Logfacies Prediction Using Self-Organizing Map for Carbonate Mishrif Reservoir in Buzurgan Oil Field, Southeastern Iraq

In this paper, the author introduces a machine learning model to classify the facies for Mishrif Formation in two wells in Buzurgan oil field, south eastern Iraq, and then testing the results with the actual facies data. The model also use data based on this field and a new model is introduced which is Self-Organizing Map (SOM), which would be best fit in case of lacking facies data or geological inexperience users. The available data of well logs for the studied wells (BU-3, and BU-13) are imported to Interactive Petrophysics software, which is used to create SOM facies classification model. The wire line logs include Gamma Ray (GR), compensated neutron tool (CNL), Sonic log (borehole compensated BHC type), and formation density compensated (FDC). Petrographic and microfacies analysis of subsurface section at well BU-3 enabled the recognition of five main environments, deep marine, shallow open marine, shoal, rudist biostrome and lagoon. Geologic logs in SOM technique are used to determine the different facies when dealing with intervals devoid of cores and cuttings and the resultant logfacies were used to interpret those intervals. The environments of well BU-13 are defined after examining available thin section of cores and cutting samples from well BU-3 and then created using SOM model. Comparison between SOM technique using Interactive Petrophysics software and petrographic analysis of subsurface section are made. The current study reaches 71.3%, providing a good calibration between log-derived facies and core-based descriptions of the Mishrif Formation.