Data-Driven Performance Optimization in Section Milling

Abstract

One of the major efforts in Oil and Gas industry's digital transformation is the increased use of data in optimizing processes. This paper focuses on optimizing the process of section-milling during well abandonment by leveraging data gathered from past section-milling cycles. Several mathematical model-based techniques have been presented in recent years for improving the rate of penetration (ROP) in section-milling. However, only a few data-driven methodologies have been adopted in this field of interest, most likely due to unavailability of data. A trainingsubset of field data from section-milling operations is used for developing a range of machine learning models. Performance of these models is then evaluated using mean absolute percentage error (MAPE) against testing subset of data.