基于长短期记忆网络的城市轨道交通系统通信延时补偿方法

A Method of Communication Delay Compensation for Urban Transit System Based on Long-term and Short-term Memory

城市轨道交通系统中会出现车地无线通信延时和丢包现象,其影响列车能量管理系统的控制实现。为此,文章利用4G无线模块采集不同司机在同一区间驾驶的18组列车功率数据并进行列车功率典型特征分析,然后提出一种基于长短期记忆网络(LSTM)的列车功率实时滚动预测方法。该方法根据列车实时功率(短期数据)及其邻近时刻功率(长期数据)对功率进行预测,有效提升了预测的准确度。通过4G通信试验,文章计算了通信延时误差,并与采用所提预测方法的计算结果进行对比。结果显示,采用该预测算法后,通信延时误差降低了21.8%,通信丢包误差降低了25.8%26.9%,可提供更准确的实时列车数据参考,使得实时改善能量流变得更为可行。

Communication delay and packet loss in train to ground wireless communication of urban rail transit system impedes the application of energy management system.18 groups of train power data,in which drivers driving in the same section,are collected via 4G wireless communication.A real-time rolling load forecasting method based on long-term and short-term memory network(LSTM)is proposed after analyzing the train power data,which improves the prediction accuracy by forecasting the power based on real-time power(short-term data)and adjacent power(long-term data).After measuring the communication parameters in a 4G communication test,the communication delay induced error is calculated and compared with the prediction method.The result shows that the prediction algorithm can reduce communication delay induced error by 21.8%and packet loss induced error by 25.8%~26.9%,which can provide more accurate real-time train power information and make real-time improvement for energy flow more feasible.