Bayesian inversion of tilt data using a machine-learned surrogate model for pressurised fractures
Saeed Salimzadeh
A * , Dane Kasperczyk A and Teeratorn Kadeethum B
A
B
Saeed Salimzadeh is a Senior Research Scientist at Subsurface Engineering and Technology team at CSIRO Energy, Clayton, Australia. He obtained his PhD in Geomechanics at University of New South Wales (UNSW Sydney) in 2014, and since has been working in international research institutes at Imperial College London; DTU, Denmark; and CSIRO, Australia. Saeed is an expert in reservoir geomechanics and hydraulic fracturing, through numerical modelling, machine learning, and inversion. He has developed the hydraulic fracturing simulator CSMP-HF and has supervised many master and PhD students. |
Dane Kasperczyk is a senior engineer working in the Energy Resources group at CSIRO, he holds degrees from University of Melbourne in Civil Engineering and Science (Earth Sciences). For the past decade he has worked on research and projects related to fracture mechanics modelling, hydraulic fracturing environmental risk probabilities and preconditioning for block cave, sublevel cave, and underground mines. Through this he has developed capability in subsurface monitoring using tiltmeters that has seen applicable use for mining, CO2 sequestration and subsurface energy storage. |
Teeratorn Kadeethum earned his bachelor’s degree in mechanical engineering from Chulalongkorn University in Thailand (2007). He then obtained a master’s degree in chemical engineering from the University of Calgary in Alberta, Canada, (2016), followed by a PhD in Applied Mathematics and Computer Science from the Technical University of Denmark (2020). Following his PhD, he was a postdoctoral associate in mechanical and aerospace engineering at Cornell University in New York, USA, from 2020–2021. He is currently a senior technical staff member at the Climate Change Security Center at Sandia National Laboratories in New Mexico, USA. Prior to this, he accumulated 4 years of industrial experience as a reservoir engineer at PTT Exploration and Production, an international oil company. Teeratorn has authored over 20 research articles, and his research interests include scientific machine learning, advanced finite element approximations, and model order reduction, particularly in the context of nonlinear partial differential equations. |
Abstract
We introduce an innovative inversion approach for deducing subsurface fractures through observations of ground surface tilt. We have constructed, evaluated, and applied a surrogate forward model, crafted using conditional Generative Adversarial Networks (cGAN), to forecast the tilts (displacement gradients) at the ground surface caused by subsurface fractures under pressure. Our findings indicate that this surrogate forward model accurately estimates the tilt vector at the surface resulting from the specified pressurised fracture. Even in complex scenarios involving multiple fractures at various depths, the model, which was initially trained on scenarios with single fractures at a fixed depth, performed well. Subsequently, we employed a Bayesian inversion algorithm to derive the optimised solution (the pressurised fracture) for a given set of surface tilt data, leveraging the surrogate forward model. The outcomes demonstrate that the inversion process with the surrogate model is both effective and significantly faster compared to the traditional finite element model that generated the training data.
Keywords: differential evolution, energy storage lenses, generative adversarial networks, ground surface monitoring, inverse analysis, machine learning, surrogate model, tiltmeters.
Saeed Salimzadeh is a Senior Research Scientist at Subsurface Engineering and Technology team at CSIRO Energy, Clayton, Australia. He obtained his PhD in Geomechanics at University of New South Wales (UNSW Sydney) in 2014, and since has been working in international research institutes at Imperial College London; DTU, Denmark; and CSIRO, Australia. Saeed is an expert in reservoir geomechanics and hydraulic fracturing, through numerical modelling, machine learning, and inversion. He has developed the hydraulic fracturing simulator CSMP-HF and has supervised many master and PhD students. |
Dane Kasperczyk is a senior engineer working in the Energy Resources group at CSIRO, he holds degrees from University of Melbourne in Civil Engineering and Science (Earth Sciences). For the past decade he has worked on research and projects related to fracture mechanics modelling, hydraulic fracturing environmental risk probabilities and preconditioning for block cave, sublevel cave, and underground mines. Through this he has developed capability in subsurface monitoring using tiltmeters that has seen applicable use for mining, CO2 sequestration and subsurface energy storage. |
Teeratorn Kadeethum earned his bachelor’s degree in mechanical engineering from Chulalongkorn University in Thailand (2007). He then obtained a master’s degree in chemical engineering from the University of Calgary in Alberta, Canada, (2016), followed by a PhD in Applied Mathematics and Computer Science from the Technical University of Denmark (2020). Following his PhD, he was a postdoctoral associate in mechanical and aerospace engineering at Cornell University in New York, USA, from 2020–2021. He is currently a senior technical staff member at the Climate Change Security Center at Sandia National Laboratories in New Mexico, USA. Prior to this, he accumulated 4 years of industrial experience as a reservoir engineer at PTT Exploration and Production, an international oil company. Teeratorn has authored over 20 research articles, and his research interests include scientific machine learning, advanced finite element approximations, and model order reduction, particularly in the context of nonlinear partial differential equations. |
References
Arjomand E, Salimzadeh S, Mow WS, Movassagh A, Kear J (2024) Geomechanical modelling of ground surface deformation induced by CO2 injection at In Salah, Algeria: Three wells, three responses. International Journal of Greenhouse Gas Control 132, 104034.
| Crossref | Google Scholar |
Bjørnarå TI, Bohloli B, Park J (2018) Field-data analysis and hydromechanical modeling of CO2 storage at In Salah, Algeria. International Journal of Greenhouse Gas Control 79(October), 61-72.
| Crossref | Google Scholar |
Bunger AP, Lau H, Wright S, Schmidt H (2023) Mechanical Model for Geomechanical Pumped Storage in Horizontal Fluid-Filled Lenses. International Journal for Numerical and Analytical Methods in Geomechanics 47, 1349-1372.
| Crossref | Google Scholar |
Gunning J, Ennis-King J, LaForce T, Jenkins C, Paterson L (2020) Bayesian well-test 2D tomography inversion for CO2 plume detection. International Journal of Greenhouse Gas Control 94(December 2019), 102804.
| Crossref | Google Scholar |
Jeffrey RG, Chen Z, Mills KW, Pegg S (2013) Monitoring and measuring hydraulic fracturing growth during preconditioning of a roof rock over a coal longwall panel. ISRM International Conference for Effective and Sustainable Hydraulic Fracturing 2013, 893-914.
| Crossref | Google Scholar |
Lecampion B, Gunning J (2007) Model selection in fracture mapping from elastostatic data. International Journal of Solids and Structures 44(5), 1391-1408.
| Crossref | Google Scholar |
Lecampion B, Jeffrey R, Detournay E (2005) Resolving the geometry of hydraulic fractures from tilt measurements. Pure and Applied Geophysics 162(12), 2433-2452.
| Crossref | Google Scholar |
Salimzadeh S, Kasperczyk D, Kadeethum T (2023) A surrogate model for predicting ground surface deformation gradient induced by pressurized fractures. Advances in Water Resources 181(May), 104556.
| Crossref | Google Scholar |
Singh H (2022) Hydrogen storage in inactive horizontal shale gas wells: Techno-economic analysis for Haynesville shale. Applied Energy 313(January), 118862.
| Crossref | Google Scholar |
