基于MPC-PMP的油电混合动力列车能量管理优化控制

Energy management optimization controls of electric-diesel hybrid trains based on MPC-PMP

合适的能量管理策略能够提高列车混合动力系统的效率、减少燃油消耗成本,针对庞特里亚金极小值原理(PMP)策略需要预知全局工况的问题,以油耗最小为目标,提出了一种基于模型预测控制(MPC)的能量管理策略,可以实现在线应用。首先,基于长短时记忆(LSTM)神经网络建立了速度预测模型;然后,基于PMP算法在速度预测的时域内进行滚动优化,每一时刻求解最优的功率控制序列;最后,基于实际运行工况,对比了MPC-PMP策略、离线PMP策略和传统阈值法的能量管理结果。仿真结果表明,MPC-PMP策略虽然比离线PMP策略的燃油经济性稍微降低,但相比传统的阈值法策略,燃油经济性提高了14.34%,而且满足实时性要求。

A suitable energy management strategy can improve the efficiency of train hybrid systems and reduce fuel consumption costs.In response to the issue of the Pontryagin's Minimum Principle(PMP)strategy requiring prediction of whole operating conditions,we proposed an energy management strategy based on Model Predictive Control(MPC)with the goal of minimizing fuel consumption,which can be applied online.Firstly,a speed prediction model was established based on Long Short Term Memory(LSTM)neural network;Then,based on the PMP algorithm,rolling optimization was carried out in the time domain of speed prediction,and the optimal power control sequence is solved at each moment;Finally,according to actual operating conditions,the energy management results of the MPC-PMP strategy,offline PMP strategy,and traditional threshold method were compared.The simulation results show that although the fuel consumption rate of MPC-PMP strategy is slightly lower than that of the offline PMP strategy,lower by 14.34%than that of the traditional threshold method strategy,and the MPC-PMP strategy meets real-time requirements.