PHD PROJECT: UNCERTAINTY QUANTIFICATION OF GEOMECHANICALLY SENSITIVE RESERVOIRS USING PHYSICS-INFUSED MACHINE LEARNING
What is the problem?
When we use the subsurface to store CO2 or Hydrogen, the rocks can deform inelastically as the reservoir conditions change – processes well described in geomechanics. But this permanent strain can lead to significant leakage and to induced seismicity risks during operation and/or long term storage. At worst, these risks pose a substantial threat to property and health – so much so that regulators and investors may halt projects before they start as we can’t correctly estimate the geomechanically-derived risks.
The key technical challenge for most storage reservoirs is to accurately quantify uncertainties and risks associated with geomechanical property changes from a few computer model forecasts. Geomechanical simulation software is complex and data-intensive, each run taking many hours to days. Such long run times make statistically thorough methods to quantifying uncertainty impractical. In most circumstances, we can’t afford the thousands of required model realisations. So such risks may be misestimated or even missed.
The longest model run times occur when we couple simulations of fluid flow (production and injection) with geomechanical simulations to predict how a development plan may alter the reservoir rocks properties and how this will impact fluid movement and field operations. To solve the system of equations for fluid flow and geomechanics together, we need to connect very different modelling approaches, the differences in the solvers typically precluding full coupling. Instead, the packages interact by running separately but simultaneously and passing data back and forth between iterations. This is a technically monumental challenge and the run times of the models are far longer than the combined times of each model run separately.
To accurately quantify uncertainties in geomechanically sensitive reservoirs we must run many more models than we do today, exploring a more diverse set of geological scenarios. But to run more models, we must significantly improve the efficiency of coupling fluid flow and geomechanical simulations. Machine learning provides one solution: once trained on an appropriate data set it can capture complex, non-linear systems very rapidly.
What you will do
As a PhD student on this project, you will explore how to speed up coupled geomechanical & flow simulation using novel machine learning, allowing better estimates of uncertainty. By training a machine learning algorithm on numerical models, we can forecast reservoir behaviour for unsimulated model outcomes at a fraction of the computational cost while preserving crucial underlying physics. Machine learning works well on many reservoir problems, from generating geological models to speeding up the solution of non-linear physical processes in fluid flow models but generally acts as a black box, losing the physics of the training set of simulations. We will explore how physics embedded machine learning tools, developed recently for other domains, can maintain good forecasts and ensure that forecasts are based on more than a sophisticated interpolation to the training data.
About the role
The project will entail you using fluid flow and geomechanical simulations, applying machine learning techniques, characterising the reservoir, and identifying key uncertainties and potentially some lab work. No one will have the perfect mix of skills (yet) so please apply if any of these skills fit your background.
The PhD is funded by an industry partner creating prospects for access to real data from real projects and interactions with industry experts. Based at the Institute of Geoenergy Engineering at Heriot-Watt University, the student will join a wide group of researchers working on the energy transition, such as the GeoDataScience group specialising in the use of machine learning and the Geomechanics Group who are ALERT members.
Who we want
We are looking for someone with some of the following attributes:
- A numerate geomechanics, geoscience, engineering or physics graduate or any graduates who have experience with machine learning. At least a 2:1 degree or equivalent at BSc level but ideally a relevant postgraduate degree
- Strong spoken and English language
- Excellent communication skills in writing and presentation
- Knowledge of MS office
- Knowledge of reservoir simulation and geomechanical simulation tools prefered
- Machine learning knowledge is highly desirable
- Great team working and collaboration skills
- Be open and interested, keen to learn and enthusiastic. Research is fun!
Funding
A full maintenance grant for 3 years. Fees are paid for UK residents. Fees are paid for the full 3 years for EU residents but for a formal start before August 2021. Start from your home country due to Covid travel restrictions is allowed. International students are welcome to apply but we cannot cover fees. The studentship is open until filled.
How to apply
Interested? The information will appear on the Heriot-Watt University website shortly:
If you have questions please contact the research team directly:
Dr Dan Arnold : D.p.arnold@hw.ac.uk
Dr Helen Lewis: h.lewis@hw.ac.uk (ALERT member)
Prof Vasily Demyanov: V.Demyanov@hw.ac.uk