instability of pump-turbines during start-up
TRANSCRIPT
-
7/30/2019 Instability of Pump-turbines During Start-up...
1/8
Instability of Pump-Turbines during Start-upin Turbine Mode
Thomas Staubli Florian Senn Manfred Sallaberger
Lucerne University of Applied Sciences and Arts (LUASA) VA TECH HYDRO AG
Technikumstrasse 21 Technikumstrasse 21 Hardstrasse 319
CH-6048 Horw CH-6048 Horw CH-8019 Zrich
Switzerland Switzerland Switzerland
Introduction
During the last decade the deregulation in the European electricity market has resulted in rapidly changingconditions on the market. Due to the growing demand for balancing power and frequency control an investmentin increased pumped storage capacity became economically feasible. Reversible pump-turbines seem to be inmany cases the most cost-effective solution.
Occasionally torque fluctuations of reversible pump-turbines are encountered in power plants during start-up inturbine mode operation. Such fluctuations can slow down the process of synchronization what is highlyundesirable when fast peak power production is required.
During start up there is practically no load on the turbine shaft and the turbine operates close to the runawaycharacteristic. The guide vanes are opened only a few degrees during this phase.
A first case study of such oscillations on a model pump-turbine was presented by Yamabe [1] and [2]. Heobserved oscillations with pronounced hysteretic behavior which interacted with unsteady cavitation patterns. Acase study and a simple cure of the problem by detuning some guide vanes are given by Klemm [3]. A linearstability analysis to predict the occurrence of the oscillations was successfully introduced by Martin [4] and [5].Also Doerfler [6] presented a case study on how stable operation could be achieved in spite of the instability atno load.
Recent experiences with single stage reversible pump turbines are published by Billdal and Wedmark [7]. Theypropagandize multiflow guide vanes (MGV) to overcome difficulties with synchronization and to obtain stablespeed after load rejection.
All authors agree that the so-called S-shape of the four quadrant characteristic of the pump turbines isresponsible for the oscillations at no load operation. An example for such a four quadrant characteristic with S-
shape near runaway is given in Figure 1. The vertical slopes of branches of the characteristics near runaway aredirectly linked to an exciting energy transfer from the flow to the oscillating system.
In the following a numerical study will be presented which focuses on the prediction of the characteristic nearrunaway and on the flow phenomena leading to the instability. To do so, tools were developed to analyze localand time-dependent flow, momentum and energy exchange in each of the runner and guide vane channels and inthe vaneless spaces.
For validation of a model of a reversible pump-turbine with a known unstable behavior and well documentedmodel test data was chosen. The four quadrant characteristic of this model turbine is given in Figure 1.
1. Numerical flow simulation
The flow near the no load operation of turbines becomes very complex in a sense that the flow is dominated by
backflow regions and vortex formations in all parts of the turbine. Furthermore, partial pumping flows start tobuild up in some or all runner channels. Additionally, the flow becomes vigorously unsteady. Recirculationzones build up and disappear, vortical flows are swept away.
-
7/30/2019 Instability of Pump-turbines During Start-up...
2/8
To predict such flows - at least qualitatively correct - grid generation must be carried out carefully. The gridsused in this study were generated using only hexaedra elements. Grid generation was done with the commercialsoftware ICEMCFD v11.0. Grid quality parameters are listed in Table 1. Table 2 gives the boundary conditionsand the types of simulations which have been carried out.
15 deg.
10 deg.
6 deg.
3 deg.
1 deg.
runaway line
(T=0 Nm)
hg
Dnku
=
2
11
hgD
Qkcm
=
2
421
Fig. 1. Measured four quadrant characteristics of the chosen pump-turbine model
casing draft tube guide vanes
(coarse)
rotor
(coarse)
guide vanes
(fine)
rotor
(fine)Elements (hexaeder) 751 424 258 055 475 000 990 522 1330520 2360160nodes 787 705 267 792 539 240 1 069 470 1455600 2502864min. angle [deg] 30.0 28.0 28.8 25.4 28.8 25.1min. det 0.32 0.35 0.56 0.46 0.58 0.52max. vol. change 41.5 11.5 12.9 43.5 12.1 16.1max. aspect ratio 351 488 133 517 125 308ave. y+/ max. y+(for bep operation)
28.0/70.4 15.9/94.8 21.9/41.9 15.0/42.9 15.7/31.8 12.4/45.6
Tab. 1. Grid quality parameters
simulation of characteristics investigation of instability
description stationary simulation transient simulation
inlet mass flow varied (operating point)
outlet average static pressure
wall no slip
speed of rotation 1000 rpm
rotor stator interfaces frozen rotor transient rotor stator
time step physical timescale 0.005 [s] 5 degrees of rotation 1-5 degrees of rotation
turbulence modell SST
simulation time 300 iterations 3 revolutions up to 55 revolutions
imbalance of mass flow convergence
-
7/30/2019 Instability of Pump-turbines During Start-up...
