In this paper, two different advanced control approaches for a pressurized solid oxide fuel cell (SOFC) hybrid system are investigated and compared against traditional proportional integral derivative (PID). Both advanced control methods use model predictive control (MPC): in the first case, the MPC has direct access to the plant manipulated variables, in the second case, the MPC operates on the setpoints of PIDs which control the plant. In the second approach, the idea is to use MPC at the highest level of the plant control system to optimize the performance of bottoming PIDs, retaining system stability and operator confidence. Two MIMO (multi-input multi-output) controllers were obtained: fuel cell power and cathode inlet temperature are the controlled variables; fuel cell bypass flow, current and fuel mass flow rate (the utilization factor kept constant) are the manipulated variables. The two advanced control methods were tested and compared against the conventional PID approach using a SOFC hybrid system model. Then, the MPC controller was implemented in the hybrid system emulator test rig developed by the thermochemical power group (TPG) at the University of Genoa. Experimental tests were carried out to compare MPC against classic PID method: load following tests were carried out. Ramping the fuel cell load from 100% to 80% and back, keeping constant the target of the cathode inlet temperature, the MPC controller was able to reduce the mismatch between the actual and the target values of the cathode inlet temperature from 7 K maximum of the PID controller to 3 K maximum, showing more stable behavior in general.
The use of model predictive control (MPC) in advanced power systems can be advantageous in controlling highly coupled variables and optimizing system operations. Solid oxide fuel cell/gas turbine (SOFC/GT) hybrids are an example where advanced control techniques can be effectively applied. For example, to manage load distribution among several identical generation units characterized by different temperature distributions due to different degradation paths of the fuel cell stacks. When implementing an MPC, a critical aspect is the trade-off between model accuracy and simplicity, the latter related to a fast computational time. In this work, a hybrid physical and numerical approach was used to reduce the number of states necessary to describe such complex target system. The reduced number of states in the model and the simple framework allow real-time performance and potential extension to a wide range of power plants for industrial application, at the expense of accuracy losses, discussed in the paper.