A nonlinear model predictive controller based on the linguistic model for biochemical continuous sterilization

Continuous sterilization is conducted on medium before fermentation to achieve destruction of microorganisms in the medium. In the process control of continuous sterilization, high control quality may not be obtained for there exist severe nonlinear property and system disturbances in the process. In this paper, a nonlinear model predictive controller based on the linguistic model is proposed. The linguistic model is employed to make predictions of the future dynamic of process, which consists of a series of fuzzy rules, whose antecedents are the membership functions of the input variables and the consequents are the predicted output represented by linear combinations of the input variables. For the linear properties of the consequent, the control actions could be obtained in a quadratic optimization structure without the calculation of on-line linearisation. Based on the field data, both the antecedent and the consequent parameters are tuned by a hybrid-learning algorithm. The data-driven determination of the antecedent and the consequent provides an optimization procedure with optimized controller parameters for nonlinear model predictive control. Experiments are conducted, and the performance of the proposed nonlinear MPC controller is evaluated in comparison with the methods of model predictive control nonlinear with successive linearisations and the nonlinear model predictive control with neural network approximator. The experimental results show that the proposed controller could provide higher control accuracy and lower energy consumption.