The Method For Fault Type Identification Of Transmission Line Considering Transient Singular Information And Imbalance Of Recorded Fault Dataset

In the field of relay protection in power system,it is necessary to identify transmission line fault type fast and accurately,in order to ensure the normal operation of relay protection equipment,guarantee the accuracy of fault location and accident analysis for transmission system,and improve the stability and reliability of power system.With the rapid development of smart grid and the gradual sophistication of information acquisition and processing technology,compared with traditional fault type identification method for transmission line,the new method for fault type identification with artificial intelligence pattern recognition technique has the advantages of high accuracy,strong robustness and good generalization.In this paper,the fault type identification methods for transmission line based on artificial intelligence pattern recognition algorithm were mainly studied,including the extraction of transient singular information for fault signal,the construction of multi-class classification model and the optimization of unbalanced recorded fault dataset.Firstly,the theory of wavelet transform and singular value decomposition were introduced,and the construction method of stationary wavelet singular value was proposed.The transient fault signals of transmission line were decomposed by stationary wavelet transform,based on wavelet multi-scale analysis theory,to obtain the stationary wavelet coefficients of different scales.The information feature matrixes were constructed by using stationary wavelet coefficients of different scales,and then they were decomposed through singular value decomposition algorithm to obtain the singular value of constructed matrixes,as a measure for transient information and information complexity of transmission line fault signal.Secondly,three pattern recognition methods for transmission line fault type were researched,including k nearest neighbor classification,support vector machine and artificial neural network.The influence of imbalanced datasets on the pattern recognition methods was analyzed,and a multi-class combinatorial classifier of support vector machine was constructed.In view of unbalanced dataset during fault type identification of transmission line,an improved heuristic synthetic minority over-sampling technique(SMOTE)was proposed to optimize the unbalanced dataset.It can avoid the over-fitting problem and data marginalization effect comparing with traditional SMOTE algorithm,and improve the recognition ability of minority samples during classification.Meanwhile,the result evaluation indexes for two-class classification of imbalanced dataset were extended to make it suitable for multi-class classification.Finally,a new fault type identification method for transmission line based on stationary wavelet singular value and improved heuristic SMOTE algorithm was proposed in this paper,and the specific recognition scheme was given.Stationary wavelet singular values of fault component signals in transmission line were extracted as the feature parameters for classification.Improved heuristic SMOTE algorithm was used to adjust the imbalance of recorded fault dataset of transmission line,and its optimization results was contrasted with traditional SMOTE algorithm.In order to reduce the number of feature parameter,the principal component analysis algorithm was used to reduce the dimension of extracted feature vector.Result of simulation in PSCAD/EMTDC and MATLAB indicates that the proposed method is feasible and effective,and the extracted feature parameters can effectively excavate the characteristic information of different fault types,and the improved heuristic SMOTE algorithm can effectively improve the recognition accuracy of minority samples and overall classification accuracy in classification method.It has strong generalization,and it is suitable for various kinds of artificial intelligence pattern recognition method.