Study on the classification method of power disturbances based on the combination of S transform and SVM multi-class classifier with binary tree

A new method based on the combination of the S transform and support vector machine (SVM) multi-class classifier for the classification and recognition of power quality disturbance signals in power system is presented in this paper. The proposed method consists of time-frequency analysis, feature extraction, and pattern classification. In the first stage, S transform is applied to extract a set of optimal feature vectors for the classification of power quality disturbance signals. Different power quality disturbances have distinct characteristics such as maximum standard deviation, local maximum, and duration time, etc. By analyzing the complex matrixes generated by S transform of signals, five features were extracted, through which six types of power quality disturbance signals can be classified accurately, therefore the dimension of the feature vectors is decreased greatly. In stage two, the power quality disturbance types are classified through the multi-class classifier based on SVM. The features extracted from S transform are used as the input to a SVM multi-classifier. Combining decision-making method of binary tree with SVM binary classifier, the SVM multi-classifier is formed. It reduces the number of SVM classifiers greatly. The simulation results show that the method presented in this paper has good performance on classification accuracy and computing speed, compared with the one-against-one model.