The Development of an Optimal Artificial Neural Network Model for Estimating Initial,Irreducible Water Saturation – Australian Reservoirs

摘要

Initial, irreducuble water saturation, Swir is an importantparameter that needs to be determined accurately whenattempting to characterize hydrocarbon reservoirs. Swir is alsoone of the key parameters in relative permeabilityrelationships. Furthermore, an unrepresentative value of Swirmay lead to invalid residual oil saturation estimates when thelatter is correlated with the former.Swi may have a dependence on several other parameters,including: absolute rock permeability, porosity, pore sizedistribution and capillary pressure. The above parameters aredirectly influenced by geological deposition and subsequentchanges, such as diagenesis effects (for example clay-filledpores). It is a common practice to measure Swir utilizingrepresentative core plugs by measuring capillary pressure witha centrifuge, at speeds equivalent to the maximumrepresentative (reservoir) capillary pressure. However, a semiempiricalmodel that could estimate Swir to a good degree ofaccuracy would be of significant value.Over the last few years, artificial neural networks havefound their application in petroleum engineering. In somecases such models have outperformed models employingconventional statistical and regression analysis. In this study,an Artificial Neural Network (ANN) model has beendeveloped for the prediction of Swi (specifically irreduciblesaturation, Swir) using data from a number of onshore andoffshore Australian hydrocarbon basins. The paper outlines amethodology for developing ANN models and the resultsobtained indicate that the ANN model developed is successfulin predicating values of Swir over the range of data used forcalibration. This neural network based model is believed to beunique for Australian reservoirs