In this paper, a fuel cell system supervisory controller is developed for a fuel cell-based hybrid electric vehicle to safely control the interactions between powertrain components, maximize efficiency and minimize the degradation of the fuel cell. The proposed fuel cell supervisory controller includes three main elements: a state machine, an optimal setpoint generator and a power limit calculator. The state machine, as the top layer of the supervisory controller, is responsible for coordinating the various subsystems of the fuel cell, including the three subsystems, anode, cathode, thermal, and the DC/DC converter. The primary purpose of the state machine is to ensure global control over these subsystems as well as facilitate communication between the fuel cell system, diagnosis system, and Vehicle Control Unit (VCU). The state machine not only allows for the appropriate transitions between states but also governs the fuel cell system operation in all its different operating states such as Start-up, Shutdown and Run. The optimal setpoint generator is responsible for determining the operating conditions of the fuel cell system that maximizes the system's efficiency. It is designed by taking into account the comprehensive model of the fuel cell stack, considering manufacturing constraints, and incorporating the compressor map which then provides the optimal setpoints for all the subsystems' local controllers. A power limit calculator is also developed to compute the stack available power and feeds this information to the energy management system in the VCU. This information is used by the VCU to split the requested power between the fuel cell and the battery. The experimentally validated stack model and the complex model of the subsystems based on the Inn-Balance project data are used in the simulation. Furthermore, the subsystems' local controllers used in the MATLAB-Simulink were validated in a real vehicle test bench. The Common Artemis 130 km/h Driving Cycle (CADC) for automotive applications is used to verify the proposed fuel cell system supervisory controller in the MATLAB-Simulink environment. The simulation results showed that the proposed control structure functioned properly in the Run mode using this CADC-based load profile.

This work presents a Nonlinear Model Predictive Controller (NMPC) strategy for the optimal operation of a fuel cell system to maximize its average efficiency and minimize degradation. Because of the different time scales associated with the different controlled variables, the proposed NMPC is divided into three distinct parts, each one customized with its specific prediction horizon and sampling time. The NMPC is based on a comprehensive model of the fuel cell stack, considering manufacturing constraints and incorporating the compressor map. The NMPC outputs are the optimal setpoints for all the subsystems’ local controllers. The studied fuel cell system incorporates the anode recirculation loop with an ejector, a water separator, and a bleed valve, and the proposed NMPC determines the optimal opening percentage of the bleed valve to prevent nitrogen accumulation and efficiency distortion. The proposed strategy yields higher efficiency than offline steady-state setpoints in the evaluated scenario, while full future-current information generally provides further improvement relative to limited preview.

This work presents a Nonlinear Model Predictive Controller (NMPC) strategy for the optimal operation of a fuel cell system to maximize its average efficiency and minimize the degradation. Because of the different time scales associated with the different controlled variables, the proposed NMPC is divided into three distinct parts, each one customized with its specific prediction horizon and sampling time. The NMPC is based on a comprehensive model of the fuel cell stack, considering manufacturing constraints, and incorporating the compressor map. The NMPC outputs are the optimal setpoints for all the subsystems’ local controllers. The studied fuel cell system incorporates the anode recirculation loop with an ejector, a water separator and a bleed valve, and the proposed NMPC determine the optimal opening percentage of the bleed valve to prevent nitrogen accumulation and efficiency distortion. The performance of the proposed NPMC in terms of efficiency is compared with an offline setpoint generator through simulation. The results show the efficiency improvement using the proposed strategy. Furthermore, considering the unpredictable changes in load current within automotive applications, the study evaluates the efficiency when a constant load is assumed after a specific time in the prediction horizon and compares it with cases in what the load current is assumed known throughout the entire prediction horizon.