Sei sulla pagina 1di 7

~ ~-

IEEE Transactions on Power Systems, Vol. 9, No. 4, November 1994 1709

HYDRO TURBINE MODEL FOR SYSTEM DYNAMIC STUDIES

E, De Jaeger N.Janssens B. Malfliet F. Van De Meulebroeke


Senior Meniber, IEEE

Laborelec (Belgian Laboratory of the Electricity Industry), Electrical Division


B-1630 Linkebeek (Belgium)

,411stract - A non-linear model for dynamic studies of hydro The second section is devoted to the modeling of a power
turbines is proposed. The model of a single turbine and its plant consisting of several turbines, supplied by a common
water supply conduit is given in the first section of the tunnel. The model presented is able to simulate large
paper, together with a paraneter estimation from field tests. transients such as load rejections and their effect on the
The second section is devoted to the modeling of multiple remaining groups, due to the hydraulic coupling through the
turbines supplied from a common tunnel, taking into common conduit.
account the hydraulic coupling effects. Comparisons
between simulation results and experimental data are Comparisons between simulation and experimental
presented. results (are presented.

keywords : modeling, hydro turbine, system dynamics 1. MODELING OF A SINGLE TURBINE

INTRODUCTI( )N 1.1 Modeling of the Water Column

This paper deals with a hydraulic model suitable for The model is based upon the assumption that water acts
dynamic studies Of hydro power plaIltS. Figure 1 shows the 21s illcolnpressihle fluid [ 11. The penstock is modeled
basic structure of a hydro turbine model. assuming a rigid conduit.

From the laws of momentum, the rate of change of flow


in the conduit is :
Mech.uiical Toique
dq
-= -(Po-P-Pf) (1)
TurblneCoiitrol g,e Turhuie Rotor
Dyiiaiiics > oyll'ullcs -3 Dyixun ics
Speed >
dt Tw

I
Assigned Power
pressure

4 coodull
Dynanllcs

Fig. 1 - Striictiire (fahydro tiii-bine


tlow
I
Electrical Torque static pressure of the water column, p the pressure at the
turbine admission and pf the pressure loss due to friction in
the conduit. T, is a characteristic constant of the conduit
which is usually known as the water time constant or the
water starting time, when expressed in a p.u. system.
The friction pressure loss is taken to be proportional to flow
squared [ 11:
In this context, we will treat more precisely the modeling
of the prune mover, including the water supply conduit.
In the first section, we present the model of a single
turbine and its water supply conduit. The general non-
linear model is given and a paraneter estimation is 1.2 Modeling of the Turl)ine
proposed. Typical results are given (they result from field
tests, conducted at the Coo pumped storage plait, Belgium). 'I'lie turbine is modeled by its valve characteristic [ 11:

94 WM 186-7 PWRS A paper recommended and approved q= Gfi (3)


by the IEEE Power System Engineering Committee of the
IEEE Power Engineering Society for presentationat the
IEEE/PES 1994 Winter Meeting, New York, New York, Where G is a fUIlcti011 O f the gate positioll Y.
January 30 - February 3, 1994. Manuscript submitted Froin experimental d a h analysis [2], this functioll has been
July 8, 1993; made available for printing
December 6, 1993. identified as the combination of a quadratic function of the
0885-8950/94/$04.00 0 1994 IEEE
1710

gate position y and a first order tilter (simple time lag). I11 Figure 2 depicts such a function of the flow,
steady-state operation, we GW write : corresponding to a 260 m head (Coo I pumped storage plant,
Belgium). The zero flow point is adjusted in order to
Go(y) = do+ d,y + d,y2 (4) accurately simulate the turbine speed decay when the gate is
completely closed, for example after a load rejection (see
while the global tr'ansfer function will be written, section 2.4 of this paper).
introducing the Laplace formalism loo0 - ' . 1 . . ' 1 . . - 1 . . .
A

where s is the L?place variable and Go(s) the hplace


msform of Go(y(t)).

