Reinforcement learning-based intelligent energy management strategy for connected hybrid electric vehicles

daha sonra doldurulacaktırNowadays, the increasing number of motor vehicles creates significant pressure on energy consumption and environmental pollution. In this context, hybrid electric vehicles offer an effective alternative for sustainable transportation systems and contribute to the targets of reducing emissions by increasing energy efficiency. Energy management strategies in hybrid electric vehicles are of critical importance in terms of effective control of multiple power sources. However, traditional energy management strategy approaches are insufficient in terms of adaptation to dynamic driving conditions, multi-objective decision making, and real-time control. In this thesis, multi-layered and learning-based energy management strategies are developed for hybrid electric vehicles. The proposed method is based on three different design scenarios: (i) multi-objective control, (ii) multi-agent system control, and (iii) incentive-based learning control, which are the main focuses of study. In the multi-objective control part, conflicting objectives such as energy consumption, battery management, emission, and speed monitoring are optimized with the DDPG algorithm. In the multi-agent structure, learning controllers are assigned to each vehicle in the platoon scenarios, and energy management strategies are integrated using the MATD-3 algorithm. In the incentive-based learning strategy, a reward function that provides a more effective control structure is designed by considering the long-term performance criteria with the DRL method utilizing the SAC algorithm. In the design scenarios, a hybrid electric vehicle model is created in the MATLAB/Simulink environment, and different approaches are tested comparatively. The developed strategies are evaluated in both simulation and human-in-the-loop simulation environments, and improvements are achieved in criteria such as energy efficiency, system stability, and emissions. This thesis aims to contribute to next-generation intelligent transportation systems by providing flexible and scalable energy management solutions that can be applied at both the individual and the platoon levels for hybrid electric vehicles.