The deployment of fuel cell hybrid electric buses (FCHEB) presents a zero-emission solution that aligns with global climate goals.
However, the performance of these systems is contingent upon advanced energy management strategies (EMS). Recent advancements spotlight predictive EMS (PEMS) which leverage future driving data for more efficient energy distribution. Yet, there’s been a pivotal shift as researchers realize the limitations of traditional methods reliant solely on vehicle speed predictions. Passenger load fluctuations, a critical component in real-world scenarios, have eluded conventional approaches but are now being integrated into the predictive framework.
Traditional approaches have utilized model predictive control (MPC) to gain foresight on vehicle speed, directly impacting energy allocation decisions among onboard systems. However, the urban transit environment poses unique challenges—particularly the variable nature of passenger numbers, which significantly alters vehicle mass and, consequently, energy requirements. Historically, both speed and passenger numbers were treated as isolated parameters rather than intertwined influences on bus operation.
This research elucidates a novel approach leveraging twin delayed deep deterministic policy gradient algorithms (TD3) within a data-driven guideline to address this oversight. The implementation of a dual prediction model employing BiLSTM (Bidirectional Long Short-Term Memory) networks is noteworthy, ensuring comprehensive forecasting of both speed and passenger numbers. This dual-focus strategy is harnessed to train a TD3-based decision-maker on datasets sourced from actual bus routes, incorporating both driving dynamics and passenger metrics. Notably, the research documented a substantial 5.92% reduction in operational costs for FCHEBs compared with antecedent models, whose static predictions neglected passenger variability.
For transportation experts, these insights represent a shift towards integrating advanced predictive analytics into EMS, underscoring the importance of multidimensional data in optimizing energy distribution. The expectation is clear—incorporating both real-time operational conditions and predictive analytics can redefine how energy management systems are engineered for next-generation public transport solutions.
