Big Data and Machine Learning Optimize Operational Performance and Drill-Bit Design

Abstract

This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 202243, “Use of Big Data and Machine Learning To Optimize Operational Performance and Drill-Bit Design,” by Simon Cornel, SPE, Baker Hughes, and Gonzalo Vazquez, Senex Energy, prepared for the 2020 SPE Asia Pacific Oil and Gas Conference and Exhibition, originally scheduled to be held in Perth, Australia, 20–22 October. The paper has not been peer reviewed. Time savings and bit longevity are major challenges in coal-seam gas (CSG) unconventional fields onshore Queensland. Maximizing rate of penetration (ROP) on the basis of optimal drilling parameters was the key to tackling these issues. A formal process for optimizing performance was developed, with a focus on optimizing polycrystalline diamond compact (PDC) bit design and drilling hydraulics and developing a drillers’ road map. As a result, ROP increased from 50 to 150 m/h. Time savings of more than 150 hours for the drilling campaign was achieved. Background The drilling campaign encompassed two areas for development. Field A is a CSG acreage 45 km southwest, and Field E 30 km northeast, of Wandoan, Queensland. Eighty vertical wells would be drilled across the two fields. This would provide the opportunity to trial the data-mapping concept to increase ROP and fine-tune bit designs to improve longevity. In addition, the rig-sensor data set provided the basis for creating a machine-learning model to help improve drilling parameters during the production-drilling phase. Drilling Challenges The three primary challenges to be addressed included the following: - Economics for unconventional fields, specifically delivering time savings in order to lower the average cost per well - Control vibrations to avoid twistoffs and bottomhole-assembly fatigue - Consistent performance The benchmark for the initial project was an on-bottom ROP of 80 m/h. Therefore, the target for this campaign was to achieve an average on-bottom ROP of over 100 m/h and drilling-time savings of 1 hour per well. Methodology Bit optimization typically has been an iterative process, taking bit-record information to make assumptions about bit performance before providing recommendations for the next well. A more-detailed approach may include the use of log data (i.e., gamma ray and sonic) to estimate rock strength and overlay the depth-based well data to identify key incidents that affected performance (both positive and negative) so such incidents can be mitigated in future wells. This approach, while exhaustive, takes time to complete and is open to subjective interpretation of the engineer performing the analysis. When drilling CSG wells, typical well duration is 3–5 days. With short turn-around times between wells to analyze the data, the goal in this study was to develop an automated approach using real-time data gathered from the rig and present it to the field and rig crew in a visualized method. A two-step approach was devised in which data from the rig was used, followed by a real-time model.