A novel hierarchical real-time energy management strategy for marine transportation based on feedforward-feedback coordinated control

Real-time energy management strategies (EMS) for marine applications often struggle to achieve optimal performance. To address this limitation, this paper introduces a three-layers hierarchical strategy based on feedforward-feedback coordinated control architecture. The upper layer utilizes the dual-state Pontryagin's Minimum Principle (DPMP) to derive globally optimal control sequences, which serve as the baseline for multi-neural networks training, and the PSO-GD is implemented for initialization convergence acceleration. The middle layer employs trained multi-neural networks (MNN) to activate a corresponding sub-network, which are categorized by low-, medium-, and high-speed mode, for the feedforward pre-calibration of the equivalent factor (EF). Finally, the lower layer implements a dual-state feedback (DSF) mechanism that performs the final real-time correction of the EF, incorporating two aspect correction based on baseline error of state-of-health (SOH) and state-of-charge (SOC). Both simulation and experiment validation are implemented to confirm the effectiveness of the proposed strategy. Simulation validation confirms the effectiveness of the proposed strategy, demonstrating 96.08% of the global optimal performance, and the end-of-life (EOL) attainment of battery is delayed by 12.13%, significantly extending operational longevity. Experimental results demonstrate close consistency with simulations, confirming both simulation accuracy and real-time implementation feasibility.