Model predictive control of non-ideal-gas heat exchangers

Model predictive control of non-ideal-gas heat exchangers

Heat exchangers with non-ideal working fluids are core components of many industrial processes and power cycles, such as the supercritical CO2 Brayton cycle. In power applications, heat exchangers are involved in several control tasks. During steady operation, heat-exchanger outlet streams must be regulated to thermodynamic and mass flow setpoints in the presence of disturbances. During load-following operation, time-varying heat transfer profiles must be accurately tracked while respecting equipment and safety constraints. Model predictive control (MPC) can improve closed-loop performance in these scenarios since it performs well for transient multivariable control problems and can routinely deal with constraints.
Implementing MPC for non-ideal-gas heat exchangers is challenging due to 1) nonlinear fluid behaviours rendering simplified modelling techniques invalid, 2) inherent numerical limitations in simulating compressible-flow systems, and 3) the unavailability of fluid property measurements inside heat exchangers. We address these challenges by developing an MPC that is implemented via successive online linearisations of an analytical reduced-order 1D heat exchanger model. The controller’s objective is to regulate the outlet state of the heat exchanger’s process stream to a specified reference state. The setpoints of the compressors that drive mass flow through each stream are used as the control variables. A nonlinear observer based on this reduced-order model is used to estimate the internal state of the heat exchanger from inlet and outlet fluid property measurements. We perform closed-loop simulations for a CO2-CO2 heat exchanger and a molten salt-CO2 heat exchanger to analyse the performance of the controller. The controller provides good performance with prompt disturbance rejection and negligible steady-state offset.