Geothermal energyProject Data Fusion and Machine Learning
Report: Data Fusion and Machine Learning for Geothermal Target Exploration and Characterisation (PDF 5MB)
This tool enables the user to build on the IT research that takes geothermal data from a wide variety of sources and uses modern statistical methods to analyse the data to assist with targeted geothermal exploration.
The ‘Data Fusion and Machine Learning for Geothermal Target Exploration and Characterisation’ proposed an ambitious objective: to fundamentally alter the approach to geothermal exploration, allowing fully probabilistic modelling of the subsurface to inform exploration and characterisation risks and decision making. To do so would require demonstrating several research firsts in the field of geophysics, collating and simultaneously interpreting data that had never previously been analysed together, and calling for new data to fill critical gaps when identified. The Measure aimed to encapsulate this research into a software system usable by industry, and to demonstrate this new methodology and implementation using a number of real case-‐studies of relevant regions in Australia.
Our success in achieving these goals relied on collaboration between ICT experts and statisticians at NICTA, geologists and geophysicists at a number of universities (Australian National University, University of Sydney, University of Melbourne, University of Adelaide), geothermal exploration companies (Geodynamics Ltd, Petratherm Ltd, Hot Rock Ltd) and State Government geological agencies (South Australia, Victoria).
The major product of the Measure was open source software that we have named ‘Obsidian’ (a volcanic rock fused at high temperature). The software allows simultaneous joint inversion of many geophysical and geological data sets, providing probabilistic outputs. The outcomes met or exceeded most objectives, presenting geothermal explorers and investors with a new paradigm for robust, probabilistic assessment of exploration and development risk.