Prediction of Reservoir Compressibility Using Subsurface Cores, Well Logs, and Seismic Data by Neural Network

Abstract

Abstract In this study, three-dimensional pore volume compressibility of a carbonate reservoir was predicted. The primary data of the model were petrophysical parameters, measured compressibility factor on core samples, conventional well logs, and three-dimensional seismic attributes. Neural network algorithms were employed to propagate the compressibility data along the well axis and to predict the distribution of compressibility within three-dimensional seismic acquisition area. A probabilistic neural network algorithm resulted in a correlation of 85% between the predicted and measured compressibility along the wells-axis. The seismic attributes were extracted to find the best correlation and minimum error between the generated and target attributes. The correlation coefficient of 78% indicates the high accuracy of the model and the optimal choice of neural network algorithms. The results of this study provide insights into the application of seismic data to field-wide prediction of reservoir compressibility.