Prediction of oil reservoir porosity based on BP-ANN

Porosity of oil reservoir rock is usually determined by Core Analysis. But this method is expensive and time consuming. Also because of lithology changes, heterogeneity of reservoir rock, and nonexistence of sufficient well cores, determination of the parameters by the usual methods is not accurate. So the best way to decrease cost, increase accuracy, and decrease time is applying advanced software such as Geolog and Back-Propagation of Error Artificial Neural Network (BP-ANN). In this paper, a BP-ANN was designed to predict the porosity of formations using the well logs data in Parsi field, located in southwest of Iran. The data of two wells (No. 33 and No. 19) that have core data were used for training, testing, validation, and generalization processes. Then the BP-ANN results were compared to evaluations obtained from Geolog Software (GS). With respect to the results, it was concluded that the BP-ANN is more accurate than GS in determining oil reservoir porosity. At the end, porosity was simulated in three other wells (No. 48, 49, and 64) that lack core data.