ISSN:
1573-8868
Schlagwort(e):
discriminant analysis
;
petroleum exploration
Quelle:
Springer Online Journal Archives 1860-2000
Thema:
Geologie und Paläontologie
,
Mathematik
Notizen:
Abstract Stepwise linear discriminant analysis and regression function techniques were used to classify and analyze the data from 224 oil, gas, and dry holes drilled to the base of the Silurian Geologic System in southeast Ohio, U.S.A., from 1968 to 1975. These are 8.7% of all wells drilled in the same geographical area over the time interval. The wells were divided into six groups depending on the amount of control the drilling company had over their ultimate completion. Thirteen environmental (geologic and engineering) factors plus location data were collected about each well. Successive groupings of oil and gas wells were used to determine an environmental index which was used to establish the hydrocarbon producing potential of the next group. Discriminant analysis was used to determine whether the wells within each group were classified properly and regression analysis was used to determine a production estimating function. Geophysical well data were used to estimate whether gas or oil was expected. With this dichotomy, production estimating functions which explain 74% of the variation for oil wells and 78% of the variation for gas wells were formulated. Using the discriminant function for these groups as a ranking function the following results were obtained: For oil wells, 18.5% Type II errors for the productive group and 24.4% Type II errors for the nonproductive group; for gas wells, 5.4% Type II errors for the productive group and 17.1% Type II errors for the nonproductive group. By accepting the above Type II errors one is significantly able to reduce the number of noneconomic wells completed that originally should have been considered dry holes. The inverse was tested with three wells and yielded 33% Type II errors. Any area in which adequate data are available is amenable to these types of analysis with their attendant diagnostic capabilities.
Materialart:
Digitale Medien
URL:
http://dx.doi.org/10.1007/BF01029422
Permalink
