[1]钱力,梁艺轩,陈历泉,等.基于数据驱动的铁路编组站驼峰推峰速度建模研究[J].高速铁路技术,2025,16(06):9-15.[doi:10.12098/j.issn.1674-8247.2025.06.002]
　QIAN Li,LIANG Yixuan,CHEN Liquan,et al.On Data-driven Modeling of Humping Speed at Railway Marshalling Yards[J].HIGH SPEED RAILWAY TECHNOLOGY,2025,16(06):9-15.[doi:10.12098/j.issn.1674-8247.2025.06.002]
点击复制

基于数据驱动的铁路编组站驼峰推峰速度建模研究()

《高速铁路技术》[ISSN:1674-8247/CN:51-1730/U]

卷:
16卷
期数:
2025年06期
页码:
9-15
栏目:
勘察设计
出版日期:
2025-12-30

文章信息/Info

Title:
On Data-driven Modeling of Humping Speed at Railway Marshalling Yards
文章编号:
1674-8247(2025)06-0009-07
作者:
钱力1; 梁艺轩2; 陈历泉1; 施莉娟3
(1. 中国铁路广州局集团有限公司, 广州 510088; 2. 同济大学, 上海 201804; 3. 同济大学上海市轨道交通结构耐久与系统安全重点实验室, 上海 201804)
Author(s):
QIAN Li1;  LIANG Yixuan2;  CHEN Liquan1;  SHI Lijuan3
(1. China Railway Guangzhou Group Co., Ltd., Guangzhou 510088, China; 2. Tongji University, Shanghai 201804, China; 3. Shanghai Key Laboratory of Rail Infrastructure Durability and System Safety,Tongji University, Shanghai 201804, China)
关键词:
编组站;  驼峰;  推峰速度;  多元线性回归;  随机森林;  XGBoost模型
Keywords:
marshalling yard;  hump;  humping speed;  multiple linear regression;  random forest;  XGBoost model
分类号:
U292.16
DOI:
10.12098/j.issn.1674-8247.2025.06.002
文献标志码:
A
摘要:
为实现铁路编组站驼峰解体提钩机器人与推峰车列的协同共速,需根据车列编组和钩计划,为每一钩溜放作业提前规划推峰策略。本文围绕提钩机器人作业中的关键参数——推峰速度的建模与预测,提出一种融合数据驱动与残差修正机制的建模方案。首先,基于江村站实测数据,开展推峰速度与车组参数(辆数、重量、测长)之间的皮尔逊相关性分析与共线性诊断,识别并筛选影响推峰速度的关键因素; 在此基础上,构建多元线性回归模型对推峰速度进行初步预测; 随后引入随机森林回归算法对原模型残差进行非线性拟合,实现误差修正与动态补偿,从而提升预测精度与泛化能力; 进一步引入XGBoost模型进行对比分析,评估不同算法的表现差异。结果表明,“线性回归+随机森林修正”方法在测试集上显著优于原始线性模型和XGBoost模型,其决定系数(R2)提升至0.862 9,MAPE 降低至6.15%,NRMSE降低至0.096 2。研究结果验证了该方法的有效性与工程适用性,在机器人自动提钩场景下为待解车列预先制定推峰策略提供推峰速度预案支持。
Abstract:
To achieve coordinated speed synchronization between the uncoupling robot and the humping locomotive during break-up operation at hump yards, it is necessary to plan humping strategies in advance based on the train composition and uncoupling schedule. This paper focused on the modeling and prediction of humping speed—a key parameter in operation of robot uncoupling, and proposed a refined modeling framework integrating data-driven approaches with a residual correction mechanism. First, using measured data from Jiangcun Yard, Pearson correlation analysis and multicollinearity diagnostics were conducted to identify and screen key factors influencing humping speed, including the number of cars, weight, and measured length. A multiple linear regression model was constructed for initial speed prediction. Based on this, a random forest regression algorithm was introduced to nonlinearly fit the residuals of the baseline model, enabling error correction and dynamic compensation to enhance prediction accuracy and generalization capability. Furthermore, the XGBoost model was applied for comparative analysis to evaluate the performance differences among algorithms. The results show that the “linear regression + random forest correction” method significantly outperforms both the baseline linear model and the XGBoost model, with the coefficient of determination(R2)increasing to 0.862 9, MAPE decreasing to 6.15%, and NRMSE reduced to 0.096 2. The findings validate the effectiveness and engineering applicability of the proposed approach, providing humping speed pre-planning support for formulating humping strategies for trains awaiting break-up in the context of robotic automatic uncoupling.

参考文献/References:

