Vous êtes sur la page 1sur 16

Experimental Investigation of a Francis Turbine during

Exigent Ramping and Transition into Total Load Rejection


Chirag Trivedi, M.ASCE 1; Einar Agnalt 2; and Ole Gunnar Dahlhaug 3
Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

Abstract: This study investigates the unsteady pressure fluctuations in a hydraulic turbine observed during a dangerous case of steep ramp-
ing interrupted by an unexpected transition into total load rejection. Although hydraulic turbines are expected to experience such events only a
few times over their lifetime, the resulting pressure amplitudes are so significant that they take a toll on a machine’s operating life. The focus
of the present study is to experimentally measure and numerically characterize time-dependent pressure amplitudes in the vaneless space,
runner, and draft tube of a model Francis turbine. To this end, 12 pressure sensors were integrated into a turbine, including four miniature
sensors mounted in the runner. Steep ramping was performed by changing the rotational speed using a frequency controller. After a few
seconds, as the load increased, total load rejection was initiated. This resulted in the generation of strong vibrations throughout the entire
structure and strong pressure fluctuations in the turbine. The data analysis shows that pressure amplitudes are in the order of 10–20% of
hydraulic energy in the vaneless space and runner, with high-amplitude fluctuations occurring at expected characteristic frequencies, includ-
ing those associated with rotor-stator interactions, water hammer travel times, and standing waves in the turbine. Various
stochastic frequencies were also observed, especially at the runner outlet. DOI: 10.1061/(ASCE)HY.1943-7900.0001471. © 2018 American
Society of Civil Engineers.
Author keywords: Francis turbine; Hydropower; Load rejection; Pressure; Rotor-stator interaction.

Introduction Uncertainties regarding the availability of wind/solar power


often lead to unpredictable variations in grid frequency (Caralis
The electricity market requires high levels of power-generation et al. 2012). The electricity market is inclined toward energy stor-
flexibility and a robust governing mechanism to maintain grid sta- age when demand is low and toward boosting power production as
bility. Variable-speed hydraulic turbines have been developed to demand increases (Bessa et al. 2017). Hydropower is playing a
allow real-time load adjustments (ramping up and down) and to growing role to meet such requirements and to maintain grid sta-
allow it to be turned off and on several times per day. During load bility (Sivakumar et al. 2014). Hydraulic turbines have the flex-
change and/or start-stop periods, turbines can experience unex- ibility to respond quickly (less than 1 s) while using synthetic
pected cases of total load rejection and/or cascade load shedding. inertia to control primary (a few seconds), secondary (less than
A review of the literature and of pressure measurements shows that 1 min), and territory (a few minutes) controls depending on re-
such conditions spur considerable levels of fatigue damage to quirements (Hell 2017; Yang et al. 2016b). Synthetic inertia con-
blades as a result of high-amplitude pressure fluctuations. Detailed trol is dependent on the inertia of rotating masses and on the
measurements are vital to understanding flow fields under critical flywheel effect (Ellingsen and Storli 2015). Under conditions
transient conditions and to allow robust turbine design. The present of primary/secondary control, grid frequency is restored within
study examines the time-dependent pressure amplitudes of the a few seconds depending on the size of the turbine and the ramp-
vaneless space, runner, and draft tube. Measurements were col- ing rate (Beevers et al. 2015; Claude 2017). For a synchronous-
lected from a model Francis turbine during steep ramping and speed turbine, the ramping rate is dependent on the rate of guide
unexpected transitions into total load rejection. Time-domain vane opening/closing (discharge). For a variable-speed turbine,
and frequency-domain analyses of the acquired pressure data are the ramping rate is a function of the discharge and rotational
presented. speed. To meet real-time demand requirements by using stored
energy, variable-speed turbines/pump-turbines are commonly used
(Béguin et al. 2017).
1
Researcher, Waterpower Laboratory, Dept. of Energy and Process Hydraulic turbines are designed to operate at the best efficiency
Engineering, NTNU—Norwegian Univ. of Science and Technology, point (BEP), at which point the flow is stable and runner blades
7491 Trondheim, Norway (corresponding author). ORCID: https://orcid experience minimal levels of fatigue (Trivedi et al. 2013b). How-
.org/0000-0002-2198-8981. E-mail: chirag.trivedi@ntnu.no ever, when accommodating real-time demand, blades experience
2
Ph.D. Candidate, Waterpower Laboratory, Dept. of Energy and Process moderate to high levels of fatigue loading as a consequence of
Engineering, NTNU—Norwegian Univ. of Science and Technology, high-amplitude dynamic stress (Coutu et al. 2004; Gagnon et al.
7491 Trondheim, Norway. E-mail: einar.agnalt@ntnu.no 2016; Yang et al. 2016a). Such stresses are related to unsteady pres-
3
Professor, Waterpower Laboratory, Dept. of Energy and Process sure fluctuations occurring during steep ramping (Dörfler et al.
Engineering, NTNU—Norwegian Univ. of Science and Technology,
2013; Melot et al. 2013; Morissette et al. 2016; Nennemann et al.
7491 Trondheim, Norway. E-mail: ole.g.dahlhaug@ntnu.no
Note. This manuscript was submitted on April 5, 2017; approved on 2014). Amplitudes of such pressure fluctuations are more than
December 7, 2017; published online on April 10, 2018. Discussion period twice the amplitude measured at BEP (Amiri et al. 2016). During
open until September 10, 2018; separate discussions must be submitted for periods of primary control, a turbine must change the required load
individual papers. This paper is part of the Journal of Hydraulic Engineer- within 5 s (times may vary depending on the grid network and
ing, © ASCE, ISSN 0733-9429. ramping rate of a turbine), and pressure amplitudes vary quickly,

© ASCE 04018027-1 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


which in turn spurs time-dependent variations in stresses (Trivedi
and Cervantes 2017).
Turbines can occasionally experience unexpected total load
rejection during steep ramping when the grid frequency fluctu-
ates beyond the prescribed limit (Guo et al. 2016; Mossinger
and Jung 2016). An example of such a dangerous event con-
cerns the cascade load shedding of 2–3 turbines/pump-turbines
that share the same penstock. The first turbine experiences total
load rejection because of unknown faults. Consequently, the
second turbine ramps steeply to balance the grid frequency. In
the meantime, the second and then third turbine experience total
load rejection. This leads to considerable variations in water
Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

