Incorporation of experts’ experience into machine learning models using well logs analysis for Priobskoye and Muravlenkovskoye brownfields

UDK: 550.8.072
DOI: 10.24887/0028-2448-2017-12-28-31
Key words: artificial intelligence, machine learning, litho-facies classification, sedimentological maps, well logs interpretation
Authors: D.V. Egorov, N.V. Bukhanov, O.T. Osmonalieva, B.V. Belozerov (Gazpromneft NTC LLC, RF, Saint-Petersburg), A.A. Reshytko, M.V. Golitsyna, A.S. Semenikhin (IBM Science and Technology Center, RF, Moscow)

Professional community of petroleum engineers and geoscientists is facing strong intervention form artificial intelligence methods; although there is still no common understanding on how exactly these methods will be integrated in exploration and production workflows. Geoscience domain is very sensitive to expert knowledge and experience. Therefore the main challenge of artificial intelligence solutions is to accumulate and assimilate expert opinions in order to reproduce decision making process. Incorporation of such a priori information into machine learning models is still a relevant research task.

Intelligent interpretation of well logs data is presented in this paper using expert knowledge of Gazpromneft NTC engineers. In particular analysis is presented for two brownfields: Muravlenkovskoye and Priobskoye using clustering and classification hierarchical approach. Accuracy of forecast using this methodology is increased due to incorporation of sedimentological maps based on geological concepts of the fields. Comparison of different machine learning algorithms is discussed and crucial role of a priori knowledge is focused.

High importance within the workflow belongs to iterative process of validation of sand fraction forecast in wells. Therefore, new thin beds, initially left out of scope, can be distinguished and spatial trend of effective thickness can be evaluated. Involving data from other oilfields would allow creating artificial digital technician which would be trained by expert in order to help in intellectual data interpretation.

References

1. Wickham H., Tidy data, The Journal of Statistical Software, 2014, Issue 10.

2. Borazjani O., Ghiasi-Freez J., Hatampour A., Two intelligent pattern recognition models for automatic identification of textural and pore space characteristics of the carbonate reservoir rocks using thin section images, Journal of Natural Gas Science and Engineering, 2016, V. 35, Part A, pp. 944-955.

3. Hall M., Hall B., Distributed collaborative prediction: Results of the machine learning contest, The Leading Edge, 2016, V. 36 (3), pp. 267–269.

4. Tengelidi D. et al., Fourier spectrums clustering for automated facies recognition of Field Y, Proceedings of 7th EAGE Saint Petersburg International Conference and Exhibition, 2016, DOI: 10.3997/2214-4609.201600253.


Attention!
To buy the complete text of article (Russian version a format - PDF) or to read the material which is in open access only the authorized visitors of the website can. .