Temperature Prediction for Subway Station Based on LSTM-SVR Algorithm with Multi Period Characteristics

Abstract By analyzing the timing characteristics of historical temperature data of the subway sensors, aiming at the problem of poor accuracy of a single temperature prediction model, combined with the long-term trend, multi period and irregular change characteristics of the temperature data, this paper proposes a combined prediction model based on the Long Short-Term Memory network (LSTM) and Support Vector Regression (SVR) theory, the prediction mean error of LSTM-SVR is lower than single model, which can predict the temperature of subway station with high accuracy, so as to provide a basis for controlling air-conditioning and ventilation equipment, also saving energy and reducing consumption.