51³Ô¹ÏÍø

Research to improve the efficiency of geothermal drilling has led to freely available software for optimisation challenges that involve uncertainty.

New software for optimisation under uncertainty has been developed from research in 51³Ô¹ÏÍø’s Department of Computing, led by Johannes Wiebe, who worked with Schlumberger New Energy as part of his PhD to help optimise drilling operations for sustainable geothermal energy.

“This has truly been an ideal collaboration”, said Dr Inês Cecílio of Schlumberger New Energy. “Johannes understood the problem, and adapted well to the dual challenge of developing academic theory and practical applications to solve an industry-wide issue."

Geothermal wells require specialised drilling, often through particularly hard rocks. When drilling a new well, there is a risk that the drill bit may break and need to be replaced. Engineers do not always have complete knowledge about rock composition or how quickly a drill bit will degrade, but they do have (imperfect) predictions.

Johannes worked with colleagues at 51³Ô¹ÏÍø and Schlumberger to define a complex mathematical problem of interest to both academic and industry communities. He then went through an iterative process of developing his theory, and applying it to Schlumberger’s needs. This led to an optimisation method for drilling in the face of incomplete information, trading off the time cost of drilling more slowly against the reduced risk.

The solution Johannes offered was nuanced, but in summary involved drilling quickly at the beginning, accepting that the drill bit may have to be replaced. Then drilling more slowly at the end when additional caution could prevent equipment failure at a stage when it’s more expensive to fix. Johannes’ work with the Schlumberger team has led to a published piece of open source software as well as an and a .

Play video

“As part of my studentship, I got to spend time in the Schlumberger office in Cambridge”, said Johannes. “I enjoyed the office atmosphere – it enabled different kinds of conversations from the ones I would usually have at university. I appreciate the opportunities I received to present my work to my industrial collaborators several times. They gave me detailed, immediate and thoughtful feedback that helped me check my thinking and research in real time.”

Johannes subsequently realised that the tools he had developed for geothermal energy might also solve broader optimisation problems involving uncertainty.

Algorithms for optimising under uncertainty

Johannes subsequently realised that the tools he had developed for geothermal energy might also solve broader optimisation problems involving uncertainty.

Significant research in recent years, at 51³Ô¹ÏÍø and elsewhere, has explored algorithms for optimising decisions in uncertain conditions. But it can be difficult for practitioners to try out the many state-of-the-art algorithms. In response to this, Johannes has gone on to develop and release  – a new online Python software package freely available to everyone. This generalises his collaboration with Schlumberger by modelling a way of representing optimisation under uncertainty, known as robust optimization, using the algebraic modelling language Pyomo. Johannes also describes the ROModel tool in another recent .

Play video

An ideal collaboration

“This has truly been an ideal collaboration with Johannes Wiebe and 51³Ô¹ÏÍø’s ”, said 51³Ô¹ÏÍø alumnus Dr Inês Cecílio, Manager of Digital Automation Systems at and one of Johannes’ collaborators on the project. “Johannes understood the problem, and adapted well to the dual challenge of developing academic theory and practical applications to solve an industry-wide issue. We are delighted that his software packages are now available to everyone and I look forward to seeing them being used to improve decision-making in all sorts of uncertain environments.”

Johannes’ work is funded by an EPSRC Industrial CASE studentship. He is a member of the , a PhD training program focusing on the multi-disciplinary aspects of computer science research.