Abstract: With the ambiguity of fault condition and the uncertainty of panoramic state during the operation of power equipment,a Bayesian network method was improved by combining fuzzy functions for fault diagnosis and state estimation.In this network,Bayesian probability was used to measure the correlation between multi-dimensional individual indicators and different faults,and a time-varying scoring function was constructed to integrate the feature information with different timeliness and to quantify the fuzzy fault condition.In addition,all faults were graded based on the hazard,and the multiple fuzzy functions were constructed and integrated in the network to measure the fuzzy importance of continuously changing fault conditions in panoramic state estimation.The comprehensive score could be calculated,and the operating state with potential failure could be inferred.The 500kV oil-immersed power transformer was taken as an example to test the effectiveness of the proposed method,and the results showed that the application accuracy of the proposed method was much higher than the existing linear evaluation methods.

Key words: Bayesian network, fuzzy functions, fuzzy fault condition, panoramic state estimation

摘要： 针对电力设备在运行过程中故障程度模糊和全景状态的不确定问题,提出一种融合模糊函数改进的贝叶斯网络故障诊断和状态评估方法。在该网络中,采用贝叶斯概率测度多维特征指标与不同故障之间的关联性,构造时变评分函数整合具有不同时效性的特征信息,量化故障发生的模糊状态。另外,基于危害性对故障进行分级,在网络中融合多个模糊函数,分别描述不同故障对应连续变化的模糊状态在全景状态评估中的模糊重要性;在此基础上,计算设备运行的综合评分值,推测其所处的状态等级和潜在故障。最后,以500kV油浸式电力变压器为例对所提方法的有效性进行实验和分析,结果显示该方法的应用准确率远高于现有的线性评价方法。

关键词: 贝叶斯网络, 模糊函数, 模糊故障状态, 全景状态评估