[1] 周琪. 驼峰解体自动提钩装置设计与关键技术研究[J]. 河南科技, 2021, 40(1): 113-116.
ZHOU Qi. Research on Design and Key Technology Analysis of Automatic Hook Lifting Device for Hump Disintegration[J]. Henan Science and Technology, 2021, 40(1): 113-116.
[2] 张红亮, 杨俊杰, 邱星, 等. 基于自动提钩的驼峰溜放部分纵断面设计优化研究[J]. 铁道学报, 2024, 46(7): 15-22.
ZHANG Hongliang, YANG Junjie, QIU Xing, et al. Research on Optimization Method for Profile Design of Hump Distributing Zone Based on Automatic Uncoupling Technology[J]. Journal of the China Railway Society, 2024, 46(7): 15-22.
[3] 张所斌. 自动摘钩机器人模型的设计与制作[D]. 成都: 西南交通大学, 2017.
ZHANG Suobin. Design and Manufacture of Robot Model for Automatic Hooking-off [D]. Chengdu: Southwest Jiaotong University, 2017.
[4] 赵赛. 驼峰解体自动摘钩系统的研究[D]. 石家庄: 石家庄铁道大学, 2013.
ZHAO Sai. Research on Automatic Unhooking System of Hump Disintegration[D]. Shijiazhuang: Shijiazhuang Tiedao University, 2013.
[5] 邱星. 基于自动提钩的驼峰溜放部分纵断面设计与间隔调速控制研究[D]. 北京: 北京交通大学, 2022.
QIU Xing. Research on Longitudinal Section Design and Interval Speed Control of Hump Sliding Part Based on Automatic Hook Lifting[D]. Beijing: Beijing Jiaotong University, 2022.
[6] 李继东. 驼峰推峰自动摘钩系统关键技术研究[D]. 北京: 中国铁道科学研究院, 2022.
LI Jidong. Research on Key Technologies of Automatic Hook-off System for Hump Peak Pushing[D]. Beijing: China Academy of Railway Sciences, 2022.
[7] BOYSEN N, FLIEDNER M, JAEHN F, et al. Shunting Yard Operations: Theoretical Aspects and Applications[J]. European Journal of Operational Research, 2012, 220(1): 1-14.
[8] 安迪, 马玉坤, 宋海锋, 等. 编组站提钩自动化智能识别设计[J]. 铁道运输与经济, 2020, 42(3): 54-60.
AN Di, MA Yukun, SONG Haifeng, et al. An Intelligent Recognition Design of Automatic Wagon Uncoupling in Marshalling Yard [J]. Railway Transport and Economy, 2020, 42(3): 54-60.
[9] 臧丽超. 铁路货车摘钩机器人设计与仿真研究[D]. 石家庄: 石家庄铁道大学, 2013.
ZANG Lichao. Design and Simulation of Railway Freight Car Unhook Robot[D]. Shijiazhuang: Shijiazhuang Tiedao University, 2013.
[10] 唐汇东. 一种铁路驼峰场推峰速度自动计算方法:中国, CN112918515A[P]. 2021-06-08.
TANG Huidong. An Automatic Calculation Method for Humping Speed in Railway Hump Yards: China, CN112918515A[P]. 2021-06-08.
[11] 林通源. 驼峰自动化系统推峰速度自动控制技术的探讨[C]//中国铁道学会2004年度学术活动优秀论文评奖论文集. 2005: 68-75.
LIN Tongyuan. On Automatic Control Technology of Humping Speed in Hump Automation Systems [C] // Collection of Outstanding Essays from the 2004 Academic Conference of the China Railway Society. 2005: 68-75.
[12] 云辉, 顾炎. 驼峰推峰速度控制系统Fuzzy模型的研究[J]. 铁道学报, 1989, 11(4): 49-59.
YUN Hui, GU Yan. A Study of Hump-yard Locomotive Speed Control System with Fuzzy Model[J]. Journal of the China Railway Society, 1989, 11(4): 49-59.
[13] 何霄. 模糊神经网络在驼峰机车推峰速度控制中的应用[J]. 机械研究与应用, 2015, 28(2): 83-85, 88.
HE Xiao. Application of Fuzzy Neural Network in Controlling the Humping Speed of Hump Engine[J]. Mechanical Research & Application, 2015, 28(2): 83-85, 88.
[14] 张健. 基于自动提钩的驼峰推峰速度控制和调车机车操纵优化研究[D]. 北京: 北京交通大学, 2024.
ZHANG Jian. Research on Hump Pushing Speed Control and Shunting Locomotive Handling Optimization Based on Automatic Hook Lifting[D]. Beijing: Beijing Jiaotong University, 2024.
[15] 郭荣昌. 基于模糊神经网络的驼峰推峰速度自动控制系统的研究[D]. 兰州: 兰州交通大学, 2011.
GUO Rongchang. Research on Automatic Control System of Hump Pushing Speed Based on Fuzzy Neural Network[D]. Lanzhou: Lanzhou Jiaotong University, 2011.
[16] TURANOV KH T, DJALILOV KH KH, SUYUNBAYEV SH. M. Calculation of the Speed of the Carriage in Different Sections of the Hump [J] International Journal of Advanced Research in Science, Engineering and Technology, 2020,7(10), 15385-15395.
[17] LIVINGSTON W L. Locomotive Speed Control System: U.S. CAD750512[P]. 1967-01-10.
[18] ZHOU Kaichen, SONG Shiji, XUE Anke, et al. Smart Train Operation Algorithms Based on Expert Knowledge and Reinforcement Learning[J]. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2020, 52(2): 716-727.
[19] 陶德高. 驼峰道岔分路概率和计算推峰速度的数学模型[J]. 铁道运输与经济, 1985, 7(9): 10-12.
TAO Degao. Mathematical Model of Shunting Probability and Peak Pushing Speed of Hump Turnout[J]. Railway Transport and Economy, 1985, 7(9): 10-12.
[20] 张觉印. 脱钩点位置的计算方法及驼峰净平台长度的合理取值范围的探讨[J]. 中国铁道科学, 1991, 12(2): 88-99.
ZHANG Jueyin. Formulas for Calculating the Position of Disconnecting Coupling Gears of Cars and Suitable Range of Lengths of the Flat Part of Hump Crest [J]. China Railway Science, 1991, 12(2): 88-99.

备注/Memo

备注/Memo:
收稿日期:2025-07-10
作者简介:钱力(1972-), 男,教授级高级工程师。
基金项目:中国铁路广州局集团有限公司科技研究开发计划(2024K034-N)
更新日期/Last Update: 2025-12-30