Capacitance Resistance Clustered Model for Mature Peripheral Waterflood Performance Prediction & Optimization

Abstract

Optimizing water injection rate distribution in waterflooding operations is a vital reservoir management aspect since water injection capacities may be constrained due to geographic location and facility limitations. Traditionally, numerical grid-based reservoir simulation is used for waterflood performance evaluation and prediction. However, the reservoir simulation approach can be time-consuming and expensive with the vast amount of wells data in mature fields.
 Capacitance Resistance Model (CRM) has been widely used recently as a data-driven physics-based model for rapid evaluation in waterflood projects. Even though CRM has a smaller computation load than numerical reservoir simulation, large mature fields containing hundreds of wells still pose a challenge for model calibration and optimization. In this study, we propose an alternative solution to improve CRM application in large-scale waterfloods that is particularly suitable for peripheral injection configuration. Our approach attempts to reduce CRM problem size by employing a clustering algorithm to automatically group producer wells with an irregular peripheral pattern. The selection of well groups considers well position and high throughput well (key well). We validate our solution through an application in a mature peripheral waterflood field case in South Sumatra. Based on the case study, we obtained up to 18.2 times increase in computation speed due to parameter reduction, with excellent history match accuracy.