Digital twins for building control

Start Date 21/09/20 - Finish Date 20/09/24

Creating a digital twin of the LSI building with the aim to use real-time control and machine learning to reduce energy consumption.

Building services, heating, cooling and ventilation, are necessary for a comfortable building but require a large amount of energy. There is potential to reduce the amount of energy needed by providing these services intelligently, i.e only when required, or when providing optimal benefit. This could be achieved by creating a digital twin of the building which utilises a  thermal model of the building to monitor, predict and optimise the energy usage of building services.

There are a number of challenges that must be completed for the project to succeed. A thermal model of the building must be made and tested, the role of this will be to predict the temperature in each room of the building. For this to work, various uncertain properties and behaviours of the building must be found and their uncertainty understood. For example, these will include material properties such as the thermal performance of walls and typical occupancy patterns. To do this historic thermal and weather data will be used to train the model and discover the probability distribution of uncertain components.

Combined with the above a multi-objective optimisation system must be created so that for set comfort criteria, for example, heated rooms must not drop below 21°C, a solution can be found which minimises the total building energy usage. Finally, this must all be done in real time with live data, a solution must be found and recalculated at intervals depending if the building performs as the model expected. 

Recent Updates

A lumped parameter thermal model of one floor of the LSI building has been initialised. Research Software Engineers Omar Jamil and Freddy Wordingham are creating methods to collect data from the LSI’s building management system, this will be used to quantify building model parameters.

Collaborators

• Professor Tim Dodwell, Academic Principle Investigator, https://emps.exeter.ac.uk/engineering/staff/td336
• Professor Jonathan Fieldsend, Co-Investigator, http://emps.exeter.ac.uk/computer-science/staff/jefields