A Multi-Layer Backpropagation Network to Estimate CO2/ brine Interfacial Tension Data: A Comparative Study

Plethora emission of Carbon Dioxide has the largest contribution to the global warming problem. CO2 Capture and Storage (CCS) is a process proposed to decrease CO2 content of the atmosphere by capturing it from the power plants and injecting it into underground reservoirs where it’s supposed remain there for a long time. One of the main parameters affecting the CCS prospect in oil and saline aquifer reservoirs is the Interfacial Tension (IFT) acting between CO2 and reservoir brine solution; thus, accurate estimation of the IFT is crucial in estimating the possible volume of the CO2 sequestered in the underground layers. Since experimental methods used to estimate IFT are associated with some problems and are also time consuming, Hence, a model is developed in this study based on the Artificial Neural Network (ANN) and is used to estimate the CO2/ brine IFT data. The developed ANN based model is trained by using a comprehensive dataset adopted from the open literature; additionally, Bayesian regularization technique is employed in the training phase of the ANN to improve its generalization capability. Performance of the ANN based model to predict the CO2/ brine IFT is, then, compared with the three most well-known correlations published in the literature. Graphical error analysis and statistical parameters comparison reveals that the ANN based model has a remarkable IFT prediction capability and outperforms all of the indicated correlations. In addition, the ANN based model can accurately predict the trend in the IFT data with respect to the changes in pressure, temperature and type of the salt present in the brine solution. Therefore, Constructed ANN based model can be used in different software packages demanding CO2/ brine IFT as an alternative to the existing correlations.

کلیدواژه ها:

CCS ، CO2/ brine Interfacial Tension ، CO2-EOR ، MLBP ، Bayesian Regularization