Economic model predictive control with learning-based terminal cost for fuel cell electric vehicles

Leveraging connected-vehicle information is pivotal for unlocking the full performance of fuel cell electric vehicles (FCEVs) through an advanced energy-management strategy (EMS). Learning-based methods create new opportunities and challenges for EMS by offering powerful learning capabilities. This paper proposes a novel economic model predictive control with learning-based terminal cost (EMPC-LTC), which is tailored to the energy management of fuel cell electric vehicles for performance enhancement. To make full use of the To make full use of the future driving conditions, preserving the desirable closed-loop economic performance while reducing the computational burden for energy management, the proposed EMPC-LTC leverages the parametric terminal cost acting as a good approximation of the cost-to-go performance of the long-horizon EMPC. A feedforward neural network is utilized to characterize the parametric terminal cost which is trained based on the sampled data. For the simulation study, the results demonstrate that the proposed method performs better than the general EMPC since it maintains the desirable closed-loop economic performance with the lower on-line computational burden.