3/8
Fig. 2. Computational domain of the pump turbine model
Figure 2 shows the computational domain. At the inlet to the computational domain the flow rate was given as
boundary condition and the average static pressure was given at the outlet.
2. Validation
For validation the numerical flow simulations for operation near the runaway point experimental data frommodel test were used. The validation was carried out in two steps. In a first step stationary simulations wereperformed. The demand with respect to computational power is much lower for stationary simulations comparedto unsteady, transient simulations. However, the expectations in the accuracy of the results of the stationarysimulations are low, since the flow is certainly not stationary near runaway. In spite of this fact, the stationarysimulated points follow well the measured points as demonstrated in Figure 3. General observations are that thestationary simulations predict well the slope of the characteristic near runaway and that the simulated pointsgenerally lie at lower ku1coefficients compared to the experiment.
On the other hand the transient simulations give results which are closer to the measured data and show a slightlyoverhanging characteristic near runaway the typical S-shape.
hg
Dn
ku = 21
1
measured operating
pointsstationary
simulations
transient
simulations
3810-3
m3/s
1000 rpm
Fig. 3. Comparison of measured points with results of stationary and transient simulations
3. Procedures to analyze fluxes
During mesh generation mesh-regions were defined for evaluation of local fluxes. This definition of mesh regionwhich can be surfaces or volumes allows the analysis of local time variations of fluxes and balances, e.g. in eachguide vane or rotor cannel. Figure 4a shows the control volumes defined for the guide vanes and the runner andFigure 4b shows surfaces defined in the vaneless space between runner and guide vanes and between guidevanes and wicket gate. In the following the fluxes through the entire, cylindrical surface A are analyzed.
scroll and stay vanes (yellow)guide vanes (red)
rotor (blue)
draft tube (grey)inlet (green)pinched outlet
-
7/30/2019 Instability of Pump-turbines During Start-up...
4/8
Fig. 4a. Control volumes within the turbine Fig. 4b. Control surfaces in the vaneless spaces
The flow rate is the simplest example of the flux through a surface. In a three dimensional space the mass flow is
defined by following equation:dAcntQ
A
=rr
)( (1)
The normal vectors nr
are orientated outwards with respect to the defined
control volumes and velocity vectorscr
are the absolute velocities on asurface fixed in space or the relative velocities on a rotating surface. Withthese vectors a sign of the flux is defined.
outin
A
QQdAcntQ +== rr
)( with: 0,0 >< inout QQ (2)
The absolute value of flow is: outin
A
abs QQdAcntQ == rr
)( (3)
The flow rates of inflow inQ and of outflow outQ can be determined as follows:
2)(
QQtQ absin
+=
2)( absout
QQtQ
= (4)
The energy flux (e.g. total energy, kinetic energy) or the components of the momentum flux can be defined inthe same manner. The total energy flux, also used in the following, is:
( ) outinA
totaltot EEdAcntptE&&
rr& +== )()( (5)
outintotalabstot EEdAcntptE&&
rr& ==
)()( (6)
2)(
totabstotintot
EEtE
&&&
+=
2)( abs
tottotouttot
EEtE
&&&
= (7)
Definition of power coefficients:
QHg
tEtK
j
Ej =
)()(
&
& (8)
4. Results
The process of energy dissipation for operating points near runaway involves in- and outflows from the runner.The high energy flow is entering the runner from the guide vanes and drives the runner up to speed where parts
of the channel start to pump flow outwards. The equilibrium of energy input and dissipation by pumping resultsto zero torque at the shaft.
surface B
guide vane channelsrunner channels
guide vanes
stay vanes
surface A
outcr
incr
nr
-
7/30/2019 Instability of Pump-turbines During Start-up...
