Ultra-short-term Railway traction load prediction based on DWT-TCN-PSO_SVR combined model

Abstract The ultra-short-term prediction of traction load is a key aspect of electric power control in electrified railroads. In order to provide more accurate traction load prediction data for the control process, this paper designs a combined prediction method combining Discrete Wavelet Transform(DWT), Temporal Convolutional Network (TCN) and Support Vector Regression based on Particle Swarm Optimization(PSO_SVR) for the characteristics of traction load with strong random fluctuation, large jump amplitude and frequent no-load. This method will use DWT model to decompose the traction load with strong random fluctuation and difficult to predict into sub-series with single fluctuation frequency and easy to predict; then, according to the frequency difference of different series, TCN model is selected to predict medium and low frequency series. The SVR model is selected to predict the high frequency sequences. Finally, the prediction results of each sub-sequence are summed to obtain the final prediction. The combined prediction model designed in this paper is used to predict the traction load of traction power supply stations. The experimental results show that the prediction method proposed in this paper has higher prediction accuracy than the existing prediction methods in the current prediction field.