The invention provides a train operation scheduling optimization method based on deep reinforcement learning. The method comprises the following steps: establishing a system model; deeply strengthening the Markov decision process; and results and analysis. According to the method, a train dynamics model and a train operation scheduling system Markov model are established according to train operation characteristics aiming at the problem that the uncertainty of emergencies can cause time delay influence on the train and cause serious social influence and huge economic benefit loss. The invention provides a train operation scheduling strategy based on a depth deterministic strategy gradient algorithm by considering the total delay time and the minimum energy consumption of a train. Noise is added in training actions, so that generalization ability and robustness of the method in an uncertain environment are improved. According to the scheduling strategy provided by the invention, the total delay time of train operation is effectively reduced, and the energy consumption is reduced.
本发明提供一种基于深度强化学习的列车运行调度优化方法,包括如下步骤:建立系统模型;马尔可夫决策过程的深度强化;结果和分析。本发明针对突发事件的不确定性会对列车造成时间延误影响,造成严重的社会影响和巨大的经济效益损失,根据列车运行特性,建立了列车动力学模型和列车运行调度系统马尔可夫模型。考虑列车总晚点时间和能源消耗最小,提出了一种基于深度确定性策略梯度算法的列车运行调度策略。通过在训练动作中加入噪声,提高了方法在不确定环境下的泛化能力和鲁棒性。本发明提出的调度策略有效地减小了列车运行总晚点时间,降低了能源消耗。
Train operation scheduling optimization method based on deep reinforcement learning
基于深度强化学习的列车运行调度优化方法
2023-08-01
Patent
Electronic Resource
Chinese
IPC: | B61L Leiten des Eisenbahnverkehrs , GUIDING RAILWAY TRAFFIC / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / G06Q Datenverarbeitungssysteme oder -verfahren, besonders angepasst an verwaltungstechnische, geschäftliche, finanzielle oder betriebswirtschaftliche Zwecke, sowie an geschäftsbezogene Überwachungs- oder Voraussagezwecke , DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES |
Freight train scheduling via decentralised multi-agent deep reinforcement learning
TIBKAT | 2021
|Train temporary parking strategy optimization method based on deep reinforcement learning
European Patent Office | 2023
|