利用AP-SSVM算法识别GIS的局放缺陷类型

针对传统方法对GIS的缺陷类型识别准确率低的问题,利用AP-SSVM算法识别GIS的缺陷类型,首先将不同缺陷的GIS局放信号进行小波包分解,对各小波包系数进行时频分析,提取信号时域和频域的信息熵,经过AP聚类后,确定最优小波包系数,并将对应的信息熵带入各分类器进行识别。试验结果表明,提出的AP-SSVM算法在识别精确度方面优于SVM和APSVM,能够达到85%以上,为现场GIS故障诊断提供了新的思路,有利于GIS的安全运行。

Identification of GIS Partial Discharge Defect Type through AP-SSVM Algorithm

Aiming at low accuracy rate of identification of GIS defect types through conventional methods,this paper uses AP-SSVM algorithm to identify GIS defect types.First,GIS partial discharge signals of different defectsgo through wavelet packet decomposition.Then,time -frequency analysis is made of various wavelet packet coefficients to extract time-domain and frequency-domain information entropy of the signals.After AP clustering,we can determine the optimal wavelet packet coefficient and bring corresponding information entropies into classifiersfor identification.Experimental results indicate that the proposed AP-SSVM algorithm is superior to SVM and AP-SVM in the respect of identification accuracy and its accuracy can exceed 85％,thus providing a new idea for on-site GIS fault diagnosis in favor of safe GIS operation.