A novel Model Predictive Control framework for replacing traditional on-board Heat Pump controls
Heat pumps are a key technology for decarbonization and enabling demand-side flexibility. Fully realizing this potential requires advanced, model-based control strategies capable of responding to specific signals and anticipating system behaviour through predictive models of both thermal demand and heat pump performance. While significant progress has been made in modelling techniques for control systems, the integration of dynamic heat pump models within this framework remains an area requiring further development. Heat pumps are often modeled using performance curves in control systems, but such representations neglect onboard controls, which significantly affect performance and are typically inaccessible to end users. Addressing this issue requires detailed models that simulate the thermodynamic vapor cycle of each heat pump component. A common approach in the literature employs state-space representations of the thermodynamic cycle, using heat exchanger models such as Moving Boundary Models. However, these models are computationally intensive and often too complex for real-time dynamic control applications. This study provides a first demonstration of how the state-space formulation of the heat pump can be adapted for control applications. A black-box modeling approach is employed to represent the thermodynamic properties of the refrigerant, while linearization techniques are used to assess their impact on computational efficiency. Based on this approach, a new Model Predictive Control framework is proposed, using a linearized vapor-cycle model with Moving-Boundary heat exchanger representations. The framework is designed to act as a lower-level controller that replaces the on-board controllers typically used in heat pumps. The new control strategy has been implemented for a case study and tested in a simulation environment to evaluate its ability to track setpoints. The results demonstrate the effectiveness of the approach and provide a preliminary assessment of its potential, limitations, and future challenges.