Search

Predictive Functional Control (PFC) is a heuristic Model Predictive Control (MPC) algorithm that offers intuitive, transparent and simple designs, along with the basic predictive control characteristics, in the cost and complexity roughly similar to that of a standard PID (Proportional-Integral-Derivative) design. But despite these advantages, its practical utilisation has largely been confined to a relatively small selection of simple industrial applications which exhibit benign first-order dynamics. This is mainly due to the use of over-simplified assumptions within the algorithm that generally work well for simpler systems but often cause undesirable tuning difficulties in slightly more complicated higher order applications. Another critical issue that limits its practicality is the lack of consistency and reliability in closed-loop performances while handling severely underdamped and/or open-loop unstable processes. Therefore, the primary objective of this research is to overcome these prominent deficiencies and hence extend the scope of PFC to a broader range of SISO applications by proposing: (i) a performance oriented controller tuning method which uniquely bases parameter selection on the expected control activity for a well-informed and more meaningful tuning decision, (ii) a new PFC algorithm based on relative measures with far simpler controller tuning as compared to the standard practices of parameter selection, and (iii) a systematic design framework integrating the concept of pre-stabilised or closed-loop predictions within the overall formulation for efficient control of challenging processes. Furthermore, the thesis also investigates a relatively unexplored application of PFC in the area of nonlinear predictive control and therefore presents an efficient and cost-effective PFC design for a class of nonlinear systems as the final contribution. The efficacy of these proposals has been investigated through numerous simulation studies which suggest marked performance improvements over the conventional PFC, and indeed the PID, in real-world scenarios. It has been observed that: (i) the new tuning proposal for conventional PFC and the proposed Relative PFC algorithm both provide approximately upto 30% faster closed-loop settling times as compared to the existing tuning methods which are either too aggressive for practical implementation or somewhat conservative to have a meaningful impact on the closed-loop behaviour, (ii) the use of the proposed pre-stabilised, or closed-loop, predictions ensure output stability and recursive feasibility under constraints where the direct utilisation of challenging open-loop predictions within PFC fail to perform reliably, and (iii) the proposed Nonlinear PFC algorithm, being inherently better at handling process nonlinearities, provides approximately 2-4 times faster closed-loop responses than the linear PFC (and the PID), and is therefore a natural choice in processes involving wider operating ranges.