Permeability Prediction from Log Data using Machine Learning Methods

In this paper, models for permeability prediction of oil reservoirs using a machine learning approach and petrophysical data are compared. Various machine learning methods, including multi-resolution graph-based clustering, conventional artificial neural networks and Extreme Learning Machines are employed to have a comprehensive comparison. RCAL data from one of Iran's oil reservoirs was used to develop and test the machine-learning approach. The results of the machine learning models employed in this paper are compared with relevant real petrophysical data and well evaluations. Seven input models of two different wells of this reservoir were considered for permeability estimation. The input logs data of models include Resistivity (RT), Effective Porosity (PHIE), Density log (RHOB), Sonic log (DT) and Compensated neutron porosity log (NPHI) logs data. The correlation coefficient and the root mean square error between the prediction data and core data in the ELM method were obtained as 0.94 and 0.06, respectively. In the MRGC method, the correlation coefficient and the root mean square error between the prediction data and core data were obtained as 0.98 and 0.09, respectively. The obtained results in this paper show that the mentioned models are well able to estimate permeability values in all parts of the studied formation and it can be concluded that the clustering method based on MRGC has more correlation with the core data, and Instead, the ELM method has the least amount of error in permeability prediction. According to the error values, ELM can be recommended as the final selected algorithm for permeability prediction in this study.