Towards real-time capable optimal control for fuel cell vehicles using hierarchical economic MPC

Fuel cell vehicles are predicted to play an important role in the electrification of heavy-duty transportation, hence the efficient operation of their powertrains is critical. Since the system dynamics of fuel cell systems are characterized by widespread time-scales (from electrochemistry to thermal processes), the implementation of optimization-based control methods poses several challenges. Control of systems with dynamics covering multiple time-scales leads to a trade-off between control performance and computational effort if solved with a conventional single-layer Model Predictive Control (MPC) scheme. One approach is to partition the problem into different time-scales and to construct a hierarchical architecture combining multiple control layers. In this paper, a two-layer hierarchical MPC structure is presented that optimizes the operation of a fuel cell truck. Its power trajectory and temperature are optimized in a planning layer MPC over a prediction horizon that captures the entire remaining driving route and can therefore be hours long. On a secondary operational layer, the operation of the fuel cell system, namely its electrical current as well as its pressure and stoichiometry at the cathode, are optimized with a sampling time that is around one order of magnitude faster than on the planning layer. In simulation, the proposed approach achieves operation that only consumes 1.4% more hydrogen than the benchmark solution, while demonstrating real-time feasibility and the benchmark can only be computed in hindsight. •Hierarchical MPC scheme is proposed for energy management and fuel cell operation.•Simulative comparison between hierarchical and single layer MPC.•Computational complexity of approaches is compared and discussed.•Variation of battery capacity shows predictive capabilities regarding constraints.