Modeling, History Matching, Forecasting, and Analysis of Shale-Reservoir Performance With Artificial Intelligence

Abstract

This article, written by Senior Technology Editor Dennis Denney, contains highlights of paper SPE 143875, ’Modeling, History Matching, Forecasting, and Analysis of Shale-Reservoir Performance Using Artificial Intelligence,’ by Shahab D. Mohaghegh, SPE, Intelligent Solutions and West Virginia University, and Ognjen Grujic, SPE, Seed Zargari, SPE, and Masoud Kalantari, West Virginia University, prepared for the 2011 SPE Digital Energy Conference and Exhibition, The Woodlands, Texas, 19-21 April. The paper has not been peer reviewed. Advances in horizontal drilling and multistage hydraulic fracturing have made shale reservoirs a focal point for many operators. Our understanding of the complexity of the flow mechanism in the natural fracture and its coupling with the matrix and the induced fracture, the effect of geomechanical parameters, and optimum design of hydraulic fractures has not necessarily kept up with our interest in these prolific hydrocarbon-rich formations. A new approach to modeling, history matching, forecasting, and analyzing oil and gas production in shale reservoirs was developed. It uses pattern-recognition capabilities of artificial intelligence and data mining as a workflow to build a full-field reservoir model to forecast and analyze oil and gas production from shale formations. Introduction A reservoir-simulation and -modeling technology called top-down intelligent reservoir modeling [referred to here as top-down modeling (TDM)] was applied to shale formations, and the full-length paper details examples for New Albany, Lower Huron, and Bakken shales. Natural fractures in the shale have the highest permeability in the reservoir and, as the main conduit, contribute significantly to production. Multistage hydraulic-fracturing procedures enable reaching and intersecting the existing natural fractures in the shale formation. Mapping of the natural fractures in the shale formations has proved to be an elusive task. Even with the most-advanced logging technologies, only the intersection of the natural fractures with the wellbore can be detected, while the extent of these fractures beyond the wellbore and how they are distributed throughout the reservoir (between wells) remain the subject of research. TDM tries to model the effects of hydraulic and natural fractures on the production from wells rather than modeling the discrete fracture networks themselves. While the development of stochastic realizations of natural fractures and their intersection with the induced hydraulic fracturing is being studied with stochastic and numerical reservoir modeling, TDM fills the existing gap for a predictive model that can be built by use of a minimum number of assumptions about the nature of the reservoir and about our understanding of its complexity. TDM starts with a solid assumption that whatever the nature of the natural-fracture distribution and its interaction with the induced hydraulic fractures may be, these factors must influence the amount of the hydrocarbon that each well is able to produce. These signatures can be used to build reservoir models, match the production history, and build a predictive model that can aid reservoir-management decisions.