Adaptive predictive control based on adaptive neuro-fuzzy inference system for a class of nonlinear industrial processes

In present paper, a novel adaptive predictive control method is proposed for a class of nonlinear systems via adaptive neuro-fuzzy inference system (ANFIS). In the proposed method, a kind of nonlinear generalized predictive controller (GPC) is utilized where the model is achieved using an adaptive intelligent system. The dynamics of the system are classified into two linear and nonlinear parts. Linear part is approximated using least squares estimation technique, and the nonlinear part is identified using an ANFIS-based identifier. Therefore, the future behavior of the system is predicted based on the intelligent identification method in order to be used for designing the controller. The controller is updated based on these two identified models of the system's parts. The proposed method has the ability of real time implementation, and also there is no need of pre-training phase of the network. The controller performance is investigated by carrying out different simulations on two nonlinear process benchmark problems. For this purpose, a liquid level control system and a continuous stirred tank reactor (CSTR) are considered. Simulation results show the fidelity of proposed method for unknown nonlinear systems in presence of noisy and disturbed conditions.