城市轨道交通车辆以太网流量的预测方法
马乐庭1伍红波2
Research on Methods of Urban Rail Transit Vehicle Ethernet Flow Prediction
MA LetingWU Hongbo
-
作者信息:1.中车长春轨道客车股份有限公司, 130062, 长春;
2.北京交通大学电子信息工程学院, 100044,
3.北京
-
Affiliation:CRRC Changchun Railway Vehicles Co., Ltd., 130062, Changchun, China
-
关键词:
-
Key words:
-
DOI:10.16037/j.1007-869x.2022.07.047
-
中图分类号/CLCN:U231+.7
-
栏目/Col:应用技术
摘要:
随着城市轨道交通云平台的部署,城市轨道交通车辆对通信带宽的需求越来越高。近年来,用以太网替代传统的车辆总线已成为共识。然而,面向城市轨道交通应用尤其基于车辆以太网的数据流量的研究还处于探索阶段。考虑车辆以太网数据流量的统计规律特性,采用3种流行的智能预测方法,即RBF(径向基函数)神经网络模型、LSTM(长短期记忆)神经网络模型和LS-SVM(最小二乘支持向量机)模型对车辆以太网流量进行预测,通过仿真试验对3种方法的预测性能进行了有效性验证。结果表明:在小样本(3000个数据量)条件下,采用LS-SVM模型对车辆以太网数据进行预测,相比其他两种方法能获得更高的预测精度。
Abstracts:
With the deployment of urban rail transit cloud platform, urban rail transit vehicle grows higher demand on the communication bandwidth. In recent years, it has become a consensus on replacing conventional vehicle buses with Ethernet. However, the research on flow data for urban applications especially on vehicle Ethernet buses is at exploration stage. Considering the statistical characteristics of vehicle Ethernet data, three popular intelligent methods for traffic flow prediction, the RBF network, LSTM, and LS-SVM, simulations and tests are carried out to validate the performance of the three methods. Results show that the LS-SVM model shows superior performance of prediction accuracy in case of small sample condition (3 000 data volume).
引文 / Ref:
马乐庭,伍红波.城市轨道交通车辆以太网流量的预测方法[J].城市轨道交通研究,2022,25(7):230.
MA Leting,WU Hongbo.Research on Methods of Urban Rail Transit Vehicle Ethernet Flow Prediction[J].Urban mass transit,2022,25(7):230.
MA Leting,WU Hongbo.Research on Methods of Urban Rail Transit Vehicle Ethernet Flow Prediction[J].Urban mass transit,2022,25(7):230.