Bayesian Network Based Time-Sequence Simulation for Power System Reliability Assessment

In this paper, an approximate inference algorithm on Bayesian networks (BN) for reliability assessment of power systems by time-sequence simulation is presented. Based on uniformly distributed pseudo random numbers, the duration time sampling for a component state can be obtained through inverse transform method. According to the logical relationship of the power system to be assessed, the sampling of minimal state cut sets of the system and the system node's sampling are obtained respectively. Thus, the corresponding sample sequence in forward sampling is available. When the amount of sample sequence is large enough, marginal and conditional statistic values are close to marginal and conditional probability of the node. Then the approximate inference results of the BN can be obtained. Numerical simulating results show the effectiveness of the proposed algorithm. It can easily identify the weaker components of a system in reliability, calculate the system reliability indices, and is suitable for a large-scale power system.