Two fundamental issues in any mature oilfield re-development are to locate remaining mobile oil and optimising placement of new wells. In offshore East-Malaysia Field BR, infill drilling has been the preferred and economically-attractive option to increase oil recovery, due to its operational simplicity, low risk and historically good results. A new integrated workflow was implemented to identify the lowest risk i.e best locations for infill wells in these multi-layer mature S-Sands oil reservoirs, which have been producing since 1969. Two approaches were integrated to quantify understanding of attractive remaining oil targets in the complex, highly commingled S-sands with 200+ layers spanning across 5 connected fault-blocks.
First, a machine-learning ensemble-based approach which integrated static and dynamic uncertainties was utilized to condition multiple geologically-plausible, equiprobable history-matched cases. Realism & risk was integrated by multi-disciplinary uncertainty-based workflow incorporating all reservoir uncertainties along with appropriate suite of analogs. This was balanced by a complementary approach using classical RE methods which manually integrated multiple methods. The approach helped identify likely locations of remaining oil by incorporating reservoir maps and correlations with current fluid contacts (from reservoir performance and logging data), pressure trending, production bubble maps, surveillance data, decline curve analysis, etc. RTA was included as part of connected volume analysis to ascertain the presence of undrained accumulation and connectivity between fault-blocks. Empirical techniques rely heavily on hard data and volumetric calculation of OOIP, however, impact of reservoir heterogeneity and continuity is at times overlooked.
Since each approach has merit, best outcomes are achieved by integrating these 2 approaches. All simulation results are rigorously checked against well performances and hard data. For some reservoirs, it was recognized that dynamic model falls shorts due to multiple reasons. In such cases, relying on classical RE is a handy method to review and identify potential unreliable estimates. A merge between classical RE and machine-learning ensemble-based modeling in this study had significantly reduced turnaround time for infill target evaluation and optimization.
This study identified, ranked S-sands undrained and appraisal targets. A total of 12 infill wells were identified from the ensemble of dynamic models. Post high-grading and derisking the simulation work using classical RE methodology, 5 high-certainty infill wells with largest incremental reserves were put forth for maturation. Henceforth, an infill redevelopment consisting of 5 accelerated wells planned through the utilization of existing facilities is proposed. This paper describes the workflow and integration of reservoir simulation and classical RE approaches to finalize infill locations and describe associated risks.
