Using Data Analytics to Improve Drilling Rate of Penetration

Abstract

Abstract It takes an average of 7 days to drill through the Travis Peak formation in the southern Haynesville basin. The objective of this paper was to improve the rate of penetration (ROP) and therefore reduce the time taken to drill through this formation. Machine learning and regression models were applied to electronic drilling record data and identified that the auto-driller ROP setpoint was the most influential variable ROP for wells drilled in the Travis Peak formation. This finding was corroborated during field visits with well site leaders and drillers with the discovery that ROP was being used as the controlling setpoint in the auto-driller system for drilling in the Travis Peak and Calvin Sand formations. After discussion with the rig provider, a new set of auto-driller parameters was proposed using weight on bit as the controlling setpoint with the objective of increasing the rate of penetration. The new auto-driller parameter roadmap was piloted on two wells. ROP performance for these two wells was significantly better than the nearest offset and other offsets in the nearby area. The drilling team estimate about two days per well time saved as a result of the improved ROP.