Investigating controller performance in hybrid SOFC systems in the presence of unknown nonlinearities

Solid oxide fuel cells (SOFCs) are energy conversion devices that offer many benefits over various other fuel cell types as a result of high operating temperatures (800-1000 °C). Unfortunately, SOFCs tend to possess poor load following capabilities due to delays along the fuel path and complex system dynamics. Maintaining safe operating conditions during changes in power demand is addressed using a controller designed to regulate the fuel cell current based on fuel flow measurement. In order to compensate for the resulting mismatch between demanded and delivered power, the SOFC system is hybridized with an energy storage device, such as an ultra-capacitor. Prior research at the HySES laboratory at RIT has led to control designs that guarantee robustness to uncertainties in system parameters such power electronics efficiencies. However, existing controllers for this system were developed under assumptions made about the unknown dynamics of the fuel supply system (FSS), such as exponential or bounded tracking. Retaining these controller designs, this thesis develops a general set of closed loop system equations in which the prior assumptions about the FSS are relaxed. The FSS behavior is treated as an unknown nonlinearity. Thereafter, concepts of absolute stability, Lyapunov stability and linear system approximation are used to evaluate the closed-loop system. The analysis leads to analytical conditions relating the controller gains and the local behavior of the FSS, predicting the onset of instability in the closed-loop system. The results are validated using simulations and using a hardware-in-the-loop test stand. Additionally, the problem of transient fuel utilization control of SOFCs is revisited and addressed by using a nonlinear observer design and an auxiliary hydrogen injection strategy. These approaches aim to compensate for fuel path delays and maintain desired operating conditions during transient loading conditions. Findings are validated using desktop simulations