pressure levels in the entire hydraulic system (including tur- Fig. 1. Open-loop hydraulic system at the Waterpower Laboratory,
bines) over a short period of time (Gagnon and Léonard 2013; Norwegian University of Science and Technology (NTNU): (1) feed
Guo et al. 2016; Moisan et al. 2014; Nicolet et al. 2008, 2009; pump, (2) overhead tank—primary, (3) overhead tank—secondary,
Zeng et al. 2016). During total load rejection, the generator is (4) pressure tank, (5) magnetic flowmeter, (6) DC generator, (7) Francis
decoupled from the grid, causing the load to the turbine to in- turbine, (8) downstream tank, and (9) basement
stantly reach a value of zero while the runner accelerates at the
runaway speed (Trivedi et al. 2016c). Available hydraulic en-
ergy is dissipated on the blades, and high-amplitude pressure
fluctuations develop. Investigated Model
Previous studies on hydraulic turbines observed under transient
conditions (e.g., load variation, start-stop, and total load rejection) Francis Turbine
showed that the amplitudes of pressure fluctuations can be four
times the amplitudes observed at the steady-state BEP (Trivedi The test rig available at the Waterpower Laboratory of the Norwe-
et al. 2014a, b, c, 2015a, b). In the runner, pressure amplitudes gian University of Science and Technology was used to collect
increase linearly with the rotational speed, and amplitudes are measurements. This test rig is capable of operating under two dif-
maximized at the runaway speed (Trivedi et al. 2016a, b, c). In the ferent configurations (closed loop and open loop) according to
vaneless space, amplitudes are even greater, at twice those of the measurements collected. The closed loop is preferred for steady-
runner. During total load rejection, guide vanes cannot be closed state measurements, whereas the open loop is preferred for transient
rapidly because of water hammer risks and draft tube surges measurements. The open loop represents a hydraulic system similar
(Pejovic and Karney 2014; Zeng et al. 2016). Certain prototypes to a prototype whereby an overhead tank represents the headrace
adopt an air injection system to limit surging effects in the draft and the downstream tank represents the tailrace during measure-
tube. Nevertheless, air injection can spur further increases in runner ment collection. Fig. 1 shows an open-loop configuration. Water
rotational speeds to 2–5% of those without air injection (Chen et al. from a large basin (indicated by Point 9 in the figure) is continu-
2008; Nishi and Liu 2013). ously pumped into an overhead tank (Point 2) acting as a reservoir,
Modern turbine runner designs are inclined toward the use of and water then flows down (gravity) to the turbine (Point 7).
lightweight models and maximizing hydraulic performance (Liu Discharge to the turbine is regulated by guide vanes. The draft tube
et al. 2015; Monette et al. 2016). The lightweight runner accelerates outlet is connected to a downstream tank (Point 8) wherein a con-
rapidly to the runaway speed even before guide vanes start to close. stant water level is maintained at atmospheric pressure, and water
Furthermore, during manufacturing, mechanical features (natural above the runner centerline is discharged to the basement (Point 9).
frequency and material stiffness) are occasionally undermined Feed pumps (Point 1) can be operated at the selected speed to
because of the involved material and blade coating costs (Hübner obtain a required head for measurements. Two pumps are driven
et al. 2016; Liu et al. 2016; Trivedi 2017). A runner may expe- by a 315-kW variable-speed motor, and the pumps can be operated
rience resonance when the natural frequency level is lower than in series or parallel depending on discharge/head requirements.
the rotor-stator interactions (RSI) frequency, and pressure ampli- The open loop can produce a head of up to 16 m, and the closed
tudes amplify (Coutu et al. 2008; He et al. 2014; Zuo et al. 2015) loop can produce a head of up to 100 m (discharge of up to
during each acceleration/deceleration. Turbines should be de- 0.5 m3 s−1 ).
signed to withstand extreme conditions found under normal, emer- An open-loop hydraulic system was used to enable configura-
gency, or catastrophic transient regimes (Dorji and Ghomashchi tions similar to a prototype for the current measurements. The Fran-
2014). cis turbine was a reduced-scale (1∶5.1) model of a prototype
Power-generation trends mean that highly flexible force hy- (4 × 110 MW) operating at the Tokke power plant in Norway.
draulic turbines must operate under challenging conditions every The model turbine included 14 stay vanes integrated within the spi-
day (Hell 2017; Luna-Ramírez et al. 2015). Turbines transition into ral casing, 28 guide vanes, a runner with 15 full-length blades and
total load rejection during power ramp up/ramp down (Bucur et al. 15 splitters, and a draft tube. The runner inlet and outlet diameters
2014; Yasuda and Watanabe 2016). This represents one of the most were 0.63 and 0.347 m, respectively. Estimated Reynolds numbers
dangerous conditions that can be experienced by a turbine. No ex- were 1.87 × 106 at the BEP. The runner and direct-current (DC)
perimental study has focused on such transient conditions or, more generator (power = 352 kW and speed = 26 Hz) rotor were coupled
specifically, on pressure loading within the turbine, although tran- on the same shaft. The test rig was customized for the governing of
sients significantly affect the operating life of the runner. The cur- the rotational speed and guide vane opening to collect transient
rent lack of studies conducted on such conditions may be attributed measurements. In-house program was developed and integrated
to risks of damaging entire test rigs and generators as runners into the governing system to obtain steep ramping by altering
quickly accelerate toward the runaway speed, spurring test rig the rotational speed to decouple the generator from the connected
resonance. load and to close the guide vanes.

© ASCE 04018027-2 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


Instrumentation tube sensors DT1 and DT2 were positioned 0.126 m downstream
from the runner and 180° circumferentially apart along the same
The Francis turbine was equipped with all instruments required to
plane. The other two sensors (DT3 and DT4) were positioned
carry out model testing according to International Electrotechnical
0.376 m downstream from the runner. An additional sensor, DT5
Commission (IEC) 60193 (IEC 1999). To acquire differential pres-
(not shown in the figure), was mounted 10 mm in from of the draft
sure levels cross the turbine, ring-type manifolds [IEC 41 (IEC
tube outlet to examine surging levels when present. Pressure data
1991)] were used at the spiral casing inlet and draft tube outlet.
were acquired at a sampling rate of 5 kHz. Natural frequencies of
A differential pressure transducer was connected to the manifolds.
the pressure sensors exceeded 25 kHz.
The shaft torque, friction torque, and axial torque were measured
using toque transducers during the measurements. A temperature
sensor was used to measure the water temperature and estimate Calibration and Uncertainty Quantification
the change in density. For the rotational speed measurements, a pho- Before the measurements were collected, calibration and uncer-
Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

tocell and circular disk with permanent grooves were mounted on the tainty quantifications of all of the measuring instruments and
main shaft. In addition, a positioning sensor was used to extract the sensors were conducted. IEC 60193 (IEC 1999) was followed to
runner angular position in real time. The frequency response time of quantify uncertainties in turbine efficiency at BEP. The flowmeter
the sensors was much higher (except that of the flowmeter) than the was calibrated using a primary method, i.e., weighing tank [ISO
maximum expected frequencies. A magnetic flowmeter was used for 5168:2005(E) (ISO 2005)]. A pneumatic deadweight tester (GE
steady-state measurements because of the very slow frequency re- P3000 Series, Alzenau, Germany) at an accuracy of 0.008% of
sponse time involved (1 Hz). During the transient period, flow-rate the measured value was used as a primary device for the pressure
data were not considered for the analysis. sensor calibrations. Torque sensors were calibrated using a primary
To acquire pressure data from different locations, 10 pressure method, standard weight, and arm. Uncertainties obtained from the
sensors within the turbine and two pressure sensors at the inlet con- calibration are given in Table 1. The total uncertainty value (^et ) of
duit were flush-mounted. Sensors IN1 and IN2 were mounted at the 0.2% included systematic (^es ) and random uncertainties (^er )
inlet conduit to monitor the water hammer, and they were posi- qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
tioned 7.2 and 0.2 m upstream from the turbine inlet, respectively. êt ¼  ê2s þ ê2r ð1Þ
The locations of the pressure sensors in the turbine are shown in
Fig. 2. Two sensors (VL1 and VL2) were flush-mounted on the qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
bottom ring of the distributor in the vaneless space. VL2 was posi- ês ¼  ê2Q þ ê2H þ ê2T þ ê2n þ ê2ρ ð2Þ
tioned 240° circumferentially away from VL1 in the clockwise di-
rection. Four sensors (R1, R2, R3, and R4) were miniature type and The systematic uncertainty is the root-sum square of uncertain-
flush-mounted onto the surface of the runner crown. The runner ties in the discharge (^eQ ), head (^eH ), torque (^eT ), runner rotational
included alternating arrangements of splitters and blades. Hence, speed (^en ), and water density (^eρ ) of the calibration. The maximum
R1 and R2 were located between a splitter suction side and a blade uncertainties of pressure sensors located in the vaneless space, run-
pressure side, whereas R3 and R4 were positioned between two ner, and draft tube were measured as 0.12, 0.26, and 0.14%,
blades following the splitter trailing edge. A slip-ring mechanism respectively. Uncertainties in guide vane angular positioning (^eα )
was used to acquire pressure data from the runner sensors. Draft were measured as 0.7%.

Fig. 2. Locations of pressure sensors in the turbine; Sensors R1, R2, R3, and R4 are in the runner; DT1, DT2, DT3, and DT4 are in the draft tube cone;
VL1 and VL2 are in the vaneless space

© ASCE 04018027-3 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


Table 1. Computed Uncertainties in the Flow Measurement through Results and Discussion
Calibration
Variable Uncertainty (%) The turbine initially operated at the steady-state load, i.e., BEP, and
then the rotational speed of the runner was swiftly decreased to
e^ H 0.17 increase the power output. In the meantime, the total load measured
e^ Q 0.12
at the generator terminal was rejected, and the turbine was set to a
e^ T 0.15
e^ n 0.035
no-load setting. Fig. 3 shows time-dependent variations in the run-
e^ ρ 0.01 ner rotational speed (n), guide vane opening (α), torque (T), and net
e^ r 0.18 head (H). The flow parameters are normalized by corresponding
e^ t 0.2 values at BEP, which are provided in Table 2. In Fig. 3, the time
e^ VL 0.12 up to 0 s shows the steady-state BEP load. The load was set to in-
e^ R 0.26 crease at 0 s (tramp ) by decreasing the rotational speed from 5.55 Hz.
Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

e^ DT 0.14 The torque was swiftly increased linearly as the speed decreased, and
e^ α 0.7 at 1.5 s (tT¼0 ), the load to the generator terminal was suddenly re-
jected. The runner was abruptly accelerated to the runaway speed
(9.5 Hz) from 3.8 Hz. The guide vanes were set to close immediately
following load rejection, and the rotational speed decreased slowly.
To determine repeatability of the test rig, measurements were The guide vanes were closed at 15 s, i.e., tGVA¼0 . The runner entered
conducted at steady-state BEP before collecting the transient mea- a steady state at approximately 80 s.
surements. The results were compared with available benchmark The net head was computed from the differential pressure and
data (Trivedi et al. 2013a), and the maximum deviation was mea- velocity across the turbine. Therefore, variations corresponded to
sured as within 0.18, i.e., random uncertainty (^er ). The BEP was the static and dynamic pressure change at the turbine inlet and out-
obtained at α ¼ 9.9°, nED ¼ 0.18, QED ¼ 0.15, and the hydraulic let. The head variation was observed approximately 16% from the
efficiency was measured as 93.1  0.2% BEP. At the turbine inlet, static and dynamic pressure levels are
dependent on the guide vane aperture/opening (GVA) and runner
nD rotational speed. As the runner accelerated to the runaway speed (at
nED ¼ pffiffiffiffi ð3Þ
E the same time the guide vanes were closing), discharge to the tur-
bine decreased, resulting in an increase of pressure at the turbine
where n = rotational speed of the runner in revolutions per second; inlet as water level in the tank increased. At the draft tube outlet,
D = reference diameter of the runner (D ¼ 0.349 m); and E = water levels were constant, and the pressure was measured as nearly
specific hydraulic energy level (J kg−1 ). The discharge factor was the same that observed under the transient condition. Pressure in the
computed using Eq. (4) draft tube was stabilized at t ≈ 100 s, i.e., following the steady run-
ner condition.
Q
QED ¼ pffiffiffiffi ð4Þ Turbine Inlet
D2 E
Two pressure sensors (IN1 and IN2) were flush-mounted at the tur-
where Q = turbine inlet discharge (m3 s−1 ). bine inlet conduit to investigate water hammer effects and standing

