Interventional Causal Identification of Dissolution-Buffered Gas Influx in Managed Pressure Drilling

• Hamza Qureshi

  University of Lakki Marwat, Tajazai Road, Lakki Marwat 28420, Khyber Pakhtunkhwa, Pakistan

  Author

• Bilal Mahmood

  Khwaja Fareed University of Engineering and Information Technology, Abu Dhabi Road, Rahim Yar Khan 64200, Punjab, Pakistan

  Author

Managed pressure drilling operates the wellbore as an actively actuated hydraulic system in which choke and pump commands continually reshape pressure and flow signals. This operating mode reduces large excursions but makes early gas influx discrimination more difficult because the dominant surface observables are simultaneously driven by control interventions, compressible storage, frictional variability, and multiphase inventory changes. In oil-based and synthetic-based fluids, part of an invading gas mass can dissolve into the circulating liquid and induce swelling, producing pit-volume and density shifts without an immediately dominant free-gas holdup. This paper formulates influx diagnosis as an interventional causal identification problem rather than as a residual-threshold task. The technical contribution is a structural causal model for annular hydraulics with thermodynamics-augmented latent inventories, paired with an invariant-causal learning objective that exploits routine control interventions as quasi-experiments. A constrained causal discovery layer identifies a minimal adjustment set for influx effects under partial observability and sensor bias, while a counterfactual simulator computes risk-relevant outcomes under hypothetical choke trajectories. The method separates actuation-induced transients from influx-driven signatures by enforcing invariance of causal mechanisms across intervention regimes and by penalizing spurious predictive paths that fail under do-operations. A practical estimation algorithm is derived using intervention-conditioned variational inference with inequality-consistent parameterizations for phase partition and swelling. Numerical studies demonstrate improved separation of swelling-consistent pit gain from free-gas-driven compressibility, stable behavior under intermittent telemetry and closure mismatch, and calibrated counterfactual risk estimates for short-horizon decision support.

• PDF