The global average Recovery Factor (RF) in oil fields is only about 20-40%. A possible reason for such a low RF might be that Enhanced Oil Recovery (EOR) techniques are not yet broadly applied. This could be for economic reasons, concerns regarding the effectiveness of EOR and potential damage to the reservoir, or the lack of reservoir-specific recommendations.

In this contribution, we introduce a methodology that selects EOR materials for specific reservoir conditions by using Artificial Intelligence (AI) methods. We investigate the consistency of the screening results with the results obtained by state-of-the-art techniques that are used to identify EOR methods only, i.e., without EOR material specificity.

Our method correlates physical and chemical representations of injection fluids, including EOR materials, with reservoir-specific information on lithology, porosity, permeability, as well as oil, water and salt conditions. We have used machine learning on the combined data set in order to provide recommendation for EOR cocktail for injection fluids. Reservoir specific data input on rock, oil, and water conditions available in well logs is transformed by the AI model into a reservoir-specific recommendation of EOR candidate materials for optimized EOR effectiveness. The screening criteria are ranked based on EOR effectiveness and the similarity of key reservoir parameters at pore scale.

Methodologically, a Naïve Bayes Classifier with 10-fold cross-validation over the full training data set classified all instances with an accuracy of up to 90%. In order to compare with the EOR method screening criteria typically used in the industry, we have created a test data set containing instances based on averaged parameter values for representing each EOR method. In this case, our method is capable of classifying the test data set with nearly 100% accuracy. Our methodology allows to produce recommendations for EOR cocktails, including concentrations of their chemical components, for specific reservoir conditions that are readily available through well logs.

This content is only available via PDF.
You can access this article if you purchase or spend a download.