Fig. 3. Time-dependent variations of runner rotational speed (n), guide vane position (α), torque (T), and head (H) during steep ramping and
transition into total load rejection; time stamps tramp , tT¼0 , tGVA¼0 , and tn¼0 correspond to the time of ramping, transition into total load rejection,
complete closing of the guide vanes, and runner at steady condition, respectively

© ASCE 04018027-4 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


Table 2. Flow Parameters Obtained at BEP Load decreased. Opposite variations can be observed after total load re-
Parameter Quantity jection. However, after total load rejection, the guide closed, so the
pressure levels in the inlet conduit increased because of surging
Head (m) 11.91
effects. The total level of pressure increase was measured as 3%
Discharge (m3 s−1 ) 0.199
Torque (N m) 623.12
at IN1 and 3.4% at IN2. High-amplitude pressure fluctuations
Speed (Hz) 5.55 can be observed for the transition into total load rejection. Normal-
Inlet pressure (kg m−3 ) 212.5 ized pressure fluctuations (p~ E ) for 1.5–2.5 s are shown in Fig. 5.
Efficiency (%) 93.1 The fluctuations are normalized using Eq. (6)
Guide vane angle (degrees) 9.9
~ − p̄ðtÞ
pðtÞ
p~ E ¼ ð6Þ
ðρEÞBEP
Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

where p̄ðtÞ = time-average (interpolated) pressure (Pa).


~ is subtracted from time-average pressure
Acquired pressure pðtÞ
p̄ðtÞ, and the selected time window for averaging is 0.2 s. High-
amplitude pressure fluctuations can be observed at 2.2 s. During
steep ramping, flows accelerated toward the turbine, and after load
rejection, the flow quickly decelerated because the turbine could
not accept more discharge as the runner approached the runaway
speed. This resulted in water hammering followed by surging in
the conduit, which induced high-amplitude and low-frequency
fluctuations at 2.2 s. Pressure amplitudes increased as the runner
accelerated to the runaway speed. High-amplitude (∼5% of ρE)
Fig. 4. Unsteady pressure variation at the turbine inlet (IN1 and IN2) fluctuations can be observed between 2.1 and 2.3 s. Frequencies
during steep ramping and transition into total load rejection; pressure changed from the start of the total load rejection period, i.e., 1.5 s,
data are normalized using hydraulic energy at BEP [Eq. (5)] and are dependent on the rotational speed, water hammer, and
standing waves associated with the rapid closure of guide vanes.
A spectral analysis was conducted to investigate time-dependent
variations in the frequencies and corresponding pressure ampli-
waves developed during periods of load change and total load tudes. Fig. 6 shows a spectrogram of unsteady pressure fluctuations
rejection. Sensors IN1 and IN2 were positioned 7.2 and 0.2 m up- observed at IN2. Amplitudes are computed using Eq. (7)
stream from the turbine inlet, respectively. Unsteady pressure var-
iations observed at both locations are shown in Fig. 4. The acquired p~
p~ E rms ¼ pEffiffiffi ð7Þ
pressure data were normalized using Eq. (5) 2

~
pðtÞ High amplitudes can be observed from the point of total load
p ¼ ð5Þ
ðρEÞBEP rejection (1.5 s) to the complete closing of the guide vanes (15 s).
Prior to the load change, the frequency of blade passing (fb ¼
~
where pðtÞ = fluctuating pressure levels (Pa); ρ = water density 166 Hz) is dominant. During steep ramping and total load rejec-
(kg m−3 ); and E = specific hydraulic energy level (J kg−1 ) at BEP. tion, amplitudes of stochastic frequencies are dominant in addition
Pressure in the inlet conduit drops during ramping, and after the to the blade passing frequency, which varies with rotational speed.
total load rejection, the pressure increases rapidly. In this turbine, at Amplitudes of the blade passing frequency vary from 0.7 to 0.9%
constant GVA, the discharge is inversely proportional to the rota- of ρE. The high-amplitude (frequency < 40 Hz) observed at the
tional speed. During steep ramping, rotational speeds decrease, the time of total load rejection (tT¼0 ) can be clearly observed and is
discharge level (velocity) is increased, and the pressure level is thus associated with the water hammer and with surging, as mentioned

Fig. 5. Normalized pressure fluctuations (factor of pressure fluctuations) at the turbine inlet (IN1 and IN2) during steep ramping and transition into
total load rejection; pressure data are normalized using Eq. (6); t ¼ 1.5 s indicates the start time of total load rejection

© ASCE 04018027-5 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


BEP load because VL1 is located close to the casing inlet, whereas
VL2 is located in the last casing section. Furthermore, the local
pressure level is dependent on the available vaneless space relative
to the guide vane position. Sensor VL1 is located close to the
guide vane trailing edge and VL2 is located in front of the trailing
edge where vaneless (radial) space around the sensor is limited. In
the vaneless space, time-dependent pressure linearly follows the
runner rotational speed. The pressure level decreases with rota-
tional speed during ramping, and it increases as the runner accel-
erates to the runaway speed after load rejection. As the rotational
speed increases, the flow to the runner decreases and pressure levels
build in the vaneless space because guide vanes are located at a
Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Spectrogram of unsteady pressure fluctuations at IN2 during fixed opening. After a few seconds, the guide vanes close and
steep ramping and transition into total load rejection; tramp , tT¼0 , the rotational speed decreases. Hence, the pressure in the vaneless
and trunaway indicate the time of power ramp-up, total load rejection, space decreases and stabilizes at 80 s, i.e., tn¼0 . High-amplitude
and runaway speed, respectively pressure fluctuations can be observed during the transition into total
load rejection (t ¼ 1.5 s) and at the runaway speed (t ¼ 3.5 s).
Extracted pressure fluctuations observed at one of the locations
(VL1) are shown in Fig. 8. For graphical clarity, fluctuations are
earlier. Other high-amplitude frequencies observed between 20 and
shown for 0.5–2.5 s in Fig. 8(a) and for 2.5–4.5 s in Fig. 8(b).
40 Hz are associated with standing waves in the inlet conduit. Such
During steep ramping, amplitudes of the fluctuations increase.
broad ranges of frequencies were obtained from previous transient
At the time of load rejection, amplitudes are approximately 15%
measurements collected from this turbine (Trivedi et al. 2014c).
of ρE and decrease as the runner accelerates to the runaway speed.
However, at the runaway speed (between 3.5 and 4 s), pressure
Vaneless Space amplitudes are 11% of ρE. High-amplitude pressure fluctuations
In the vaneless space, the flow field is dependent on complex in- correspond to the blade passing frequency, which varies with the
teractions between guide vanes and runner blades. The interaction runner rotational speed.
time can be computed from Eq. (8) The model Francis turbine is equipped with 28 guide vanes and
30 blades, including splitters. Two blades (at 180°) and two guide
 
