基于稀疏表示算法的GIS局部放电识别

为提高GIS绝缘缺陷的识别正确率,针对GIS出现的绝缘缺陷以及产生的局部放电特点,设计了4种典型绝缘缺陷物理模型,对获得的局部放电灰度图谱,用稀疏表示分类算法进行缺陷类型识别。该算法首先用最小--范数方法计算稀疏表示系数,运用压缩感知将低维观测信号恢复到高维原始信号,通过计算各类缺陷局部放电灰度图的最小残差来进行图像匹配,避开了一般模式识别分类算法中较为复杂的特征提取。测试结果表明该方法对GIS各类模拟缺陷的正确识别率较高。

Partial Discharge Pattern Recognition of GIS Based on Sparse Representation algorithm

Four kinds of typical artificial defect models of GIS and partial discharge (PD) detection system were established, in order to improve the recognition rate. According to the characteristics of PD and gradation spectrum obtained, have used sparse representation algorithm to distinguish the type of defects. First, the algorithm based on minimum-norm method to calculate the sparse representation coefficient, the low-dimensional signals observed were restored to the high dimensional original signal by compressive sensing. Then, for image matching through calculating the minimum residual of various type defects of PD, avoid complex feature extraction of general pattern recognition classification algorithm. Results demonstrate the method is effective.