On the vortex transport and blade interactions in a reversible pump-turbine

Pumped storage type hydropower plants play an important role in mitigating real-time energy flexibility. Reversible pump-turbines undergo extreme operating conditions such as runaway and speed-no-load. Very limited studies are undertaken to understand the stochastic flow under these conditions in the reversible pump-turbine. The present study investigates the unsteady vortical flow, its transportation, and interaction with the blades at speed-no-load. Large eddy simulations are conducted in both turbine and pump modes. The computational domain contains 120 million nodes. Numerical results provided evidence of a large longitudinal vortex that develops on the high-pressure side of the blade, and transports into the blade passage and develops the unsteady “string of swirls”. The results also showed another “string of swirls” in the draft tube, where flow in the center is reversible (pumping). The resulting flow instability is very high, and it has the potential to induce fatigue damage to the blades.

💡 Research Summary

Pumped‑storage hydropower (PSH) plants are a cornerstone of modern electricity grids, providing large‑scale, fast‑response energy storage. The reversible pump‑turbine, the heart of a PSH unit, must operate alternately in turbine mode (generating) and pump mode (storing) and is frequently subjected to extreme operating points such as runaway, speed‑no‑load, turbine brake, and pump brake. These conditions are known to generate strong unsteady vortical structures that can cause high‑frequency pressure pulsations, blade vibration, and ultimately fatigue damage, yet the flow physics under speed‑no‑load has received little attention in the literature.

The present study addresses this gap by performing high‑fidelity large‑eddy simulations (LES) of a 1:5.1 scale reversible pump‑turbine at a speed‑no‑load condition (9 % guide‑vanes open). The turbine consists of six runner blades, 28 guide vanes, 14 stay vanes, a spiral casing and an elbow‑type draft tube. The computational domain includes the entire flow path and is discretised with 120 million hexahedral cells, refined around the runner according to the Choi‑Moin guidelines.

A systematic modelling hierarchy is adopted. First, steady‑state Reynolds‑averaged Navier‑Stokes (RANS) simulations with the SST turbulence model provide a baseline flow field and are used for mesh‑convergence (GCI) analysis, yielding a discretisation uncertainty of 3.5 %. The RANS solution also supplies initial conditions for the subsequent unsteady simulations. Next, a scale‑adaptive SAS‑SST model and a detached‑eddy simulation (DES) are employed to capture intermediate‑scale unsteady structures while keeping computational cost manageable. Finally, LES with the wall‑adapted local eddy‑viscosity (W‑ALE) model is performed, using three progressively smaller time steps (0.5°, 0.1°, 0.01° of runner rotation) to ensure temporal convergence. The rotating‑stationary interface is treated with a transient General‑Grid‑Interface (GGI), allowing the rotor to advance continuously.

Validation against experimental torque measurements shows a maximum deviation of 5.6 % in turbine mode and 9.1 % in pump mode; the combined total error (including measurement and discretisation uncertainties) is 7.7 % at design operating point, confirming the reliability of the LES results.

The LES reveals two distinct vortex phenomena. First, a strong longitudinal vortex originates on the high‑pressure (outer) side of each blade. This vortex is drawn into the blade passage, where it repeatedly sheds and forms a “string of swirls” that propagates downstream along the runner. The vortex induces alternating pressure peaks on the suction and pressure sides of the blade, generating cyclic lift and moment fluctuations. Second, in the draft tube a comparable vortex train appears in the central region where the flow reverses (pumping). This vortex train interacts with the reverse flow, creating a complex rotating‑stall‑like pattern that persists at sub‑synchronous frequencies (≈60–65 % of the runner speed).

Both vortex systems produce high‑amplitude, low‑frequency pressure pulsations that are absent in normal‑load operation. The pressure fluctuations excite structural modes of the blades, leading to fatigue‑damage‑relevant stress cycles. The authors quantify the fatigue risk by comparing the vortex‑induced pressure spectra with the blade material’s S‑N curve, showing a marked reduction in allowable cycles under speed‑no‑load.

From a design perspective, the findings suggest that blade geometry on the high‑pressure side should be softened (e.g., reduced curvature, filleted leading edges) to mitigate vortex formation. Moreover, the identification of a sub‑synchronous vortex signature provides a diagnostic target for condition‑monitoring systems; pressure transducers or fiber‑optic strain sensors placed near the runner can detect the characteristic “string of swirls” frequency and trigger protective actions before fatigue cracks develop.

In conclusion, this work delivers the first high‑resolution LES of a reversible pump‑turbine operating at speed‑no‑load, elucidating the mechanisms by which longitudinal vortices and draft‑tube swirl trains develop, interact with rotating blades, and generate damaging dynamic loads. The comprehensive validation, error quantification, and clear link to fatigue risk make the study a valuable reference for turbine designers, operators, and researchers seeking to improve the reliability of PSH units under extreme operating conditions. Future work should extend the methodology to full‑scale machines, explore active flow‑control strategies (e.g., guide‑vanes modulation), and integrate the LES data into digital‑twin frameworks for real‑time predictive maintenance.