1 1 1 vanes (at 180°) interact with one another at the same time. A com-
tRSI ¼ − ðsÞ ð8Þ plex flow condition is developed after load rejection. As the runner
n zgv zb
accelerates to the runaway speed, the interaction time decreases. At
where zgv and zb = number of guide vanes and blades, respectively. the same time, the amount of available vaneless space increases
The expected characteristic frequency in the vaneless space is the because of the rapid closure of the guide vanes. The blade and guide
blade passing frequency (fb ), which is dependent on the instanta- vane interaction sequence is shown in Fig. 9. After load rejection, a
neous rotational speed (n) of the runner blade interacts with the corresponding guide vane at time t1 , at
which point the runner instantaneous speed is ω1 and the guide vane
f b ¼ nzb ðHzÞ ð9Þ position is α1 while a pressure wave develops, which propagates
toward the turbine. Subsequent interactions occur at times t2
To acquire time-dependent variations of the pressure fluctua-
and t3 , and the respective rotational speeds are ω2 and ω3 . The pres-
tions and frequencies, two sensors, VL1 and VL2, were flush-
sure wave that developed during t1 interactions interferes with the
mounted onto the bottom ring of the guide vanes. Unsteady
pressure wave developed from t2 interactions as the rotational
pressure variations observed at VL1 and VL2 locations during steep
speed increases. Similarly, pressure waves from t1 and t2 interac-
ramping from BEP and during transition into total load rejection
tions interfere with t3 interactions. Pressure amplitudes vary de-
are shown in Fig. 7. The average pressure level observed at
pending on levels of energy generated from the pressure waves.
VL1 is ∼2.5% higher than that observed at VL2 as the steady-state
Furthermore, when the wave propagation speed coincides the tim-
ing of neighboring blade interaction, resonance results. Hence, to
support safe design, proper computations of the wave speed and
RSI for maximum possible rotational speeds are vital.
To investigate the RSI frequency and harmonics, a spectral
analysis of the acquired pressure data was conducted. A spectro-
gram of unsteady pressure fluctuations observed at VL1 is shown
in Fig. 10. The blade passing frequency of 166 Hz can be clearly
observed during the steady-state BEP load period (t < 0 s). The fre-
quency decreases (after tramp ) as the rotational speed decreases. The
amplitude variation is identical to the variation shown in Fig. 8. The
spectrogram shows three deterministic frequencies, i.e., 0.5f b , fb ,
and 2fb . Harmonics 0.5fb and 2fb appear only when the rotational
speed is high at 3–15 s. Harmonics observed at high runner
rotational speeds may be associated with the interference and re-
Fig. 7. Unsteady pressure variation in the vaneless space (VL1 and
flection of hydroacoustic waves (Arpe et al. 2009; Nicolet 2007;
VL2) during steep ramping from BEP and transition into total load re-
Nicolet et al. 2010) from the wall of the guide vanes and to the
jection; pressure data are normalized using hydraulic energy at BEP
rotating pressure field of the vaneless space (Botero et al. 2014;
[Eq. (5)]
Hasmatuchi et al. 2011; Pacot et al. 2016; Widmer et al. 2011).

© ASCE 04018027-6 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Normalized pressure fluctuations (factor of pressure fluctuations) in the vaneless space (VL1) during steep ramping from BEP and transition
into total load rejection; pressure data are normalized using Eq. (6); t ¼ 1.5 s in plot (a) indicates the start time of total load rejection; t ¼ 3.5 s in plot
(b) indicates the time when runner attained the runaway speed

Fig. 10. Spectrogram of unsteady pressure fluctuations in the vaneless


space (VL1) during steep ramping from BEP and transition into total
load rejection; tramp , tT¼0 , and trunaway indicate the time of ramping,
total load rejection, and runaway speed, respectively; f b is the blade
passing frequency in the turbine
Fig. 9. Rotor-stator interaction and wave propagation during runner
acceleration toward runaway condition after the load rejection; time
t1 indicates the reference (start) time of interaction; t2 and t3 are the (frequency) in time at the same relative locations from guide vanes
subsequent interactions as runner rotates; interaction time is dependent and runner blades. Global phenomena (e.g., vibration of the entire
on the instantaneous rotational speed of the runner [Eq. (8)] turbine, hydroacoustic waves, and RSI) can be captured by both
sensors in the vaneless space, whereas local phenomena (e.g., local
flow conditions, noise, and vibrations from a nearby guide vane)
are captured by the closest sensor.
The higher harmonics may induce resonance in the turbine as the A coherence analysis of VL1 and VL2 in the vaneless space is
rotational speed approaches the natural frequency of the runner. shown in Fig. 11. A high coherence value (∼0.75) for a blade pass-
This may result in the catastrophic failure of the turbine component. ing frequency of 166 Hz was captured by both sensors. In addition,
Such risks may become more pronounced when turbines share the frequencies of 210–290 Hz show high coherence (0.2–0.7) values
same penstock and when cascade load shedding occurs. representing frequencies at the runaway speed (time 3–4 s), at
In addition to the deterministic frequencies (i.e., RSI), stochastic which point global vibrations of the turbine are dominant. Low co-
frequencies of a broad range were obtained, specifically during the herence values denote the presence of stochastic/local noise at VL1
transition to total load rejection and the runaway speed. High am- and VL2, where the frequencies are random and out-of-phase. For
plitudes (8% ρE) of the stochastic frequencies are shown in the further investigations, the time-dependent coherence between VL1
spectrogram (Fig. 10) at 1.5 and 3–7 s. Such high amplitudes and VL2 is estimated as shown in Fig. 12. The blade passing fre-
may be associated with strong vibrations in the turbine and with quency shows coherence at above 0.75 that varies with the runner
local flow conditions. A pressure signal includes two frequency rotational speed. However, at approximately 5 s, the coherence is
components, i.e., deterministic and stochastic. The stochastic very low, illustrating the dominant effects of local noise/vibrations
and deterministic frequencies can be determined from a coherence in the turbine. The high-amplitude frequencies (shown in the
analysis, which determines the occurrence of specific phenomena spectrogram [Fig. 10)] observed during the transient period are

© ASCE 04018027-7 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


velocity triangles observed under the BEP=tramp , tT¼0 , and trunaway
conditions are shown. At tramp , the turbine was operating at the syn-
chronous speed at which the inlet and outlet velocity triangles cor-
responded to BEP. From tramp , the rotational speed decreased, in
turn decreasing the tangential velocity value (u1 ). In this turbine,
discharge is inversely proportional to the rotational speed when
guide vanes are located at a fixed position. This increases the flow
velocity (cm1 ) to the runner while decreasing the relative velocity
(w1 ). The tangential component of absolute velocity (cu1 ) is slightly
increased. At the runner outlet, the direction of cu2 is shifted to the
other side as the rotational speed is reduced. Thus, during ramping,
both discharge (Q) and Δcu are increased, increasing the torque
Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. Coherence between VL1 and VL2 pressure data from the va-
output [Eq. (10)]
neless space; pressure sensors VL1 and VL2 are 240° circumferentially
apart in the clockwise direction T ¼ ρQðcu1 r1  cu2 r2 Þ ðN mÞ ð10Þ

where cu = tangential velocity (m s−1 ); and r = radius (m).


The change in velocity vectors at tT¼0 is shown in Fig. 14. An in-
crease in c1 and w1 during ramping can cause a pressure drop in the
blade path.
After tT¼0, pressure increases with rotational speed because dis-
charge to the runner decreases and hence the velocity is measured
as cm1 . At the runaway speed (trunaway ), velocities c, cm , and cu are
small, whereas tangential (u) and relative velocity (w) values are
very high (for example, the velocity triangle in Fig. 14 at trunaway ),
and kinetic energy dissipates along the blade surfaces. The turbine
is not connected to the load; the dissipation of such high energy
induces high-amplitude pressure fluctuations. The extracted pres-
sure fluctuations (p~ E ) at tT¼0 and trunaway are shown in Fig. 13.
High-amplitude fluctuations can be seen at the R1 location of
the blade path. Amplitudes increase as rotational speeds decrease
from the synchronous speed. At 1.5 s, amplitudes increase swiftly
Fig. 12. Time-dependent coherence between VL1 and VL2 pressure to 20% of ρE because of an unexpected transition into total load
data acquired from the vaneless space; times of 0 and 1.5 s indicate the rejection. Such amplitudes are even (∼5%) larger than those
start time of steep ramping and transition into total load rejection, re- obtained in the vaneless space. Similarly, at the R2, R3, and R4
spectively locations, amplitudes are measured as 11.74, 8.2, and 6.9% of
ρE, respectively. At the runaway speed (t ¼ 3.5 s), amplitudes
measured at R1 and R2 are similar to those obtained during the
stochastic. Such short-term stochastic frequencies may be strong transition into total load rejection.
enough to spur fatigue/cracking in the runner blades. However, at the R3 and R4 locations, amplitudes are 40%
smaller than those obtained during the transition into total load re-
jection. Pressure fluctuations correspond to low and high frequen-
Runner cies in the runner. To visualize pressure fluctuations in detail, a
short-time window at R1 is shown in Fig. 15. Both low-frequency
The pressure field in the runner is driven by a rotational speed and
and high-frequency fluctuations are clearly shown, and the frequen-
flow rate (i.e., GVA). From the runner inlet to the outlet, the pres-
cies vary with time. Low-frequency fluctuations are found between
sure field is constantly varied along the blade passage. From pre- 1.4 and 1.9 s. The high-frequency fluctuations correspond to a
vious investigations (Trivedi et al. 2016c), strong swirling flows deterministic frequency, i.e., the guide vane passing frequency
have been obtained in the middle of the blade passage, and particu- (fgv ), and other fluctuations may correspond to stochastic frequen-
larly along the crown side at the runaway speed. To investigate such cies, i.e., random noise or the vibration of the runner
flow conditions along the blade path, four miniature pressure sen-
sors, denoted R1, R2, R3, and R4, were mounted as shown in Fig. 2. f gv ¼ nzgv ðHzÞ ð11Þ
Pressure data were acquired (sampling rate of 5 kHz) using a slip-
ring mechanism. Time-dependent pressure loading levels observed High-frequency amplitudes are dominant at the runaway speed
at R1, R2, R3, and R4 are shown in Fig. 13. Figs. 13(a–d) show (time 3–4 s), unlike frequencies observed during (1–2 s) the tran-
acquired pressure levels normalized from Eq. (5). Figs. 13(e–h) sition into total load rejection. A time-dependent spectral analysis
show extracted fluctuations [factor of pressure fluctuations (p~ E )] on pressure data for the R1, R2, R3, and R4 locations was con-
normalized from Eq. (6). In the runner, pressure levels observed ducted to examine characteristic frequencies and amplitudes
at all locations linearly follow the rotational speed. However, (Fig. 16). Before the load change (t < 0 s), a guide vane passing
the absolute change is not the same for all locations. The pressure frequency of 155 Hz and a second harmonic of 310.8 Hz is ob-
change is approximately 9, 2.5, 2, and 1.8% of ρE for the R1, R2, served. In contrast to the vaneless space, the harmonic of the RSI
R3, and R4 locations, respectively. Pressure stabilizes after the frequency (0.5f gv ) is not present in the runner. During steep ramp-
runner reaches the steady-state condition, i.e., t ¼ 80 s. ing and the transition to total load rejection (1.5–3.5 s), pressure
Pressure changes with respect to the rotational speed may be fluctuations of deterministic and stochastic frequencies are well-
explained by velocity triangles, as shown in Fig. 14. Approximate captured. Amplitudes of the stochastic frequencies are dominant

