Customizable Fuzzy Expert System for Regional and Local Play Analysis
The objective of this project is to create a user-definable and customizable fuzzy expert system tool to dramatically speed local and regional play analysis and reduce subsequent drilling risk. This general tool will not require significant knowledge of computer programming and will guide users through the process of building a successful expert system to evaluate plays from field to basin scale using public and/or private data and their own knowledge. To demonstrate the effectiveness of the tool, a secondary objective of analyzing the Pennsylvanian play of southeast New Mexico will be performed. Public data will be organized for analyzing this outstanding, bypassed-pay play, which will provide an example of the usage of the system while simultaneously providing a significant opportunity for identifying new reserves.
Petroleum Recovery Research Center (PRRC), New Mexico Institute of Mining and Technology, Socorro, NM
This project has reached the end of the software development stage and a functional prototype of the software has been implemented. The software is now ready for beta testing with outside users, which was scheduled for the third year of the project. The project was ended at the end of the second project year due to DOE Natural Gas and Oil Program budgetary constraints. The project, while on task to produce expected results including deliverable software to aid and speed prospect generation using state-of-the-art computational intelligence technology, has been put on hold indefinitely
The system comprises four major components: a graphical user interface (GUI) for compiling and organizing expert knowledge, a GUI for organizing project data, a GUI for analyzing results, and a fuzzy inference engine for processing rules. The knowledge input into the system is to be expressed generally as “if-then” rules. Fuzzy rules will be employed when inputs are imprecise and factual data are insufficient or scarce. Local databases, such as the New Mexico Production Data System (ONGARD) located at the PRRC, and remote databases, such as “An Oil and Gas Development System,” supported by DOE, were to be linked to the customizable system through the Internet and allow users to search for additional data. Statistical tools and methods of interpretation have been designed to allow users to input raw data and calculate appropriate rules and analyses quickly and efficiently. This software is fully integrated and allows seamless integration of user data in the expert system design process. A new customizable fuzzy inference has been designed to provide the relational computations required to relate scarce local data with sparse regional data.
Collection and processing of test data, the Pennsylvanian Carbonates of SE New Mexico, has been completed. This test data set was collected to aid in finalizing the work flow operations and wizards necessary for accurate and efficient expert system design. While that aspect of the project was delayed, the dataset by itself represents a valuable resource for anyone interested in the play and will be made available (see the contact information below). During the first project year geologic data acquisition and analysis was completed on the Upper Pennsylvanian and Lower Permian carbonate reservoirs within the project areas in the Permian Basin of southeastern New Mexico. During 2006, work on the Bough intrashelf project area was completed. Geologic data was acquired, mapped and analyzed on a second project area, the shelf-margin project area that includes the Upper Pennsylvanian Dagger Draw oil reservoirs. The second project area was added to diversify the expert system’s approach to prospecting for oil in carbonate stratigraphic traps. Although reservoirs in both areas are formed primarily by phylloid algal mounds that have seen multiple episodes of diagenesis, the reservoirs within the Bough intrashelf area grew on bathymetrically high paleostructures and the reservoirs within the shelf-margin area grew primarily as bioherms on a constructional shelf-margin and their geographic location does not appear to be related to paleostructures. Both of the project areas have significant production. The 58 Permo-Pennsylvanian reservoirs that are at least partially present within the Bough intrashelf project area have produced a combined total of 329 million bbls oil. The 24 Upper Pennsylvanian reservoirs within the shelf-margin project area have produced a combined total of 83 million bbls oil.
Successful development of this customizable fuzzy expert system would bring great computational power for regional play analysis to companies of all sizes—benefiting large companies interested in recompletion opportunities but having reduced domestic exploration budgets and small companies without the resources to maintain full-time multidisciplinary exploration staff. Ultimately, widespread use of this technology can increase efficiency and production from domestic oil and gas sources. Analysis of the Pennsylvanian carbonates, a significant by-passed pay play, was to showcase the abilities of the software and provide immediately useful information to companies interested in developing production from that play.
Future oil and gas reserves discovered in the United States will be dominated by an improved understanding and extension of existing plays and improvements in secondary/tertiary recovery. A powerful and recently tested technology for finding reserves in mature and virgin basins are expert systems: artificial intelligence tools that store and implement expert opinions and methods of analysis. Such systems, expanded and made completely user-customizable, would have a dual role in exploration and development—first, in storing and accessing the wealth of expert knowledge that is lost through job changes, attrition, and retirement; and second, through formalizing exploration methodologies in robust and repeatable processes.
