基于支持向量机的励磁涌流识别算法

New Algorithm for Inrush Current Identification of Transformer Based on Support Vector Machine

为提高变压器差动保护识别励磁涌流的能力,将支持向量分类机应用于励磁涌流识别,提出了一种基于支持向量分类机的变压器励磁涌流和内部故障识别新方法.基于励磁涌流和内部故障电流的特点,充分考虑电流互感器饱和的特点提取电流互感器二次侧间断角和二次谐波等特征,并对励磁涌流和内部故障电流的识别方法进行了分析;用EMTDC程序进行仿真,生成训练样本和测试样本,对支持向量机进行了训练和测试.结果表明,应用支持向量分类机对励磁涌流和内部故障进行识别,识别率平均可达99%以上.

In order to improve the ability of transformer differential protection, a new algorithm based on SVM （support vector machine） was proposed for the identifications of inrush current and internal faults of transformers by using the pattern recognition function of SVM for inrush current identification. Based on the characteristics of inrush current and internal fault current and the full consideration of mutual inductor saturation, the second harmonic and dead angle characteristics were extracted, and the implementation details were discussed. EMTDC, a software, was used to generate training samples and testing samples, and SVM was trained and tested by utilizing these samples. The test results show that with the help of the proposed algorithm, the average rate of identification between inrush current and fault current is above 99%.