Evolution from Water-Oil Ratio to Tree Based Classifier - A Novel Methodology for Effective Diagnosis of Water Production Mechanism in Oil Wells

Abstract

Water an important component of any life process plays a vital role in oil and gas production. As hydrocarbon reservoir depletes water production is on an increase. The management of the excessive water to minimize the environmental and economic impact is a critical issue for all Oil and Gas companies. Such companies set aside lot of resources for managing produced water however due to lack of methods to diagnose the underlying water production mechanism (WPM) the best of the water shutoff technologies cannot be used. This paper presents a novel methodology for diagnosing WPM by extracting key features from water/oil ratio (WOR) graphs and integrating them with static reservoir parameters. Non-parametric tree-based random forest ensemble classifiers are constructed for predicting the WPM. Further their interpretably is enhanced by generating a single logistic Model tree by incorporating smeared cases classified using ensemble classifier. Our results show high prediction accuracy rates (over 90%) and easy to implement workflow. By adoption of this methodology oil and gas companies can accurate and timely diagnose WPM saving considerable time and money.