1 3 Mechanical Power

The net mech,ulicaI turbine power P , is evaluated from


the available hydraulic power P,,,,, by subtracting the
friction losses.
et
8cr 200

0 20 40
+
y

60
unit 2
unit 3

80
FLOW ( m3/sec )
The hydraulic power is proportional to flow times Fig. 2 - Frictionfactor Kbq)
pressure, while the friction losses are the product of the
friction torque (function of flow and speed) and the speed : 1.5.2 Transfer function G(s)

Quasi-steady-state tests have been performed at the Coo


pumped storage plant (Belgium), in order to identify the
transfer function G(s)(equation (3)) and other parameters of
I The friction torque can be expressed as : the turbine - conduit model.
Only one turbine was operating at a time and these tests
consisted of modifying the assigned generation (power set-
point) and measuring important variables as gate position y,
where Kf (9) is a quadratic function of the flow q (flow pressure at the turbine admission p and generated electrical
depending friction factor) [2]: power Pe,. Figures 3 to 5 show typical recorded data; letters
a to e mark the v,arious stages of the test :
a = opening of the spherical valve
b = opening of the gate and acceleration of the turbine
1.4 Non-linear Model of a Single Hydro Unit c = synchronizing of the alternator
d = power set-point modifications
Grouping equations (1) to (8) leads to the global block e = opening of the main circuit breaker (disconnecting the
diagram shown in appendix,'giving the non-linear model of alternator from the grid)
a single hydro turbine (fig. A l l .

1.5 Model parameter determination

1.5.1 Friction factor Kkq)

The turbine friction factor is obtiined as function of the


flow by processing efficiency measurements, as given in [3,
41. These data list the net turbine mechanical power, the
flow and the efficiency of the units, measured for a large
number of steady-state operating conditions, with constant
d
. . . . F.........,.........,......"..
head.
The coefficients of the polynomial (8) are estimated by curve - 100 0 100 200 300
TIME ( 5 )
fitting.
Fig. 3 - Measured gate position
1711

l a b c d

w
5m 20::- -

v )
w ::
C T
CT -
e, 1 0 1 -

0 - . . . . . . . . .I . . . . . . . . . 1 . . . . . . . . . 1 . . . . . . . . . -

-100 0 100 2 00 300 0 10 GATE POSITION


20 ( cm ) 30 40
TIME ( s )
Fig. 4 - Measured pressitre ut the trrrbine admission Fig. 6 - Meusrrred trunsjierflinctions G and Go(y)

1.5.3 Parameters T, and f


P

The knowledge of the measured pressure, the computed


flow (9) and its derivative with respect to time allows the
simultaneous identification of Tw and fp, from equation (1)
(law of momentum).

2. MODELING OF MULTIPLE PEN STOCKS AND


TUKIjINES SUl'P1,IED FKOM A COMMON TUNNEL

The above proposed model must be improved to take into


-50 . . . . . . . . .F . . . . . . . . . l d. . . . . . . . . l . . . . . . f . . account the coupling effect of pressure variations, existing
-100 0 100 200 300
TIME ( s ) bctweeii turbines supplied froin a coininon tunnel, as
pictured in figure 7.
Fig. 5 - Measitred generated electricul power Turbine 1

Froin the experiinenbd electrical active power P,, and


n
pressure p, and using the friction factor Kdq), one can
compute the flow by solving the following equation
(assuming that P,,, = P,, during the test) Turbine 2

Common tunnel

(Note that the speed rernains constant during these tests)

Application of ( 3 ) yields
h Turbine 3

G = -
k Fig. 7 - Midtiple fiirbines sripplied from coninion trtnnrl
1712
2.1 Modeling of the Conduits

Let us introduce the following symbols :

T, = water time constant of the common conduit (upper


tunnel) where (i ,j, k) = cyclic permutation of (1,2, 3).

Twi= water time constant of the individual penstock related This equation describes the flow dynamics in each
to machine i (i = 1.2 or 3) individual penstock, including the hydraulic interaction
(coupling effect) with the other turbines.
fpc = friction coefficient of the common conduit
The analysis of the results of load rejection tests
fpi = friction coefficient of the individual penstock related to (performed when the three turbines were in service) shows
machine i (i = 1,2 or 3) that the pressure variation (i.e. the difference between
steady-state pressures before and after the test) is
pc = pressure at the bifurcation between the common tunnel significantly greater for the disconnected unit than for the
and the individual penstocks. remaining machines. It suggests that the head loss due to
friction in the penstocks is much more important than in the
pOc= static pressure of the water column at the bifurcation upper conduit. This empirical conclusion led us to simplify
equation (14) by neglecting the fpc coefficient.
poi= static pressure at the turbine i admission (i = 1, 2 or 3)

