Modelling is an important element in the field of IoT and heat pumps, as it turns data into information. In three Deep Dive sessions consisting of presentations by the participants on completed, ongoing and planned research work in the field of IoT and heat pumps, different aspects of modelling were analysed:
Semantic models contain information on how data relates to the real world, an example is Building Information Modelling (BIM). It was developed to consolidate the large number of different information sources throughout the lifecycle of a building. Current research projects focus on the integration of HVAC components in BIM, such as heat pumps or air handling units, the combination of real time construction information with BIM to create a digital twin for deviation detection and analytics, the use of BIM information to set up building performance simulations and the use of BIM for fault detection for facility management. It was concluded that standardisation is key for efficient data exchange and interoperability and that integrated BIM is not fully realized yet.
Data driven models and machine learning approaches make use of large quantities of data, e.g. by supervised or unsupervised learning, clustering, classification, etc. In the presented research projects, these methods are used for different applications, such as adaptive parameter selection and fitting for online calibration of a heat pump model during performance testing, automatic generation of regression models are created automatically e.g. for the prediction of room temperature depending on supply air temperatures, valve settings, etc, as well as the prediction of the heating value of heterogenous waste based on weather data.
Hybrid models or grey box models combine analytical models with detailed physical relations with data driven models based on statistics to find a good compromise on complexity, accuracy and calculation time. The different presentations provided insight in several applications of grey box models, such as advanced system monitoring, and fault detection based with model-based interpretation of system alerts for supermarket refrigeration systems and identification of incorrect settings in heat pumps for domestic heating. It was concluded that it is important to correctly record environmental influences on the observed system to be able to distinguish a fault from a change in the environment.
To the issue: