【摘要】 为了削弱城轨交通系统牵引网电压波动并实现再生制动能量的回收利用,城市轨道交通系统一般会采用超级电容加锂电池组成的混合储能系统来实现稳压和节能的双重目标。混合储能系统的功率分配问题是一个关键问题,因此提出了一种基于强化学习在线序列决策的功率分配策略。为了达到较好的稳压节能效果,对强化学习中的PPO算法进行了改进,将原算法中固定不变的学习率改为随时间变化的学习率,使得强化学习中智能体(Agent)的训练效果更好且算法收敛更快。将这种改进后的D-PPO算法应用在城轨混合储能系统上,能够使混合储能系统更好地去削弱或填补牵引网电压的波峰和波谷并实现节能目的。为验证所提方法的有效性,在MATLAB和Python上进行联合仿真实验,结果表明所提方法削弱了城轨牵引网电压的波动并且实现了节能。更多还原

【Abstract】 In order to reduce the voltage fluctuation of the traction network of urban rail transit system and realize the recycling and utilization of regenerative braking energy,the hybrid energy storage system composed of supercapacitors and lithium batteries is generally adopted in urban rail transit system to achieve the dual purposes of voltage regulation and energy saving.The power allocation problem of hybrid energy storage system is a key problem.In this paper,a power allocation strategy based on reinforcement learning online sequential decision is proposed.In order to achieve better voltage regulation and energy saving effect,this paper improves PPO algorithm in reinforcement learning by changing the fixed learning rate in the original algorithm to the learning rate that changes with time,so that the training effect of Agent in reinforcement learning is better and the algorithm convergence is faster.Applying the improved D-PPO algorithm to HESS,HESS can better weaken or fill the voltage peaks and troughs of traction network and achieve energy saving.In order to verify the effectiveness of the proposed method,a joint simulation experiment is carried out on MATLAB and Python.The experimental results show that the proposed method weakens the voltage fluctuation of urban rail traction network and achieves the purpose of energy saving.更多还原