用粗糙集理论和贝叶斯网络诊断SF_6断路器故障

Fault Diagnosis of SF_6 Circuit Breaker Using Rough Set Theory and Bayesian Network

为了在断路器故障时能快速、准确地找出故障原因,提出了一种基于粗糙集理论和贝叶斯网络的高压SF6断路器故障诊断的方法。该方法首先根据断路器的故障样本集找出征兆集合和故障集合之间的关系以建立断路器故障诊断决策表,然后利用粗糙集理论属性约简中的区分矩阵算法对决策表进行约简,剔除冗余知识,简化专家知识得到最小诊断规则进而构建贝叶斯网络可以有效降低网络结构的复杂性,最后利用贝叶斯网络的概率推理实现了对断路器故障原因的快速分析。经过实例证明,该方法用于高压SF6断路器的故障诊断是可行有效的,并且最后给出的结果还可以为断路器的状态检修提供依据。

In order to find the reason quickly and accurately why the circuit breaker faults occur, an HV SF6 fault diagnosis method based on rough sets theory （RST） and Bayesian network （BN） is presented. First, according to the fault sample sets of the circuit breaker, the relationship between fault symptom sets and fault reason sets are found. Thus, the fault diagnosis decision table of circuit breaker is established. Second, the discernible matrix algorithm of attribute reduction in the RST is used to reduce the attribute of this fault diagnosis decision table and to eliminate the redundancy knowledge. Then the minimal diagnostic rules can be obtained after simplify the expert knowledge. The complexity of BN structure can be decreased effectively based on the minimal rules. Finally, probability reasoning can be realized by BN, which can be used to analyze fault reasons of circuit breaker fleetly. This method is proved to be feasible and effective for the HV SF6 circuit breaker by the result of practical fault diagnosis examples. The resuits can provide maintenance basis for the circuit breaker maintenance.