CAJ | 학술논문

针对传统的局部放电模式分类器存在的不足,提出了一种基于统计特征参数与相关向量机( RVM )的变压器局部放电类型识别的新方法。首先针对4种变压器局部放电实验模型的二维图谱提取出表征图谱特征的16个统计参数,然后设计一对一RVM多分类模型,将统计参数作为输入向量送入RVM分类模型,实现放电类型识别。测试结果表明,RVM分类器具有较好的放电识别效果,与支持向量机( SVM)相比具有计算复杂度低、相关向量少、训练及测试时间短等优点,两者识别精度相当,均高于BPNN。
침대전통적국부방전모식분류기존재적불족,제출료일충기우통계특정삼수여상관향량궤( RVM )적변압기국부방전류형식별적신방법。수선침대4충변압기국부방전실험모형적이유도보제취출표정도보특정적16개통계삼수,연후설계일대일RVM다분류모형,장통계삼수작위수입향량송입RVM분류모형,실현방전류형식별。측시결과표명,RVM분류기구유교호적방전식별효과,여지지향량궤( SVM)상비구유계산복잡도저、상관향량소、훈련급측시시간단등우점,량자식별정도상당,균고우BPNN。
To overcome the defect of traditional partial discharge pattern classifier, a novel method is proposed based on statistical parameters and RVM for partial discharge type recognition. 16 statistical parameters are extracted which represent partial discharge 2-dimension diagram. One against one multiple RVM classifier is designed. And then the extracted parameters are sent to RVM model for partial discharge type recognition. Experiment results demonstrate that RVM classifier can get good recognition effect. Compared with SVM, RVM has lower complexity, less relevance vec-tors, shorter training and testing time. The partial discharge type recognition accuracy of RVM and SVM is better than that of BPNN.