The primary goals of developing such a tool would be to expand the knowledge base through which industry brings new oil resources to market, to accelerate new prospect generation, and to identify bypassed pay in the mature producing regions of the United States. This project continues development of a hybrid fuzzy expert system, which has shown great success in reducing exploration risk in specific regional plays by qualifying drilling risk at new prospects prior to drilling, into a customizable tool that can be populated easily with rules by experts working anywhere in the United States, without those same experts needing additional specialization in computer science.
Four major tasks were originally planned in the project plan. Task 1 is the development of a customizable fuzzy expert system, with the ability to self-generate software and fully customize integral components of the expert system for non-computer programmers. Task 2 is the development of interfaces to simplify the goals of Task 1. The creation of wizards to aid in the workflow process makes expert system development both quicker and easier. Task 3 calls for the system to be validated using the Pennsylvanian carbon-ate play of SE New Mexico. Development of this bypassed pay play aids in the creation of wizards to assist users in Task 2.
An ongoing effort in technology transfer comprises Task 4, as the ultimate test of the successful program will require that it is widely available and utilized.
A complete prototype system has been developed, comprising:
- An adaptable inference engine.
- Automatic management of expert knowledge.
- Automated management of user data.
- Preset and initial fuzzy curves from data sets based on statistical analysis.
A fully integrated system prototype has been implemented. This system would have been adjusted based on the analysis of the Pennsylvanian data and have user friendly wizards added during software testing scheduled for the projects during the third year, however the Pennsylvanian data does represent a significant stand alone data set.
The prototype system has recreated results from earlier prototype expert systems using two previous data sets used to create expert systems for oil exploration (Risk Reduction with a Fuzzy Expert Exploration Tool, DOE project No. DE-AC26-99BC15218).
Current Status (July 2007)
The project completed 26 months of 36 months planned in the original contract. The final 12 month task list will not be started due to reduced DOE funding in FY2007. All core software components are in place and the process of integrating them into one overall interface is complete. A large preliminary sample knowledge base for the Permian basin was developed through expert interviews and collection of knowledge using the Delphi method. The database for the Pennsylvanian carbonates is complete and a prototype of the software has been developed. However, testing and analysis of that data using the current version of the software not be completed under the reduced project scope. The final report is near completion.
This project was awarded under DOE solicitation DE-PS26-04NT15450-2B.
Project Start: January 1, 2005
Project End: May 31, 2007
DOE Contribution: $523,500 (reduced from $799,997 original anticipated contract amount)$799,997
NETL – Rhonda Jacobs (firstname.lastname@example.org or 918-699-2037)
PRRC - Robert Balch (email@example.com or 505-835-5305)
Balch, R. S., Ruan, T., and Schrader, S.M., Fuzzy Expert Systems in Oil Exploration, SIAM Conference on Computational Science and Engineering, Orlando, FL, February 12-15, 2005.
Schrader, S.M., Balch, R.S., and Ruan, T., Using Neural Networks to Estimate Monthly Production: A Case Study for the Devonian C1+, Southeast New Mexico, SPE 94089, presented at the SPE Production and Operations Symposium, Oklahoma City, OK, April 17-9, 2005.
Ruan, T., Balch, R.S., Schrader, S.M., and Hart, D.M., A Web-Based Fuzzy Ranking System and Application, presented at the 9th World Multiconference on Systemics, Cybernetics, and Informatics, Orlando, FL, July 10-13, 2005.
Ruan, T., Balch, R.S., and Schrader, S.M. A Fuzzy Expert System for Oil Prospecting in the Lower Brushy Canyon of Southeast New Mexico, presented at the IEEE International Conference on Information Reuse and Integration, Las Vegas, NV, August 15, 2005.
Schrader, S.M., Balch, R.S., and Ruan, T., Knowledge Management, Collection, and Storage in Expert System Development, Upstream CIO, September 2005, pp. 22-24.
Schrader, S. M. and R. S. Balch., Automated Analysis of Gridded Geologic Map Data. Paper presented at American Association of Petroleum Geologists Annual Convention, Houston Texas, April 9-12, 2006, 6pp.
A screenshot of the Question Management Interface.
A close-up of the Fuzzy Set Definition module.
A screenshot of the Question Management Interface.
A screenshot of the Fuzzy Set Definition Interface.