In recent years, it has become increasingly important to develop hard-to-recover reserves concentrated in reservoirs with complex structures. Such reservoirs cannot be identified by standard open porosity boundary values due to the need to substantiate the pore space model, the porosity estimation model based on well logs and the boundary value for different types of reservoirs. Nuclear magnetic logging enables to identify all reservoir types in a section because its readings depend primarily on the free fluid content of the rock, and it directly estimates the effective porosity in the section. The article proposes a method of effective porosity modeling according to the data of radioactive complex of geophysical surveys of wells using machine learning methods, on the data of actual measurements of nuclear magnetic logging, in the section of the Achimov strata. The application of machine learning methods enables to take into account the variability of initial data and provides a more detailed description of the areas of log curve extrema compared to linear models. The peculiarity of the approach is the use of ensemble of neural network models and decision tree scaffolding. The approach enabled the authors to further identify reservoir intervals, re-evaluate effective thicknesses and consider the prospects of the studied strata.
References
1. Khanin A.A., Porody-kollektory nefti i gaza neftegazonosnykh provintsiy SSSR (Reservoir rocks of oil and gas of the USSR petroliferous provinces), Moscow: Nedra Publ., 1973, 304 p.
2. Bulgakov R.B., Issledovaniya geologicheskogo razreza skvazhin (Geological section studies of wells), Part 2, Ufa: Informreklama Publ., 2010, 240 p.
3. Dakhnov V.N., Geofizicheskie metody opredeleniya kollektorskikh svoystv i neftegazonasyshcheniya gornykh porod (Geophysical methods for the determination of reservoir properties and oil and gas saturation of rocks), Moscow: Nedra Publ., 1985, 310 p.
4. Coates G.R., Xiao L., Prammer M.G., Nuclear magnetic resonance. Principles and application, Houston: Halliburton Energy Service, 1999, 346 p.
5. Privalova O.R., Gadeleva D.D., Minigalieva G.I. et al., Well logging interpretation for Kashir and Podolsk deposits using neural networks (In Russ.), Neftegazovoe delo, 2021, V. 19, no. 1, pp. 69–76, DOI: http://doi.org/10.17122/ngdelo-2021-1-69-76
6. Mitchell T.M., Machine learning, McGraw-Hill, 1997, 414 p.
7. Hamada G., Ahmed E., Chao N., Artificial neural network (ANN) prediction of porosity and water saturation of shaly sandstone reservoirs, Advances in Applied Science Research, 2018, no. 8, pp. 26-31, URL: https://www.primescholars.com/articles/artificial-neural-network-ann-prediction-of-porosity-and-water-saturation-of-shaly-sandstone-reservoirs.pdf
