Principal Investigators: Carlos Tores-Verdin (Dept of Petroleum Eng.) and Mrinal K. Sen

Funded by: Department of Energy

Project Summary

We propose a new computer algorithm for the efficient and consistent integration of 3-D seismic data, well-log data, geological information, and reservoir production data, into a petrophysical model that can describe and predict the static and dynamic behavior of a hydrocarbon reservoir. The integration of measurements will be achieved in the form of a quantitative 3-D cellular petrophysical model that will be readily amenable to numerical simulation of multi-phase fluid flow under specified production regimes. The 3-D cellular model will be updated and modified via a feedback control loop as new measurements are acquired through the reservoir production cycle. Data integration and model update will be performed with a stochastic Bayesian inversion procedure driven by a set of joint probability density functions (PDF’s) that will govern the link between the underlying petrophysical model and the seismic acoustic and elastic parameters. The estimated petrophysical variables will also be made to honor the borehole and core data within the accuracy allowed by their intrinsic measurement uncertainties. The petrophysical model will comprise space- and/or space dependent distributions of petrophysical variables such as porosity, absolute permeability (tensorial behavior), relative permeability, and fluid saturation, for instance. The joint PDF’s will be inferred by petrophysical analyses of the available well-log and core data, exhaustive statistical analysis, measurement consistency checks, and laboratory analyses of elastic and acoustic properties of core data. Moreover, the joint PDF’s will be adjusted to respond to transitional scales in spatial resolution and hence will allow the natural integration of measurements with different lengths of spatial support. We describe the mechanics of the proposed stochastic Bayesian inversion algorithm, including its numerical implementation on serial and massively parallel computer architectures. The Bayesian formulation allows the natural introduction of petrophysical constraints into the inversion, of lithology-based petrophysical relations, and it readily provides local estimates of uncertainty for the final 3-D cellular petrophysical model. Moreover, because the joint PDF’s take into account the different lengths of spatial support available from the data, there is no need to apply mathematical fluid-flow uspcaling prior to performing the numerical simulations of reservoir behavior. In effect, having used the 3-D seismic data as ground truth to the petrophysical variables, basically does away with the need to perform mathematical upscaling of the fluid-flow parameters. We will develop practical ways in which the proposed stochastic Bayesian inversion will yield uncertainty estimates of hydrocarbon production data. The proposed computer algorithm for effective and efficient reservoir data integration will be developed, tested, refined, and transferred to the industry with the use of available data sets provided to us by various oil producing companies. Although we will be making explicit use of both post- and pre-stack 3-D seismic data, the main thrust of this proposal will be the development of an efficient stochastic inversion algorithm that can fully exploit the lateral amplitude variations of 3-D pre-stack seismic data, and the high vertical resolution available from well-log data. To our knowledge, currently there is no such algorithm available to the oil industry.