Author:

Abstract:

The energy transition towards a sustainable energy landscape with decentralised energy production and consumption units requires a combination of several technologies to keep electricity grids operational, secure and affordable. Among them is the integration of energy flexibility services in which grid assets are able to increase/decrease their energy production/consumption for a certain time. Within such a sustainable and energy flexible context, heat pumps are seen as a key technology, thanks to their renewable energy resource character and their potential to couple the heating and cooling sector to the electricity sector. Realising such a sector coupling enables grid users and the electricity grid in general to tremendously extend the sources for energy flexibility services. If coupled to an energy storage in the form of a physical buffer tank or by exploiting the thermal inertia of the building structure, the energy supply can be further decoupled from the energy demand. While literature already showed the associated potential of heat pumps, the effective implementation remains limited and is related to the lack of appropriate heat pump communication control interfaces, a lack of communication interface standardisation among heat pump manufacturers and the absence of fully understanding the default heat pump control logics. Regarding the heat pump control logics, heat pump manufacturers generally equip their heat pump systems with internal control strategies to ensure safety and reliability. These strategies can generally not be bypassed without requiring tailor-made solutions, thus limiting the ability to fully control the heat pump behaviour via external energy management systems. Contrarily, many works in literature neglect the importance of these internal control strategies by adopting large modelling time steps or assuming that the heat pump control response always meets its setpoint value. To this end, the first goal of this research work focusses on developing a short time-step, dynamic simulation heat pump model, capable of drawing a full picture on the internal control strategies and aimed to evaluate the importance of each individual control strategy. It has been found that incorporating these strategies influences both the operational short-term heat pump behaviour and long-term energy-related indicators, including its energy flexible behaviour. The causes could be related to many internal control strategies, in which the compressor speed control, domestic hot water timing constraints and pump control can be listed among the most influencing control strategies. To further investigate the operational heat pump behaviour and its ability to take part in demand side management actions, an experimental hardware-in-the-loop set-up has been built. Comparison of the available heat pump control interfaces under several energy flexible control strategies indicates that the options for directly controlling the electrical heat pump power consumption remain limited and show only limited functionalities without meeting high control accuracies. Alternatively, indirect heat pump control by controlling thermal-related variables with the aim of achieving a certain heat pump response proves to be more effective, but at the cost of an additional level of complexity by having to convert the required electrical heat pump response to thermal-related setpoint variables. A main shortcoming of all interfaces is the inability of the interface to bidirectionally communicate with an energy management system on the ability of the heat pump system to follow a certain control signal. Furthermore, the internal control strategies implemented in the energy flexible control interfaces, both for direct and indirect power control, have been showing improper functioning for several experimental trials in which the flexibility potential retrieved via simulations with a validated model could not be retrieved in the experimental set-up. Several recommendations are given, including revising domestic hot water temperature sensor positioning, pump control logics, adjustable power step-sizes and allowing energy management systems to adjust the compressor controller ramping rate. With the energy transition requiring an increasing demand for energy flexibility services with a high power control (prediction) accuracy, a digital twin of the heat pump system is developed. The digital twin allows to take into account the latest measurements from the real system and to predict its future states. With the aim to reach a general heat pump digital twin modelling approach, the potential of artificial intelligence is investigated by the adoption of artificial neural networks. However, feedforward neural networks have been showing difficulties to provide reliable predictions of the heat pump behaviour for longer time horizons than only one time step as these networks cannot take into account previous control actions from for instance the compressor and pump control. To this end, the potential of recurrent neural networks and more specifically long-short term memory networks is further investigated. With these networks being able to process a sequence of data, results prove their ability to take previous control actions from the internal heat pump control into account and to provide reliable predictions. An optimal prediction performance is reached when supplying the digital twin with data from the previous 60 minutes. However, converting the defined model architecture from solely providing space heating services to providing both space heating and domestic hot water services reaches difficulties with correctly predicting the heat pump operation for each service. This has ultimately led to developing a grey box heat pump digital twin model in which the control decision process on providing space heating or domestic hot water services is implemented by a simplified rule-based control approach, while the remaining control aspects are learned by the developed recurrent neural network. While an improved prediction accuracy is reached, the obtained model experiences oscillations in its power and pump flow rate predictions and recommended directions for future heat pump digital twin modelling research are given. Finally, this work proposes and experimentally validates a design energy flexibility characterisation procedure. This procedure aims to support end-users and energy system designers in estimating the energy flexibility of their heat pump and thermal energy storage system combination already during the design stage of the building. This should allow an improved integrated, energy flexible system design.