Dynamic programming solutions extracted SOC-trajectory online learning generation algorithm based approximate global optimization control strategy for a fuel cell hybrid electric vehicle

The trajectory of the battery state of charge (SOC) optimized by using dynamic programming (DP) is the global optimization solution to enhance the economy performance of the fuel cell hybrid electric vehicles under various driving cycles, however, this method requires prior knowledge of the future driving cycles. To utilize the solutions of DP, a SOC-trajectory online learning generation algorithm based approximate global optimization energy management control strategy is proposed. Initially, the global optimality of DP is used to extract the optimal SOC gradients for diverse driving scenarios. Real-time generation of optimal gradient factors for SOC trajectories is facilitated through the training of a backpropagation neural network with DP solutions. Subsequently, the deterministic rules are designed to plan SOC under actual driving conditions, with a dynamically updated threshold by the trained agents. Finally, based on the above, the optimal calculation of energy allocation is performed by combining sequence quadratic programming. Numerical verification, inclusive of hardware-in-the-loop experiments, show the effectiveness of the proposed strategy. The results demonstrate that the proposed strategy improves fuel economy by 7.39% compared to ECMS. Additionally, it reduces the cost of fuel cell life loss by 32.09% and achieves over 90% optimization of global driving cost.