S变换模矩阵和最小二乘SVM在雷电及操作过电压识别中的应用

Application of S-transform modular matrix and LS-SVM in identification of lightning and switching overvoltages

提出一种基于S变换模矩阵和最小二乘支持向量机(SVM)的雷电及操作过电压识别方法。通过对零序过电压信号的S变换模矩阵进行奇异值分解,将过电压信号的特征信息分解到不同的时频特征子空间,提取奇异值的5类统计特征参量作为过电压识别的特征向量,并将其输入最小二乘SVM分类器,实现雷电及操作过电压的类型识别。过电压实测数据表明:所提特征方法的特征量维数低,抗干扰能力强;采用的识别方法训练次数少,识别率高,可较好地应用于雷电及操作过电压的识别。

S-transform modular matrix and LS-SVM（Least Square-Support Vector Machine） are applied to identify the lightning overvohage and switching overvoltage. The singular value decomposition decomposes the S-transform modular matrix of zero sequence overvohage signals into different time-frequency characteristic subspaces,from which five statistical features are extracted and used as the input vectors of the LS-SVM classifiers to identify the lightning and switching overvoltages. Results of the test with measured overvohage data indicate that,the characteristic dimension is small,the features are immune to electromagnetic noises ,the training times is low,and the recognition rate is high. The proposed method can be well applied in identification of lightning and switching overvohages.