The complete paper describes an alternative solution for identifying the presence of natural fractures, classifying them into fracture-quality-related flowability, and distributing them vertically within the well interval, and proposes a lateral distribution method for reservoir modeling. The proposed approach, using the machine-learning technique of self-organizing-map (SOM) clustering, effectively assists recognition of fracture presence and quality along the well-depth interval.
Field Overview and Data Used
The case study was conducted in an oil field discovered in the early 1980s in southwestern Hungary. Thirty-six wells penetrated the naturally fractured carbonate in the Triassic formation. The main lithology of this reservoir is limestone and dolomite associated with faults and exhumation breccia and marl/shale.
The reservoir is saturated oil (with gas cap) with unlimited aquifer (strong water drive). The gas cap, however, is mainly composed of 85% carbon dioxide and up to 1,800 ppm of hydrogen sulfide.
