基于油气参数分析的电力变压器故障分步式诊断算法

发布时间：2019-06-11 09:38

【摘要】：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。
[Abstract]:In order to improve the diagnosis speed and accuracy of power transformer fault diagnosis, a step-by-step fault diagnosis algorithm based on transformer oil and gas parameters is proposed. In the first step of the algorithm, the quantum behavior support vector machine (SVM) fault diagnosis algorithm is adopted, that is, SVM is used to classify the faults of large power transformers. In the process of classification, an improved genetic algorithm with quantum behavior is used to optimize the parameters of SVM. On the basis of the first step, the samples existing in suspicious areas are classified by K-nearest neighbor clustering analysis algorithm in the second step. The simulation results show that the improved quantum genetic algorithm only needs 50 generations to reproduce to get the best classification model, while the ordinary genetic algorithm needs 170 generations to get the optimal classification model. At the same time, the organic combination of cluster analysis and support vector machine improves the classification accuracy from 97.5% to 100%. It can be seen that the proposed fault step-by-step diagnosis algorithm can effectively improve the diagnosis speed and accuracy of fault diagnosis, and can be widely used in power transformer fault diagnosis.
【作者单位】：重庆大学通信工程学院;重庆工商大学会计学院;
【基金】：国家重点基础研究发展计划(973计划)(2012CB21520) 第四届国家大学生创新性实验项目(101061118)~~
【分类号】：TM41

【相似文献】