Estimation of Uncertainty of a Volumetric Reservoir Model and Optimization of Numerical Inversion in Well Log Data Interpretation

Abstract

This study considers an approach to joint inversion of logging data in oil and gas wells for finding the volume fractions of reservoir components. The inversion is reduced to a linear least squares problem with additional iterations to account for physical constraints. To analyse the quality of inversion, uncertainty evaluation tools are developed for both the model responses and inversion results – volume fractions of the reservoir components. The tools include algorithms for construction of depth-related “confidence corridors” for individual properties and standard deviation ellipses to analyse cross-correlations. An optimization problem of improving the quality of inversion by tuning specific readings of logging methods for individual reservoir layers (or facies/lithotypes) is also posed and solved by minimizing a mismatch criterion between the actual and simulated log curves. Thus, the developed toolkit provides a means for optimized joint inversion of logging data with visual control of uncertainty and cross-correlation in the inversion results.