The paper intends to share the ideas and methodology on how to improve history matching and manage uncertainties by integrating the deterministic method and stochastic Experimental Design (ED) approach in the most practical way. The integrated method was applied in the re-development study of an offshore clastic oil Field. In the conventional deterministic approach, history matching is conducted by manually tuning the key parameters to achieve pressure and saturation matching iteratively. The main advantage of the deterministic approach is in providing the opportunity to understand the details of the well and reservoir local behaviors during the history matching exercise. The disadvantages are the workflow yields only a single outcome hence poses difficulty in quantifying the effect of multiple uncertainties and is dependent on the of engineer's subjective. Meanwhile, history matching stochastic workflow assisted by ED is a more systematic procedure built to capture the effect of uncertainties on the range of outcomes. However, the parameters used in the stochastic process is typically of global scale, thus the engineer may lose insight into the localized reservoir behaviors. Furthermore, proxies built in ED, although statistically consistent; do not always conform to the reservoir physics as most of the reservoir mechanisms are too complex for reliable modeling by ED-based proxy equations. Candidate Field re-development study aimed to further improve recovery by infill drilling. Analysis of well production data suggested that the remaining oil was entrapped to the crestal region of the field. Full field simulation work was then conducted to target and quantify the infill opportunity. The integrated approach began with a structured deterministic history matching adopting the well-known stratigraphic method. The deterministic history matching involves mostly local modifications of the aquifer strength, fault sealing and well's PI multiplier to achieve satisfactory match on the well and reservoir level. This was then used as the starting base case for the ED workflow. ED History Match led to multiple best matched cases, which were then run in forecasting mode and yielded a range of outcomes. ED History Matching also proved to improve the match quality of the initial deterministic history matched model. Sweet-spotting exercise was performed based on multiple History Matched models and suggested 3 infill potential locations in the major reservoirs, providing three concepts of infill drilling: unpenetrated fault blocks, crestal oil accumulation and sub-optimum drainage area. After considering the risk and confidence level of each potential location, the crestal oil accumulation was selected as the most attractive. Having multiple best match models helped to provide the range of incremental UR that is consistent with the range of uncertainties. The study demonstrated that integration of deterministic and stochastic history matching method has been instrumental in managing uncertainties and identifying redevelopment opportunities of a mature waterflood field.

The field under the study is located offshore Sabah, Malaysia. It is a mature field with 55 wells, and has been on production for more than 30 years. Despite continual efforts to improve the reservoir description and subsequent placement of infill wells, the average ultimate recoveries of the production wells in the field remain stubbornly low. In terms of reservoir continuity, the field represents the "Perfect Storm" in that the reservoir sands are thin, interbedded with shales, and appear to be extensively faulted. Furthermore, the problems of characterization are exacerbated by poor seismic imaging over the crest of the field, and an unreliable well-to-well correlation due to a lack of clearly distinguishable marine shales. As a structural modelling subject, the field represents a complex problem which exposes the limitations of pillar gridding for geo-cellular model construction. The recovery factor for the field stands at 9% indicating significant recoverable reserves remain, driving the need to better understand the reservoir geology and how this relates to production behavior. A novel reservoir modelling approach was required which would retain enough granularity in the model grid to represent the complex structural & stratigraphic compartmentalization, but without the penalty of excessive simulation run times. The complex stacked stratigraphy and fluid distribution was delineated through integrated analysis of the seismic (in so much as this was possible), the fluid contacts at the wells and the reservoir pressure data, resulting in an initial model of 835 discrete structural/stratigraphic compartments. Within this structural framework, stochastic modelling of net sand facies was used to create the heterogeneous sands and shale which form the fabric of the reservoir which reproduced the baffled connectivity required to emulate the severe pressure depletion and subsequent recovery that is commonly seen. Oil in place volume estimates were re-confirmed by the history matched dynamic model via experimental design, which also resulting multiple best match cases. The multiple static and dynamic models permitted a consistent approach to identifying infill drilling targets and assessing redevelopment feasibility, and the associated uncertainties. With the focus on regions that are currently inaccessible from the existing infrastructure, the study ultimately recommended the location of two new drilling platforms to provide optimal access to the remaining oil. The novel reservoir modeling, which includes the integration of seismic, fluid contact and pressure data to better define reservoir correlation and compartmentalization, was successfully applied to quantify the size of the prize of this highly compartmentalized reservoir. The experimental design approach was then instrumental in managing the principle uncertainties in a consistent way to develop a range history matched subsurface models used to identify future development options.

No abstract