The article deals with the problem of non-uniform optimal well placement automation in a field 3D hydrodynamic model. A brief overview is given to describe existing methods of well placement with multivariate resource-intensive calculations for finding optimal options. The task is to develop a new tool that is able to offer a non-uniform well placement for small and medium-sized fields in an acceptable time. Based on neural network algorithms and machine learning methods, smart assistant is developed in the form of a calculation tool for selecting optimal development option based on maximizing NPV. The tool accepts hydrodynamic model, conditions of development, economic parameters of the oil field as an input, and generates a schedule file (schedule section) of hydrodynamic model as an output according to the specified conditions: placement of a given number of wells, determination of well commissioning order, specifying or determination of well completion and hydraulic fracture parameters, selection of candidate wells for infill drilling, horizontal drilling, drilling of targeted injection wells, transfer to injection. The optimal NPV variant of the well placement is generated in a time comparable to hydrodynamic modeling without the participation of specialists. The methodology and results of the developed tool testing are presented. Comparison of the calculations results with the options recommended by the design and technical documentation for 12 fields in terms of cumulative oil production and NPV was performed. Possible directions of development of the considered tool are shown.
References
1. Mayorov K.N., Chebkasov D.S., Antipin D.V. et al., On the application of the Alpha Zero algorithm to optimize the placement of an irregular grid of production wells (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2021, no. 3, pp. 76-78, DOI: https://doi.org/10.24887/0028-2448-2021-3-76-78.
2. Silver D. et al., A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play, Science, 2018, V. 362, no. 6419, pp. 1140–1144, DOI:10.1126/science.aar6404
3. URL: https://rn.digital/rnkim/
4. Patent RU2783031C1, Method for oil reservoir development, Inventors: Karachurin N.T., Chebkasov D.S., Mayorov K.N., Vakhrusheva N.O.
5. Maiorov K., Vachrusheva N., Lozhkin A., Solving problems of the oil and gas sector using machine learning algorithms, Acta Montanistica Slovaca, 2021, V. 26, no. 2, pp. 327-337.