Investigation of Pantograph Wire-Following Contact Force in Electric Locomotives Based on Multi-Agent Proximal Policy Optimization

doi:10.16628/j.cnki.2095-8188.2026.04.004

Abstract

Abstract:

The pantograph-catenary system, as the key interface for traction power transmission in electric locomotives, determines current-collection quality and operational reliability through its coordination performance. At high speeds, the sliding of pantograph along the contact wire excites coupled vibrations in the pantograph-catenary system, undermining contact stability and reducing current-collection efficiency. Regulating the wire-following contact force(WFCF) is an effective remedy. A multi-agent proximal policy optimization(MAPPO) method augmented with a long short-term memory(LSTM) network is proposed to achieve dynamic WFCF control. Within a centralized-training, decentralized-execution(CTDE) framework, multiple agents are trained to learn speed-specific control policies, enabling precise force regulation across operating conditions. Experimental results show that the proposed approach substantially suppresses WFCF fluctuations and enhances contact wire-following performance(CWFP).

Key words: wire-following contact force, MAPPO, pantograph, contact wire-following performance

CLC Number:

TM501

MA Zhenhao, GUO Fengyi, HAN Congxin. Investigation of Pantograph Wire-Following Contact Force in Electric Locomotives Based on Multi-Agent Proximal Policy Optimization[J]. LOW VOLTAGE APPARATUS, 2026, 0(4): 32-40.

Figures/Tables 10

References 17

[1]	郭凤仪, 陈明阳, 陈忠华, 等. 弓网滑动电接触摩擦力特性与建模研究[J]. 电工技术学报, 2018, 33(13):2982-2990.
[2]	于治国, 刘茂盛, 张婷婷, 等. 动车组受电弓升降弓拉弧试验研究[J]. 电器与能效管理技术, 2017(17):15-20.
[3]	张东, 鲁超, 母婷佑, 等. 弓网电弧作用下碳-铜材料转移特性分析[J]. 电器与能效管理技术, 2020(6):15-19.
[4]	陈忠华, 李兵红, 陈明阳, 等. 弓网滑动电接触电磁热力耦合效应研究进展[J]. 电工技术学报, 2023, 38(10):2777-2793.
[5]	GUO F, WANG X, WANG Z, et al. Method and System for Assessing Sliding Electrical Contact Performance of Electric Locomotive Pantograph-catenary System:US 12493086 B2[P].2025-12-09.
[6]	MORALES E F, MURRIETA-CID R, BECERRA I, et al. A survey on deep learning and deep reinforcement learning in robotics with a tutorial on deep reinforcement learning[J]. Intelligent Service Robotics, 2021, 14(5):773-805. doi: 10.1007/s11370-021-00398-z
[7]	ABDIKARIM I. Applications of multi-agent deep reinforcement learning:Models and algorithms[J]. Applied Sciences, 2021, 11(22):10870-10870. doi: 10.3390/app112210870
[8]	WANG H, LIU Z, WU Y. A hierarchical control strategy for active pantograph in high-speed railway[C]// 2023 International Annual Conference on Complex Systems and Intelligent Science.Shenzhen, China,2023:826-833.
[9]	吴延波, 韩志伟, 王惠, 等. 基于双延迟深度确定性策略梯度的受电弓主动控制[J]. 电工技术学报, 2024, 39(14):4547-4556.
[10]	WANG H, HAN Z, LIU W, et al. A reinforcement learning-based pantograph control strategy for improving current collection quality in high-speed railways[J]. IEEE Transactions on Neural Networks and Learning Systems, 2024, 35(5):5915-5928. doi: 10.1109/TNNLS.2022.3219814
[11]	WANG H, LIU Z, HAN Z. HO2RL:A novel hybrid offline-and-online reinforcement learning method for active pantograph control[J]. IEEE Transactions on Industrial Electronics, 2025, 72(6):6286-6296. doi: 10.1109/TIE.2024.3477002
[12]	ZHOU N, ZOU H, ZOU D, et al. Investigation on the applicability of pantograph and catenary model urban railway system[J]. Journal of Southwest Jiaotong University, 2017, 52(2):408-415.
[13]	宋洋, 刘志刚, 鲁小兵, 等. 计及接触网空气动力的高速弓网动态受流特性研究[J]. 铁道学报, 2016, 38(3):48-58.
[14]	LIU Z, JÖNSSON P A, STICHEL S, et al. Implications of the operation of multiple pantographs on the soft catenary systems in Sweden[J]. Proceedings of the Institution of Mechanical Engineers Part F Journal of Rail & Rapid Transit, 2016, 230:971-983.
[15]	GUAN C, CHEN F, YUAN L, et al. Efficient communication via self-supervised information aggregation for online and offline multiagent reinforcement learning[J]. IEEE transactions on neural networks and learning systems, 2025, 5(36):9044-9056.
[16]	LYU G, LI M. Multi-agent cooperative control in neural MMO environment based on MAPPO algorithm[C]// 2023 IEEE 5th International Conference on Artificial Intelligence Circuits and Systems(AICAS). Hangzhou, China, 2023:1-4.
[17]	WANG Z, WANG H. GAN-based PPO-LSTM high-speed railway line optimization[C]// 2023 3rd International Conference on Digital Society and Intelligent Systems(DSInS). Chengdu, China, 2023:125-130.

算法	参数	数值
LSTM参数	隐藏层大小	256
LSTM参数	层数	2
PPO参数	学习率	0.000 4
	折扣因数γ	0.98
	GAE_lambda	0.95
	策略截断范围	0.3
	熵正则系数	0.02
	值函数损失系数	0.18
	批量更新样本数	512
	每回合步数	4 096
	总训练回合	10 000

参数	DSA200
弓头质量/kg	6.52
上部框架质量/kg	4.3
下部框架质量/kg	10.72
弓头刚度/(N·m^-1)	6 350
上部框架刚度/(N·m^-1)	10 900
下部框架刚度/(N·m^-1)	80
弓头阻尼/(N·s·m^-1)	0
上部框架阻尼/(N·s·m^-1)	0
下部框架阻尼/(N·s·m^-1)	70
设计时速/(km·h^-1)	200
静态接触压力/N	70±10

速度/(km·h^-1)	平均绝对误差		标准差
	未控制	MAPPO	未控制	MAPPO
	30	1.86	1.54	2.21	1.77
50	2.03	1.65	2.38	1.86
70	2.22	1.72	2.57	2.12