运用时序贝叶斯知识库的电网故障诊断方法

A Power System Fault Diagnosis Method Using Temporal Bayesian Knowledge Bases

电网故障时有大量报警产生,充分利用报警信号及其时序信息,处理好保护与断路器误动、拒动、信息缺失等不确定性情况,对于电网故障诊断显得非常重要。时序贝叶斯知识库(temporal Bayesian knowledge bases,TBKB)能够清晰表达多个事件之间的时序约束关系,并具备贝叶斯网络的推理能力。建立了基于TBKB的电网故障诊断模型,提出了元件故障与保护动作、保护动作与相应断路器跳闸等之间的时序因果关系(TCR)表达、时序约束一致性检查方法。根据电网结构,可先在线搜索出疑似元件,再对它们自动构造TBKB模型。针对信息缺失节点的状态进行假设,形成假设状态组合。针对这些状态组合,通过贝叶斯反向、正向推理,可判断故障元件,误动与拒动的保护与断路器。多个算例验证了该方法的有效性。

After the fault occurs in power grid, many alarm messages are generated. For power system fault diagnosis, it is important to utilize the alarms and their temporal information, and deal with the uncertainty such as mal-function, rejection and incompletion. The theory of temporal Bayesian knowledge bases （TBKB） can clearly express the temporal constraint relationship among multiple events, and possess Bayesian network’s reasoning ability. The TBKB-based power system fault diagnosis models were studied. The expression of temporal casual relationship （TCR） among fault components and protection operations and related breakers tripping are proposed. The consistency checking of TCR was studied, as well as on-line searching algorithm of suspicious components and automatic generating method of TBKB models. For the states of information missing, the state assumption is adopted to create hypothetic state combinations. For these states, Bayesian backward and forward reasoning is made to detect the fault component and identify mal-function and rejection of protections and breakers. The given examples have illustrated that the proposed fault diagnosis method is effective.