© ASCE 04018027-8 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. Unsteady pressure variation in the runner (R1, R2, R3, and R4) during steep ramping from BEP and transition into total load rejection;
pressure data in plots (a–d) and (e–h) are normalized using Eqs. (5) and (6), respectively; tT¼0 and trunaway indicate the time of total load rejection and
runaway speed, respectively

across the deterministic frequency in the runner, as shown in the pertaining to RSI is dominant and clearer compared with other op-
spectrograms marked with dashed lines. Stochastic frequencies erating conditions of a turbine. The increase of random noise may
reach up to 96 Hz at the R1 and R2 locations, whereas at the be attributed to flow separation from the blade leading edge as flow
R3 and R4 locations frequencies reach up to 20 Hz. These stochas- angle changes from the design point (Li et al. 2017). Signal R4
tic frequencies may be associated with local flow conditions devel- shows the lowest value across the locations because of the influence
oped from high rotational speeds and from the rapid closure of of the draft tube flow field. This indicates that the power of the
guide vanes following load rejection. Strong vibrations of the run- deterministic frequency is weak relative to the other locations.
ner may be a source of such frequencies. The deterministic signal of RSI weakens with distance from the
To further investigate characteristic (deterministic/stochastic) point of RSI (vaneless space). Dominant effects of vortex shedding
frequencies of the runner, a signal-to-noise ratio (SNR) (Dolecek from the splitter trailing edge and of unsteady swirls in the draft
2013; Engelberg 2008) analysis of the blade passage was con- tube cone spurred stochastic pressure fluctuations around R4.
ducted. SNR is the ratio of the deterministic signal (in this case, During steep ramping (from 0 to 3.5 s), the stochastic noise levels
the guide vane passing frequency) to the stochastic noise. A value increased in all locations, and at the runaway speed (t ¼ 3.5 s)
above zero denotes that the power of the deterministic signal is SNR increased because of strong effects of RSI frequency, as
dominant in pressure data acquired from the corresponding loca- shown in Fig. 16. After the closure of the guide vanes, the deter-
tion. Fig. 17 shows the time-dependent SNR for the R1, R2, ministic and stochastic frequencies showed the same amplitudes of
R3, and R4 locations in the runner. For the analysis, a discrete time pressure fluctuations. Overall, in the runner during the steep ramp-
span of 0.1 s was used to obtain reliable SNR values under steep ing, random fluctuations observed at the R3 and R4 locations were
pressure change conditions at t ¼ 1.5 and 3.5 s. As expected, high high relative to those of the R1 and R2 locations. As stated earlier,
SNR values and high-quality deterministic signals are shown at this may be related to vortex shedding from the splitter trailing edge
BEP prior to load change (t < 0 s). At BEP, random noise in or to effects of the draft tube flow field.
the turbine runner is low because flow from the guide vane passages To investigate similarities between the R3 and R4 locations
approaches the runner smoothly. Thus, the deterministic signal in the runner, a coherence analysis was carried out (Fig. 18).

© ASCE 04018027-9 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 14. Guide vane openings at BEP (tramp ) and runaway (trunaway ), and approximate velocity triangles at the runner inlet and outlet of the investigated
Francis turbines at tramp , tT¼0 , and trunaway ; R1, R2, R3, and R4 are the locations of the pressure sensors in the blade passage; runner inlet and outlet
radius are r1 and r2 , respectively

Fig. 15. Normalized pressure fluctuations (factor of pressure fluctuations) in the runner (R1) during steep ramping and transition into total load
rejection; pressure data are normalized using Eq. (6); t ¼ 1.5 s indicates the start time of total load rejection; t ¼ 3.5 s indicates the time when runner
attained the runaway speed; low-frequency fluctuations are dominant between 1.4 and 1.9 s

The frequency range of 10–100 Hz shows low coherence levels, de- R3 and R4 is shown in Fig. 19. Most of the frequencies show high
noting the dominant effect of local flows in this range at correspond- levels of coherence prior to the load change. However, during the
ing locations. Furthermore, high coherence values observed for the load change, stochastic noise/vibrations are dominant, and coherence
frequency range of 150–200 Hz show that both sensors captured the levels are very low. After t ¼ 3 s, frequencies showing high coher-
same frequency flow phenomena in the runner at the same time. This ence values are guide vane passing and harmonic. No stochastic
may be attributed to vibrations of the runner and to the guide vane noise/vibrations are observed when guide vanes close completely
passing frequency. The time-dependent coherence observed between (t ¼ 15 s), and the runner decelerates gradually.

© ASCE 04018027-10 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 16. Spectrogram of unsteady pressure fluctuations in the runner


(R1, R2, R3, and R4) during steep ramping and transition into total load
rejection; f gv is the guide vane passing frequency in the runner; t ¼ 0 s, Fig. 19. Time-dependent coherence analysis of the pressure data from
t ¼ 1.5 s, and t ¼ 3.5 s indicate the start time of total load change, total runner (R3 and R4) during steep ramping and transition into total load
load rejection and runaway speed, respectively rejection

DT4, were positioned further downstream (0.25 m from DT1/DT2).


Fig. 20 shows unsteady pressure variations in the draft tube (DT1,
DT2, DT3, and DT4). Pressure levels in the draft tube increased
during ramping. After load rejection, pressure levels suddenly
dropped to approximately 5% of the initial value and gradually
recovered as the rotational speed decreased from the runaway
speed. Similar trends were observed at all locations of the draft
tube. Pressure levels stabilized in the draft tube at approximately
150 s. The normalized pressure fluctuations are shown in Fig. 21.
Prior to load change, the pressure amplitudes were very low be-
cause the turbine was operating at BEP. Amplitudes gradually in-
creased as the rotational speed decreased during ramping. Soon
Fig. 17. Time-dependent signal-to-noise ratio of the pressure data from after the transition to total load rejection, the amplitudes increased
runner (R1, R2, R3, and R4) during steep ramping and transition into swiftly. Peak-to-peak amplitudes observed at the time of total load
total load rejection rejection, and runaway speeds were measured as 4.8 and 4.2% of
ρE, respectively. However, amplitudes observed at the DT3 and
DT4 locations were approximately 1.5% lower than those of the
DT1 and DT2 locations observed under the runaway condition.
In the draft tube, high-frequency pressure fluctuations correspond
to the blade passing frequency. In addition, standing waves of ap-
proximately 42 Hz are often obtained from the draft tube under
transient conditions (Trivedi et al. 2014c). In the test rig, the draft
tube is connected to a large tank, at which point a constant water
level is maintained. The diameter and length of the tank are 2.5 and
4.32 m, respectively. Standing waves travel from the runner outlet
to the tank outlet. Thus, the blade passing frequency and standing-
wave frequency represent global/deterministic frequencies in the
draft tube. The blade passing frequency varies with the rotational
speed of the runner, whereas the standing-wave frequency is nearly
constant. Both frequencies with high amplitudes are shown in
Fig. 18. Coherence analysis of the pressure data from runner (R3 and Fig. 22. Variations in the blade passing frequency are marked as
R4) during steep ramping and transition into total load rejection; f gv is fb in the figure. The box with dashed lines (0–20 s) shows the fre-
the guide vane passing frequency in the runner quency related to the standing wave, which is dominant during load
change and load rejection. In addition, high-amplitude stochastic
frequencies were obtained from all locations in the draft tube, as
shown for 0 to 5 s in the spectrogram marked as fstochastic . Unlike
Draft Tube
those of the runner and vaneless space, stochastic frequencies are
The examined Francis turbine is equipped with an elbow-type long dominant during load change and transitioning to total load rejec-
draft tube, through which the outlet is positioned 4.37 m from the tion (0–3 s).
runner axis. Pressure variations observed in the draft tube were To investigate the range of stochastic frequencies, a coherence
quite different from those of the vaneless space and runner. Four analysis was conducted on the pair of pressure sensors in the draft
pressure sensors (DT, DT2, DT3, and DT4) were mounted to tube. Fig. 23 shows results of the coherence analyses conducted on
the wall of the draft tube cone. DT1 and DT2 were positioned along pairs of DT1–DT2, DT3–DT4, and DT1–DT3 during load change
the same plane 0.126 m from the runner outlet and 180° circum- and transition into total load rejection (0–3 s). DT1–2 shows high
ferentially apart from one another. The other sensors, DT3 and levels of coherence relative to those of the DT3–4 and particularly

