A machine learning aided high-fidelity urban building modelling approach for energy-efficient, low-carbon and climate-responsive buildings

Project ID: 2531ad1476

(You will need this ID for your application)

Research Theme: Energy and Decarbonisation

UCL Lead department: Bartlett School of Environment, Energy and Resources (BSEER)

Department Website

Lead Supervisor: Rui Tang

Project Summary:

Climate change presses the need to decarbonise buildings and empower them more resilient. Urban-building modelling can help understand the impact of energy-efficient and climate-responsive technologies across a large part of building stock; however, such tools are incapable of modelling the realistic and local urban microclimate and consequently have to be simplified with static and pre-generated weather files. This may cause modelling results to mislead the subsequent decision-making process. The United Nations’ Sustainable Development Goal reiterates that sustainable cities are expected to be resilient under extreme weather, but this modelling challenge will trouble the identification of vulnerable buildings of overheating under heatwaves and the development of effective mitigation strategies.

This project aims to develop a novel urban-building modelling approach to precisely model the effects of microclimate on urban buildings. The open-source data (e.g., LiDAR, satellite images) will be used to model the 3D morphology of urban buildings by deep learning techniques. The weather variables of urban-building model will be determined by an urban canopy microclimate model, which is formulated based on the identified urban-building morphology. A hybrid physics-machine learning modelling approach developed previously by the supervisory team will be used to establish the urban-building model and compute the dynamic response of individual buildings. The student will have access to the data of UCL new Here East campus, where the supervisory team is working on to develop a digital twin platform (Digibuild), to demonstrate the performance of the proposed modelling approach and analyse its ability to support practical scenarios under the current circumstance and future climate change for reducing energy consumption and mitigating overheating risks of buildings.
This project will benefit policymakers, local authorities, urban planners and building owners. The student will deepen the understanding on sustainable, low-carbon and resilient buildings and grasp skills in energy system modelling and machine learning techniques.