Current challenges in oil and gas exploration and field development require the use of all the available information for enhancing resolution and quantification of the resources. Hydrocarbon exploration typically presents problems related to resolution of the geological targets to enable prospect generation for drilling. Other aspects are related to characterization of fluids in reservoirs to further de-risk the prospects and increase success in drilling. On the field development side, significant value is provided by the monitoring of the reserves over time to enable optimization of reservoir management and history matching with reservoir simulators.
While each geophysical discipline is traditionally evolving on separate paths, there is vast potential in the integration of various geophysical measurements sensing the same targets using different physical properties. Joint inversion is one method to solve complex 3D problems by relying on some measure of similarity among different parameters. We explore the application of joint inversion techniques to several exploration problems and reservoir fluid monitoring. Several aspects need to be taken into account in approaching multi-physics inversion that span from comparing data with different resolution at different depths, different sensitivity to different parameters, scaling problems and validity of the similarity measure assumptions. These considerations suggest that algorithm development is as important as the implementation strategies where various degrees of constraints can be imposed. The problem then becomes how to seamlessly and quantitatively integrate geological interpretation with multiple geophysical datasets each one ill-posed, non-unique and subject to various degrees of uncertainty. Joint inversion methods were developed and applied to address this problem and approach the global minimum of the solution through data integration. Bayesian inversion methods provide the mathematical framework for describing uncertainties by means of probability density functions, but experience suggests that our a-priori knowledge of parameters and error distributions is typically insufficient. For the integration of complex 3D multi-physics datasets, we often find it advantageous to rely on scaling through inversion dimensions (1D, 3D), proceed from data-driven to model-driven approaches, single-domain to multi-domain inversions and constrained or hierarchical joint inversion.
Copyright. All Rights Reserved.