考虑小波奇异信息与不平衡数据集的输电线路故障识别方法

Method for Fault Type Identification of Transmission Line Considering Wavelet Singular Information and Unbalanced Dataset

鉴于输电线路故障识别中数据集的非均衡性问题,提出一种基于小波奇异信息和改进合成少数类过采样(synthetic minority over-sampling technique,SMOTE)算法的输电线路故障识别方法。首先,通过PSCAD/EMTDC仿真构造输电线路故障不平衡数据集,结合平稳小波变换(stationary wavelet transform,SWT)与奇异值分解(singular value decomposition,SVD)技术提取相电流及零序电流的故障分量的小波奇异值作为特征参数,然后采用改进SMOTE算法在少数类的样本中心邻域进行插值再抽样处理,调整数据集的不平衡度,利用优化后的数据集训练支持向量机(support vector machine,SVM)组合分类器,对不同故障工况下的10种输电线路故障类型进行分类识别。仿真结果表明,该文的方法能有效地提高分类算法在样本数据不平衡的情况下对少数类的识别能力和整体的识别准确率,具有较好的泛化性和较强的鲁棒性,并且对多种分类算法同样适用。

In view of unbalanced dataset during fault type identification of transmission line, a fault type identification technique based on wavelet singular information and improved synthetic minority over-sampling technique （SMOTE） was proposed. Firstly, unbalanced fault dataset of transmission line was established in PSCAD/EMTDC. Wavelet singular values of phase current fault component and zero-sequence current fault component were extracted as feature parameters based on stationary wavelet transform （SWT） and singular value decomposition （SVD）. Furthermore, improved SMOTE algorithm was used to adjust the imbalance of dataset through random interpolation and resample in the vicinity of minority class center. Then, the support vector machine classifiers were trained by optimized dataset, to classify and identify 10 fault types of transmission line under various fault condition. Result of simulation indicates that the proposed method can effectively improve the identification ability of minority class and identification accuracy of classification algorithm under the condition of unbalanced sample dataset. It has better generalization performance and stronger robustness, and it is suitable for various classification algorithms.