TY - JOUR

T1 - Hybrid modeling of well killing fluid filtration in the conditions of fractured-porous reservoirs based on physico-mathematical modeling and machine learning

AU - Gumerov, R. R.

AU - Kalinin, S. A.

AU - Roshchektaev, A. P.

AU - Karmushin, S. R.

AU - Neverov, V. V.

AU - Kozhukhov, A. S.

AU - Katser, Yu D.

AU - Ippolitov, M. S.

AU - Kuchendaeva, E. M.

AU - Novikov, E. V.

AU - Besov, A. S.

AU - Mullyadzhanov, R. I.

AU - Golovin, S. V.

PY - 2024/12

Y1 - 2024/12

N2 - The purpose of this work is to increase the efficiency of well killing operations in carbonate fractured porous reservoirs with high gas factor, presence of hydrogen sulphide and abnormally low formation pressure. Various technologies are used to conduct well killing operations in such conditions, including those using injection in a certain sequence of different volumes of non-Newtonian viscoelastic and emulsion blocking compounds, as well as salt solutions in order to prevent oil, gas and water shows. This result is achieved by preventing the absorption of technological compositions into the bottomhole zone of the formation and providing back pressure by a column of fluid in the borehole to the formation. The major challenge is the difficulty in selecting the optimal composition and sufficient volume of well killing fluids, while ensuring a minimal number of unsuccessful operations. Hybrid modeling, which combines machine learning techniques with classical methods of physical and mathematical modeling, is chosen as a means to solve this problem. The hybrid approach enables to capture complex and non-intuitive dependencies in the data and to rely on the physical principles lying behind the mathematical models of fluid flow in fractured porous media. The developed models provide accurate calculation of the volumes of technical fluids necessary for successful operations, with an error range between 2 and 50 cubic meters depending on the specific technical fluid and the well in question. The coefficient of determination R2 reaches 0,7 which indicates a high level of accuracy in the regression models used in the calculations.

AB - The purpose of this work is to increase the efficiency of well killing operations in carbonate fractured porous reservoirs with high gas factor, presence of hydrogen sulphide and abnormally low formation pressure. Various technologies are used to conduct well killing operations in such conditions, including those using injection in a certain sequence of different volumes of non-Newtonian viscoelastic and emulsion blocking compounds, as well as salt solutions in order to prevent oil, gas and water shows. This result is achieved by preventing the absorption of technological compositions into the bottomhole zone of the formation and providing back pressure by a column of fluid in the borehole to the formation. The major challenge is the difficulty in selecting the optimal composition and sufficient volume of well killing fluids, while ensuring a minimal number of unsuccessful operations. Hybrid modeling, which combines machine learning techniques with classical methods of physical and mathematical modeling, is chosen as a means to solve this problem. The hybrid approach enables to capture complex and non-intuitive dependencies in the data and to rely on the physical principles lying behind the mathematical models of fluid flow in fractured porous media. The developed models provide accurate calculation of the volumes of technical fluids necessary for successful operations, with an error range between 2 and 50 cubic meters depending on the specific technical fluid and the well in question. The coefficient of determination R2 reaches 0,7 which indicates a high level of accuracy in the regression models used in the calculations.

KW - classification

KW - fractured-porous reservoir

KW - gradient boosting

KW - machine learning

KW - mathematical modeling

KW - non-Newtonian fluid

KW - regression

KW - rheological tests

KW - well killing

KW - well killing fluid

UR - https://www.mendeley.com/catalogue/2a807b78-d9fa-30d7-98a1-12f9e15ff048/

U2 - 10.24887/0028-2448-2024-12-46-52

DO - 10.24887/0028-2448-2024-12-46-52

M3 - Article

VL - 2024

SP - 46

EP - 52

JO - Neftyanoe khozyaystvo - Oil Industry

JF - Neftyanoe khozyaystvo - Oil Industry

SN - 0028-2448

IS - 12

ER -