On Infinite-Horizon State-Constrained Optimal Control - With Application to a Fuel Cell APU

This thesis presents a method for computation of optimal control policies for state-constrained, time continuous, infinite-horizon systems. It is applicable to input-affine systems where the stochastic contribution can be modeled as a vector of white noise signals that is added to the control input, with the intensity of the noise being inversely proportional to the control cost. A distinctive advantage of the method is that it can treat problems where the interaction between stochastic disturbances and hard constraints are of importance to the optimal control policy. The method is demonstrated with the example of control of a fuel cell driven auxiliary power unit. This is a unit that would produce on-board electricity for a truck or other vehicle by reforming the hydrocarbon fuel and feeding the obtained hydrogen to a fuel cell stack. Since energy is stored in several locations in the system, control is required to ensure that the stochastic power demand can be met at optimum fuel efficiency