© ASCE 04018027-11 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 20. Unsteady pressure variation in the draft tube (DT1, DT2, DT3, and DT4) during steep ramping from BEP and transition into
total load rejection; pressure data are normalized using Eq. (5); tT¼0 and trunaway indicate the time of total load rejection and runaway speed,
respectively

Fig. 21. Normalized pressure fluctuations in the draft tube (DT1, DT2, DT3, and DT4) during steep ramping and transition into total load rejection;
pressure data are normalized using Eq. (6); t ¼ 1.5 s indicates the start time of total load rejection; t ¼ 3.5 s indicates the time when runner attained
the runaway speed

for frequencies above 100 Hz. The axial distance between the
planes of DT1–2 and DT3–4 is 250 mm, and the radial distance be-
tween DT1/DT3 and DT2/DT4 is 360 mm. For further analysis, the
coherence between DT1 and DT3 is shown in the figure. Coherence
from 200 to 300 Hz frequencies shows a similar trend as that
obtained for DT3–4. Similarly, coherence from 800 to 1,000 Hz
is high, indicating that both locations experienced similar flow
conditions.
The results of an SNR analysis conducted on the DT1, DT2,
DT3, and DT4 locations are shown in Fig. 24. Samples equivalent
to the time window of 0.1 s were used, and the total window size
was set to 30 s. The SNR analysis results show similar patterns as
those obtained from the runner. During load change, SNR values
Fig. 22. Spectrogram of unsteady pressure fluctuations in the draft decrease to −4 dB. However, during total load rejection and tran-
tube (DT1) during steep ramping from BEP and transition into total
sition to the runaway condition, the SNR value increases remains
load rejection; fb is the guide vane passing frequency in the runner,
positive for up to 25 s. Low-frequency cyclic variations of the SNR
and fstochastic is the frequencies of random pressure fluctuations
value from 5 to 25 s may be associated with time-dependent

© ASCE 04018027-12 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


Fig. 25. Unsteady pressure variation at the draft tube outlet during
Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

power ramp-up from BEP and transition into total load rejection; pres-
sure data are normalized using Eq. (5)

Conclusions

Experiments were conducted on a model Francis turbine during


steep ramping with abrupt transitions into total load rejection. A
challenging scenario involving a variable-speed prototype turbine
was recreated using a model turbine. The load to the turbine was
Fig. 23. Coherence analysis of the pressure data from the draft tube increased from the BEP by reducing the rotational speed. The total
(DT1, DT2, DT3, and DT4) during power ramp-up from BEP and load to the generator terminal was rejected as the load increased,
transition into total load rejection and the turbine was set to a no-load setting. Consequently, the run-
ner swiftly accelerated to the runaway speed from the deceleration
mode. Unsteady pressure measurements were collected under tran-
sient turbine conditions.
During steep closing of guide vanes, pressure rise in the inlet
conduit was 3.4% because of a water hammer effect. The sudden
shift from water acceleration to deceleration spurred steep surging
in the conduit. Furthermore, amplitudes of the blade passing fre-
quency were obtained far, i.e., 7.2 m, from the turbine. In the vane-
less space, pressure linearly followed the runner rotational speed.
Amplitudes of the blade passing frequency were dominant (15% of
ρE), and the highest amplitudes were obtained at the time of total
load rejection. Two harmonics (0.5f b and 2fb ) of the blade passing
frequency were obtained for a particular time, i.e., 3–10 s (roughly
the runaway speed). This may be associated with the strong reflec-
tions of hydroacoustic waves and with the interference of pressure
Fig. 24. Time-dependent signal-to-noise ratio of the pressure data from waves from subsequent interactions observed at high rotational
the draft tube cone (DT1, DT2, DT3, and DT4) during power ramp-up speeds.
from BEP and transition into total load rejection To investigate the pressure field along the blade passage in the
runner, four pressure sensors (R1, R2, R3, and R4) were flush-
mounted. Pressure variations were measured as approximately 9,
2.5, 2, and 1.8% of ρE at the R1, R2, R3, and R4 locations, respec-
variations in the pressure amplitudes. Such low-frequency pressure tively. The highest pressure amplitudes (up to 20% of ρE) in the
fluctuations are shown in the spectrogram (Fig. 22). turbine were obtained from the R1 location in the runner. Ampli-
To investigate unsteady variations observed at the draft tube tudes were maximized across all measurement locations in the
outlet, a pressure sensor, DT5, was mounted 10 mm in front of turbine. In addition to the guide vane passing frequency, strong
the draft tube outlet. Time-dependent pressure variations and fluctuations in stochastic frequencies were obtained from the run-
normalized pressure fluctuations (0–4 s) observed are shown ner. The conducted coherence analysis showed dominant ampli-
in Fig. 25. Moderate pressure variations were observed during tudes of the frequency range (10–100 Hz and over 300 Hz).
load change and transition to total load rejection. Pressure lev- Hence, estimations of the fatigue loading on blades during the de-
els dropped rapidly as the runner accelerated to the runaway sign phase based on expected deterministic frequencies may not be
speed, and the maximum level of pressure drop was measured sufficient for robust turbine design.
as 10% at ∼24 s. Thereafter, pressure levels slowly increased Unlike that of the vaneless space and runner, pressure in the
and stabilized at 106 s. Normalized pressure fluctuations are draft tube inversely followed the rotational speed. After total load
shown in the magnified window for 1–4 s, in within the maxi- rejection, pressure levels dropped by approximately 5% from the
mum amplitudes are 2% of ρE during the load change and tran- initial value and gradually recovered as rotational speeds decreased
sition to total load rejection. Effects of the runaway condition at from the runaway speed. Peak-to-peak amplitudes of unsteady
the draft tube outlet are very minor relative to those of the other pressure fluctuations were measured as 4.8%. In addition to the
locations in the turbine. Pressure amplitudes are similar to those blade passing frequency, stochastic frequencies with similar ampli-
obtained at BEP. tudes were obtained during the transient period.

