Energy-Efficient Peltier System Control Using Adaptive MPC and a Unscented Kalman Filter

The paper focuses on a novel nonlinear control application - ice clamping of workpieces using Peltier cells which is based only on available measurements of the hot-side temperature. A discrete-time model predictive control (MPC), which is adapted to Taylor linearizations on desired state and input values, takes advantage of estimates from an Unscented Kalman filter (UKF) for the temperatures on both the hot and cold sides. Favorable reference values are obtained from the nonlinear system model to minimize the energy consumption. The time-varying MPC is implemented with the UKF, and comprehensive simulation studies are performed using different prediction horizons. The simulation results are compared with each other in terms of both tracking precision and estimation accuracy. They clearly demonstrate the effectiveness of this control methodology and indicate the robustness w.r.t. parameter uncertainties and disturbances.