Título: Applications of neural networks and fuzzy logic in geological modeling of mature hydrocarbon reservoir

Autor: Paygun, B.; Luthi, S. M.; Bryant, I. D.

Evento: Artificial Intel. in the Petrol. Ind. Int. Conf.

Lugar y fecha del Evento: Lillehammer, Norway, Sep. 13-15

Descriptores: Geologic Model; Artificial Intelligence; Computer Programming; Correlation; Data; Depth Correlation; Expert Systems; Fuzzy Logic; Mathematical Analysis; Mathematics; Model; Programing; Research; Reservoir Study; Study; Well Logging Data

Año: 1995

País de publicación: Norway

Tipo de documento: Conferencia

Resumen

Neural networks are used for well-to-well correlation, and a fuzzy classifier is used for delineation of geological objects in a mature field with many wells. A tedious task, well-to-well correlation also poses challenges to the production geologist because of lateral changes in facies, pinchouts and depth shifts due to faulting. These geological complications give rise to different patterns in the well logs across the field. Neural networks are trained on one or more key wells to discover the patterns that are indicative of particular markers. The neural networks are then applied to all the wells to locate the same markers. The boundaries of geological objects in the field using well logs give distinct patterns in the well logs, and a fuzzy clustering algorithm classifies the well log patterns. Interpolation of membership values to a particular class into the interwell region produces a map of the geological object represented by that class.