Contribution à la commande prédictive non linéaire.

Abstract

In this thesis, two fuzzy Generalized Predictive Control (GPC) methods are proposed for discrete-time nonlinear systems via Takagi-Sugeno system based Kernel methods. In the first approach, which is based on Kernel ridge Regression strategy (TS-KRR), the unknown nonlinear systems is approximated by learning the Takagi-Sugeno (TS) fuzzy parameters from the input-output data. Two main steps are required to construct the offline TS-KRR approach: the first step is to use a clustering algorithm such as the clustering based Particle Swarm Optimization (PSO) algorithm that separates the data into groups and obtains the antecedent TS fuzzy model parameters. In the second step, the consequent TS fuzzy parameters are obtained using a Kernel ridge regression algorithm. Furthermore, the TS based predictive control is created by integrating the TS-KRR into the Generalized Predictive Controller. Next, an adaptive, online, version of TS-KRR is proposed and integrated with the GPC controller resulting an efficient adaptive fuzzy generalized predictive control methodology. In the adaptive TS-KRR algorithm, the antecedent parameters are initialized with a simple K-means algorithm and updated using a simple backpropagation algorithm. Then, the consequent parameters are obtained using the sliding-window Kernel Recursive Least squares (KRLS) method. For each control structure (the control strategies based on the online and offline strategies), two simulation studies were presented to justify the validity of the proposed approaches and the results were compared with other techniques cited in references. Furthermore, another adaptive fuzzy Generalized Predictive Control (GPC) is proposed for discrete-time nonlinear systems via Takagi-Sugeno system based Kernel Least Squares Support Vector Regression (TS-LSSVR). The proposed adaptive TS-LSSVR strategy is constructed using a multi-kernel lest squares support vector regression where the learning procedure of the proposed TS-LSSVR is achieved in three steps: In the first step, which is an offline step, the antecedent parameters of the TS-LSSVR are initialized using a fuzzy cmeans clustering algorithm. The second step, which is an online step, deals with the adaptation of the antecedent parameters which can be done using a backpropagation algorithm. Finally, the last online step is to use the Fixed-Budget Kernel Recursive Least Squares method to obtain the consequent parameters. Furthermore, an adaptive generalized predictive control for nonlinear systems is introduced by integrating the proposed adaptive TS-LSSVR into the generalized predictive controller (GPC). The reliability of the proposed adaptive TS GPC controller was investigated by controlling two nonlinear systems: A surge tank and continuous stirred tank reactor (CSTR) systems. The proposed controller has demonstrated good results and efficiently controlled the nonlinear plants. Furthermore, the adaptive TS GPC controllers have the ability to deal with disturbances and variations in the nonlinear systems.