Creating predictive maintenance models with machine learning
2 November 2018
EPCC is working with Edinburgh-based Oil & Gas SME Artificial Lift Performance (ALP) to develop a predictive model of electrical submersible pump (ESP) operations. Using machine learning, the project will develop models based on historical well data that will predict future ESP operating performance and define optimal intervention schedules.
The use of ESPs to maximise hydrocarbon recovery is widespread in the oil & gas industry but the pumps are notoriously difficult to manage, and the ability to predict performance and potential failure would be of huge benefit. ESP failure causes loss of production (downtime) plus equipment replacement costs and rig time to re-install a new pump. Consequently, being able to predict ESP failure is crucial as a planned, proactive pump replacement schedule is far more cost and time-efficient than dealing with an unexpected failure. Along with improved production and reduced operating costs, predictive modelling is expected to provide technology that will lead to the adaptation of related techniques to further improve devices and assets employed across the sector.
Sandy Williams CEO and ALP FounderOur goal is to predict impending failure of the pump by drawing on real-time data from hundreds of ESPs, which would be a game changer for oil companies.
Ian Phillips chief executive of OGICThere is a focus across the industry on moving to predictive maintenance as companies recognise the benefits it can bring to the sector. The ability to predict the need for repairs can reduce non-productive time on a project, which in return can deliver significant cost savings. Predictive maintenance also brings major safety benefits as it can determine the condition and performance of equipment before failure occurs. ALP’s development of a predictive model of ESP operations is an excellent example of how this technology can be implemented.
Funded by the Oil & Gas Innovation Centre (OGIC), this project started in October 2018 and will run until March 2019.