AAPG Annual Convention and Exhibition

Datapages, Inc.Print this page

Relative Permeability Curve Calculation From Core-Plug Scale to Reservoir Scale Constrained by Facies Pattern


In current conventional reservoir simulation practice, the input relative permeability curve data is derived from an average of different laboratory measurement results using several core plugs. Each core plug is only about 10cm which is far smaller than real reservoir. Thus, the relative flow characteristics in core plug will have a big difference comparing the flow characteristic in real reservoir which is usually in a scale of several kilometers. And the average of relative permeability curve from several core plugs makes the heterogeneity results from different facies disappear. A new methodology used to calculate the relative permeability curve for different facies based on the core plugs laboratory data is proposed. In first step, the entire laboratory measured relative permeability curves are grouped according the measured porosity and permeability of each core plug. Then, different conceptual geological facies models which represented using different porosity and permeability rhythms are established. After that, based on Buckley-Leverett theory, the oil and water relative permeability at different average water saturation will be calculated using the reservoir simulator data at natural depletion and water injection development stage using the geological facies model as input for flow simulation. By that flow simulator and calculation, different relative permeability curves for different facies will be calculated. Thus, a distribution of relative permeability curve spatial cell model can be obtained through transform the three dimensional facies model into the corresponding relative permeability curves data. Then, in the final flow simulation, each cell in the flow model will have its relative permeability curve data which represents a more closer heterogeneity reality of underground reservoir. The principle and implementation steps of the proposed method are illustrated using a real data set. It is shown that the simulation results using the proposed method improve the history matching significantly.