A New Method for Identification and Classification of Power Quality Disturbance Based on Modified Kaiser Window Fast S-transform and LightGBM

Abstract

Since the large amount of calculation and low time-frequency resolution of S-transform in power quality disturbance detection, and the power quality disturbance data sets often has the problem of unbalanced categories, a new method of power quality disturbance identification and classification based on modified Kaiser window fast S-transform (FMKST) and Light Gradient Boosting Machine (LightGBM) was proposed. Firstly, sampled signal spectrum was obtained through fast Fourier transform. Secondly, iterative loop filtering interval positioning algorithm was used to determine the disturbance frequency interval, and then the window parameters were determined according to the frequency range of the disturbance frequency interval and transform the corresponding interval; Finally, the feature vector was extracted from the FMKST modulus time-frequency matrix of the sampled signal and the modified LightGBM classifier was constructed for classification. Simulation and test results show that the proposed method has higher recognition accuracy and faster diagnosis speed, and is suitable for rapid identification and classification of massive power quality disturbance data.