. Analytical and simulation methods have been proposed to solve such kinds of problems, among which sampling methods feature prominently. Recently, the authors extended widely used structural reliability algorithms, subset simulation, cross-entropy-based importance sampling as well as uncertainty quantification methods built from particle integration methods and exact confidence, all for efficient reliability analysis in discrete spaces. This paper tests the performance of these algorithms for static network reliability assessment. In particular, we compare these methods for optimal power flow problems in various IEEE benchmark models. Overall, the cross-entropy-based method outperforms the other methods in all benchmark models except the largest IEEE 300, while the adaptive effort subset simulation and particle integration methods are more suitable for handling high-dimensional problems. By building up the benchmark models, we provide unified examples for comparing different emerging methods in static network reliability assessment and also to support improvement or combination of these methods.