pi = pressure at the turbine i admission (i = 1 , 2 or 3) 2.2 Modeling of the Turbines

qi = flow in individual penstock and turbine i (i = 1, 2 or 3) The simple turbine model (3) is not able to simulate very
large transients of the turbine :
qc = flow in the common conduit (forced to be equal to the - it cannot simulate accurately the turbine speed when this
sum of the flows in the individual penstocks, by the one undergoes large (positive or negative) variations (such
continuity equation) operating conditions occur in case of load rejection, for
example),
The flow dynamics in the common conduit is established - it leads to pressure indetermination and numerical
by applicatioR of (1) : problems when the gate reaches complete closure.

On empirical basis, the following improved


representation is thus proposed, in order to introduce a
speed deviation effect, which is a function of the gate
On the other hand, we can write, for each individual position :
penstock
q + (D(G-G&) A o ~+ qaff 1
G
=J;; (15)
A q is the speed deviation ftom synchronism, p m s s e d by
(i = 1,2 or 3) a simple-lag filter :

Extracting (pot - pc) from (11) and substituting in (12) 1


yields Ao~= -( 0 - 00) (16)
1 + Trii S
The constant offset quantities GM and qdf are
introduced in order to avoid indetermination, when the gate
reaches complete closure.
D is a constant parameter and the transfer function G is still
Recalling that the flow in the corninon conduit is equal to given by (5). but Go(y) m m have a sornewhat more
the sum of the individual flows, we obain finally : complicated expression: in fact, introducing
C;oi(y) = db + d;y + d;y2 (17)

Go(y) is now defined by


1713

Figure 8 shows the speed of the disconnected machine


(group 3) while figure 9 gives the pressure at the admission
of one of the remaining units (group 2).

Note that G, and ,q are not independent : they must be


chosen in such a way that the pressure at the turbine
admission is equal to the static pressure, when the gate is
completely closed i.e. when y = 0 and qi = 0 (remember that
the f coefficient has been neglected in (14)). In these
p4:
conditions, according to the preceding definitions, (15)
reduces to

(19)
GOff
Note also that when the speed remains construit, (15)
reduces to
Fig. 8 - Lolid rejection test :speed of the disconnecied
mnckine
(20)
which is similar to (3) 32 " ' I " ' I " ' I ~ " I ' ' ' I ' ' ~

- Experimental
......... Simulated
2.3 Non-linear Model of Multiple l u r i h e s

The general block diagram giving tlie non-linear model


of a hydro-turbine, including hydraulic interaction with the
other turbines supplied by a coininon tunnel, in shown i n
appendix (fig. A2).

Tlie transfer function C; (defined by (5), (17) aiid (18) )


can bc identilied from tests a s described i n 1.5.2. These
tests are also useful to estimate parameters such as ti,, and 4
T, = T,, + T, (separation of the two parmeters TN,cand -20 0 20 40 60 80 100
T,, can be made, knowing the relative lengths and areas of TlhlE ( s )
the coininon conduit and the individual penstocks, for Fig. 9 - Lorid rigeciion te.st :pressure (it ilre admission of
example). one of the remining tnachines
Other parmeters such as D or TI,, are to be tuned to
accurately sirnulate large transients, among which load
rejection (being one of the most practically interesting and
constraining transients).
A hydraulic model suitable for dynamic studies of hydro
power plants has been proposed. In this context, we have
The rotor dynamics is described by tlie classical rotating
inore precisely dealt with the modeling of the prime mover,
inasses equalion, relating the inertia torque to the resultant
including che water supply conduit.
of rhe torques 011 the rotor.
The general non-linear model has been given and a
p:u,unetcr estimation has been proposed. Typical results
were givcn (hey result from field tests, conducted at the Coo
2.4 Application
pumped srorage plant, Belgium).
Tlie inodeling of a power plant consisting of several
Load rejection tests have been performed at the Belgian turbines, supplied by a coinInoii tunnel has been investigated
pumped storage plant in Coo and the above presented rnotlel and we have shown that the presented model is able to
lins been used to simulate Ihcse transients. Comparisons simulate large transicnts as load rejections and their effect
1xtween measured and coinputed (lata are given in figures 8 on the remaining groups, due 10 the hydraulic coupling
and 9. tlirough the coinInoii conduit. Comparisons bctween
The initial loading of each machine W;L\ 0.8 p.u and load simulation ;uid experimental results have been presented
rejeclion was perfornied for one turbine out of tliree. ancl show the sufficient accuracy of the model.
1714

