Model Predictive Control (MPC) of System Identified Continuous Stirred Tank Reactor (CSTR) with Constraints

The continuous stirred tank reactor (CSTR) is a chemical system used in a wide variety of industrial applications that need the addition and removal of reactants and products continuously. This paper discusses the design of model predictive controller for unknown linear CSTR. The model of the CSTR is estimated by system identification method using random time series input and output data with constraint limits. The system identification method, estimated transfer model coefficients determined by using Gauss newton algorithm and validated. Then discrete model predictive control is utilized to discover the future state variables and output variables, and system adjusts the input flow rate(W1) to keep the output concentration (Cb) at the desired setpoint position. The objective function is developed and the optimization problem is addressed by using Quadratic programming (QP) with constraints. The simulation results with varying the parameters of MPC reveal the effectiveness of the proposed method.