Research on the Energy Management Strategy for a Hydrogen Fuel Cell Hybrid Excavator Based on an Improved DDPG

To address the multi-objective energy management problem that balances power performance, fuel economy, and fuel cell lifespan, this paper proposes an energy management strategy based on an improved Deep Deterministic Policy Gradient (DDPG) algorithm, tailored to the operational characteristics of excavators. While the DDPG algorithm is introduced to optimize the energy management strategy, the traditional DDPG still exhibits insufficient adaptability when handling dynamic nonlinear systems. Building on the conventional DDPG framework, the following enhancements are implemented: First, historical state-space information is incorporated to expand the spatial dimensions. Second, the reward function is optimized by integrating power tracking and adjustment factors based on actual vehicular engineering requirements, guiding the agent to achieve optimization more efficiently. Finally, the critic network integrates batch normalization and dropout layers to accelerate training and enhance generalization, while the actor network incorporates additional hidden layers with ReLU activation functions, along with normalization and scaling processes, enabling more flexible and precise action outputs for improved power distribution. Simulation results demonstrate that compared to the traditional DDPG algorithm, the improved DDPG energy management strategy achieves over 20% faster convergence speed with superior stability. In terms of energy management efficacy, under standard cycle conditions, the hydrogen consumption of the improved DDPG strategy deviates by only 4.5% from the offline optimal allocation trajectory of dynamic programming (DP), while outperforming the deep Q-network (DQN) strategy by reducing hydrogen consumption by 3%.