Numerical Code Comparison Project - A Necessary Step towards Confidencein Geothermal Reservoir Simulators

Abstract

Numerous computer codes capable of solving the general geothermal reservoir engineering problem involving mutiphase flow of fluid and heat have been developed over the past several years. At the present time, the validity of results obtained using these simulators to model the performance and predict the response of real geothermal systems is the subject of concern both to those involved in developing and managing geothermal resources and to those involved with financing such enterprises. Do these codes accurately simulate the physical processes believed to be important? Can simulator users develop realistic conceptual models and appropriate values of input parameters to insure that numerical models of particular reservoirs yield valid results?

Experience gained in several attempts to model the Wairakei
geothermal reservoir in New Zealand (Mercer and Faust 1979; Garg, Rice, and Pritchett, 1979) demonstrates that problems involved in applying these codes to real systems are significant (Sorey and Fradkin, 1979; Donaldson and Sorey, 1979), and that results obtained from various models can be quite differences and in evaluating the usefulness of numerical simulators for geothermal reservoi r analysis is the comparison of simulator results for a set of well-specified problems involving processes applicable in reservoir analysis.

Under the direction of DOE'S Geothermal Reservoir Engineering Management Program (GREMP), a set of six test problems has been developed in an attempt to meet this need. The problem set covers a range of reservoir situations including single- and two-phase flow under 1, 2, and 3 dimensional conditions. Each problem has been test run to insure that the parameter specifications will yield workable solutions, and in several cases analytical solutions are available for comparison. Brief Experience gained in several attempts to model the Wairakei geo- A necessary first step in resolving such differ descriptions of the problems are given below in each problem, the desired grid and time-step sizes were specified to minimize differences in results due to numerical discretization.