摘要
混合动力汽车作为新能源汽车在现阶段过渡状态的发展主力,其电池效率和经济性至关重要。为了实现实时优化并且改善混动汽车的部分参数,提出了基于深度强化学习的能量管理策略,并在仿真过程中引入神经网络对工况进行预测。以混合动力汽车模型以及参数作为支撑,搭建了混合动力汽车仿真环境,与建立的能量管理模型进行迭代交互。应用深度强化学习中的不同算法对电池功率进行控制和改善,得到了不同算法下的优化结果,验证了所建立模型的有效性和可靠性,对电池的优化以及功率调控具有一定的实际意义。
Hybrid electric vehicles,as new energy vehicles,are the main force in the development of the transitional state at this stage,and their battery efficiency and economy are crucial.In order to realize real-time optimization and improve some parameters of hybrid vehicles,an energy management strategy based on deep reinforcement learning was proposed,and a neural network was introduced in the simulation process to predict operating conditions.Based on the hybrid vehicle model and parameters,a hybrid vehicle simulation environment was built,and then iteratively interacted with the established energy management model.Different algorithms in deep reinforcement learning are used to control and improve battery power,and the optimization results under different algorithms are obtained,which verifies the validity and reliability of the established model,which has certain practical significance for battery optimization and power regulation.
作者
苏明亮
姚方
Su Mingliang;Yao Fang(School of Electric Power Civil Engineering&Architecture,Shanxi University,Taiyuan Shanxi 030013,China)
出处
《电气自动化》
2023年第4期115-118,共4页
Electrical Automation
基金
国家自然科学基金项目(U1509218)
山西省电力公司科技项目(SGTYHT/18-JS-202)。
关键词
混合动力汽车
深度强化学习
能量管理策略
电池荷电状态
工况预测
hybrid electric vehicle
deep reinforcement learning
energy management strategy
state of charge
driving condition prediction