Accurate Permeability Estimation in a Heterogeneous Middle East Carbonate Field through Facies Classification and Core-Log Integration

摘要

In this case study, we examine permeability estimation in a Middle East carbonate field where oil is mainly produced from the Cretaceous limestone reservoirs. Due to the complex depositional and diagenetic processes, the reservoir rocks exhibit significant heterogeneity in petrophysical properties. In the industry, it is a common practice to estimate permeability with porosity vs. permeability (poroperm) relationships derived from core data. However, in our study field, the semi-log crossplots of core porosity and permeability generally exhibit a wide spread. As a result, the poroperm models determined from these crossplots can make quite inaccurate predictions about permeability. In sedimentary rocks, variations in pore geometrical attributes define distinct flow units. Within each flow unit, the rocks exhibit similar fluid-flow characteristics and consistent petrophysical properties. Therefore, core samples belonging to the same flow unit generally exhibit better porpoperm correlations than those from the entire well. Based on this principle, we first classify reservoir rocks into a number of facies and then define a poroperm relationship for each facies based on core measurements. The method requires a set of well logs sufficient to classify the reservoir rocks into the distinct facies. In some cases basic logs such as GR, density and neutron porosity will be sufficient, but in other cases additional logs will be required to correctly differentiate the facies. Core measurements are only needed in a key well penetrating the reservoirs under study. The following is the detailed workflow: 1. Apply a clustering algorithm to well log curves to assign facies to cored intervals. 2. For core samples in each facies, develop a poroperm relationship based on measured core porosity and permeability in this form: logK = A*phi+B. 3. Train a self-organizing map with well log patterns associated with each facies at the cored intervals and propagate the facies classification to un-cored intervals using select log curves. 4. Use the poroperm relationships defined for different facies to calculate a continuous permeability curve for the entire well. In our study field, wireline triple combo logs and core data were collected in 7 wells. The clustering algorithm identified 5 facies from cored intervals in one key well. The facies classification was then propagated to un-cored intervals in the 7 wells using well logs. Based on core data from the key well, 5 poroperm relationships were established for the 5 facies using regression and continuous permeability curves were calculated from these relationships for the 7 wells. There is an excellent match between predicted and core permeability in all 7 wells. In contrast, a single poroperm relationship that ignores rock facies produces permeability predictions that fail to reflect the full variation in the core measurements in each well. In this report, we show the interpretation results from two wells as validation of the proposed workflow.