Search

Introduction Gandhar oilfield, Cambay Basin, Gujarat is one of the Oil and Natural Gas Corporation Limited's (ONGC) major brownfields and a pilot candidate for India’s first large-scale CO2 sequestration project. The field which has been produced from multilayered sand bodies of Hazad Member, Ankleshwar formation has undergone various phases of production, and currently, it has reached a highly matured stage of its production life. ONGC has planned to implement the CO2 EOR technique in this field to recover an extra 15% of residual oil based on recent studies of source-sink matching, petrophysical properties analysis, current reservoir pressure, minimum miscibility pressure (MMP), and other laboratory experiments. The field is being studied for the assessment of CO2 storage potential and the feasibility of seismic techniques to monitor injected CO2 in Hazad sands (reservoir) and overlying brine saturated Ardol sands of Ankleshwar formation. For such a study, an integrated geophysical approach is performed with seismic, petrophysical, and geophysical well-log data provided by the National Data Repository-DGH (Govt. of India) and ONGC. Prediction of unrecorded P-sonic logs using Gradient Boosting Regressor algorithm in Gandhar oilfield A total of 15 wireline geophysical well log data from a 50 sq. km. block of the Gandhar oilfield for this study were provided but only 12 out of the 15 wells had P)-sonic logs. The sonic log (DT) is one of the essential logs required to perform petrophysical analysis, to aid in missing check-shot profiling, fluid substitution modeling, and seismic modeling & inversion studies. Therefore, the prediction of unrecorded P-sonic logs was required to reduce the uncertainty in reservoir characterization at the field scale by reducing the scarcity of sonic log data over an area. To overcome this problem, we implemented the Ensemble machine learning technique named Gradient Boosting Regressor (GBR) (J.H. Friedman, 2001, 2002) algorithm to predic