Extending reservoir property prediction with pseudo-wells

Abstract

We discuss a procedure called the ‘Hit Cube’ which assigns spatial positions to stochastically simulated pseudo-wells, i.e., 1D stratigraphic columns with attached well logs, with the aim of quantifying probable reservoir properties and their quality throughout a seismic data volume. We generated two sets of pseudo-wells (‘hit targets’, e.g., gas-filled units; and ‘false hits’, e.g., brine-filled units) using real well data, geological knowledge and Monte Carlo statistics. From these models, we constructed scaled synthetic seismic traces using a composite statistical wavelet extracted from the seismic data. We then matched the models with the real traces throughout the seismic volume. At any given location within the seismic cube, we assume that the rock properties defined in the pseudo-wells are similar to those represented by the seismic data if the similarity between the real and synthetic traces satisfies a predefined hit criterion. Final results of this process are likelihood cubes, which can easily be interpreted or used for further analysis. We have used this technique to correctly predict hydrocarbon presence and distribution within a deltaic sequence containing gas-filled and brine-filled channel sand deposits.