Seismic data integration in geostatistical reservoir modeling

Abstract

Seismic data integration in reservoir modeling workflows is the one of the fastest-growing fields in the Earth Sciences. The actual geostatistical methods (co-kriging, stochastic simulation) can use seismic data as a secondary variable if there is a well-determined linear correlation between well log data and seismic attribute. Seismic interpreters must often increase this correlation. The application of multi-attributes via neural network may help in this case. A neural network type, called multi-layer perceptron, and its application in 3D porosity distribution prediction in a Hungarian natural gas reservoir, are described in this paper.