Prediction and classification of voltage sag trend based on support vector machine with parameter optimization

This paper uses support vector machine with parameter optimization to predict the variation trend of voltage sag waveform, intensively extracts its features, and proposes a source identification method for voltage sag. Firstly, the time-domain and time-frequency-domain characteristics of voltage sag waveform is extracted by S-transform, and the initial indicators is formed. The correlation between different initial indicators is analyzed by Pearson correlation coefficient method. Random forest optimization algorithm is used to select the features of the initial indicators, and the subjective and objective weight coefficients is introduced to optimize the calculation results of the importance of the initial indicators, so as to solve the problem of the decline of identification accuracy caused by the wrong selection of features. The identification model of voltage sag source is improved by using support vector machine, and the improved particle swarm optimization algorithm is introduced to determine the key parameters and of the identification model, so as to improve the accuracy of the identification model. The simulation results of an example show that the proposed method may effectively reduce the initial indicator system, and provide an effective idea for further improving the accuracy of source identification of voltage sag.