Use of Inverse Modeling for Conditioning Geostatistical Models to Vertical Tracer Profiles

Abstract

Abstract A new tracer test for estimating reservoir heterogeneity called vertical tracer profiling (VTP) is presented. This interwell tracer test is performed by injecting and/or collecting tracer samples at different locations along the wellbore. The problem in analyzing this and conventional interwell tracer data is finding the solution to an inverse problem, that is, the inference of the reservoir heterogeneity from the tracer production data. In this paper two inverse modeling techniques are presented and compared in terms of speed and quality of results. The first technique uses a tracer simulator as a forward model in a simulated annealing geostatistical algorithm for conditioning the generated model to tracer history. The second technique uses a least square history matching technique applied to the tracer history coupled with a geostatistical modeling method to solve the inverse problem. These techniques were used to investigate the effectiveness of vertical tracer profiling for determining reservoir zonation compared to traditional sampling of the tracer data. Both 2D and 3D models were used to test the efficiency of the inverse modeling algorithms, the use of tracer data for estimating horizontal correlation, and the advantages of the vertical tracer profiling. The 2D permeability field used in this study was measured on a sample of Antolini sandstone and the 3D permeability field was generated using a Gaussian simulation technique for a fining upwards geologic sequence. The geostatistically generated 2D models were tested by comparing model simulation results with experimental waterflood. The 3D models were also compared using waterflooding simulations.