Application of a Sustainable Cognitive Twin Solution for Drilling Hazard Avoidance and Optimization

Abstract

Abstract Model-based methods such as torque and drag, hydraulics, hole cleaning, and downhole dynamics have improved over the past years for pre-job design and post-analysis of drilling operations. However, there is a still a large gap in extending these traditionally manual processes and scaling them up for model-based, real-time decision support systems. Significant opportunities exist in applying these towards drilling hazard avoidance, performance diagnostics, and optimization. It is desirable to have an automated solution that orchestrates data ingestion and model calculations based on physics-based, data driven, or hybrid methods, to analyze real-time field data and enable valuable insights into operations. In this context, cognitive twins refer to the ability to deliver actionable insights for real-time optimization and hazard avoidance through systematic automation of data and model pipelines. In this work, a sustainable real-time solution is proposed comprising of data aggregation, cleansing, analytics, and cognitive twins to enable the automated orchestration, analysis, and insight delivery from drilling models. This process drives the scalability of the system – ensuring evergreen models and analysis are always available across all rigs and wells in a drilling fleet. Methods of modeling and analysis combine data driven methods with physics-based constraints to compute certain key quantities (e.g., friction factors, pressure deviations, and cuttings bed height). These quantities contain results and intermediate features which can be used for machine learning methods to drive fleet performance optimization above and beyond well-level analysis. We applied these methods to both onshore and offshore wells in real-time to identify and automatically detect several early warning signs of common drilling problems such as stuck pipe, washouts, and poor hole cleaning. Issues related to friction, hydraulics, cuttings transport, general drilling efficiency, and combinations of the previous categories are translated into actionable insights, which can be immediately implemented in the field to enhance safety, reduce non-productive time, and increase performance. In addition, the insights were also analyzed at a fleet-wide scale to identify patterns and drive operational best practices. Common challenges and solutions relating to implementing such a system at scale are addressed. This methodology addresses and overcomes many of the challenges involved in applying digital twins in a sustainable, real-time mode. It extends functionality into cognitive twins that can analyze modeled behavior to real-time results and enables automated insights to be generated for single models such as torque and drag in isolation and combinations of models such as the insights derived from interpreting torque and drag, hydraulics, and hole cleaning results simultaneously.