Adaptive predictive control : analysis and expert implementation

A generalized predictive controller has been derived based on a general state-space model. The case of a one-step control horizon has been analyzed and its equivalence to a perturbation problem has been emphasized. In the case of a small perturbation, the closed-loop poles have been calculated with high accuracy. For the case of a general perturbation, an upper bound on the permissible perturbation norm has been derived. A functional analysis approach has also been adopted to assess the closed-loop stability in the case of nonlinear systems. Both the plant-model match and plant-model mismatch cases have been analyzed. The proposed controller has proven to be so robust that an adaptive implementation based on Laguerre-filter modelling has been motivated. Both SISO and MIMO schemes have been analyzed. Using a sufficient number of Laguerre filters for modelling, the adaptive controller has been proven to be globally convergent. For low-order models, the robustness of the adaptive controller can be insured by increasing the prediction horizon. The convergence and robustness results have been extended to other predictive controllers. A comparative study has shown that the proposed controller would be superior to the other predictive controllers if the open-loop system is stable, well-damped, and of unknown order or time delay. To achieve a reliable control without deep user involvement, the adaptive version of the proposed controller has been implemented using the expert shell, G2. The resulting expert system has been used to orchestrate the operation of the controller, provide an interactive user interface, adjust the Laguerre-filter model using AI search algorithms, and evaluate the performance of the controller on-line. Based on the performance evaluation, the tuning parameters of the controller can be adjusted on-line using fuzzy-logic rules.Applied Science, Faculty ofElectrical and Computer Engineering, Department ofGraduat