Fuzzy predictive control applied to an air-conditioning system

A method of designing a nonlinear predictive controller based on a fuzzy model of the process is presented. The Takagi-Sugeno fuzzy model is used as a powerful structure for representing nonlinear dynamic systems. An identification technique which enables the acquisition of the fuzzy model from process measurements is described. The fuzzy model is incorporated as a predictor in a nonlinear model-based predictive controller, using the internal model control scheme to compensate for disturbances and modeling errors. Since the model is nonlinear, a non-convex optimization problem must be solved at each sampling period. An optimization approach is proposed, that alleviates the computational burden of iterative optimization techniques, by using a combination of a branch-and-bound search technique, applied in a discretized space of the control variable, with an inverted fuzzy model of the process. The algorithm is applied to temperature control in an air-conditioning system. Comparisons with a nonlinear predictive control scheme based on iterative numerical optimization show that the proposed method requires fewer computations and achieves better performance. Real-time control results are presented. Copyright (C) 1997 Elsevier Science Ltd.