Prediction of compressional wave velocity by an artificial neural network using some conventional well logs in a carbonate reservoir

Abstract

As vital records for the upstream petroleum industry, compressional-wave (Vp) data provide important information for reservoir exploration and development activities. Due to the different nature and behaviour of the influencing parameters, more complex nonlinearity exists for Vp modelling purposes. Therefore, formulating a prediction tool that can accurately estimate the lacking log data, such as Vp, is of prime importance. Therefore, an attempt has been made to develop a prediction model for Vp as a function of some conventional well logs by using an artificial neural network (ANN). The obtained results are compared to those of multiple linear regression (MLR) models. A total of 2156 data points from a giant Middle Eastern carbonate reservoir, derived from a conventional wire line and a dipole sonic imager log were used in this study. The efficiency of the employed approach, quantified in terms of the mean squared error correlation coefficient (R-square), and prediction efficiency error, is evaluated through simulation and the results are presented. The result showed that an ANN outperforms MLRs and was found to be more robust and reliable.