Hierarchical intelligent energy-saving control strategy for fuel cell hybrid electric buses based on traffic flow predictions

Vehicles' perception capability of environmental situations has been enhanced in the connected environment. However, the utilization of abundant and diverse traffic information poses a challenge in the formulation of energy-efficient strategies for current connected vehicles. To address this challenge, a hierarchical intelligent energy-saving control strategy for fuel cell hybrid buses based on traffic flow prediction is proposed. Diverging from traditional approaches that rely on surrounding vehicles status information, a more macroscopic perspective is introduced by incorporating traffic flow prediction information into the formulation of energy-saving control strategies for the first time, which enhances the adaptability of strategies. At the upper layer, a multi-objective intelligent eco-driving control strategy is designed, encompassing driving safety, energy consumption costs, traffic efficiency, and ride comfort. At the lower layer, an intelligent energy management strategy is developed to reduce hydrogen consumption and maintain stable battery state of charge. Simultaneously, action variables serve as the information bridge between the upper and lower layer strategies, and comprehensive analyses and validations of strategic performance are conducted from various perspectives. The research findings indicate that the introduction of traffic flow information enhances the cognitive capabilities of the intelligent system towards the traffic environment with little impact on the convergence process. The model excels in energy efficiency, driving smoothness, and passenger comfort compared to the baseline model, while also having the opportunity to surpass the traffic efficiency of the IDM model. The developed energy management strategy demonstrates an energy-saving benefit of 92.07 % of the offline optimal benchmark, and closely approaches the theoretical optimal performance of MPC with a prediction horizon of 5s. Additionally, the proposed strategy demonstrates high computational efficiency and effectively mitigates the degradation of the vehicle lifespan, making it conducive to real-time online applications.