--> Abstract: Geostatistical Quantification of Geological Information for a Fluvial-Type North Sea Reservoir, by Jef K. Caers; #90914(2000)

Datapages, Inc.Print this page

Jef K. Caers1
(1) Stanford University, Stanford, CA

Abstract: Geostatistical quantification of geological information for a fluvial-type North Sea reservoir

Accurate predictions of petroleum reservoir performance require reliable models characterizing the complex reservoir heterogeneity. These reservoir models have to be properly conditioned to all data, available at different scale/precision, and should be consistent with the underlying reservoir geology. Geostatistical simulation techniques generate multiple equiprobable reservoir models that depict the uncertainty due to lack of information. Traditional to geostatistics, geological continuity is represented through the variogram. The variogram is limited in describing complex geological structures as it measures correlation between rock properties at two locations only: it is a two-point statistic. Reservoir analogs such as outcrops can serve as training images depicting the interpreted geological structure. Due to scarcity of well data, the variogram models are often borrowed from these training sets. However, the same training images can be utilized to extract multiple point statistics, i.e. dependency between multiple locations. Recently, a neural network based approach was proposed for modeling and reproducing such multiple-point statistics in geostatistical simulations. This novel technique anchors the complex geological structure to the actual subsurface data. This paper applies the neural network algorithm on a complex fluvial-type North Sea reservoir. This 500 foot thick reservoir is characterized by a trend of upward increasing sandiness. Well-defined fluvial channels of sandstones embedded in mudstone matrix occur towards the base, while inter-stratified channels occur towards the top. The reservoir is informed by twelve vertical wells, producing over 600,000 barrels/day. Detailed, conceptual geological models are available as training sets. As a basis for comparison between the two stochastic approaches (neural based versus variogram based), we use one of the deterministic geological models, which is conditioned to the well data, as a reference set. All fine scale reservoir models have 300,000 grid cells. Flow simulation is performed on the geostatistical reservoir models produced with the neural network technique.

AAPG Search and Discovery Article #90914©2000 AAPG Annual Convention, New Orleans, Louisiana