Research on load-resistant enhancement control optimization of composite energy power supply system based on deep reinforcement learning

Abstract

To solve the problem of insufficient inertia in the power electronics components under the development trend of power electronics in the transmission system of heavy-duty electric vehicles, combined with the composite energy power supply mode and control method, voltage-based feedback channels and virtual capacitor branches are designed for both engine-generator sets and DCDC control loops. These innovations compensate for system inertia and introduce a TD3 deep reinforcement learning-based adaptive regulation component to manage electrical systems in response to DC voltage fluctuations. This approach enhances the stability of the entire system in the face of load disturbances. Strategy validations demonstrate that this method ensures stable system operation, significantly mitigates DC voltage fluctuations during sudden load changes and is highly adaptable without relying on system models.