Probabilistic history matching of dual-porosity and dual-permeability korolev model using three discrete fracture models

Tengizchevroil (TCO) operates two giant carbonate oil fields, Tengiz and Korolev, located on the northeast shore of the Caspian Sea in Kazakhstan. Both fields are Middle Devonian to Upper Carboniferous isolated carbonate platforms. Since 1991, the fields have produced around 2B barrels of oil. As more geologic and dynamic data becomes available, an updated history match of the existing dynamic field models is required to provide more accurate simulation results for estimating reserves, optimizing production and assessing future field development opportunities. In this paper, we present a case study on the use of brown-field design of experiments (DoE) on dual-porosity and dual-permeability Korolev field simulation model. Korolev is a highly fractured reservoir, and the physics of fluid movement is mainly controlled by the fracture network. In order to capture the uncertainty in the extent of fracture region at Korolev, discrete low, mid and high fracture models were created. The history matching process was split into three separate DoE studies, one for each of the fracture models, to obtain good quality proxies. After history matching the three fracture models separately, a combined proxy was created, and the final probabilistic P10, P50, P90 models were selected from the suite of all low, mid and high fracture extent models. The history matching workflow consisted of selecting uncertainty parameters and ranges which honor geological data, identifying parameters having high impact on the history match quality, conducting brown-field DoE of historical and prediction periods, developing proxies for EUR objective functions and history match mis-match functions, and model selection. The majority of history matching effort was spent on the static well pressure match, followed by minimizing the modular dynamic test (MDT) pressures, production logging tool (PLT) profiles and water cut mis-match functions. Approaches in key areas which helped to improve the quality of Korolev history matched models will be discussed in detail. The history matching workflow with low, mid and high fracture models described in this paper is believed to be superior to approaches that use a single fracture realization with fracture porosity, fracture permeability and sigma (fracture-matrix interaction term) history match modifications within the simulator. This is because the use of three fracture realizations allows for adjustment of intrinsic fracture properties (fracture density, aperture size, fracture extent, etc.). In addition, the use of three separate DoE studies allows for more accurate proxy model development.