Dynamic modelling and nonlinear model predictive control of a reversible solid oxide fuel cell

Abstract

The primary objective of this work is to research and develop a dynamic model and an advanced control strategy for a reversible solid oxide fuel cell in a grid to ensure load tracking whilst maintaining fuel utilisation and temperature dynamics within a safe range. This report presents the successful development of the dynamic model and corresponding controller. To do so, modeling and control are discussed in two separate chapters where model requirements, specifications and conception are presented followed by the control techniques, control objectives and controller synthesis. A RSOFC model has successfully been developed and thoroughly validated against other literature. To do so, a set of fitting parameters has been distilled from literature to create a good overlap with other studies and are combined in a manner that they can be implemented in other work. Steady state dynamics and the transient dynamics have shown a good match to the available literature. Additionally, the model offers useful insights into RSOFC (transient) dynamics with relation to temperature effects, fuel composition and cell support structure, which have not been documented before. At last the RSOFC stack, together with the developed model has been built in a plug-and-play manner that it can be implemented and adjusted by others, to be used in combination with other models. For control an output-feedback adaptive nonlinear model predictive controller has been developed, which is an advanced version of the well established (non)linear model predictive controller. The development of the temperature controller is given wherein the structure of the MPC is presented together with its trajectory, adaptive constraints and tuning variables. After completing the development of the controller, the controller was simulated together with the dynamic model as part of a micro-grid. The simulations were split up into short and long term scenarios and showed satisfactory results as all the set control objectives were achieved.