Application of back-propagation neural networks in the distribution of oil sand reservoirs—a case study from the western slope of the Songliao Basin, China

Abstract High-quality hydrocarbon source rocks are present in the upper Cretaceous layer western slope of Songliao Basin. Oil and gas have accumulated these at shallow edge basin, which has led to formation oil sand resources. This study uses back-propagation (BP) neural network method predict distribution reservoirs is first its kind China. First, based on basic data collected by core sample, well log geochemical analyses, reasonable selection samples, cores divided into mudstone, siltstone, fine sandstone, medium sand, according lithology. Second, a three-layer BP model constructed with two hidden S-type layers one linear output layer. Third, through comparison effect different numbers training sessions sample data, we demonstrate that accuracy can be increased 90% after 100,000 times. Then, log-derived unknown lithologies input whether they contain sands. We show target horizon area, results consistent research corresponding sedimentary facies. Thus, conclude it feasible use reservoirs.