The article discusses the technology of automatic pattern recognition of standard GR logging graphs with the aim of quickly building facies models of terrigenous depositions. Well logs provide information on the change in the particle size distribution of sediments over time and, depending on the sedimentation environment, are characterized by individual forms. This allows us to preliminarily determine the genesis of sediments and establish their facies affiliation with its further refinement by core materials.
The algorithm for pattern recognition of well logs is developed using neural networks on the example of a delta reservoir, which includes two standard facies - channel and bar facies with cylindrical and funnel-shaped forms of GR, respectively. Neural network training was carried out on a control sections consisting of 15 wells, for which different facies sand bodies of the delta reservoir were identified by an expert sedimentologist.
To increase the reliability of the recognition of logging forms, additional procedures have been used into the operation of the neural network, including: 1) correction of the training library (transfer of part of non-standard curves to the class of uncertain ones); 2) expanding the training library by adding several smoothed options to the existing log images; 3) simplification the way images are stored - instead of color coded images in a neural network, black and white began to be used, which accelerated its training.
The developed technique of automated pattern recognition of logging curves will optimize the processes of sedimentological analysis and will ensure the construction of a realistic geological model of the field with a facies options. Its use will make it possible to quickly identify of individual sand bodies with the spread of reservoir properties within the boundaries of each facies, which will allow for spatial monitoring of the geological heterogeneity of the development objects.
References
1. Rykus M.V., Rykus N.G., Sedimentologiya terrigennykh rezervuarov uglevodorodov (Sedimentology of terrigenous hydrocarbons reservoirs), Ufa: Mir pechati Publ., 2014, 324 p.
2. Muromtsev V.S., Elektrometricheskaya geologiya peschanykh tel – litologicheskikh lovushek nefti i gaza (Electrometric geology of sand bodies - lithological traps of oil and gas), Leningrad: Nedra Publ., 1984, 260 p
3. Reading H.G., Sedimentary environments: processes, facies and stratigraphy, Blackwell Publishing Limited, Second edition, 1986.
4. Baykov V.Ya., Bakirov N.K., Yakovlev A.A., Matematicheskaya geologiya (Mathematical geology), Part 1. Vvedenie v geostatistiku (Introduction to geostatistics), Moscow – Izhevsk: Publ. of Institute of Computer Science, 2012, 228 p.
5. Voskoboynikov Yu.E., Gochakov A.V., Kolker A.B., Fil'tratsii signalov i izobrazheniy: fur'e i veyvlet algoritmy (s primerami v Mathcad) (Signals and images filtration: Fourier and wavelet algorithms (with examples on Mathcad)), Novosibirsk: Publ. of SIBSTRIN, 2010, 188 z.
6. Baykov V.A., Afanas'ev I.S., Sergeychev A.V., Asmandiyarov R.N., Ivashchenko D.S., Sakhanenko A.I., Automatic well test data processing: a time series wavelet analysis approach (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2012, no. 11, pp. 34-37.
7. Douglas D., Peucker T., Algorithms for the reduction of the number of points required to represent a digitized line or its caricature, The Canadian Cartographer, 1973, no. 10(2), p. 112 – 122.
8. LeCun Y., Bottou L., Haffner P., Gradient-based learning applied to document recognition, Proceedings of the IEEE, 1998, V. 86(11), pp. 2278 - 2324.
9. Nikolenko S., Kadurin A., Arkhangel'skaya E., Glubokoe obuchenie (Deep Learning), St. Petersburg, Piter Publ., 2018, 480 p.
10. Vorontsov K.V., Mashinnoe obuchenie (Machine learning), URL: http://www.machinelearning.ru/wiki/images/0/0d/Voron-ML-Compositions.pdf
