Comparison of Deterministic and Probabilistic Simulation Models of Channel Sands in the Statfjord Reservoir, Brent Field

Abstract

Listen

Translate article

ABSTRACT The Brent Field has an estimated STOIIP of 3,440 MMstb and a GIIP of 6.9 Tscf. The field is operated by Shell on behalf of a Shell – Esso joint venture. It consists of two separate reservoirs, the Brent and Statfjord reservoir. Both reservoirs contain light oil, overlain by condensate rich primary gas caps. Definition of an optimal long-term development plan for the Brent Field required the construction and application of the so-called Brent Full Field Simulation Model. A key issue in the construction of reservoir simulation models of the field is the description of the reservoir geology. The description of channel dominated sands in the Statfjord reservoir was anticipated to pose major difficulties in the history matching process. Therefore, the merits of using more detailed geological models, in particular probabilistic models, were investigated. Well data were used to generate three probabilistic models of the geological most complex middle reservoir units of a part of the Statfjord reservoir. To incorporate vertical crossflow, these models were embedded between the deterministically modelled over- and underlying reservoir units. Extensive software was developed to translate the extremely detailed models (>150,000 blocks) into coarser simulation models. To allow for a valid comparison between the probabilistic approach and the conventional deterministic approach, a completely conventional model with identical grid definition was built for the same part of the reservoir. Simulation of the displacement characteristics revealed that the incorporation of the more detailed geology led to a substantially improved description of the historical well performances. The implicit definition of shales in the probabilistic models proved to be an important characteristic: the production performance in the over- and underlying non-probabiiistically modelled layers was significantly affected by the inclusion of the detailed geology in the probabilistically modelled layers. In view of the promising results, the use of more detailed geological models in future generations of the Brent Full Field Simulation Model is being reviewed.