d,+ d,Y + %U2

1 I
G

Fig. A1 - Non-linear model of a single hydro turbine

Goffl

. ~ .* . .
a,+ ay *a2y-
3

GOff 4+ I I I

-
Fig. A2 Non-linear model of a hydro turbine taking into account the hydraulic coupling with the other turbines supplied
from the same tunnel
1715

REFERENCES
FClicien Van De Meulehroeke (SM'89) was born in 1940.
[l] IEEE Working Group on Prime Mover and Energy He received the Electrical Civil Engineer degree
Supply Models for System Dynamic Pcrfonnance Studies (Electronics) from the Catholic [Jniversity of Leuven (KUL),
"Hydraulic turbine and turbine control models for Belgium, in 1963.
system dynamic studies" Since 1964, he has been with Laborelec, the Belgian
IEEE Trans. on Power Systems, vol. 7, No 1 (feb. 1992) Laboratory of the Electricity Industry. From 1964 to 1987,
pp. 167-179 he was Chief Engineer, in ch'uge of turbogenerators
automation and system dynamics, in the section 2
[2] E. De Jaeger, B. Maltliet, N. Janssens, "ModiYe (Electronics and Automatics). In 1987, he became head of
hydraulique pour la centrale de Coo I. Essais du 26 juin the department"System dynamics and protections" in the
1991.", Laborelec Report EM-405-91-016/EDJ/BM/NJ. section E (Electrical Power). Since decernber 1988, he is
head of the Electrical Division of Laborelec.
[3] "Coo I, Groupe 3 : dktennination du rendement de la
turbine - poinpe kquipke d'une nouvelle roue.", Laborelec
Report 4-GD/5 - 194 R2, 14 dec. 1976.

[4] "Centrale de Coo - Trois-Ponts, Groupe 2: determination


du rendernent de la turbine-pompe kquipke de la nouvelle
roue.", Laborelec Report 4 - GD/S - 203 R2, 8 avr. 1977.

Emmanuel De Jaeger was born in Narnur, Belgium, on


September 25, 1962. He studied at the Catholic IJniversity of
Louvain (IJCL, Louvain-la-Neuve, Belgium) and graduated
as Electrical Civil Engineer in 1985. He received the 1'h.D.
degree in Electrical Engineering from the sane IJniversity,
in 1991.
Since 1991, he is with the system dynamics department of
the Belgian Laboratory of the Electricity I i i d ~ t r y
(Laborelec). His main fields of interest include theory and
parameters estiination of electrical machines, modeling,
simulation, analysis and control of power system.

Noel Janssens was born in Louvain, Belgium, on December


22, 1948. He studied electrical engineering at the [Jniversity
of Louvain and graduated in 1971. Research activities for
obtaining the Ph.D. degree (1981) were in the field of the
Inodeling of magnetic hysteresis and the study of
ferroresonance. From 1981 to 1983, he worked at ACEC
(Charleroi) as head for R & I) in the (.hi Load Tap Changer
department. From 1978 to 1981 and since 1984 he is with
Laborelec, first in the Mathematics and Data Processing
department and since 1989 i n the Electricity Division where
his main fields of interest are the modeling, simulation and
control of Power Systems. He is also teaching at the
[Jniversity of Louvain in the Electrical Engineering
department.

Bernard Malfliet was born i n Halle, Belgium on August


25, 1965. He received tlie diploma i n electrical engineering
from the University of Brussels in 1988. The sane year he
joined Laborelec at the system dynamics department. His
main interests are automated control and simulation of the
power grid. He is currently working on centralised voltage
and speed control for the Belgian grid and on turbogroup
simulation.

Potrebbero piacerti anche