Académique Documents
Professionnel Documents
Culture Documents
Schleif
Engineering and Research Center
Bureau of Reclamation
February 1971
Governor Characteristics for Large Hydraulic Turbines --
-- - - -
-E.
114198 71A C- 1
F. R. Schleif
9. P E R F O R M I N G O R G A N I Z A T I O N N A M E A N D ADDRESS
2 . SPONSORING A G E N C Y N A M E A N D ADDRESS
Same
15. S U P P L E M E N T A R Y NOTES
16. A B S T R A C T
In considering appropriate parameters for large hyc
characteristics to satisfv Dower svstem needs stror '
fundamental to control .characteristics which can be
flywheel effect and penstock time constant. To aic
combinations of parameters t o satisfy the power systt
interrelation was made. Results of the study are given
with several new features for better speed control wc
unit speed control and area load control. A guide for p
MAY 1 3 1971
Bureau of Reclamation
Denver, Colorado
-
17. K E Y WORDS A N D D O C U M E N T A N A L Y S I S
C . COSATI Field/Group 1 3 ~
18. DISTRIBUTION S T A T E M E N T 19. S E C U R I T Y C L A S S 2 1 . NO. O F P A G E
( T H I S REPORT)
Available from the National T e c h n i c a l Information Service, Operations 19
UNCLASSIFIED
D i v i s i o n , Springfield. V i r g i n i a 2215 1. 20. S E C U R I T Y C L A S S 2 2 . PRICE
1 ( T H I S PAGE)
UNCLASSIFIED $3.00
GOVERNOR CHARACTERISTICS FOR
LARGE HYDRAULIC TURBINES
by
F. R. Schleif
February 1971
These studies were conducted with the close cooperation and assistance
of many engineers, both in the field and in the Denver Office,
particularly C. G. Bates, B. G. Seitz, G. H. Johnson, and D. C. Erickson
of Hydraulic Machinery Section, Mechanical Branch, Division of
Design .
Page
The Problem. . . . . . . . . . . . . . . . . . . . . . . . . 1
Selectionof OverallControl Characteristics. . . . . . . . . . . . . . . . . . 1
GovernorTypeand Control Characteristics . . . . . . . . . . . . 2
Refinementof GovernorSystem ,.... 3
General. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . 3
TemporaryDroop Governor. . . . . . . . . .
. . . . . . . . . 4
DoubleDerivativeGovernor. . . . . . . . . .
. . . . . . . . . . . . 4
EquivalentHigh-OrderGovernors . . . . .
. . . . . . . . . . . . 6
Alinement ,. . . . . . . . . . . . 8
Refinementof ServomotorControl. . . . . . . . . . . . . . . . 9
Verificationby FieldTests . . . . . . . . . . . . . 10
SynchronizingPerformance. . . . . . . . . . . . . . . . . . . . . . . 14
Conclusions . . . . . . . . . . . . . . . . . . . . . . . 15
Applicability. . . . . . . . . . . . . . . . . . . . . . . . . 15
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Glossaryof Symbols . . . . . . . . . . . . . . . . . . . . 16
LIST OF FIGURES
Figure
Figure Page
10 Servosystemrepresentation. . . . . . . . . . . . . . . . . . . 9
11 Responses of closed loop servosystem of Figure 10c to
step signals within proportional range of valves. . . . . . . . . . . 9
12 Influence of gain and damping in servosystem upon
control of speed transients by derivative governor of
Figure 3 with a=0.05, T2 = 0.1, T4 = 0.1, T7 = 1.5,
T w = 1.85, T m = 10.5 10
""""""""""
13 Field test of closed loop servosystem on Unit G-4 in
existingGrandCouleePowerplant . . . . . . . . . . . . . . . . 11
14 Field test of governor systems on Unit G-4 in existing
Grand CouleePowerplant . . . . . . . . . . . . . . . . 11
15 Startup with gate limit at 0.4 per unit. . . . . . . . . . . . 12
16 Startup with gate limit at 0.2 per unit. . . . . . . . . . . . . . . 12
17 Computed dynamic performance for rejection of rated load . . . . 13
18 Computed performance for partial load rejection. . . . . . . . . . . 14
19 Response of speed to speed change signal under
conditions for synchronizing. . . . . . . . . . . . . . . 15
APPENDIX
ii
THE PROBLEM steam turbines, for the benefit of slightly higher
efficiency, develop a high percentage (roughly up to 80
In considering appropriate parameters for large percent) of their total output through the reheat cycle
hydropower generating units, of which those for Grand which introduces an appreciable delay, with a time
Coulee Third Powerplant are a prime example, control constant in the order of 16 seconds, in this large
characteristics to satisfy power system needs strongly component of their response. This yields somewhat
influence economics of the design. Parameters sluggish governing characteristics. These factors
fundamental to control characteristics, such as the combine to relegate the thermal sources, so far as
penstock water starting time and the mechanical inertia practicable, to the role of baseload operation with a
or flywheel effect, are subject to control by the minimum of system regulating activity. Consequently,
designer but if abnormal values are required the system needs are best served by exploiting the
increase is obtainable at appreciable incremental cost. potential controllability of hydropower. This potential
To aid the designers with a basis for the most is high insofar as the energy of stored water is available
economical combination of the parameters to satisfy with no more delay than that of opening the turbine
the requirements, an analysis of the requirements and gates and the associated time to accelerate the water
their interrelation was undertaken. In this approach the column.
