Christine Groves

Supersonic ejectors offer a promising solution for extending low-pressure well life and increasing total power output in geothermal plants by entraining low-pressure fluid using a high-pressure primary flow. While steam supersonic ejectors are widely used in industrial applications such as refrigeration, their deployment in geothermal power generation remains largely unexplored, leaving a critical gap in field-validated numerical models for utility-scale two-phase systems. To address this, this study presents a 3D Computational Fluid Dynamics (CFD) framework validated against industrial-scale field tests conducted at the Theistareykir Geothermal Power Plant in Iceland (connecting wells ThG-11 and ThG-15). Four RANS turbulence models (Standard $k-\epsilon$, RNG $k-\epsilon$, Realizable $k-\epsilon$, and $k-\omega$) were evaluated in ANSYS Fluent using a homogeneous Eulerian wet-steam formulation. The Realizable $k-\epsilon$ model demonstrated superior accuracy, achieving the lowest absolute error (6.4%) against field data. While non-equilibrium thermodynamic relaxation caused a systematic 8.9–12.6% overprediction in primary motive flow, secondary entrainment predictions closely tracked physical performance, with entrainment ratio errors reaching 0.00% under stable operation. Crucially, the model identifies operational boundaries: while field data places the physical backflow limit at an inlet pressure ratio of $2.2$ ($9.9\mathrm{bar}$ difference), numerical divergence near zero-entrainment establishes a conservative modelling threshold at $2.6$. By defining these physical and numerical limits while quantifying nozzle-sizing safety margins, this work provides a verified benchmark for scaling up CFD models for full-capacity geothermal ejector networks.