5/8
The discharge being pumped out of the runner has to reenter the runner. This increases the inflow into the runnerabove the flow rate given at the inlet to the turbine scroll. This process of pumping seems to be an unsteadyprocess for the investigated model turbine for an operating point slightly above runaway.
Figure 5 shows a typical time history of flow rate fluctuations through the surface A in the vaneless spacebetween guide vanes and runner over three revolutions of the runner. The normal vector is defined here aspointing radially outwards. The mass conservation is satisfied, as the evaluated flow rateQ(t) through the surface
Aremains constant in time at the given valueQ at the inlet. Accordingly, fluctuations of the outflowQout(t) andfluctuations of the inflow into the runner Qin(t) have to be in equilibrium.
surface A
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
50.0 50.5 51.0 51.5 52.0 52.5 53.0
Revolutions [-]
Coefficient[-
]
Q = 38103
m3/s; n =1000 r m
Q out(t)
Q
Q in (t)
Q
Q(t)
Q
Fig. 5. Time varying flow rates through the surface A of the vaneless space
These flow rate fluctuations seem to be localized to the vaneless space between guide vanes and runner, since atthe outlet of the turbine no fluctuations are observed as shown in Figure 6. Here, the normal vector is defined indirection of the draft tube. The backflow Qin(t) in direction of the runner amounts to about one third of thedischargeQ at the inlet to the turbine. All simulated flow rates are almost constant in time.
-2.0
-1.5
-1.0
-0.5
0.0
0.5
50.0 50.5 51.0 51.5 52.0 52.5 53.0
Revolutions [-]
Coefficient[-]
Q =38103
m3/s;n =1000 rpm
Q in (t)
Q
Q(t)
Q
Q out(t)
Q
Fig. 6. Flow rate fluctuations at the outlet of the turbine
The periodical flow rate fluctuations through the surface A in the vaneless space are linked to the torquefluctuation on the shaft. In phases in time where a partial pumping flow builds up in the runner channels, thetorque increases simultaneously, Figure 7.
-
7/30/2019 Instability of Pump-turbines During Start-up...
6/8
-1.5
-1.0
-0.5
0.0
0.5
50.0 50.5 51.0 51.5 52.0 52.5 53.0
Revolutions [-]
Co
efficient[-]
Q = 38103
m3/s;n = 1000 r mKPmech =
T(t)2n/60
gHQ Q out(t)
Q
Fig. 7. Analogy between outflow fluctuations through the surface A and the rotor torque fluctuations
When analyzing the energy fluxes through the surface A of the vaneless space, we see that the power coefficients
)(tKintotE&
and )(tKouttotE&
fluctuate out of phase and at the same frequency as the mechanical torque
fluctuations )(tKPmech .The resulting power coefficient )(tK totE& fluctuates with a certain phase shift with the
mechanical power indicating that inertia effects of the rotating water masses might be involved.
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
50.0 50.5 51.0 51.5 52.0 52.5 53.0
Revolutions [-]
Coefficient[-]
Q =38103
m3/s; n =1000 rpm
KPmech (t)
Ktot in (t)
Ktot out(t)
Ktot(t)
Fig. 8. Energy fluxes on the surface A
The question arises now how these flows lead to energy transfer to the vaneless space and how the in- andoutflows look like in detail. Figure 9 clearly demonstrates the existence of enhanced vortices transporting fluidoutwards. These vortices exit the runner channels in front of the leading edges of the runner vanes into thevaneless space. The vortex strength varies in time and space. For the chosen operating point, which is slightlyabove the runaway point, the variation in time is dominant, which results in the global flow rate fluctuationthrough the surface A. It can be assumed that with decreasing flow rate Q at the inlet to the turbine the effect ofthe spatial variation of the vortex formation will more and more dominate and that rotating stall will be observedfor operating points below runaway, as it was experimentally observed for a pump turbine e.g. by Staubli [8].
The difference between the in- and out-energy fluxes through the surface A indicates that a large amount of theenergy dissipation occurs in the vaneless space between guide vanes and runner for operating points nearrunaway.
-
7/30/2019 Instability of Pump-turbines During Start-up...