© ASCE 04018027-13 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


The examined model Francis turbine was of a reduced scale Acknowledgments
(1∶5.1) relative to prototypes operating at the Tokke power plant
in Norway. It is clear from the obtained measurements that pressure These experiments were conducted through a research project
amplitudes can reach up to 20% of ρE when unexpected load entitled “High head Francis turbine.” The project was finan-
rejection occurs during load change. In addition to deterministic cially supported by the Research Council of Norway and by the
frequencies of RSI, amplitudes of stochastic frequencies were Norwegian hydropower industry.
found to be persistent and critically high over short periods.
One or more of such events may be sufficient to spur fatigue crack-
ing in the prototype runner because stochastic frequencies are Notation
spread across a range of 100 Hz. This broad range may coincide
with natural frequencies of the runner and may generate resonance. The following symbols are used in this paper:
Unsteady pressure measurements collected under such conditions c = absolute velocity (m s−1 );
Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

are vital even when a turbine is designed for the safest conditions. cm = flow velocity (m s−1 );
However, in regard to prototype design, a key challenge concerns cu = tangential component of absolute velocity (m s−1 );
scaling from laboratory measurements to real turbines. As sug- D = diameter (m);
gested in IEC 60193, trustworthy scaling is achieved only under DT1, DT2, DT3, and DT4 = locations of pressure sensors
steady-state operating conditions around the BEP. The present in the draft tube;
study involved the collection of transient measurements under E = specific hydraulic energy (J kg−1 ); E ¼ gH;
time-dependent pressure loading and difficult scaling conditions. e^ = uncertainty (%);
Three key parameters must be addressed in detail prior to proto- f = frequency (Hz); f ¼ f=n;
type scaling: (1) speed of sound and hydroacoustic effects; (2) cav- g = gravity (m s−2 ); g ¼ 9.821465 m s−2 ;
itation and surface roughness levels; and (3) standing waves. The H = head (m);
speed of sound varies significantly from a model to a prototype IN1, IN2 = locations of the pressure sensors on the conduit;
because it is dependent on the hydraulic system and conduit in-
n = runner angular speed (revolution per second);
volved. In addition, the presence of air content and the combined
nED = speed factor;
responses of mechanical structures contribute more scaling com-
plexities. Regarding the second parameter, it is well known that p = pressure (Pa);
cavitation and surface roughness play important roles in scaling. p~ E = factor of pressure fluctuations; p~ E ¼ p=ρE;
~
During cavitation, the presence of gas pockets reduces pressure am- ρE ¼ 116.84 kPa at BEP;
plitudes, but cavitation spurs random fluctuations as cavitation bub- Q = discharge (m3 s−1 );
bles collapse. After load rejection, the runner accelerates and then QED = discharge factor;
decelerates, and cavitation behavior is no longer known. However, R1, R2, R3, and R4 = locations of pressure sensors in the
future investigations on the relationship between runner rotational runner;
speed and cavitation may prove useful. Regarding the third param- r = radius (m);
eter, effects of standing waves are largely dependent on hydraulic T = torque (N m);
systems functioning near turbines upstream and downstream, t = time (s);
which affect wavelengths and frequencies. Variations in conduit u = tangential velocity (m s−1 );
lengths and/or penstock joints alter the wavelengths and resonance VL1, VL2 = locations of pressure sensors in the vaneless space;
frequencies of standing waves. For such cases, laboratory test data w = axial/relative velocity (m s−1 );
are irrelevant. z = number of blades/guide vanes;
α = guide vane angle (degrees);
β = blade angle (degrees);
Future Work
γ = coherence;
It is essential to apply transient pressure loadings derived from η = efficiency;
laboratory tests to real machines. To address this issue, two inter- ρ = water density (kg m−3 ); and
dependent approaches must be considered: (1) detailed experiments ω = angular speed (rad s−1 ).
on more pressure sensors focused on the speed of sound measure-
ments; and (2) numerical simulations considering levels of
Subscripts
flow compressibility. Compressible flow simulations conducted at
BEP have recently been completed (Trivedi 2018). Such simula- b= blade;
tions will help elucidate the standing waves and hydroacoustic phe- gv = guide vane;
nomena in turbines. Similar approaches can be applied to prototype h= hydraulic;
turbine simulations. r= relative;
Future studies will further investigate stochastic pressure fluc- rms = root-mean square;
tuations through the use of different tools such as rainflow count- s= systematic; and
ing diagrams and other time-dependent techniques. A numerical
t= total.
approach presented by Li et al. (2017) may be useful for
transient-flow analysis numerically. Such analysis will help to
understand the complex flow pattern in blade passages during
References
shutdown. Future studies must also focus on stochastic fluctua-
tions and on effects on runner life spans based on two to five Amiri, K., Mulu, B., Cervantes, M. J., and Raisee, M. (2016). “Effects of
cases of exigent ramping and based on daily transitions to total load variation on a Kaplan turbine runner.” Int. J. Fluid Mach. Syst.,
load rejection. 9(2), 182–193.

© ASCE 04018027-14 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


Arpe, J., Nicolet, C., and Avellan, F. (2009). “Experimental evidence of Hübner, B., Weber, W., and Seidel, U. (2016). “The role of fluid-structure
hydroacoustic pressure waves in a Francis turbine elbow draft tube interaction for safety and life time prediction in hydraulic machinery.”
for low discharge conditions.” J. Fluids Eng., 131(8), 081102. Earth Environ. Sci., 49(7), 072007.
Beevers, D., Branchini, L., Orlandini, V., De Pascale, A., and Perez-Blanco, IEC (International Electrotechnical Commission). (1991). “Field accep-
H. (2015). “Pumped hydro storage plants with improved operational tance tests to determine the hydraulic performance of hydraulic
flexibility using constant speed Francis runners.” Appl. Energy, turbines, storage pumps and pump-turbines.” IEC 41, Geneva, 430.
137(1), 629–637. IEC (International Electrotechnical Commission). (1999). “Hydraulic
Béguin, A., Nicolet, C., Hell, J., and Moreira, C. (2017). “Assessment of turbines, storage pumps and pump-turbines: Model acceptance tests.”
power step performances of variable speed pump-turbine unit by means IEC 60193, Geneva, 578.
of hydro-electrical system simulation.” Proc., HYPERBOLE Conf., IOP ISO. (2005). “Measurement of fluid flow-Procedures for the evaluation of
Publishing, Bristol, U.K., 5. uncertainties.” ISO 5168:2005(E), Geneva, 65.
Bessa, R., Moreira, C., Silva, B., Filipe, J., and Fulgêncio, N. (2017). “Role Li, Z., Bi, H., Karney, B., Wang, Z., and Yao, Z. (2017). “Three-
Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

