Uncertainty characterization in robust MPC using an approximate convex hull

Abstract

Model Predictive Control (MPC) of processes when uncertainty is involved is the topic of this thesis. Specifically, a method to characterize parametric uncertainty for robust model predictive control is studied. The goal is to reduce the computational complexity of robust MPC and robust Moving Horizon Estimation (MHE). The main element of this method is the computation of an approximate convex hull that approximately covers the system uncertainty in a new output prediction mapping. Given the complete uncertainty set, an approximate convex hull is computed to determine an efficient set of extreme points to represent this set. The calculated set is a subset of the uncertainty set and can be significantly smaller than the original set, which results in decreasing the computational complexity. A measure of the approximation to the original set is used to provide robust output prediction that is guaranteed to hold for all systems in the original set. In other words, the introduced method provides a dynamic guaranteed bound on all possible output trajectories. The control performance of the proposed method will be investigated on the methane reforming process as a key element of fuel cells, and on a DC-DC floating interleaved boost converter.