Beruflich Dokumente
Kultur Dokumente
I. INTRODUCTION
Vibrational problems of hydraulic generating units have
long been recognized as a cause of decreased availability
and increased production cost. Sometimes vibration level is
so high that the operator has to reduce the power output
below optimum. Large Francis turbines have been exceptionally subject to vibrational difficulties due to flow induced
structurdl vibrations, [I]. When operated at partial load
Francis turbines may experience severe vibrational problems due to the presence of intense pressure surges in the
draft tube, [Z]. Electromechanicaloscillations may also arise
leading to power swing, vibrations and noise.
A hydroelectric plarit may be considered as a system
consisting of several subsystems : the penstock including
any surge chamber, the hydraulic machine, the speed
governor, the tailrace and, finally, the generator and electrical network. Excitation of vibrations may arise from various
sources, such as flow variation in the penstock, part load
vortex in the draft tuble, load rejection, etc. The present
state of knowledge perrnits to provide remedies to vibration
problems in the field, after commissioning of the plant.
Nevertheless the main cause is to predict vibration in the
design stage.
Simulation of hydro turbines has traditionally been
performed neglecting water column elasticity, [3]. Although
such models have been1 extensively used for system stability
studies, they are not suitable for vibration analysis of
hydroelectric plants, especially when long penstocks are
implemented. The water column elasticity principle, [4], has
recently been used for the identification of higher order
hydraulic turbine models, [5],as well as for design purposes,
161 & [71.
Speed governor dynamics greatly influences hydro plant
dynamic response in the significant frequency range of up to
about 100 rls. Detailed hydro governor models are necessary to account for its filtering action on high frequencies of
the penstock -turbine transfer function, [3] & [8].
Pressure surges in the draft tube have been extensively
studied, and a great deal of knowledge has been gained and
published, [I], [9] & [IO]. Information is nowadays available
about the frequency, intension and inception of vortex
phenomenon.
Technical literature lacks papers making use of the recent
contributions to the study of the dynamic response of the
integrated hydro plant, [Ill, [I21 & [13].
In the present paper a transfer function approach is used
for vibration analysis of hydro plants with Francis turbines.
Full equations are implemented for the accurate simulation
of the turbine - penstock subsystem, taking into consideratiin water column elasticity. Torsionals of the shaft system
are also taken into consideration, while speed governor is
420
.,
B = a11 - a21a12/a22
As it can be seen from (4) the gain and phase of these
transfer functions depend upon the operating point, by
means of parameters a,]. Water starting time constant T,
depends upon the length L and cross section area S of the
penstock :
T=
,
L Qb/g s Hb
where Q, and H, are the base values of flow rate and head
respectively. Penstock elastic time constant T, is the time
taken for the pressure wave to travel the length of the
penstock at velocity a :
(1)
1
, = Lla
where :
h(s) is the differential head at the turbine inlet (P.u.),
q(s) is the differential flow rate (P.u.),
T, is the water starting time constant (s),
T, is the penstock elastic time constant (s).
It is worth noting that the relationship between the turbine
rate of flow and the turbine head depends only on the
penstock, being independent of turbine's characteristics. If
friction losses are to be taken into account, a second term
4, has to be subtracted from the right hand side of (I).
The
p.u. value of this term is in the order of I x I O - ~ ,proportional
to the steady state flow rate, and is usually neglected in
frequency response analysis, [5] & [Ill.
For small variations around an equilibrium point the
turbine can be represented by the following linearized
equations :
4 = all
a12 Wt
a13
G,w
(3)
where the transfer functions G&S) & G,(s) relate the mechanical torque with the gate opening and speed respectively
G&s) = a23[1+A(T,/T,)tanh(T,s)]/[l+a~l(Tw/Te)tanh(Tes)]
(6)
(7)
(8)
m = G,g
(5)
=G
M = Q H (n/n,)/n
(9)
where capital letters have been used for the absolute p.u.
