Machine learning used for reservoir pressure predictionValhall@AkerBP

Machine learning used for reservoir pressure prediction

Pressure prediction from 4D seismic can be improved by machine learning: an example from the Valhall Field.

At #DigEx 2019, the first conference about digitalization especially for the G&G community next week in Oslo, Cegal will show a case study of the use of machine learning on Aker BP’s Valhall Field.  In the Subsurface session Hilde Tveit Håland, senior geophysicist with Cegal will talk about “Using Machine Learning to Predict Pressure Changes in a Chalk Reservoir”.

Tveit Håland says that machine learning can help improve predictions of pressure and/or reservoir thickness from 4D seismic data. Accurate prediction of reservoir properties can help to define good drilling targets to maintain and grow production from the Valhall oil field. The field has an extensive 4D seismic database but using 4D seismic to understand reservoir properties has many challenges.

Challenges result from complex relationships between production and injection, reservoir pressures and saturations, and reservoir hardness and compaction. The 4D seismic is sensitive to both hardness and compaction. This adds a lot of uncertainty when used to either predict or interpret changes in reservoir pressures, saturations, and thicknesses.

By machine learning large datasets from a producing field can be combined using all relevant geological and geophysical data as a training model. This model can then be used to predict a set of target features, pressures in this case.

Isabel von Steinaecker, Geomodeller/Geologist with Cegal is one of the speakers in the Data Management and Visualization Session with a presentation about “Using Digitalization to Solve the Asset Data Challenge”.