Real-time incident detection in geothermal drilling operations using machine learning solutions

Abstract

Geothermal energy, while a reliable low-carbon resource, represents only a small fraction of global renewable capacity due to high upfront costs and resource risks. Drilling wells accounts for a substantial portion of geothermal project costs; successful drilling is thus crucial for geothermal projects. Finishing wells on-time and on-budget has always been a challenge for operators and mostly has modest success rates.
Overall drilling efficiency can be improved by increasing the rate of penetration (ROP) and reducing non-productive times (NPT). Geothermal systems present unique challenges like fault structures, severe lost circulation, and high temperatures that lead to potentially expensive incidents. Leveraging machine learning technologies offers promise in addressing these drilling challenges and optimising operations.
The drilling incident detection system using systematic and automated feature engineering on time-series data was able to successfully predict incidence of hole-related NPT. This system can potentially be deployed and assist drilling personnel in making tactical decisions during drilling, in the future, thereby reducing operational risks and enhancing overall drilling efficiency and cost performance.