Phase Mobility Evaluation of Naturally Fractured Reservoirs Using Well Logs

Abstract

Worldly known for their huge potential for hydrocarbons production, naturally fractured reservoirs have a high recovery rate, which is essential to improve extraction efficiency. This study represents the implementation of machine learning (ML) in the oil and gas field, where it dived deeply to build a ML model that evaluates fluid mobility of naturally fractured reservoirs (NFRs) using well logs of the HGA field. And maybe it could somehow replace the coring analysis in predicting permeability which is the crucial part of evaluating mobility. The HGA field was taken as the data source to train and test the model, the chosen input features are conventional well logs (GR, DT, PE, NPHI, RHOB, RT), and toward predicting the output feature (Permeability) in this case the predicted permeability is then used to evaluate the phase mobility. The succuss of this study finalized with a resilient ML model with an accuracy of 92% and an absolute error of 53mD which great in the range of data used, this model also reached an accuracy of 84%, and 90% with an error of 47mD in the blind set test. With the proper improvement we can generalize this model to evaluate mobility of any naturally fractured reservoir, therefor it could be an excellent help in reservoir modeling, characterization and real time reservoir monitoring.