A team of engineers and computer scientists at Heriot-Watt University has developed algorithms that help the petroleum industry extract more oil, more cost-effectively. In one North Sea field, a leading oil company has used Heriot-Watt technology to create a field development plan yielding three million more barrels of oil than their original plan. A software tool using these algorithms offered by young spin-out Epistemy is attracting significant interest across the oil industry.

Global competition for depleting reserves impels oil companies to increase finding and recovery rates. It’s no surprise that uncertainty quantification (UQ) – the science of characterising and reducing uncertain outcomes – is of great interest to the world of oil exploration and production, where encountering multiple unknown variables is common.

The software is a practical and simple tool for history matching real field data and optimising field developments; it has given us a significant speed improvement over our previous algorithm
Divisional Director, JOGMEC

The UQ group in the Edinburgh Research Partnership in Engineering and Mathematics (Heriot-Watt and Edinburgh universities) has received public and industry funding to research into uncertainty quantification for oil reservoir modelling since 2000. The research has aimed to develop algorithms that improve on those routinely used in the oil industry. It is also focused on developing techniques to ensure that forecast uncertainty ranges are robust.

For company engineers at the sharp end of reservoir modelling to benefit, research outcomes have to be practical and usable. Computer code using ‘multiple-objective’ algorithms developed by the researchers has proved its worth on real-world problems. Not only that, it can also offer speed advantages over code using single objective algorithms.

Making these leading-edge developments available across the oil industry is Heriot-Watt spinout Epistemy. Its software tool Raven, when applied to a company’s reservoir simulator, produces enhanced predictive models, using past performance data for calibration (‘history matching’), and quantifying uncertainty. Launched in 2012, the software can already help companies optimise their well-placements and increase profits.