Model Predictive Control approaches for centrifugal compression systems

Model Predictive Control (MPC) techniques are considered for industrial centrifugal compression systems with nonlinear dynamics. We consider the torque provided by an external drive and a recycle valve as control actuators for the system. Closed-loop stability in the presence of control constraints is studied via the contractive control Lyapunov function method. We solve the nonlinear MPC problem to assess a performance benchmark, and then design a Sequential Quadratic Programming (SQP) MPC approach which is computationally affordable. We show in several numerical simulations based on a realistic centrifugal compressor case study that the SQP MPC technique outperforms the classic linearized MPC and performs similarly to the nonlinear MPC approach.