of pump hydro in electric power systems.” Proc., HYPERBOLE Conf., dimensional transient simulation of a prototype pump-turbine during
IOP Publishing, Bristol, U.K., 7. normal turbine shutdown.” J. Hydraul. Res., 55(4), 520–537.
Botero, F., Hasmatuchi, V., Roth, S., and Farhat, M. (2014). “Non-intrusive Liu, X., Luo, Y., Karney, B. W., and Wang, W. (2015). “A selected literature
detection of rotating stall in pump-turbines.” Mech. Syst. Signal review of efficiency improvements in hydraulic turbines.” Renewable
Process., 48(1–2), 162–173. Sustainable Energy Rev., 51(11), 18–28.
Bucur, D. M., Dunca, G., Cervantes, M. J., Călinoiu, C., and Isbăşoiu, E. C. Liu, X., Luo, Y., and Wang, Z. (2016). “A review on fatigue damage mecha-
(2014). “Simultaneous transient operation of a high head hydro power nism in hydro turbines.” Renewable Sustainable Energy Rev., 54(2),
plant and a storage pumping station in the same hydraulic scheme.” 1–14.
Earth Environ. Sci., 22(4), 042015. Luna-Ramírez, A., Campos-Amezcua, A., Dorantes-Gómez, O., Mazur-
Caralis, G., Papantonis, D., and Zervos, A. (2012). “The role of pumped Czerwiec, Z., and Muñoz-Quezada, R. (2015). “Failure analysis of
storage systems towards the large scale wind integration in the Greek runner blades in a Francis hydraulic turbine—Case study.” Eng. Fail.
power supply system.” Renewable Sustainable Energy Rev., 16(5), Anal., 59(1), 314–325.
2558–2565. Melot, M., Monette, C., Coutu, A., and Nenneman, B. (2013). “Speed-
Chen, C., Nicolet, C., Yonezawa, K., Farhat, M., Avellan, F., and no-load operating condition: A new standard Francis runner design
Tsujimoto, Y. (2008). “One-dimensional analysis of full load draft tube procedure to predict static stresses.” 18th Int. Conf. on Hydraulics,
surge.” J. Fluids Eng., 130(4), 041106. Water Resources, Coastal and Environmental Engineering, HYDRO,
Claude, J.-M. (2017). “Performances achieved to the grid by a full power Innsbruck, Austria, 8.
converter used in a variable speed pumped storage plant.” Proc., Moisan, E., Giacobbi, D.-B., Gagnon, M., and Léonard, F. (2014). “Self-
HYPERBOLE Conf., IOP Publishing, Bristol, U.K., 6. excitation in Francis runner during load rejection.” Earth Environ. Sci.,
Coutu, A., Proulx, D., Coulson, S., and Demers, A. (2004). “Dynamic as- 22(3), 032025.
sessment of hydraulic turbines-high head Francis.” Proc., Hydrovision, Monette, C., Marmont, H., Chamberland-Lauzon, J., Skagerstrand, A.,
Quebec, 13. Coutu, A., and Carlevi, J. (2016). “Cost of enlarged operating zone
Coutu, A., Roy, M. D., Monette, C., and Nennemann, B. (2008). for an existing Francis runner.” Earth Environ. Sci., 49(7), 072018.
“Experience with rotor-stator interactions in high head Francis runner.” Morissette, J., et al. (2016). “Stress predictions in a Francis turbine at
Proc., 24th IAHR Symp. on Hydraulic Machinery and Systems, no-load operating regime.” Earth Environ. Sci., 49(7), 072016.
International Association for Hydro-Environment Engineering and Re- Mossinger, P., and Jung, A. (2016). “Transient two-phase CFD simulation
search, Madrid, Spain, 10. of overload operating conditions and load rejection in a prototype sized
Dolecek, G. J. (2013). Random signals and processes primer with Francis turbine.” Earth Environ. Sci., 49(9), 092003.
MATLAB, Springer, New York. Nennemann, B., et al. (2014). “Challenges in dynamic pressure and stress
Dörfler, P., Sick, M., and Coutu, A. (2013). Flow-induced pulsation and predictions at no-load operation in hydraulic turbines.” Earth Environ.
vibration in hydroelectric machinery, Springer, London. Sci., 22(3), 032055.
Dorji, U., and Ghomashchi, R. (2014). “Hydro turbine failure mechanisms: Nicolet, C. (2007). “Hydroacoustic modelling and numerical simulation of
An overview.” Eng. Fail. Anal., 44(9), 136–147. unsteady operation of hydroelectric systems.” Ph.D. thesis, Ecole Poly-
Ellingsen, R., and Storli, P.-T. (2015). “Simulations of the dynamic load in technique Federale de Lausanne, Lausanne, Switzerland.
a Francis runner based on measurements of grid frequency variations.” Nicolet, C., Alligné, S., Kawkabani, B., Koutnik, J., Simond, J.-J., and
Int. J. Fluid Mach. Syst., 8(2), 102–112. Avellan, F. (2009). “Stability study of Francis pump-turbine at run-
Engelberg, S. (2008). Digital signal processing an experimental approach, away.” Proc., 3rd Meeting of IAHR Workgroup on Cavitation and
Springer, London. Dynamic Problems in Hydraulic Machinery and Systems, International
Gagnon, M., and Léonard, F. (2013). “Transient response and life assess- Association for Hydro-Environment Engineering and Research,
ment: Case studies on the load rejection of two hydroelectric turbines.” Madrid, Spain, 14.
Surveillance, 7(1), 1–11. Nicolet, C., Alligné, S., Kawkabani, B., Simond, J.-J., and Avellan, F.
Gagnon, M., Nicolle, J., Morissette, J. F., and Lawrence, M. (2016). (2008). “Unstable operation of Francis pump-turbine at runaway: Rigid
“A look at Francis runner blades response during transients.” Earth En- and elastic water column oscillation modes.” Proc., 24th IAHR Symp.
viron. Sci., 49(5), 052005. on Hydraulic Machinery and Systems, International Association for
Guo, W. C., Yang, J. D., Chen, J. P., Peng, Z. Y., Zhang, Y., and Chen, C. C. Hydro-Environment Engineering and Research, Madrid, Spain, 13.
(2016). “Simulation of the transient processes of load rejection under Nicolet, C., Ruchonnet, N., Alligné, S., Koutnik, J., and Avellan, F. (2010).
different accident conditions in a hydroelectric generating set.” Earth En- “Hydroacoustic simulation of rotor-stator interaction in resonance con-
viron. Sci., 49(5), 052016. ditions in Francis pump-turbine.” Earth Environ. Sci., 12(1), 012005.
Hasmatuchi, V., Farhat, M., Roth, S., Botero, F., and Avellan, F. (2011). Nishi, M., and Liu, S. (2013). “An outlook on the draft-tube-surge study.”
“Experimental evidence of rotating stall in a pump-turbine at off-design Int. J. Fluid Mach. Syst., 6(1), 33–48.
conditions in generating mode.” J. Fluids Eng., 133(5), 051104. Pacot, O., Kato, C., Guo, Y., Yamade, Y., and Avellan, F. (2016). “Large
He, L. Y., Wang, Z. W., Kurosawa, S., and Nakahara, Y. (2014). “Reso- eddy simulation of the rotating stall in a pump-turbine operated in
nance investigation of pump-turbine during startup process.” Earth En- pumping mode at a part-load condition.” J. Fluids Eng., 138(11),
viron. Sci., 22(3), 032024. 111102.
Hell, J. (2017). “High flexible hydropower generation concepts for future Pejovic, S., and Karney, B. (2014). “Guidelines for transients are in need
grids.” Proc., HYPERBOLE Conf., IOP Publishing, Bristol, U.K., 7. of revision.” Earth Environ. Sci., 22(4), 042006.

© ASCE 04018027-15 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027


Sivakumar, N., Das, D., and Padhy, N. P. (2014). “Variable speed operation Trivedi, C., Cervantes, M. J., and Dahlhaug, O. G. (2016b). “Numerical
of reversible pump-turbines at Kadamparai pumped storage plant—A techniques applied to hydraulic turbines: A perspective review.” Appl.
case study.” Energy Convers. Manage., 78(2), 96–104. Mech. Rev., 68(1), 010802.
Trivedi, C. (2017). “A review on fluid structure interaction in hydraulic Trivedi, C., Cervantes, M. J., and Gandhi, B. K. (2016c). “Numerical in-
turbines: A focus on hydrodynamic damping.” Eng. Fail. Anal., vestigation and validation of a Francis turbine at runaway operating
77(7), 1–22. conditions.” Energies, 9(3), 149.
Trivedi, C. (2018). “Investigations of compressible turbulent flow in a high Trivedi, C., Gandhi, B., and Cervantes, M. (2013b). “Effect of transients
head Francis turbine.” J. Fluids Eng., 140(1), 011101. on Francis turbine runner life: A review.” J. Hydraul. Res., 51(2),
Trivedi, C., and Cervantes, M. (2017). “Fluid structure interaction in 121–132.
hydraulic turbines: A perspective review.” Renewable Sustainable Trivedi, C., Gandhi, B., Cervantes, M., and Dahlhaug, O. (2015b). “Exper-
Energy Rev., 68(1), 87–101. imental investigations of a model Francis turbine during shutdown at
Trivedi, C., Cervantes, M., Dahlhaug, O., and Gandhi, B. (2015a). “Exper- synchronous speed.” Renewable Energy, 83(11), 828–836.
imental investigation of a high head Francis turbine during spin-no-load Widmer, C., Staubli, T., and Ledergerber, N. (2011). “Unstable character-
Downloaded from ascelibrary.org by Chalmers University of Technology on 01/20/20. Copyright ASCE. For personal use only; all rights reserved.

operation.” J. Fluids Eng., 137(6), 061106.


istics and rotating stall in turbine brake operation of pump-turbines.”
Trivedi, C., Cervantes, M., Gandhi, B., and Dahlhaug, O. G. (2013a).
J. Fluids Eng., 133(4), 041101.
“Experimental and numerical studies for a high head Francis turbine
Yang, W., Norrlund, P., Saarinen, L., Yang, J. D., Guo, W. C., and
at several operating points.” J. Fluids Eng., 135(11), 111102–1111017.
Zeng, W. (2016a). “Wear and tear on hydro power turbines—
Trivedi, C., Cervantes, M., Gandhi, B., and Dahlhaug, O. (2014a). “Exper-
imental investigations of transient pressure variations in a high head Influence from primary frequency control.” Renewable Energy, 87(3),
model Francis turbine during start-up and shutdown.” J. Hydrodyn. 88–95.
Ser. B, 26(2), 277–290. Yang, W., Norrlund, P., and Yang, J. (2016b). “Analysis on regulation strat-
Trivedi, C., Cervantes, M., Gandhi, B., and Dahlhaug, O. (2014b). egies for extending service life of hydropower turbines.” Earth Environ.
“Pressure measurements on a high-head Francis turbine during load Sci., 49(5), 052013.
acceptance and rejection.” J. Hydraul. Res., 52(2), 283–297. Yasuda, M., and Watanabe, S. (2016). “How to avoid severe incidents at
Trivedi, C., Cervantes, M., Gandhi, B., and Dahlhaug, O. (2014c). pumped storage power plants.” Earth Environ. Sci., 49, 112002.
“Transient pressure measurements on a high head model Francis turbine Zeng, W., Yang, J., Tang, R., and Yang, W. (2016). “Extreme water-
during emergency shutdown, total load rejection, and runaway.” J. Flu- hammer pressure during one-after-another load shedding in pumped-
ids Eng., 136(12), 121107. storage stations.” Renewable Energy, 99(12), 35–44.
Trivedi, C., Cervantes, M. J., and Dahlhaug, O. G. (2016a). “Experimental Zuo, Z., Liu, S., Sun, Y., and Wu, Y. (2015). “Pressure fluctuations in the
and numerical studies of a high-head Francis turbine: A review of the vaneless space of high-head pump-turbines—A review.” Renewable
Francis-99 test case.” Energies, 9(12), 74. Sustainable Energy Rev., 41(1), 965–974.

© ASCE 04018027-16 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(6): 04018027

Vous aimerez peut-être aussi