TY - JOUR

T1 - Methodology for selecting analogs of reservoir fluid PVT models and rapid estimation of PVT parameters for new assets

AU - Kim, V. V.

AU - Matroshilov, N. O.

AU - Pechko, K. A.

AU - Afanasev, A. A.

AU - Simonov, M. V.

PY - 2023

Y1 - 2023

N2 - In the oil and gas industry in the process of field development it is an urgent task to create PVT models capable of describing changes in reservoir fluids in such nodes as reservoir, well and surface gathering and transport net-work. The cost of error in the PVT model is very high and at facilities with different types of oil, the planned NPV for the year may not reach the eco-nomic limit of 0.5-2.9%. Therefore, it is important to reproduce the properties of hydrocarbon mixtures reliably already at the early stages of field de-velopment. The use of high-quality PVT models early in field development will also reduce the cost of additional fluid testing and analysis, as the models can provide sufficiently accurate data for decision making. A characteristic feature of new assets is the lack of laboratory fluid results required for PVT modelling. In such cases, the value from such oil and gas projects car-ries a high degree of uncertainty, and the process of making important strategic decisions takes a long time. To solve this problem it is proposed to implement a completely new approach in the selection of PVT model analogues and operational creation of PVT model of Black Oil using machine learning algorithms, as well as the creation of a unified database of created PVT metamodels. This approach will allow the engineer to solve the problem with PVT section in the reservoir simulation model in an opera-tive mode and at the same time retain a high degree of its predictive ability.

AB - In the oil and gas industry in the process of field development it is an urgent task to create PVT models capable of describing changes in reservoir fluids in such nodes as reservoir, well and surface gathering and transport net-work. The cost of error in the PVT model is very high and at facilities with different types of oil, the planned NPV for the year may not reach the eco-nomic limit of 0.5-2.9%. Therefore, it is important to reproduce the properties of hydrocarbon mixtures reliably already at the early stages of field de-velopment. The use of high-quality PVT models early in field development will also reduce the cost of additional fluid testing and analysis, as the models can provide sufficiently accurate data for decision making. A characteristic feature of new assets is the lack of laboratory fluid results required for PVT modelling. In such cases, the value from such oil and gas projects car-ries a high degree of uncertainty, and the process of making important strategic decisions takes a long time. To solve this problem it is proposed to implement a completely new approach in the selection of PVT model analogues and operational creation of PVT model of Black Oil using machine learning algorithms, as well as the creation of a unified database of created PVT metamodels. This approach will allow the engineer to solve the problem with PVT section in the reservoir simulation model in an opera-tive mode and at the same time retain a high degree of its predictive ability.

KW - PVT

KW - analogs

KW - fluid

KW - greenfield

KW - integrated model

KW - machine learning

KW - metamodel

KW - reservoir simulation model

UR - https://www.scopus.com/record/display.uri?eid=2-s2.0-85199284110&origin=inward&txGid=a3e1d7720fee96e79bb4674ad37973d5

UR - https://www.elibrary.ru/item.asp?id=60365650

UR - https://www.mendeley.com/catalogue/6b42cd39-63c1-3a61-b200-40911a1563b6/

U2 - 10.24887/0028-2448-2023-12-36-39

DO - 10.24887/0028-2448-2023-12-36-39

M3 - Article

VL - 2023

SP - 36

EP - 39

JO - Neftyanoe khozyaystvo - Oil Industry

JF - Neftyanoe khozyaystvo - Oil Industry

SN - 0028-2448

IS - 12

ER -