سال انتشار: ۱۳۸۵
محل انتشار: یازدهمین کنگره ملی مهندسی شیمی ایران
تعداد صفحات: ۱۰
Tabatabai – Petrophysics section ,E&P Department ,RIPI,Tehran,Iran
Boozarjomehri – Chemical and Petroleum Engineering Department, Sharif University of Technology, Tehran, Iran
Badakhshan – Chemical and Petroleum Engineering Department, the University Of Calgary,Calgary, Alberta, Canada
In recent years, neural network applications have increased in many aspects of different branches of engineering, and petroleum engineering hasn’t been an exception. Asmari formation forms the reservoir rock of many Iranian carbonate fractured oil fields. Assessment of the petro-physical characteristics of such reservoir rock especially their dual permeabilities are very difficult. Rag-e-sefid reservoir which is located in south west of Iran is a highly heterogeneous fractured carbonate reservoir. Its average matrix permeability is low, but the presence of fractures compensates its poor matrix permeability giving high production rates potentialities. Data available from core analysis show that permeability ranges from less than 1 md. in some parts up to 600 md. in very thin zones. Fracture frequency is high, thus the magnitude of permeability is scattered and total permeability (matrix plus fracture permeability) cannot be predicted accurately. An innovative method is developed which can
predict permeability from geological logs using artificial neural networks. The proposed method is based on neuromorphic representation of permeability with a neural network whose inputs and output are well logs and core permeability, respectively .The performance of proposed method has been validated through its application in prediction of permeability of uncored wells. A conventional feed forward neural network has been used as neuromorphic model which maps input data (logging data ) to output data (core permeability) .