Permeability and RRT Estimation from Conventional Logs in a Middle East Carbonate Reservoir Using Neural Network Approach

摘要

A complex carbonate reservoir of a giant on-shore structure incentral Abu Dhabi consists of a porous limestone sectiondivided by several stylolitic dense layers. The reservoir isheterogeneous. High and low permeability rock types co-existwithout much variation in log characteristics. Permeabilitychanges several orders of magnitude with no appreciablechange in porosity. Reservoir rock types and permeabilitymeasurements are available from 30 cored wells. The reservoiris penetrated by more than 100 vertical wells where open holelogs of various vintages are available. Estimation ofpermeability using conventional logs in the intervals and wellswhere core data is not available has been a challenge for manyyears. In this study, relationships were built between core dataand the open hole log suite using the Artificial NeuralNetwork approach and were used in estimating reservoir rocktypes and permeability.A typical character of the reservoir is that the logsignatures are similar for several dissimilar rock types. Due tolimitation on the vertical resolution conventional open holelogs alone could not differentiate these rock types. A syntheticvariable cementation exponent ‘m’ curve, gradient curvesfrom Gamma Ray, Bulk Density and Neutron porosity logssupplemented the open hole logs in the neural networkanalysis in resolving the permeability variations. Nine neuralnetworks were constructed to handle the complexity. Robustreservoir rock type and permeability estimation was achievedin two iterations. In the initial pass, fractional facies curveswere estimated using a supervised classification network.These fractional facies curves were then used in initialpermeability estimation with the help of a supervisedestimation network. In the next pass, this permeability curve was used to estimate the reservoir rock types. In the final step,to fine-tune the magnitude of the permeability, the log derivedreservoir rock types were included in network analysis and afinal permeability curve with improved magnitude wasgenerated. The initial permeability estimation was madeseparately on two vertical sub divisions of the reservoirinterval and the reservoir rock type estimation was made onfour separate vertical sub divisions based on the occurrence ofvarious RRTs.Examples are provided showing a satisfactory predictionof both RRTs and permeability in cored and non-cored wells.The estimated permeability and RRTs were incorporated in the3D geological model supplementing core data.