Data-Driven Approach to Quantify Uncertainty in Wellbore Management through Temperature Logs

Summary We demonstrate the use of data-driven machine learning model to automate the process of analysing temperature logs to aid with production management in mature fields. Temperature log analysis has many applications such as evaluating formation productivity ( Bird et al., 1965 ), Estimating thermal conductivity (Seto et al., 1991), and skin damage determination ( Schindler et al., 2015 ). We are specifically targeting the use of temperature logs to quantify the risk of a wellbore leak. We built a machine learning pipeline that autonomously quantify the risk associated with every temperature log. The model is trained with hundreds of labelled historical temperature logs. This is motivated by the high number of acquired temperature logs in mature fields and the growing number of aging wells. In addition, Automating the analysis of temperature logs enables us to utilize the rich history of acquired logs to fine-tune well selection for future surveys (e.g. Zangel et al., 2016) and to quantify the risk associated with a given spatial coordinate and operational condition. (e.g. AlAjmi et al., 2015 ). This methodology contributes to a sound well production management strategy in mature fields with aging wells.