自发电式轨道车辆轴箱温度预测方法研究

Research on Prediction Method of Axle Box Temperature of Self-power Railway Vehicle

针对上海地铁所采用的自发电式轴温监测系统中的轴箱温度预测问题,通过长短时记忆网络(LSTM)、BP神经网络以及支持向量回归机(SVR)三种预测方法进行轴箱温度预测并将结果进行对比分析。最后,利用实测数据,对各方法的预测结果进行对比,结果表明,SVR的预测结果均优于其他预测方法,且预测精度可达到98.59%。

Aiming at the axle box temperature prediction problem in the self-powered axle temperature monitoring system used by the Shanghai metro,the axle box temperature prediction is carried out through three prediction methods,which are long short-term memory(LSTM),BP neural network and support vector regression(SVR),and the results are compared and analyzed.Finally,the measured data is used to compare the prediction results of each method,the results show that the prediction results of SVR are better than other prediction methods,and the prediction accuracy can reach 98.59%.