PREDICTIVE CONTROL DESIGN FOR LARGE SCALE SYSTEMS

Abstract

The design of model-based Generalised Predictive Controller (GPC) for large scale systems is reported. A two-level decentralised Kalman filter is devised using the MAP approach. An optimal co-ordination strategy is developed for this filtering solution. This filter provides an output predictor for the GPC control design formulation. A two-level optimisation strategy is then employed to decompose the global GPC problem with the input/output constraints. The GPC solution for each subprocess is independently found at the lower level. This solution is sent to a higher level to update the values of an optimal co-ordinator. This procedure is repeated until an optimal solution is found. Simulation results demonstrate the improvements achieved by using the technique.