SPE Members
This paper presents a new approach to automated history matching. Instead of the usual, "mathematically-oriented" approach in which the strongly non-linear reservoir responses are locally linearized, this paper describes a "reservoir engineering-oriented" approach which attempts instead to use a combination of reservoir responses that is truly linear.
Two accurate linear models of a numerical reservoir simulator's pressure and production response to permeability chances are presented. These models, which are derived from the general hyperbolic decline curve relationship, are used in a history matching scheme to estimate the permeability distribution from historical production permeability distribution from historical production and pressure information. Three cases are presented which test the methods presented in presented which test the methods presented in matching for isotropic and anisotropic permeability distributions. The cases are based upon permeability distributions. The cases are based upon history generated from a 72 block 7 well model, and include both anisotropic and isotropic history matching problems.
A method of simultaneous determination of absolute and relative permeabilities is also presented. The method is based upon the isotropic presented. The method is based upon the isotropic linear permeability model developed and tested in the paper.
The history matching technique presented requires only two or three simulation experiments to accomplish the match. This is approximately one-tenth the CPU cost required for the optimal control theory automatic history matching methods described in the literature.
Reservoir simulation is widely used to determine a reservoir's production response to various recovery schemes. However, the accuracy of the results are directly dependent upon the ability of the numeric model to emulate the reservoir. The first requirement of any simulation is a precise description of reservoir parameters over a large number of grid blocks. In parameters over a large number of grid blocks. In most cases, field data does not permit such fine scale resolution of the parameters. Seismic and well log information are currently interpreted to estimate 2 reservoir's areal extent, thickness and porosity distribution; permeability and relative porosity distribution; permeability and relative permeability data from core and well tests may permeability data from core and well tests may also be incorporated into the model. Based upon this limited selection of data, the simulator model can then be tested by comparing simulated results to historical production data.
History matching is the process of selecting the properties of each gridblock in a simulator model that result in simulated well pressures and production data that match as closely as possible production data that match as closely as possible those measured during production. This process is generally accomplished with numerous simulation runs, each run evaluating a different set of model parameters. It can be time-consuming and costly. parameters. It can be time-consuming and costly. History matching is an inverse method of obtaining a reservoir description, as is well test analysis, which attempts to determine parameters such as porosity and permeability by matching a single-phase, homogeneous, radial flow model to the-test response (history). In history matching with numerical simulators, adjustments to the properties assigned to each block are made until a properties assigned to each block are made until a set of properties are found which allow the numerical model to reproduce the historical data.
P. 443
