Artificial Intelligence (AI) Natural Language Processing (NLP) Applications For Identifying Suspended Job Root Cause

Abstract

One of the characteristics of Big Data is data variety. This applied to work-over rig operations, which deliver reports to inform daily and summary rig program and activities, known as tour-report. Tour-report contains unstructured data, which requires extensive, time-consuming effort to read, digest, and extract meaningful information for further handling and analysis. The root causes of suspended jobs are critical to determine remediation strategy. In some cases, suspended jobs with severe problems are forced to be idle wells or long-term closed (LTC). Here are a few examples of suspended job causes, i.e., killing problem or high casing pressure, wellbore assembly overpull or stuck, fishing left in the wellbore, and well mechanical completion issue. This information can be taken by digesting unstructured data available in tour-report to come up with a remediation program, i.e., rework with rig capable of handling high casing pressure, fishing job, and wellbore repair. To gain insight and knowledge  discovery of tour report, PT Pertamina Hulu Rokan (PHR) Heavy Oil Asset Optimization Team (HO AOT) and Integrated Optimization Decision Support Center (IODSC) are collaborating to deliver pilot of Artificial Intelligence (AI) Natural Language Processing (NLP) approach to identify suspended job root causes from idle wells (~800 wells in Heavy Oil fields) to come up with a remediation strategy, mainly in supporting the reactivation of LTC wells in the Rokan field program. This study shows the impact of fuzzy string-matching in handling data preprocessing. In this study, several algorithms are combined with the fuzzy string-matching method. Based on this study, fuzzy string-matching can improve the performance of classification by more than 100% in most all algorithms used.