Predicted Results of Numeric Grid Models Compared with Actual Field Performance

Abstract

Two Alberta reservoirs-one a carbonate and the other a sandstonewere studied with a computer model and the results compared very favorably with field data accumulated later. An interesting by-product was the prediction of a gas cap that later was revealed by drilling in the sandstone reservoir. Introduction The reservoir simulation model is a relatively new tool, designed to assist the reservoir engineer in predicting reservoir performance. Despite its newness, predicting reservoir performance. Despite its newness, its significance as a predicting tool cannot be overstated. The rapid development of large digital computers has led to the widespread use of the reservoir simulator to solve secondary recovery problems in which fluid movement is very complex. The relative accuracy of various numerical techniques has been discussed, and tested in detail with theoretical solutions. The results of these tests have verified various numerical approaches to solving the partial differential equations governing fluid flow in partial differential equations governing fluid flow in porous media. The question we shall attempt to porous media. The question we shall attempt to answer here is "Will a reservoir perform as predicted by a grid-type model?"To answer this question we shall compare the results of a grid-type mathematical model with field observations for two large Alberta undersaturated oil reservoirs - a carbonate reservoir (the Kaybob Beaverhill Lake) and a sandstone reservoir (the Mitsue Gilwood). The model studies were carried out in May, 1965, and Jan., 1968, for the carbonate and sandstone, respectively. The comparisons are made on field data gathered since the studies and cover the following aspects of reservoir engineering: adjustment of measured rock and fluid properties; optimization of secondary recovery schemes; and maximization of sweep efficiencies. Although it has been only a few years since these predictions were made, their close agreement with predictions were made, their close agreement with field observations over this time interval gives a reasonable insight to the grid-type model's ability to accurately predict heterogeneous reservoir performance. performance. Model Description The studies of the reservoirs were divided into two main segments, a history-matching segment and a waterflood-prediction segment. History-Matching Segment The history matching involved determining the numerical values that best represented the physical properties of the reservoir system. This phase was properties of the reservoir system. This phase was accomplished by duplicating pressure-production behavior with a reservoir simulator, the TRAN III. This simulator is a mathematical model consisting of a set of equations that are solved simultaneously to give the transient pressure distribution of a reservoir-aquifer system. The equations are solved for as many as 4,000 distinct but interconnected cells representing the reservoir-aquifer configuration. Boundary conditions may be specified as "no-flow" or "constant-pressure". The equations used in solving the transient pressure distribution throughout the reservoir are the material-balance equation, the equation of state for the fluid, and the hydrodynamics force equation. JPT P. 1390