Skip to main content

Physics informed machine learning applied to geothermal energy

Academic lead
Phil Livermore, Earth and Environment
Amir Khan, Civil Eng, Chrysothemis Paraskevopoulou​, Earth and Environment, Sandra Piazolo​, Earth and Environment
Project themes
Energy and Transport, Environmental Flows, Underpinning Methods for Fluid Dynamics

Decarbonising space heating is a central focus of the UK’s goal to attain net zero emissions by 2050. One promising source of zero-carbon heat is geothermal energy derived from water-stored energy in abandoned mines, that is particularly applicable in areas such as Leeds with a large number of abandoned mining works. This project seeks to model the flow of heat within the complex geometries of water-filled mine workings, using physics informed machine learning techniques. In these methods, a model which is typically based on a neural network is fit not only to a sparse data set and the geometry, but the physical equations are imposed as additional constraints often with unknown material parameters. The aim is to use such methods not only to quantitatively assess the viability of mine-water heating but also to provide insights into the physical processes of heat flow gleaned through inference of the unknown material parameters.