values and n = noin, is the ratio of efficiencies at the current
operating point over that of base values. For deviations
around rated speed and head, the partial derivatives versus
421
load are deduced:
0.01
0,l
FREcaJwcy (W)
Fig.1 Transfer functicln G (s) of an ideal turbine at full load
(T, = 0.384 s, T, = 1.637 s)
RATE
LIMITS
va;wE
POSITW
LlMIT!j
OISTRIBUTOR vncE
AN0 GATE
SERVOMOTOR
SERVOMOTOR
TRANSIENT DROOP
COMPENSATION
PERMANENT D K W
COMPENSATION
50
40
-20
30
-40
C. Shah+torsionals
z.
z< O
20
3 0
J p2d K d = I
f
(11)
where :
J = diag [Jj, Jg] is the diagonal inertia matrix,
K i s the stlffness matrix, and
I=[T, -T,JT is the column matrix of excitation torques.
The eigenfrequency ft of this subsystem can be easily
computed from the charaderistio equation :
?lo
-120
-140
-20
-30
-160
-40
-180
0,Ol
0.1
10
FREQUENCY (Hz)
qI
422
40
20
h
n
U
-23
-40
om
01
FRE
I
(W
-I
(12)
(13)
Table I
D. Network interaction
(14)
"
SQ.&ecl#YeLw-data
~=0,02,6=0.387,T,=5.0,T,=0.04,T,=0.2
(15)
in the field, since it is critical in order to obtain close agreement between the model and the unit. Fig.5 contains
information about the values of turbine characteristic
parameters a,, in the whole loading range. Since turbine gain
aZ3gets its maximum value of about 1.2 p.u. in the range of
0.4 to 0.6 p.u. loading, it is evident that the unit responds the
strongest while partially loaded.
423
QO
Q
Q2
l a3 (14
a5 06 a7 Q8 Q9 10 11
Q2
I-mD@.U)
06
0.8
12
l-cm @.U)
Fig5 Turbine parameters a,, vs. power output (at rated head)
where :
Fig.6 Damping constant of poles ( A ) & zeros (.) of the turbine transfer
function (Kastraki HPP, unit I)
B. Vibrationanawsis
B, = 4a,,T,/Te2,
Q4
8, = A, 6,= 24allT,/Te2,
Bo = A,
In order to investigate the problem, vibration measurements were undertaken in all four units for the whole
operating zone, Table II. It can be clearly seen that vibration
level is higher in unit I, while unit IV seems to have best
vibration performance. Frequency analysis of vibration
signal showed that the vortex component increased considerably in the loading range where vibrations were maximum. Vortex frequency was found to vary from 0.7 Hz, at
0.5 p.u. output, to almost 1 Hz, at 0.75 p.u. These values
agree with measurements conducted in other Francis units,
as well as with theoretical predictions, [IO]. Fig.7 shows
units I shaft vibration spectrum at 50 MW. Main frequency
components lie at vortex frequency, synchronous frequency
as well as at 1.7 and :3.8Hz.
Given a value of about -0.8 p.u. for the turbines self
regulation it is concluded that the maximum gain of the
transfer function GJs) in the same loading range equals
about 65% of G,(s). This means that speed variations
i n f l u e n c e on mechanical torque is substantially lower than
424
trum, it is strongly amplified in the penstock due to resonance. For generator angles between 20" and 60" the electromechanical oscillations frequency varies between 1 and
1.7 Hz. At the high end of this range, elecromechanical
oscillations are expected to influence penstock cross section
vibrations by their second harmonic, in the same manner as
torsional vibrations do. This is the reason why these vibrations are higher at higher unit loadings.
IV. CONCLUSIONS
10
Table II
unit
0.25
0.5
V. REFERENCES
Vibration Phenomena in Francis Turbines :
Trans. of rhe 7dh IAHR
Symposium, Tokyo, 1980.
[2] A.H. Glattfelder, H. Grein and P.K. Dorfler, "System
Vibrations in Hydro Plants", Escher Wyss News, 111982,
pp.27-31.
[3] IEEE Committee Report, "Dynamic Models for Steam
and Hydro Turbines in Power Systems ", IEEE Trans. PAS,
vol. PAS-104, no. 4, pp.1904-4914, 1973.
[I] H. Grein,
"
425