A Rapidly Trainable Data-Driven Real-Time Energy Management Strategy for Fuel Cell Hybrid Electric Tractor

This study proposes a real-time energy management strategy that can be swiftly trained and applied for a fuel cell hybrid electric tractor. The strategy learns from the optimal solution of power allocation control sequences in past real operating conditions with the same scene characteristics, enabling real-time power allocation control to be achieved in a single matrix computation. Due to its light computational load, it exhibits high computational efficiency in simulations while also demonstrating good fuel economy. Specifically, hydrogen consumption is only 2.18% higher than that of an energy management strategy based on dynamic programming, and it reduces equivalent hydrogen consumption by 4.74% compared to a traditional model predictive control strategy.