CSUG/SPE 149485 SeTES: A Self-Teaching Expert System for the Analysis, Design, and Prediction of Gas Production From Unconventional Gas Resources

SeTES is a self-teaching expert system that (a) can incorporate evolving databases involving any type and amount of relevant data (geological, geophysical, geomechanical, stimulation, petrophysical, reservoir, production, etc.) originating from unconventional gas reservoirs, i.e., tight sands, shale or coalbeds, (b) can continuously update its built-in 'public' database and refine the its underlying decision-making metrics and process, (c) can make recommendations about well stimulation, well location, orientation, design, and operation, (d) offers predictions of the performance of proposed wells (and quantitative estimates of the corresponding uncertainty), and (e) permits the analysis of data from installed wells for parameter estimation and continuous expansion of its database. Thus, SeTES integrates and processes any available information from multiple and diverse sources on a continuous basis to make recommendations and support decision making at multiple time-scales, while expanding its internal database and explicitly addressing uncertainty. It receives and manages data in three forms: public data, that have been made available by various contributors, semi-public data, which conceal some identifying aspects but are available to compute important statistics, and a user's private data, which can be protected and used for more targeted design and decision making. SeTES can be a vital and easy-to-use tool in gas production from unconventional gas resources, and presents a new paradigm for communicating research and technology to the public and distributing scientific tools and methods. It is expected to result in a significant improvement in reserve estimates, and increases in production by increasing efficiency and reducing uncertainty. Introduction Background. Unconventional gas resources (UGRs) are best described as gas accumulations that are hard to characterize and commercially produce by common exploration and production technologies. These resources are typically located in very tight, heterogeneous, extremely complex, and often poorly understood geologic systems, often easy to find but difficult to produce. Thus, while it is not difficult to find extensive UGRs in many basins, it is very difficult to determine their flow properties from petrophysical well surveys and to design effective completion procedures. Furthermore, because of their very low permeability, establishing gas flow at commercial rates requires costly well stimulation operations. These considerations (low permeability and complex geology) are responsible for the high risk factors and unpredictable results often associated with UGR exploration and development projects, and hamper industry investment in these resources despite their potentially vast magnitude. The problem of tight gas production has inherent and multifaceted uncertainty (e.g., in reservoir properties or gas prices) coupled with overwhelming complexity in making decisions at all phases of gas field development. While the problem of making decisions for complex systems in an environment of high uncertainty appears in many engineered systems (e.g. chemical plants, managed ecosystems, transportation systems) gas production from very tight reservoirs has its own intricacies, which make the application of standard methods and tools far from trivial, if at all feasible. Nevertheless, recent work (Saputelli et al., 2005) has demonstrated that a multi-scale approach to the overall problem of field development can tame complexity, mitigate uncertainty, and coordinate information processing from a multitude of