Authors: N. Jelic, T. Kolsek, A. Bergant, J. Duhovnik
Background
The water flows through a long pipeline to the power house with water turbines
to generate power of approx 100 MW. To ensure safe operation during sudden
turbine shut-down, the kynetic energy of the water just before the power house
has to be dissipated, until the water is stopped. Two vertical cylindrical
pressure chambers have been foreseen, each equipeed with Johnsons needle type
valve (see Fig. 1). The water flows through the needle valve, forms a jet
and the energy is dissipated into turbulence.
Fig 1: General setup of the power dissipation system
This system is characterized by a variety of parameters that have to be
determined before cost-effective design and proper safe and reliable
operation can be implemented. We have conducted a study to determine
optimal chamber height for the predicted water power to be dissipated
discharge coefficient for various valve openings
assessing oscillations at both model and prototype scales (similarity rules)
comparison with preliminary measurements at the model scale
Description of the system
The system in question is shown in Fig 2. The needle valve (Fig. 3)
transforms the potential energy of the water into kinetic energy
in order to be further injected by a high speed of several tenths of m/s into
the chamber. This energy is then dissipated into thermal energy.
Fig 2: The chamber geometry
Fig 3: The valve geometry
The interval of the design head H of the selected prototype valve was
between 100 and 200 m. The nominal power which was dissipated was 60MW.
The diameter of the pressure chamber was dch = 7 m and the nominal height
was h= 17 m. The valve opening was adjusted continually between 10%
and 100% for various pressure heads.
Method
CFD calculations were performed using ICCM-Comet computer code based
on Reynolds averaged Navier-Stokes (RANS) equations. The flow was assumed to
be viscous and turbulent. The system of RANS equations was closed using the
well-known k-epsilon turbulence model. Geometry modeling and grid generation
were performed using SDRC I-DEAS software, which was also used for additional
structural investigations of the mechanical valve and chamber properties.
The typical number of cells we used was 100.000 for the entire system.
Due to excessive vibrations of the chamber used in the experiment we decided
to perform transient calculations. When a water jet exits a pipe with sudden
expansion, self-sustaining oscillations can be expected. The static pressure
was observed both in numerical and laboratory experiments at several
characteristic locations inside the chamber (top, cylindrical wall, bottom).
The most representative results were obtained at the measuring points at
the bottom of the chamber.
The results
The chamber height was varied in order to find an optimum value that gives
maximum power dissipation at an acceptable amplitude of pressure oscillations.
The static pressure oscillations at the bottom of the chamber for different
chamber heights are shown in Fig 4.
Fig 4: Oscillation of static pressure at the chamber bottom for several heights
Fig 5: The peak to peak pressure variation depending on chamber height
The underlying physics of the observed oscillations is related to the feedback
effect of confinement and the consequent formation of recirculation water
pockets inside the chamber. Fig 6 and Fig 7 show that the velocity field
distribution qualitatively exhibited large changes over time.
The water jet periodically changed direction.
Fig 6: Velocity vectors in time 1
Fig 7: Velocity vectors in time 2
In Fig. 8 we present the discharge curve obtained by experiment as compared
with the numerical results for the valve openings of 33, 67 and 100%. It can
be seen that the computed discharge coefficient was slightly overestimated
(by about 8%). This can be explained by geometric details that were not part
of the numerical model (inner supports holding the valve and decreasing the
inflow surface by about 5 %).
Fig 8: Computed and experimental discharge coefficients
We investigated the similarity of the prototype to the model. Linear scaling
at lambda=14 was determined to be appropriate for both building the experimental
setup and measuring the estimated total pressure of the prototype at model scale.
The velocity and pressure distribution over time were observed and compared for
model and prototype. The frequency spectrum of the pressure variation is shown
in Fig. 9. Preliminary experiments were performed at model scale at the
Institute of Hydraulic Research, Ljubljana, Slovenia. Oscillations were
observed with a pronounced peak slightly below 0.2 Hz. The FFT spectrum is
shown in Fig. 10.
Fig 9: FFT of pressure oscillations obtained in model scale computations
Fig 10: FFT of oscillations obtained in model scale experiment
