Robust MPC of an unstable chemical reactor using the nominal system optimization

Abstract

Abstract The continuous stirred-tank reactor with uncertain parameters was stabilized in the open-loop unstable steady state using the robust model predictive control. The gain matrices of the robust state-feedback controller were designed using the nominal system optimization and the quadratic parameter-dependent Lyapunov functions. The controller was verified by simulations using the non-linear model of the reactor and compared with the robust model predictive controller designed using the worst-case system optimization. The values of the quadratic cost function and the consumption of coolant were observed. Both robust model predictive controllers stabilized the reactor despite constrained control inputs and states. The robust model predictive control based on the nominal system optimization improved control responses and decreased the consumption of coolant.