Machine learning prescriptive well log quality analysis determination of casing effects

Well logging represents a key discipline to maximize information about the subsurface reservoir from the wellbore information. Determining reservoir properties from limited wellbore information represents a major challenge that is further exacerbated by wellbore effects. Specifically, casing may have a considerable effect on the well logging measurement performance quality, and affect interpretation. Reinterpretation of existing well logs and information has become more profound in order to enhance geosteering and utilize the well log data in order to evaluate the formation in real time. Determining casing effects from the well log data via an intelligent way, and providing prescriptive approaches to overcome the potential impact, represents a key challenge for this attempt. In this study, we offer a novel prescriptive artificial intelligence framework for the categorization of casing vs non-casing effects on well logs. The framework integrates well log data including gamma ray (GR), caliper (CA), and neutron (NEUT) to determine whether the well logs exhibit any casing effects or not. We utilized the decision tree classification framework with a CART split criterion and compared several ensemble classifier algorithms to enhance the casing impact assessment accuracy. The framework demonstrates a good performance on classification with strong accuracy scores. Specifically, it provides prescriptive recommendation on the determination of the quality of the various well logs and the effect caused by the casing. The new method offers a reliable prescriptive well log data quality analysis and interpretation for reservoir characterization.