Stuck Pipe Detection in Geothermal Operation with Support Vector Machine

Abstract

One of the biggest problems during drilling operation is a stuck pipe in which the drill string would stick or freeze in the well. This challenge leads to a significant amount of remedial costs and time. Many researchers have investigated different factors regarding the stuck pipe. These factors include poor hole cleaning, improper mud design, key seating, balling up of bit, accumulation of cutting and caving, poor bottom hole assembly configuration, differential pressure, etc. (Chamkalani, Pordel Shahri, and Poordad 2013). Since geothermal drilling targets lost circulation zones at reservoir depth, the chance of getting stuck pipe events becomes higher. Many publications reported that lost circulation events that lead to stuck pipe events have become the top non-productive time (NPT) contributor to costs in many geothermal drilling projects. The consequences of a stuck pipe are very costly, that include lost time when releasing the pipe, time and cost of fishing out the parted Bottom Hole Assembly (BHA), and efforts to abandon the tool(s) in the hole. Despite many observations that have been done to develop a system in avoiding stuck pipe incidents in oil and gas drilling operations using artificial intelligence (AI), few works have been developed for geothermal drilling operations. In this research, we propose a method to build an early warning system model for stuck pipe conditions based on a Support Vector Machine. Based on the experiment result Support Vector Machine Algorithm showed good performance with 89% accuracy and 81% recall for limited training dataset.

Keywords

• Stuck Pipe, Machine Learning, Geothermal Drilling, Support Vector Machine, Artificial Intelligence