Expert systems are artificial intelligence tools that store and implement expert opinions and methods of analysis. The goal of this project was to test and prove the ability of expert systems to enhance the exploration process and to allow the rapid, simultaneous evaluation of numerous prospects. The project was designed to create two case-study fuzzy expert exploration (FEE) tools, one for the Lower Brushy Canyon formation of the New Mexico portion of the Delaware Basin, and the second for the Siluro-Devonian carbonates of southeast New Mexico.
Petroleum Recovery Research Center (PRRC)
New Mexico Institute of Mining and Technology
Incomplete or sparse data such as geologic or formation characteristics introduce a high level of risk for oil exploration and development projects. Expert systems developed and used in several disciplines and industries have demonstrated beneficial results when working with sparse data. Tools of this type can be beneficial in many regions of the United States by enabling risk reduction in oil and gas prospecting as well as decreased prospecting and development costs. In today's oil industry environment, many smaller exploration companies lack the resources of a pool of expert exploration personnel. Downsizing, volatile oil prices, and scarcity of domestic exploration funds also have affected larger companies, and will, with time, affect the end users of oil industry products in the United States as reserves are depleted.
The stratigraphic Brushy Canyon play of the Delaware Basin of southeast New Mexico was the initial target for the project. Analysis of 60,478 forty-acre potential drilling sites by the expert system identified about 212 million barrels of new recoverable reserves in 4,481 undrilled prospects. Reduced finding costs resulting from use of the system make the pool an enticing play for both new exploration and recompletions.
The structurally complex Siluro-Devonian Carbonates of Southeast New Mexico was the second target of this project. These deep, often prolific reservoirs typically require expensive 3-D seismic data as a basis for exploration. The expert system has identified about 1,500 sections, out of an area of more than 15,000 square miles, with optimal production potential. The analysis can significantly reduce exploration costs and allow more-focused seismic surveys, therefore reducing wildcat risk.
The FEE tools enable a more efficient use of scarce exploration funds, thus contributing to efforts to reduce dependence on foreign oil and provide lower product prices for consumers.
To accomplish the development of these expert systems, massive databases of public-domain information for both plays were compiled, and additional geological, engineering, and production data were generated during the course of the project, creating a knowledge base for both formations. Models employing human expert knowledge of each play were developed, along with intuitive graphical user interfaces and fuzzy inference engines to process those expert rules, resulting in fast, multi-tiered systems that can be customized for personal or corporate philosophies while maintaining the integrity of proprietary information. Both tools were designed using the Java programming language to allow easy use through a browser window over the Internet. Stand-alone versions were developed concurrently. Both systems were extensively tested using statistics and by exclusion of blind test data.
Both expert systems offer a very good simulation of expert human explorationists. They also provide a quick-look tool for prospect analysis, enabling a faster and more consistent exploration process and the ability to rapidly evaluate well recompletion opportunities.
Support software developed for this project includes PredictOnline, an easily used neural network program; FuzzyRank, a program for selecting relevant variables using a fuzzy ranking algorithm; and Web-based Data Management System. All software can be used by anyone with access to the Internet.
The software developed for this project:
The project is complete. Based on these case studies, work on a customizable fuzzy expert system has begun.
$833,351 (28.5% of total)
The project generated six annual reports and six semi-annual reports to the DOE. The final report is available at http://ford.nmt.edu.
Balch, R.S., Ruan, T., and Schrader, S., Fuzzy Expert Systems in Oil Exploration, SIAM Conference on Computational Science and Engineering, Orlando, FL, Feb. 12-15, 2005.
Balch, R.S., Ruan, T., Weiss, W.W., and Schrader, S.M., Simulated Expert Interpretation of Regional Data to Predict Drilling Risk, paper SPE 84067, presented at the SPE Annual Technical Conference and Exhibit, Denver, CO, October 4-8, 2003.
Balch, R.S., Hart, D.M., and Weiss, W.W., Regional Data Analysis to Better Predict Drilling Success: Brushy Canyon Formation, Delaware Basin New Mexico, paper SPE 75145, presented at the Symposium on Improved Oil Recovery, Tulsa, OK, April 13-17, 2002.
Balch, R.S., Ruan, T., and Schrader, S.M., Drilling Risk Reduction with a Fuzzy Expert Exploration Tool, presented at the American Association of Petroleum Geologists' Southwest Section Annual Meeting, El Paso, TX, March 8-9, 2004.
Broadhead, R.F., and Justman, H.A., Regional Controls on Oil Accumulations, Lower Brushy Canyon Formation, Southeast New Mexico, published in The Permian Basin: Proving Ground for Tomorrow's Technologies, West Texas Geological Society, No. 00-109 (October 2000) 9.