Discriminant analysis and neural nets: Valuable tools to optimize completion practices
- Shell International Petroleum Mij., The Hague (Netherlands)
- Koninklijke/ Shell E and P Laboratorium, Rijswijk (Netherlands)
This paper describes the application of multi-variate statistical techniques, discriminant analysis and neural networks in identifying drilling and other completion practices that impact on well productivity. Discriminant analysis determines whether a well can be assigned to a group of wells, on the basis of a number of common characteristics and using linear multivariate correlations. Neural nets enable the use of nonlinear correlations for such a classification. In this study, 47 gas wells from two fields were classified into three groups: Group 1 -- no production; Group 2 -- production below 5,900 std m{sup 3}/h (5 MMscf/D); Group 3 -- production over 5,900 std m{sup 3}/h (5 MMscf/D). The variables used in the discriminant analysis included parameters such as completion type, total height of the perforated interval, mud weight, drawdown during perforation, type of mud and perforation size. This study has identified and, to some extent, quantified those parameters that either adversely or favorable affect well productivity. The results can be used to adjust operational procedures to maximize well productivity. The parameters identified as increasing productivity reflect, for the most part, sound engineering practices. Application of neural nets enables further quantification of the effects of petroleum engineering parameters on well productivity and is being developed to make it possible for the most economical preventive and remedial measures to be selected. However, statistical techniques are applicable only when a sufficiently large data base is available, i.e., they are suitable for reasonably large and fairly mature fields and/or areas.
- Sponsoring Organization:
- USDOE
- OSTI ID:
- 93070
- Journal Information:
- SPE Drilling and Completion, Vol. 10, Issue 2; Other Information: PBD: Jun 1995
- Country of Publication:
- United States
- Language:
- English
Similar Records
Analysis and testing of high-frequency regional seismic discriminants. Final report, 27 August 1991-31 July 1992
A theory of quark vs. gluon discrimination