基于多层支持矢量机的金属氧化物避雷器在线监测系统

Monitoring System of Metal Oxide Arresters Based on Multi-Layer Support Vector Machine

研究了基于多层矢量机(MLSVM)的金属氧化物避雷器在线状态监测系统,并对新投运的表面清洁的避雷器、紫外辐射老化避雷器、表面染污避雷器及压敏电阻老化避雷器进行试验研究,再提取特征量ir1/it1、it3/it5和ir3/ir5。其中比值ir1/it1可表征金属氧化物避雷器外部染污故障;比值it3/it5可表征金属氧化物避雷器压敏电阻老化故障;比值ir3/ir5作为避雷器紫外辐射老化故障衡量指标。将60%的测试数据用于训练MLSVM分类器,在分类器经训练后,将其余测试样本输入到MLSVM中,对样本进行分类。将分类结果与实际故障情况进行比对可知,MLSVM分类器具有良好的性能,可区分避雷器的不同故障状态,准确诊断避雷器的运行状况。

Research on the monitoring system of metal oxide arresters has been made by using multi-layer support vector machine( MLSVM). Tests have been performed on the clean surface arresters newly put into operation,ultraviolet radiation aging arrester,and superficial polluted arresters as well as the varistors degradation arresters. Then feature quantities ir1/it1,it3/it5 and ir3/ir5 are extracted. The ratio ir1/it1 can characterize the external contamination of the arresters;the ratio it3/it5 can characterize the aging fault of the varistors;the ratio ir3/ir5 is used as the indicator for the ultraviolet radiation aging fault of the arresters. 60% of the test data is used to train the MLSVM classifier. Then,the remaining samples are put into the MLSVM to classify. Comparing the classification results with the actual fault conditions,the MLSVM classifier has a good performance,which can distinguish different fault status of the arresters,and can accurately diagnose the operating condition of the arresters.