基于油气参数分析的电力变压器故障分步式诊断算法

Distributed Diagnosis Algorithm for Transformer Fault by Dissolved Gas-in-oil Parameters Analysis

为提高电力变压器故障诊断的诊断速度和准确率,提出了一种以变压器油气参数为诊断依据的电力变压器故障分步式诊断算法。该算法第1步采用量子行为的支持向量机(SVM)故障诊断算法,即采用SVM对大型电力变压器的故障进行分类,在分类的过程中采用改进的具有量子行为的遗传算法对SVM的参数进行寻优。在完成第1步的基础上,第2步再对存在于可疑区域的样本采用K-近邻聚类分析算法分类。仿真结果表明:改进的量子遗传算法只需要50代繁衍就能得到最佳分类模型,而普通遗传算法则需要通过170代才能得到;同时聚类分析与支持向量机的有机结合将分类准确率由97.5%提高到了100%。可见,所提出的电力变压器故障分步式诊断算法能有效地提高故障诊断的诊断速度和准确率,可广泛应用于电力变压器的故障诊断。

To improve the speed and accuracy of fault diagnosis for power transformer, we proposed a distributed diagnosis algorithm based on the hydrocarbon parameters in transformer oil. The first diagnosis step utilizes the quantum behavior of support vector machine （SVM）, namely, the support vector machine （SVM） is used to classify power trans- former faults, and the SVM parameters are optimized by an improved genetic algorithm with the behavior of quantum in the classification process. Moreover, adopting results obtained from the first step, we used K-nearest cluster analysis to further classify the samples in the suspicious areas. Simulation results show that the proposed algorithm requires only 50 generations to obtain the best classification model, in contrast the ordinary genetic algorithm needs 170 generations. Plus, the combination of cluster analysis and SVM increases the accurate classification rate from 97.5% to 100%. It is concluded that the proposed algorithm can effectively improve both speed and accuracy of the fault diagnosis, and it is worth being applied to the fault diagnosis of power transformers.