Fluid Discrimination Using Bulk Modulus and Neural Network

Abstract

Hydrocarbon prediction using the rock physical parameters is a common technique in the oil and gas industry. However, the rock physical parameters are controlled by porosity, the volume of clay, pore-filled fluid type and lithology simultaneously. Many methods are proposed to predict the existence of hydrocarbon. This paper proposes a new method ΔK which is the difference between the real bulk modulus and the bulk modulus in the brine- substitute case. The algorithm is validated through stochastic numerical modelling. The brines are separated by the ΔK, and the gas can be detected with acceptable accuracy. Furthermore, a model using deep learning approach is trained to predict the ΔK. The trained model is effective that the predicted values using this model have a strong correlation with the original ΔK. The ΔK can be applied to the data which contains Vp, Vs and density using this approach model. In this study, the ΔK is applied to the Marmousi II dataset to examine the performance and yields a good result. The combination of the deep learning and the ΔK improves our ability in hydrocarbon prediction.