首页|基于ARIMA-TCN混合模型的高速铁路时间同步方法

基于ARIMA-TCN混合模型的高速铁路时间同步方法

扫码查看
原文链接
  • 万方数据
  • 维普
列控系统作为高速铁路的核心系统,保持其系统的时间同步对于行车安全至关重要.针对现有时间同步方法易受时变上下行传输时延、随机时钟跳变等影响,导致主从时钟偏移估计不准确的问题,提出一种基于差分自回归移动平均-时域卷积神经网络(ARIMA-TCN)混合模型的高速铁路时间同步方法.首先,根据上下行链路传输速率的不对称比,建立高速铁路时钟的数学理论和实际观测模型.然后,使用拉依达准则识别处理跳变异常值,完成实际时间序列的预处理.再次,使用ARIMA模型平滑时间序列中不确定时延带来的噪声抖动,获得平稳的时间序列.最后,通过提出的注意力增强TCN模型进行预测补偿,完成时钟偏移的补偿校正.通过实验仿真,得到基站区间内位置、基站间距以及车速对高速铁路时间同步的影响性分析.实验结果表明:与对比方法相比,所提方法补偿后的均方根误差较最小二乘法减少了 75％、较最大似然估计方法误差减少了 44.4％,较BP神经网络方法误差减少了 16.7％,验证所提方法具有更低的同步误差和更高的同步精度.
Time Synchronization Method for High-speed Railway Based on ARIMA-TCN Hybrid Model
Maintaining the time synchronization of the train control system,as the key system of high-speed railways,is important for the safety of train operation.Aiming at the problem of inaccurate estimation of master-slave clock offset caused by time-varying uplink and downlink transmission delays and random clock jumps in existing time synchronization methods,a time synchronization method was proposed for high-speed railway based on differential autoregressive integrat-ed moving average model and temporal convolutional network(ARIMA-TCN)hybrid model.Firstly,the mathematical theory and practical observation model for high-speed railway clocks were established according to the asymmetry ratio of uplink and downlink transmission rates.Then,the jump outliers were identified and processed using the PauTa criterion to complete the preprocessing of the actual time series.Next,the ARIMA model was used to smooth out the noise jitter caused by the uncertain delays in the time series to obtain a stable time series.Finally,the attention enhanced TCN model was used to predict and compensate the clock offset.Through the experimental simulation,the influence of the lo-cation of the base station,the distance between the base stations and the train speed on the time synchronization of the high-speed railways was analyzed.The experimental results show that the compensated root mean square error of the pro-posed method is 75％less than that of the ordinary least squares method,44.4％less than that of the maximum likeli-hood estimation method,and 16.7％less than that of the back propagation neural network method,verifying the lower synchronization error and higher synchronization accuracy of the proposed method.

time synchronizationprecision time protocoldifferential autoregressive integrated moving average modelat-tention enhancement temporal convolutional networktime compensation

陈永、詹芝贤、张薇

展开 >

兰州交通大学电子与信息工程学院,甘肃兰州 730070

兰州交通大学交通运输学院,甘肃兰州 730070

时间同步 精确时钟协议 差分自回归移动平均模型 注意力增强时域卷积网络 时间补偿

国家自然科学基金兰州交通大学基础研究拔尖人才项目兰州交通大学重点研发项目

619630232022JC36ZDYF2304

2024

10.3969/j.issn.1001-8360.2024.06.010

铁道学报
中国铁道学会

铁道学报

CSTPCD北大核心
影响因子：0.9
ISSN：1001-8360
年,卷(期)：2024.46(6)
  • 27