(212h) An Integrated Approach to Tight Gas Production

Despite many recent technological advances and spectacular success stories, natural gas production from unconventional gas resources (mainly reservoirs with very low permeability) can be described as ?hit-or-miss? in practically all aspects of exploration and resource development. A typical tight-gas development scenario starts with the creation of reservoir description (physical properties) as a result of geological and seismic studies, followed by use of such information for well placement and fracturing design, which eventually lead to production. However, the industry relies largely on experts to accomplish many on these tasks. While experts can be highly successful, they are in short supply. Furthermore, while experts excel at many tasks, mainly revolving around visual pattern recognition, they may lag behind computers at others (such a processing of multivariate numerical data). The aim of this work is to replicate and augment human expertise in the design of production systems for tight gas. The proposed methodology relies on the following: (a) Novel approach to estimation of reservoir properties based on Bayesian networks (through collaboration with UC Berkeley/Lawrence Berkeley Lab), (b) Simplified modeling of the effect of decision variables ? such as well and fracturing placement and geometry ? on natural gas production (through collaboration with Texas A&M University, and (c) integration of the previous two entities in continuous optimization using a moving horizon approach. The term "continuous optimization" refers to making decisions about production using all available data up to that point in time and adapting models used to make future predictions. A prototype of the proposed strategy will be discussed based on simulations on a gas reservoir model calibrated on the basis of data provided by an industrial collaborator.