摘要
为优化混合动力丘陵山地移动作业底盘在梨耕工况下的燃油经济性与电池SOC维持性能,并解决强化学习能源管理策略中超参数选择的不确定性问题,提出了一种基于贝叶斯超参数优化的双延迟深度确定性策略梯度算法(BO-TD3)控制策略。首先,以双延迟深度确定性策略梯度算法(TD3)为核心,通过贝叶斯优化(BO)算法对学习率、折扣因子等关键超参数进行寻优,确定最优参数组合;然后,将优化后的参数应用于TD3算法,对梨耕工况数据进行训练,通过发动机与电机协同控制,实现燃油消耗最小化与电池利用优化。仿真结果表明,所提出的BO-TD3策略在燃油经济性和电池性能方面表现优异,与传统TD3和深度确定性策略梯度算法(DDPG)控制策略相比,燃油经济性分别提高了2.08%和10.37%,提升了车辆的综合能源管理效率。实时在线控制策略验证中,相较于TD3、DDPG和基于规则的耗电-维持充电控制策略(CDCS),BO-TD3的燃油经济性分别提高了4.75%、7.13%、28.71%,电池SOC维持性能良好,验证了策略的优越性与适用性。研究可为混合动力移动作业底盘和其他农业机械的能量管理提供新的解决思路。
To optimize the fuel economy and battery SOC maintenance performance of hybrid power mobile operating platforms in hilly terrain during plowing conditions,and to address the uncertainty in hyperparameter selection in reinforcement learning-based energy management strategies,a control strategy based on Bayesian optimization of the Twin Delayed Deep Deterministic Policy Gradient(BO-TD3)algorithm was proposed.Firstly,the Twin Delayed Deep Deterministic Policy Gradient(TD3)algorithm was used as the core,and Bayesian optimization(BO)was employed to search for the optimal hyperparameters,such as the learning rate and discount factor.The optimized parameters were then applied to the TD3 algorithm,where the plowing condition data was trained to achieve fuel consumption minimization and battery utilization optimization through the cooperative control of the engine and the motor.Simulation results demonstrated that the proposed BO-TD3 strategy significantly outperformed the conventional TD3 and Deep Deterministic Policy Gradient(DDPG)algorithms in terms of fuel economy,with improvements of 2.08%and 10.37%,respectively,thereby enhancing the comprehensive energy management efficiency of the vehicle.Further validation of the real-time online control strategy under other operating conditions showed that,compared with TD3,DDPG,and the rule-based Charge-Depleting and Charge-Sustaining(CDCS)strategy,the BO-TD3 strategy achieved 4.75%,7.13%,and 28.71%improvements in fuel economy,respectively,while maintaining excellent battery SOC performance.This study can provide a novel approach to energy management strategies for hybrid power mobile operating platforms and other agricultural machinery,offering significant application value and broad promotion prospects.
作者
师国靖
许恩永
林长波
展新
李骏
蒙艳玫
Shi Guojing;Xu Enyong;Lin Changbo;Zhan Xin;Li Jun;Meng Yanmei(College of Mechanical Engineering,Guangxi University,Nanning 530004,China;Dongfeng Liuzhou Motor Co.,Ltd.,Liuzhou 545005,China)
出处
《农机化研究》
北大核心
2026年第7期243-252,共10页
Journal of Agricultural Mechanization Research
基金
国家自然科学基金项目(52365001)
广西科技重大专项(桂科AA23062040)。
关键词
混合动力
移动作业底盘
能源管理
贝叶斯优化
TD3
燃油经济性
电池SOC维持
丘陵山地
hybrid power
mobile operation chassis
energy management
Bayesian optimization
TD3 algorithm
fuel economy
battery SOC maintenance
hilly and mountainous areas
