An example of how to treat a complex fluvio-lacustrine reservoir, characterized by both continuous and discontinuous multi-scaled heterogeneities is presented. A detailed geological description was prepared, the heterogeneities were described and quantified, and discriminated in correlatable or discontinuous levels. The discontinuous levels were geostatistically modelled and up-scaled, whilst the correlatable ones were deterministically represented. The history matching process was very rapid, allowing greater time and confidence insensitivity studies. The whole study is clearly a multi-disciplinary project, involving geology, geostatistics, up-scaling and simulation.
During the second phase of field development, the reservoir department has to optimize the hydrocarbon reserves and to minimize the costs. A full-field or a local fluid-flow simulation model allow us to propose the best solution after running sensitivity studies on such a model after the history matching process. In these studies, which are generally carried out with a fixed budget, a rapid matching process allows a greater time for sensitivity studies and gives much greater confidence in the results.
The goal of this study was to build a 3D geological model in a complex fluvio-lacustrine environment, characterized by both continuous and discontinuous multi-scaled heterogeneities in a sandy matrix. The model is used to enable first the matching of five years production history in six vertical wells and a PLT for a horizontal well, and secondly to run horizontal well productivity studies as a function of different parameters, such as well spacing, well length and the productivity index.
Besides the heterogeneous sedimentary environment of the reservoir, the problem was also to resolve a complex geometry (unconformity, pinch-out)and production involving gas and water coning in a 46 meter oil column.
In order to address these problems, the geological heterogeneities were described and quantified at a fine scale in about 30 wells. The final model was a combination of deterministic elements (correlatable barriers) and stochastic heterogeneities. The geostatistically generated realization was up-scaled and simulated.
In the first simulation run, four (out of six) vertical wells were matched on pressure, GOR and WOR. Three local minor modifications to the deterministic petrophysical values of the layers enabled the history and PLT matching of all the wells by the fourth run.
The reservoir can be divided into two units (figure 1), with a complex lower fluvio-lacustrine section, several hundred meter thick, partially eroded by a thinner (5 to 10 meters) unconformable upper unit composed of homogeneous transgressive sands.
The lower unit is made up of a vertical succession of a 20 meter thick elementary sequence which shows from bottom to top fluviatile braided deposits, fluvial to flood plain sediments, and lacustrine shales corresponding to the distal part of an alluvial fan.
P. 83^