7/8
Fig. 9. View fromthe hub side on channel vortices exiting at the leading edge
Fig. 10. Outflow into vaneless space between guide vanes and runner
5. Conclusions
The characteristics of the pump turbine close to runaway could be well predicted with transient flow simulations.Unstable flow fields were predicted for the simulations in the so called S-shaped portion of the characteristic.This simulated instability shows time-varying in- and outflow from the runner into the vaneless space. For the
investigated operating point, slightly above runaway, the band of the fluctuations corresponded to about 50percent of the main inflow to the turbine.
The existence of unstable operation is confirmed by the model test where also instability was observed in thisrange of operation.
surface A coloured withradial velocity
streamlines coloured withkinetic energy
channel vorticescoloured withkinetic energy
surface A coloured withradial velocity
radial velocity component
-
7/30/2019 Instability of Pump-turbines During Start-up...
8/8
With detailed information available in the simulated flow field local flow effects could be analyzed. It could beconcluded that local vortices forming in the runner channels close to the leading edge is the source for theunsteady in- and outflow from the runner into the vaneless space between guide vanes and runner. Therefore, thevortices and the induced outflow can be considered as the origin of the instability.
Most of the energy dissipation for operating points near runaway occurs in the vaneless space between guidevanes and runner.
Acknowledgement
This study was made possible by a grant of the Swiss Commission for Technology and Innovation (CTI) andswisselectricresearch. Industrial funding was provided by VA TECH HYDRO.
References[1] Yamabe, M.,Hysteresis Characteristics of Francis Pump-Turbines When Operated as Turbine, Trans.
ASME, J . Basic Engineering, Vol. 93, pp.80-84, March 1971[2] Yamabe, M., Improvement of hysteresis characteristics of Francis pump-turbines when operated as turbine,
Trans. ASME, J. Basic Engineering, pp. 581-585, September 1972[3] Klemm, D., Stabilizing the characteristics of a pump-turbine in the range between turbine part-load and
reverse pumping operation, Voith Forschung und Konstruktion, Vol. 28, 1982[4] Martin, C. S., Stability of pump turbines during transient operation, 5th Intl. Conf. On Pressure Surges,
BHRA, Hannover, September 1986, pp. 61-71[5] Martin C. S., Instability of pump-turbines with S-shaped characteristics, Proc. 20th IAHR Symp. Hydraulic
Machinery and Systems, Charlotte, NC, 2000[6] Doerfler, P., Stable operation achieved on a single-stage reversible pump-turbine showing instability at no-
load, XIX Symposium of IAHR Section on Hydr. Machinery and Cavitation, Singapore, 1998[7] Billdal, J .T., Wedmark, A., Recent experiences with single stage reversible pump turbines in GE Energys
hydro business, Paper 10.3, Hydro 2007, Granada[8] Staubli, T., Some Results of force measurements on the impeller of a model pump-turbine, IAHR Work
Group on the Behavior of Hydraulic Machinery under Steady Oscillatory Condition, 3rd Meeting, Lille,
1987, P. 8, pg. 1-11
Authors
Thomas Staubli graduated in Mechanical Engineering from the Swiss Federal Institute of Technology (ETH) inZrich. After two years of post-doctoral research in the field of flow induced vibration at Lehigh University,Pennsylvania, he worked in experimental fluid mechanics at Sulzer Hydro (now VA TECH HY DRO) in Zrich.He then headed the Hydromachinery Laboratory at the ETH Zrich. During this period he directed researchprojects in the field of hydraulic machinery. Since 1996 he is professor and heads the Competence Centre FluidMechanics & Hydro Machines at the LUASA.
Manfred Sallaberger is head of hydraulic development for radial machines at VA TECH HYDRO. He graduatedin mechanical engineering from the Technical University of Graz, Austria, in 1986. Until 1993 he was researchassistant at the department for hydraulic machinery and received his PhD 1994. He joined Sulzer Escher WyssZurich in 1994 as a CFD specialist and research engineer in the hydraulic laboratory. He was involved in thedesign of stationary and rotating components of a number of upgrading projects of Francis turbines andreversible pump turbines. Since 1998 he is responsible for the hydraulic design of radial machines at VA TECHHYDRO.
Florian Senn graduated in Mechanical Engineering at the University of Applied Sciences of Konstanz, Germany.From 2006 to 2008, he was research assistant at the Competence Centre Fluid Mechanics & Hydro Machines atthe LUASA. Since October 2008 he is master student at the Technical University of Munich, Germany.