A glossary of the most relevant expressions associated with digital twins for large-scale heat pump and refrigeration systems.

The glossary provides an explanation of the most relevant expression associated with digital twins for large-scale heat pump and refrigeration systems.

Many of the expressions are interdisciplinary expressions and may be found in various other fields. Even though the definitions are aligned as much as possible with other fields of application, it may be that the exact definitions differ from application to application. Therefore, it may be noted that the definitions presented here refer to the context of the project and the respective application to heat pump and refrigeration systems.


Digitalization occurs in probably any field of application and accordingly broad are the main expressions. In the following, an overview of the main definitions is given, with a main focus on applications of heat pump and refrigeration systems.

  • Digital Twin: Digital twins are a virtual representation of a physical system and consist of one or more numerical models, which are continuously adapting to the current status of the system.
  • Artificial Intelligence: Artificial intelligence (AI) is the overall discipline associated with the capability of machines to perceive its environment as a basis for improving their decision process and developing more successful actions. Machine learning and neural networks are sub-disciplines for artificial intelligence.
  • Machine Learning: Machine Learning describes the discipline of artificial intelligence in which the learning is based on data analyses and used as forecasting tool and decision support. Machine learning algorithms are not limited to a pre-defined description of a physical system but imply the possibility to develop and adjust the numerical system based on measurement data.
  • Neural network: Neural Networks are often referred to as the ‘brain’ in an AI, consisting of interconnected neuron layers – an input layer, one or more hidden layers and an output layer. Each connection in the network is assigned with a weight describing the importance of that connection. The nature of neural networks is simple to understand, but the training of the weights can be done in multiple complex ways and there exists a large variety of network architectures.

Numerical modelling

Numerical modelling describes the application of mathematical models that are used to describe a certain process. The models may be of different complexity and different approaches. An overview of the main expressions is given in the following.

  • Analytical models: Analytical models are numerical models, which are to a large extent based on physical relations, such as mass and energy balances. The physical relations are selected in an analytical way to mimic the real behavior of the real system in a most accurate way. However, analytical models do often also comprise a smaller amount of data-driven correlations, which are used to fit the analytical expressions to measured data.
  • Data-driven models: Data-driven models (or grey-box models) contain only a limited amount of physical relations. They are instead based on general approaches that are linking input to output values by general differential-algebraic equations. Furthermore, physical considerations may be used to derive the general structure or to supplement the black-box models, which yields the term grey-box models.
  • Stochastic models: Stochastic models are grey-box models which are extended by a term accounting for the uncertainty. 

Heat pump and refrigeration technology

This project focuses on applications to large-scale heat pump and refrigeration systems. An overview of the most relevant expressions in this context is given in the following.

  • Coefficient of performance (COP): The COP defines the ratio of the useful effect, e.g. supplied heating or cooling, to the consumed power. The COP is a key performance indicator for heat pump and refrigeration systems.
  • Heat pump system and refrigeration system: Both systems are machines that are extracting heat from a lower temperature reservoir and providing it to a higher temperature reservoir. Bringing the heat to a higher temperature level requires work, which is often provided by electricity-based compressors. For heat pump systems, the focus is on providing heat to a process, such as district heating, while the heat is often recovered from the ambient of an excess heat stream. Refrigeration systems are operating similarly but at retrieving the heat from a reservoir to be cooled, such as supermarket display cabinets, while rejecting it at higher temperatures to the ambient or possibly a useful application, such as district heating.
  • Heat recovery: The main purpose of supermarket refrigeration systems is the supply of cooling to the goods. However, the heat has to be rejected to the environment, while it also may be supplied to higher temperatures suitable for district heating purposes or other heating purposes. This yields highest performances, since two services are supplied at the same time and by the same equipment.
  • Compressor rack: In supermarket refrigeration systems, the compressors are often combined of several smaller units, which enables effective part load modulation. The compressors are typically installed on a compressor rack.