Full Field Behind Casing Opportunity Through Machine Learning: A Case Study in a Giant Offshore Brown Field

Abstract Previously, due to its sheer size and complexity, the approach towards a giant brown field located offshore East Malaysia has always been piecemeal and area specific. Furthermore, with the field having undergone several decades of data acquisition with multiple successions of petrophysicists in the subsurface team, extensive minor differences have been observed in the petrophysical interpretation of the well logs. These slight inconsistencies arose from the different parameters used by the different petrophysicists at different times, with each having different experiences. With the field currently in production decline, one of the methods undertaken to arrest this decline is through identifying Behind Casing Opportunities (BCO) for add perforation. However, finding a time-effective method for evaluating all possible BCO candidates within a field this large, together with correcting for the inconsistencies in the interpreted petrophysical logs have proven to be a big challenge. As a solution, the team has embarked on the adoption of machine learning through an in-house well log data artificial intelligence workflow. This workflow is a multi-stage field-based machine learning model that is developed to predict reservoir characteristics in logged wells. Apart from reservoir properties, this machine learning model also outputs a BCO flag which is used for identifying possible add perforation candidates. Running this well log data artificial intelligence workflow has allowed the subsurface team to generate a full inventory of BCO add perforation candidates in the field for the first time in recent years. Subsequent tabulation of each well according to its cumulative BCO pay thickness has greatly helped the team in applying a systematic targeted approach for BCO execution instead of painfully going through well-by-well review which could be very time consuming. With the workflow trained on wells with the highest confidence on its petrophysical evaluation, the predicted results gave a more consistent field wide reservoir properties, greatly removing localized human bias and reducing the petrophysical interpretation uncertainty. With the subsurface team now able to focus its attention on wells with the highest potential for add perforation instead of a blind shotgun approach, and with consistencies in petrophysical evaluation logs also in hand, the BCO screening and maturing process was made leaner and faster. As a result, several add perforation candidates have been identified for execution in an upcoming campaign to help arrest the field's production decline.