Artificial Neural Network (ANN)-Based Predictions of Bulk Permittivity of CO2-Water-Porous Media System

In this chapter, the underlying principles and the effectiveness of machine learning-based techniques in the study of carbon dioxide (CO2) removal process is demonstrated though an artificial neural network (ANN) that was developed to predict the bulk relative permittivity of CO2-water-porous media system as an aid for monitoring CO2 migration in the subsurface. Different configurations of the ANN were examined to achieve the best prediction output with the minimum error. The input parameters for the ANN consist of the water saturation, CO2 saturation, porosity, CO2 relative permittivity, water relative permittivity, sand relative permittivity, injection pressure and temperature with corresponding output as bulk relative permittivity obtained for different porous media, namely carbonate sand, silica sand and basalt. Good generalisation of the developed ANN was achieved from the data obtained from in-house laboratory experimental investigations. It was also found from the performance values that the ANN configuration with two hidden layers performs better than those with a single hidden layer.