The Application Research Of Bayesian Networks On Distribution System Reliability Evaluation

The reliability of Distribution System has direct relation with the user's interests for it is an important link of contact between the whole power system and users which can supply and distribute power for them. At present,Analytical Method and Monte Carlo Simulation Method are mainly used in power system reliability evaluation,but they can only analyze sensitivity and calculate reliability indexes, can not effectively identify the weak link of power system reliability. And Importance Degree Analysis Methods used in normal reliability engineering are difficult to be applied in power system reliability evaluation. This paper reviews several common methods in Distribution System Reliability Evaluation firstly, and educe the Bayesian Network Reasoning Algorithm by comparing the advantages and disadvantages of every method. This algorithm not only can evaluate the system reliability effectively and accurately, but also effectively identify the weak link of power system reliability. So it can provide some specific information which has an important guidance purpose for practical engineering decision-making such as improving system reliability, identifying faults and making maintenance scheduling.According to the own characteristics of Distribution System Bayesian Network, this article further perfects the Bayesian Networks Sequence Simulation Reasoning Algorithm, put forward a concept"User Distribution Coefficient"and optimize the distribution of system nodes'get and loss of power, so as to obtain the system nodes'parameters such as interruption frequency. In this paper a variance reduction technique is been put forward which is suit for Bayesian Networks Sequence Simulation Reasoning Algorithm through a comparative analysis of the Importance Sampling Technique which is used to reduce the variance of Monte Carlo Simulation Method, namely Bayesian Networks Sequence Simulation Reasoning Algorithm Fused Importance Sampling Technique. This algorithm can reduce the variance of system simulation, improve the simulation convergence rate and low the computational cost.This paper developes a distribution system reliability evaluation software. This software chooses Microsoft Visual C# as programming language, SQL Server 2000 Relation Database Management System to access dates. This article selects IEEE-RBTS bus2 Distribution System as experimental system and estimates its reliability respectively using the traditional Bayesian Networks Sequence Simulation Reasoning Algorithm and Bayesian Networks Sequence Simulation Reasoning Algorithm Fused Importance Sampling Technique. By analyzing and processing the experiment date, it is proved that these two methods both can get accurately the experimental system reliability indexes and latter convergence rate obviously higher than the former.