Introduction
Built environments of future cities will need to cope with climate change, and also more intricate energy demands. Avant-gardist systems that present synergies, typically via interactions and feedbacks (e.g. close adaptation of energy supply to peoples’ activities, heat recovery from infrastructure, greening of the city) are natural candidates. However, their evaluation, before implementation, can only be achieved by simulation models that accurately represent these complex, interdependent, time-dependent stochastic feedbacks. Such simulation models are only nascent at the building-scale and do not exist at all at the district and city scales.
Explaining the science
Current research advances in applied mathematics, statistical science, computer science, and engineering provides a timely confluence of developments where the formal coupling of mathematical models of systems to their physical realisations (or vice versa) introduces the opportunity to ‘twin’ the physical system with an appropriately abstracted model via measured and curated data from the actual physical system. The vision of this project is to bring the capability of digital twinning to energy modelling of built environments in manners that are transformative and have lasting impact.
Energy models of built environments are known to be poor representations of their actual counterparts owing to a combination of human actions and physical forces interacting in highly varying, partially understood, and unique manners. In other words, simulating the energy use of built environments is inherently a multiscale and multi-physics modelling challenge involving asynchronous sub-processes. Furthermore, built environments are subject to both global and local forcings which leads to large amounts of variabilities in their energy behaviours. Other challenges include variability in the information content of data streams, mismatch of resolutions between models and data, and many more.
Project aims
The specific aims are to:
- Develop novel techniques for the integration of data science and simulation models, to address specific challenges posed by stochastic energy models of built environments
- Infer inadequacies of models in their representation of the real-world processes driving energy use in built environments
- Investigate information gains from data at multiple scales in order to understand spatio-temporal patterns of energy use
- Propagate uncertainties in energy use models in order to support both short-term and long-term decision-making
Applications
The project is currently focusing on the following applications, with each presenting challenges that require adept and new forms of coupling between heterogeneous data, expert knowledge, and simulation models.
Data-centric models of end-use energy demand
The current challenge is the lack of established methods to represent stochastic (random) time-variations of energy demand in buildings at sufficient spatial resolution. This is a key source of incorrect energy predictions and it prohibits the understanding of dynamics within and across buildings. Using high resolution observations from buildings, this work will investigate 'functional data analysis' models to develop realistic (and stochastic) time-series of energy demand to be used for optimizing energy systems.
Subsurface environments
Recent research on urbanization has highlighted the significance of temperature anomalies within the shallow subsurface of the ground due to human activities. However, the spatial variabilities of ground temperature elevations at a city-scale are not well understood. This is due to lack of sufficient information (observations) and modelling complexity at such large scales. This work will investigate novel calibration approaches using sparse observations and mixed fidelity simulation models to estimate spatial variations of subsurface temperatures.
Urban farming
Integrating urban farms within dense environments has the potential to utilise waste infrastructure and resources within cities with wider societal benefits. In a unique partnership, this work will track the environment of the world’s first underground farm in London, located in tunnels designed as a WW2 air raid shelters in the 1940s. A long-term monitoring program on the site provides an opportunity to improve understanding of the thermodynamic processes governing the environment of the farm and develop new predictive models to support expansion of the farm’s activities.
Digital tools for resilient energy planning
This work will develop probabilistic master-planning tools to support effective implementation of energy policies in cities. It will integrate high-dimensional data that span environmental features, physical form and land-use, socio-demographics, and economics to analyse cities spatially. These kind of multi-dimensional analysis helps in understanding why certain energy policies are more or less successful than others, geographically, socially, and physically. It also enables sustainable infrastructure planning that is aligned with UK’s carbon targets and is, at the same time, locally robust.
Organisers
Researchers and collaborators
Dr Monika Kreitmair
Visiting Researcher
Rebecca Ward
Turing Research Fellow
Melanie Jans-Singh
Data Scientist, Zero Carbon Farms t/a Growing Underground
Mingda Yuan
Doctoral Student, University of Cambridge
