Improving Drilling Performance Through Intelligent and Automated Operating Parameter Planning

Abstract

Abstract Drilling operations are subject to drilling dynamics dysfunctions like bottom hole assembly (BHA) shocks and vibration (S&V) that can lead to premature downhole tool failures, poor hole quality and reduced rate of penetration (ROP). This is addressed by domain experts by selecting optimum drilling parameters based on their knowledge and offset performance. A hybrid approach was developed that blended machine-learning algorithms and domain knowledge to automate and improve the design of an optimum operating window for the best drilling parameters (weight-on-bit (WOB), top drive rotary speed (RPM) and flow rate) using historical offset data. The philosophy for optimum operating parameter intelligent planning is based on evolution and continuous learning from historical data. The workflow for the hybrid approach begins with the automated selection of offset well data, then an algorithm is used to align data by formation tops and automatically detected drilling zones which enables parameter statistics evaluation along the depth of the subject well. Based on offset well S&V data and computed mechanical specific energy (MSE), a machine-learning algorithm is applied to divide the intervals into drilling zones, such as benign drilling zone, hard stringer zone, interbedded zone, and severe S&V zone. When the optimum operating window is generated, more aggressive parameters are explored for benign drilling zones to maximize ROP, whereas for the other zone types, the best parameters are automatically identified through the usage of sweet spots in self-generated S&V heatmaps in offset wells. If no stable operating window is found in the heatmaps, domain mitigation logics are used based on the observed S&V modes (axial, lateral, or torsional) in offset data. The parameters’ roadmap is then executed and monitored in real-time, and the post-run data are collected to feed into another evolution cycle for continuous learning and improvement. We implemented this intelligent parameter planning workflow into a mature drilling data analytics platform for deployment to domain users. It was used by drilling engineers in four different drilling environments for field testing to replace the legacy manual drilling roadmap preparation process. Considerable time savings have been reported for offset well data collection and analysis for parameter roadmap generation. More importantly, jobs executed with the new roadmaps have demonstrated reduction in both drilling time and downhole dysfunctions. In this work, we leveraged data analytics, domain knowledge, and physics and machine-learning models to generate a depth-based optimum operating parameter roadmap automatically. This workflow enables ROP improvement and the reduction of drilling risks, becoming a key element for the evolution of autonomous drilling operations and well cost reduction.