Double-layered Model Predictive Control Algorithm In Industrial SvstemBased On First Principle Model

Combined with the features of model predictive, rolling optimization and feedback correction, Model Predictive Control(MPC) has become one of the valid control algorithms to deal with complicated multi-variable constraint problem. Based on the First Principle Model, we conduct a research on offset-free control, multi-objective optimization in steady-state and design a double-layered model predictive controller with an estimator in this thesis. Besides, we analyze the performance of this controller based on the simulation result of a distillation column model.The main work and contributions of the thesis are listed as follows:Based on the feature of the First Principle Model Predictive Control, an offset-free model predictive control algorithm is proposed. A disturbance model is built to extend the state of the original system in the algorithm. In addition, an extended Kalman filter(EKF) is designed to estimate the extended state. Thus offset-free control can be achieved by using the estimation of the disturbance state. The dynamic optimal control problem is discretized and transformed into a non-linear programming problem using the Orthogonal Collocation Finite Elements(OCFE) method. To ensure the accuracy of discretization, the Radau point is chosen as the interpolation point.Aiming at the requirements in the industry, we propose an multi-objective and layered steady-state optimization algorithm for the First Principle Model based on the structure of linear two-layered model predictive control. Feasibility analysis, economic optimization and operation point optimization are sorted and completed in different layers depending on their priorities. The optimization results of the high layer form the constraints of the low layer. Thus the optimization goal at the high layer can always be met. At last, to prove the validation of our algorithm, the Continuous Stirred Tank Reactor(CSTR) model is used for simulation.Based on the characteristics of distillation column model, a double-layered model predictive control strategy is proposed to achieve the double-composition control in distillation process. An EKF is designed to estimate the quantities of the two compositions based on the information of temperature. Moreover, high frequency temperature correction and slow frequency compositions correction are combined together to correct the quantities of the compositions in the strategy. Simulations with disturbances in feed composition and mismatches in model are used to analyze the performance of the strategy. The experimental results show that this control strategy can not only achieve offset-free control but also has the good performance of disturbance resistance and robustness.