Abstract

A fuzzy expert system designed to minimize exploration risk in the Lower Brushy Canyon Formation is being developed by the REACT group at the Petroleum Recovery Research Center at New Mexico Tech. Central to this or any expert system is a knowledge base where expert knowledge is stored, often in the form of rules. This paper examines the development and structure of the knowledge base and discusses its application in the expert system.

The knowledge base is developed through consulting with experts having experience prospecting for oil in the Delaware Basin. Both the information the experts consider relevant and the thought processes they use in the evaluation of the potential of a location to produce oil are studied. The process by which this knowledge is codified and developed into the rules is described. Large amounts of geophysical, production and well log data collected for this region are used in addition to this expert knowledge to support and refine these rules.

The knowledge base is structured into categories or subsystems, including a trap assessment, a formation assessment and a regional assessment. Each of these categories contains a set of either crisp or fuzzy rules that determine an initial estimate based on a parameter central to the type of assessment. The remaining rules in each branch of the knowledge base are used to modify and refine the estimate, mimicking the thought process of a knowledgeable person considering a potential well location.

The expert system receives data on the potential location, either through user input, or through data available to the system. This data is then processed through the knowledge base, producing estimates from each category that are combined to provide the user an overall risk assessment.

Using a crisp model of the expert system, data from recent producing wells and dry holes are used to verify that the knowledge base replicates the expert’s decision-making process to an acceptable degree of accuracy.