基于数据不确定性的变压器故障诊断研究

Study on fault diagnosis of the transformer based on data uncertainty

针对采集的变压器油中溶解气体数据的波动性和不确定性,通过引入蒙特卡洛算法进行数据处理.根据油中溶解气体试验,将处理后的变压器油中溶解气体数据进行两两比值.分别根据变压器运行规程、DGA知识以及模糊c均值聚类法对气体含量和比值进行离散化处理,再进行属性约简,并将约简结果作为神经网络的前置输入,对神经网络训练及故障进行诊断.实验结果表明,该方法可以对变压器故障进行准确判定,具有更好的工程实用性.

For the collection data of dissolved gas in transformer oil,the paper fully considers the uncertainty of the collection data,using Monte Carlo algorithm to process the data uncertainty.For the ratio of the expansion,the paper respectively uses operating procedures,DGA knowledge and fuzzy c-means clustering method to discrete the data according to different situations.Secondly,using rough set attribute reduction algorithm gets the minimum decision table and which as a result of the input trains the neural network.The neural network after training makes for the fault diagnosis and using the testing data test the neural network.The test results indicate that the method of the paper has a higher accuracy rate for transformer fault diagnosis.