Value of Downhole Drilling Data for Real-Time Formation Evaluation and Early Decision Making

Abstract

Abstract Under the environment of complex wellbore trajectory, deviated wells, or lateral drilling, be it for oil and gas, geothermal or carbon-sequestration, the newly developed downhole sensors, data-transmission in real-time together with data-analytics has enabled utilize all information contained in weight-on-bit, rate-of-penetration, rotational speed, torque, effective mud-weight used, and drilling vibrations. The high-resolution drilling data helps us mitigate drilling dysfunction and complement formation logs. In some cases, when conventional log acquisition is not possible, downhole drilling data may become the only source for formation properties. The paper presents our investigations in the standalone use of downhole drilling data. The process evaluates formation properties and integrates with other petrophysical logs to improve real-time data quality and interpretation. Case studies are included from both lab experiments and field examples. The current research efforts are encouraged by technological developments, in downhole sensors, which can now be included with the bottom-hole drilling assembly and used in in-situ acquisition of drilling data. We use downhole drilling data and consider both rule-based and machine-learning (ML) methods, to evaluate formation for lithological and geomechanical heterogeneities. A well from Gulf of Mexico is selected to apply the concept. Results are verified against established log-based formation properties estimation. In wells with high deviations, or complex trajectories, downhole drilling data is found to be more reliable compared to use of similar measurements made at surface followed by surface-to-downhole conversions. In the present work, we trained the system using 75% of downhole drilling data, together with measured vibrational and bending moment. The model was then applied in the remaining 25% of untrained intervals and demonstrated ability to predict formation properties with good correlation. The evaluation was done on memory-based data received after drilling, however, can also be implemented as automatic processing to support real-time operation. Drilling data is always available, whenever a well is being drilled. Real-time formation evaluation, based on this information, provides the drilling engineers and geoscientists an additional resource which can be either standalone or complementary to log data. Downhole drilling data is also available early in time compared to LWD sensors typically placed several tens of feet behind the bit. Ongoing improvements in sensor specifications, data quality and interpretation methods, is promising especially for environments where conventional real-time logging is not feasible. For robust ML application, training data from a range of geographical regions and reservoir is included to ensure correct prediction in all scenarios.