Predictive control for capacity controlled heat pumps

Nilay Saraf

Wednesday, April 29, 2015, 10:00 - 11:00

Room 01-012, Georges-Köhler-Allee 102, Freiburg 79110, Germany

Residential heating systems are one of the major consumers of energy globally. Minimizing this energy consumption and costs calls for energy efficient operation of heating systems and, increasing use of renewable energy. The integration of renewable energy sources in the electricity network is challenging due to the fluctuation in their generation. The use of thermal storage to decouple heat demand and electricity supply provides the possibility to integrate power from renewable energy sources and demand side management. Capacity controlled heat pumps provide efficient heating and high flexibility in operation when coupled with thermal storage. This allows using innovative optimal control strategies to minimize energy consumption and electricity costs.
Model predictive control (MPC) for heat pumps has been identified as one of the possible solutions to this challenge. MPC offers properties such as constraint handling, multivariable control and optimal performance with conflicting objectives. It has been shown to outperform conventional control strategies. For the best performance, the optimization problem formulation must accurately represent the control objectives and the problem should be computationally tractable.
The original resulting optimization problem is mixed-integer non-linear due to the system characteristics for instance, input dead-zone and non-linear dependency of the coefficient of performance of the capacity controlled heat pump on system inputs. By neglecting the dead-zone and approximating the nonlinearities in the problem, simplified convex optimization problems are obtained which guarantee global optimal solution. In this thesis, different problem formulations are studied through simulations to investigate the impact of different simplifications on the quality of the controller performance, load shifting, energy efficiency and costs against a reference non-predictive control strategy in multiple tariff scenarios.