ESP Multiphase Boosting Mechanisms and Condition Diagnosis

Research / Area 2

When free gas enters an electrical submersible pump (ESP), boosting pressure degrades, then surges, and may end in gas lock. Traditional homogeneous-flow design applies only at very low gas fractions. The group studies the chain from bubble breakup and coalescence in the rotating impeller to in-situ gas void fraction, flow-pattern transition and boosting degradation, extending to high-viscosity, emulsion and sand-laden conditions.

Experiments and mechanistic models. On the TUALP multistage two-phase ESP loop (14-stage TE-2700 with stage-by-stage taps) and the CUP centrifugal-pump visualization rig, high-speed imaging captures impeller flow patterns and surge onset. Critical bubble diameter, in-situ gas fraction and flow-pattern boundaries are modeled; a “best-matching rate” concept supports unified boosting models for single-phase, viscous, emulsified and gas–liquid conditions, validated on three ESP types (typical error within ±20%, average ~10%).

CFD. Eulerian two-fluid CFD with interfacial forces captures multistage gas–liquid flow and cross-stage pattern evolution; dynamic mode decomposition extracts dominant impeller modes. An NSFC General Program project addresses deep-sea mixed-flow ESPs.

Wide-range offshore ESP hydraulic design and automated CFD optimization. For mid-to-late-life offshore wells with large rate swings and high gas fractions, the group designs wide-passage mixed-flow ESP impellers/diffusers (400/538 series): similarity scaling, Euler-head correction and specific-speed checks seed the base model; Python drives CFturbo–Workbench–PyFluent for geometry batching, mesh update, batch solve and surrogate-model search (Latin hypercube + Gaussian process). First-round sensitivity (Spearman) shows impeller exit width b₂ dominates BEP location (ρ≈0.74), cutting the design space from 10-D to 5-D.

Condition diagnosis. Mechanistic boosting models are fused with machine learning into a real-time ESP management framework (condition recognition, fault diagnosis/early warning, remaining life), delivered as the “Intelligent ESP Management System V1.0” algorithm module.