基于贝叶斯网络的水电站多源信息故障诊断

为提升水电站电力设备故障诊断的智能化水平,特别是解决缺乏状态监测系统的设备的故障诊断问题,提出了一种基于贝叶斯网络(BN)的水电站多源信息故障诊断方法。该方法对拥有状态监测系统的设备,采用机器学习法构建BN模型。对于缺乏状态监测系统的设备,则提出一种改进专家经验法构建BN模型:使用故障树(FTA)模型转化贝叶斯网络结构,Noisy-Or模型简化条件概率表(CPT),模糊综合评价法(FCE)获取条件概率。最后,融合两个模型建立了多元信息故障诊断模型,并分别采用数据对比分析与受试者工作特性曲线(ROC)对模型进行验证。结果表明,在1000组验证数据中,发电机、电力传输系统的准确率分别为81. 7%、81. 3%,并且ROC曲线的AUC值达到0. 8099,说明该方法具有较高的准确性和实用性。

Fault diagnosis of multi-source information of hydropower station based on Bayesian network

In order to improve the intelligent level of power equipment fault diagnosis in hydropower station，especially the problem that the fault diagnosis of equipment lacking state monitoring system，a method of multi-source information fault diagnosis for hydropower station based on Bayesian Network ( BN) is proposed. The machine learning method is used to construct the BN model for the equipment with state monitoring system. For the equipment lacking of state monitoring system，an improved expert experience method is proposed to construct the BN model，and the steps are as follow. The FTA model is used to transform the BN structure，the Noisy-Or model is used to simplify the conditional probability table ( CPT) ，and the fuzzy comprehensive evaluation ( FCE) is used to obtain the conditional probability ( CPT) . The multi-source information fault diagnosis model is established by combining the two models，and validated by data comparison analysis and the receiver operating characteristic ( ＲOC) curve. The result shows that in the 1000 groups of data validation，the accuracy of generator and power transmission system is 81. 7% and 81. 3% respectively. The AUC value of ＲOC curve is 0. 8099，which means that the method has a pretty high accuracy and practicability.