Reliability Assessment Based On Bayesian Network Time Sequence Simulation For Distribution Systems

Operation and planning of distribution systems are inseparable from thereliability assessment, the existing distribution system reliability assessmentmethods have deficiencies, some are only suitable for the simple distributionsystem, some can not solve all of the distribution system reliability indices,also cannot effectively identify the weak links of the system, and somesimulation methods have not convergence criterion or have not rationalconvergence criterion. In recent years, with the global environmentalprotection and energy saving problem attention increasingly, more and moredistribution systems with distributed generations(DGs) are coming, therefore,it is also necessary to study reliability assessment of distribution systemswith DGs. In view of the above problems, the reliability assessment basedon Bayesian network timing sequence simulation(BNTSS) for distributionsystems are researched. the main research results are as follows: (1) A search strategy that starting from load points, first forwardTraversal then backward Traversal components relationship table ofdistribution system is proposed, according to failure mode and effectanalysis(FMEA) and conditional probability algorithm, A“or”relations table,A“union” relations table and A“causal” relations table that describe BNstructure are established, and the conditional probabilities of childnodes of theBN are calculated.The example results indicate the feasibility of the proposedmethod.(2) Distribution system reliability indices computing model based onBNTSS for unchanged loads and time-varying loads are separated established,Corresponding inference algorithm is proposed, that sove not only allreliability indices but also conditional probabilities of every component onsystem failure, thus,both overall assessment for distribution systems andIdentifing weaknesses in the distribution systems.(3) According as central limit theorem and laws of large numbers, inferout a conclusion that sample capacity is inversely proportional to square ofabsolute error. In order to give consideration to both simulation time andcalculation accuracy, how to selecting independent event is studied,independent event concept is expanded, independent generalized events are structured,which are random seeds,the same cycle number of times.(4) In view of distribution system having multiple reliability assessmentindices,and some indices are well-known value, some indices are per-unitvalue, the per-unit value indices as the basis indices of convergence criterionis put forwarded. As per-unit values have stability character, the per-unit valueindices as the basis indices of convergence criterion has wider applicability. Itis suitable for simple system and complex system.(5) Since uncertainty of simulation methods, confidence interval ofpopulation mean of per-unit value indices on a given confidence probability isanalyzed and expressed. It is concluded that length of confidence interval andsample mean maximum absolute error is related, that half the length ofconfidence interval is equal to the sample mean maximum absolute error, andthen main practical convergence criterion with absolute error is proposed thatthe sample mean maximum absolute error is less than the given calculationaccuracy. The sample results verify the validity and feasibility of the proposedmethod.(6) DGs are treated as end nodes, A tree structure described distributionsystems with DGs is established. On the basis of isolated island dividingprinciples, constraint conditions and target function mathematical model are expressed, isolated island dividing for distribution systems with DGs iscompleted by depth-first search strategy on rooted tree. The example resultsindicate the feasibility of the method.(7) The switch operating process are analyzed when components ofdistribution systems with DGs is failure,A method establishing BN fordistribution systems with DGs is proposed. reliability assessment exampleseparated by simulation inference algorithm and precise inference algorithmfor distribution systems with DGs are gived, the example results indicatefurther the validity and superiority of the method based on BNTSS.