Abstract:
Deviated oil and gas wellbores were drilled by competent and experienced directional driller (DD) to deliver the wellbore's trajectory on target and on time. The variable human factor and rapid changes of various surface and downhole parameters present challenges for consistent performance every time. A dynamic physics-based machine learning (ML) algorithm leverages on digital twin data analytics and real-time downhole measurement to autonomously steer Rotary-Steerable System (RSS) drilling assembly was employed. This case study describes the methods, the algorithm learning, assimilates the data in real-time, and autonomously steers. Operator's in-field redevelopment plan consists of drilling two oil producers and one water injector side-tracking from donor wells. Due to congested nature of the platform with many producing wells through multi-stacked reservoirs, the wellbore profiles are complex 3D to tap into the various reservoir target sands and avoiding close approach to nearby drilled well. Meticulous well engineering and Bottom-Hole-Assembly (BHA) analysis were performed during the pre-drill planning stage to ascertain the directional performance of the RSS BHA, sensitivity study, offset directional performance and risks were thoroughly assessed. The autonomous steering algorithm models the directional performance tapping into the vast database of digital twin, expected directional performance, evaluates past yields, projecting ahead and constantly adjusting parameters such as steering aggressiveness, dogleg severity, turn rate, whilst staying within safety margin of anti-collision if any to deliver the wellbore to target. The speed of the computation from downhole sensor measurement, coupled with high-speed telemetry of the data to the surface allows for systematic increased in speed of real-time data processed, culminating to ML autonomous steering for RSS BHA to deliver a smoother wellbore that is not possibly with human manual calculation. The complex 3D well profile entails building wellbore angle from 40 degrees to 80 degrees before dropping to 67 degrees while turning azimuth from 190 degrees to 85 degrees with 3.5 degree per 30 meters dogleg severity. After initially side-tracking performed manually by the Directional Drillers at the rig site, autonomous steering algorithm was executed to directionally drill the wellbore through all the planned geological targets till well's total depth. 1224 meters were successfully drilled in autonomous mode without any DD intervention for 3 days of drilling with average 30-45 meters per hours rate-of-penetration. This resulted in 97 percent of wellbore autonomously steered and placed optimally through all planned geological targets, with 32 percent faster drilling compared to offsets. Eighty-six autonomous closed loop steering command flawlessly executed downlink saved fourteen hours of rig time, eliminating invisible loss time translating to faster on bottom drilling. The digital transformation with advances in ML and artificial intelligence, provided impetus drilling automation, to a paradigm shift on how we traditionally drill directional wellbores. Β© 2023, Society of Petroleum Engineers.
