Quantitative prediction of fluvial sandbodies by combining seismic attributes of neighboring zones

The geological and geophysical characterization of hydrocarbon-bearing sandstones fluvial origin is a challenging task. Channel sandbodies occurring at different stratigraphic levels (i.e., in reservoir interval interest as well overlying underlying intervals) but overlapping planview usually cause significant seismic interference due to limitations resolution: this can produce error the prediction sand location thickness using attributes. To mitigate effect interferences by zones neighboring target interval, new method proposed that combines multiple attributes its interfering zones, implemented supervised machine learning algorithm support vector regression (SVR). Since intervals causing has constant value quarter wavelength (1/4 λ), stratal slice corresponding with top horizon taken base window 1/4 λ calculate for zone; similarly, bottom zone. was applied subsurface dataset (including 3D 255 wells) Chengdao oilfield, Bohai Bay Basin (China). located Neogene Guantao Formation, whose successions are interpreted origin. This application demonstrates how results remarkably improved sandstone prediction, consideration further improves accuracy predicted values thickness.