基于主成分分析法和贝叶斯网络的智能变电站故障诊断方法

摘要

Applying the principal component analysis (PCA) to Bayesian networks to simplify the failure sources and compress the fault features and to obtain the minimum diagnosis scale and reduce the complexity of the operating system was proposed.Through considering the substation's operation mode and protective measures,a corresponding Bayesian network model was established and the Bayesian network's variable elimination method was adopted to analyze a 220kV substation and ratiocinate its fault probability.The experimental results show that,this method can describe the change of fault characteristics and the analysis of the substation fault causes is direct and reliable.%提出将主成分分析法应用于贝叶斯网络,通过对故障源的简化与故障特征的压缩,得到最简故障特征表,降低操作系统的复杂性.根据变电站运行模式和保护措施建立相应的贝叶斯网络模型,利用贝叶斯网络的变量消元法对220kV变电站进行实例分析并对其故障概率进行推理.实验结果证明:该方法利于描述故障特征的变化,对变电站故障原因的分析直观可靠.