A transfer-based reinforcement learning collaborative energy management strategy for extended-range electric buses with cabin temperature comfort consideration

Electric vehicles (EVs) have received extensive attention as an environmentally friendly and sustainable mode of transportation. To address "range anxiety" issues, extended-range electric vehicles (EREVs) have gradually gained popularity as a solution. However, current research on energy management strategies (EMS) for EVs often overlooks the energy consumption of the air conditioning (AC) system, resulting in suboptimal energy allocation. Therefore, this study focuses on the extended-range electric bus (EREbus), an extended-range electric bus, and incorporates the AC system into its EMS, enabling coordinated optimization with the powertrain system. First, the study embeds a control-oriented cabin thermal management model based on the powertrain model. Next, representations transfer-based reinforcement learning (RTRL) transfers the learned policy representations from the AC-off state to the EMS in the AC-on state. Furthermore, the study analyzes the impact of different representation transfers on powertrain performance and thermal comfort. The results demonstrate that the proposed EMS can improve the convergence rate and stability of training. Compared to direct learning methods, RTRL exhibits clear advantages in reducing operating costs and improving cabin thermal comfort, achieving reductions of 8.3%-12.6 % and 5.2%-27.0 % in operating costs for driving modes with different battery levels. Moreover, setting the transfer layer appropriately promotes the utilization of the global optimal potential of RTRL. This research provides support for energy management and holds the potential for promoting the development of EVs.