A Learning Model Predictive Controller for Energy Management in Fuel Cell Hybrid Electric Vehicles

In fuel cell hybrid electric vehicles (FCHEVs), the fuel cell (FC) system serves as the primary power source and a secondary energy storage device such as a battery acts as the energy buffer. However, challenges such as high hydrogen prices and short FC lifespans underscore the need for advanced energy management strategies. This paper introduces a learning model predictive controller (LMPC) to optimize the power allocation between the FC and the battery used in FCHEVs to minimize the hydrogen consumption, enhance the FC lifespan, and maintain a reasonable battery state of charge (SoC). The proposed LMPC is reference-free and leverages the benefits of model predictive control (MPC) and iterative learning control (ILC). The LMPC design outlines the iterative construction of convex terminal sets and terminal costs based on past states and input trajectories. The effectiveness of the proposed LMPC is validated through hardware-in-the-loop (HiL) experiments.