Heat exchanger control using model predictive control with constraint removal

    We applied MPC with a constraint removal approach to a laboratory plate heat exchanger. By detecting and removing inactive constraints before solving the underlying optimization problem, this variant of MPC reduces the computational effort associated with solving the optimization problem. The bounds indicating the inactive constraints were modified to overcome the challenges arising in the control setup. In this paper, two experimental case studies were investigated to analyze the properties of the proposed control method--control of the laboratory heat exchanger plant, and the implementation on a microcontroller.

    In the real-time experiments on the laboratory plant, the constraint removal approach was able to reduce the number of constraints to be considered in the optimization problem by up to 60% compared to conventional MPC. The results further confirmed that the approach does not affect the control performance in terms of performance losses, resulting in comparable trajectories of the control inputs.

    Based on the experimental data, we further implemented and solved the optimization problems corresponding to MPC with constraint removal and to conventional MPC on a 32-bit microcontroller. Both, the computation time and the associated energy consumption decreased by approximately 68% for MPC with constraint removal in contrast to the conventional variant of MPC.

    Future research will be focused on the application of nonlinear MPC with constraint removal for the control of the heat exchanger plant and modifications towards robust MPC.