Power Quality Disturbance Feature Selection and Pattern Recognition Based on Image Enhancement Techniques

In the existing research of power quality disturbance (PQD) identification, the efficiency of signal processing is low and cannot meet the needs of practical application analysis. Furthermore, due to the lack of effective analysis of features, the complexity of classifiers is increased, and the efficiency of classification reduced by the redundant features. In this paper, in order to overcome these shortcomings, a PQD recognition method based on image enhancement techniques and feature importance analysis is proposed. First, PQD signals are converted into gray images, and three image enhancement techniques include gamma correction, edge detection, and peaks and valley detection are used to enhance the disturbance features. Then, the disturbance features are extracted from the binary images, and the original feature set is constructed, the classification ability of each feature is measured by Gini importance. Based on the descending order of the Gini importance, the sequence forward search (SFS) method is used for feature selection to determine the optimal feature subset. Finally, random forest (RF) classifier is constructed by the optimal feature subset to identify the PQD signals. The results of the simulation and contrast experiments show that the new method can determine the optimal classification subset, which recognizes the PQD signals effectively in different noise environments. Furthermore, the new method has higher signal processing efficiency compared with the EMD and ST methods.