starting point was a determination of the minimum or
most economical control characteristics which could be In more definitive terms, the controllability for such
accommodated by the power system into which the large hydropower units for daily peaking service should
units are to operate. Following this, the governing be such as to permit each unit to be loaded within 30
characteristics were studied to determine a range of minutes while on regulation. Bypassing of the
combinations of inertia and water starting time to yield governing dash pot or its equivalent, as has often been
the necessary control characteristics. done with smaller units to expedite response to the
control signal, cannot be accepted without degradation
At this point the scope of the problem expanded of system performance. It has been adequately
considerably. When the relations for flywheel effect demonstrated by system experience in the Northwest,
and water starting time were established, on the basis prior to 1964, that widespread bypassing of governor
of a conventional temporary droop type of governor, dash pots leads to instability of system speed
to meet the required overall control characteristics, the regulation. Analyses of the tieline oscillation problem
amounts of flywheel effect and water starting time subsequent to that time have also shown that bypassing
required were clearly excessive. Any workable of governor dash pots would contribute substantially to
combination of these parameters on this basis the lower frequency tieline oscillations. It is therefore
represented an excessive amount of added expense. important that large hydropower units, such as the
Because of the important influence such large units 600-mw units for the Coulee Third Powerplant which
would have on the power system the sacrifice of speed would strongly influence the power system, be capable
stability through bypassing of their governor dash pots of completely stable speed regulating capability under
to obtain workable responsiveness could not be any operating condition. This is particularly important
considered acceptable. Relief to this constraint was during the load pickup period of the day while
found in special refinement of governing characteristics generation schedules and loadings are changing rapidly.
to better suit the requirements in this problem. Since
this departed from conventional practice, the studies A review of the control characteristics of the wide
are recorded in this report. range of hydropower units of Bureau projects yielded
the criteria that reasonable response to system control
signals is represented by a response time constant of 30
SELECTION OF OVERALL seconds. Slow response is characterized by a response
CONTROL CHARACTERISTICS time constant of 50 seconds and a response time
constant of 75 seconds is very sluggish. For the Coulee
Several important influences are placing stricter Third Powerplant units the response should be no
requirements upon the controllability of large slower than that of the existing units, in other words
hydropower generating units. Principal among these is no more than 30 seconds. There is a corresponding
the changing role of hydropower from baseload type of need for sensitivity to speed deviations, rapidity of
operation to daily peaking service. This is a result of speed correction, and linearity of response maintained
integration of thermal and hydrosources through for small speed deviations. These characteristics will be
system interconnection. The large thermal sources discussed under "Refinement of Governor System" but
require appreciable time to change boiler output. they are sufficiently reflected in the gate response time
Efficiency is degraded by manipulation. The large constant for this to be a simple and useful criterion.
GOVERNOR TYPE AND desired performance an excessive departure of Twand
CONTROL CHARACTERISTICS T m from normal would be required.
Control characteristics of the conventional temporary A more promising governing system for securing the
droop-type governor may be obtained from the desired response is a "double derivative" governor.2
empirical formulae offered by Paynter1 , that is, for the This system in preceding investigations had yielded a
optimum response: reduction of response time to about one-fourth that of
the temporary droop type of governor. An
approximate expression of response obtainable with it
Tw Tw
temporary droop, [) = = 2.5- (1 ) would thus be
0 .4 T m Tm
and,
Tw T g ~ 67 -TW2
Tm
recovery time, Tr = (2)
0.17 = 5.9 Tw
where For the preliminary values of Tw = 2 and T m = 8, Tg
would be about 33 seconds. This was considered to
. . ~LV yield satisfactory prospect for accomplishing the
T w = water starting time =
~ desired response, T g = 30 seconds, by a minor
and adjustment of T w or T m either of which is within
practical range for the designer. A plot of this relation
. .. N2WR2 between Twand T m for three degrees of
T m = mec h amca I starting time =
1.6 x hp x 10 6 responsiveness is shown in Figure 1. Although intended
":' 2.6
[)Tr ~
(3) UJ"2.4
Tg~a ::e
f-
Substituting (1) and (2) into (3) ~ 22
f-
a:
<1
2.5 Tw 5.9 Tw 14.7 Tw2 ~ 2.0
Tg ~-x-~- - a:
UJ
Tm a a Tm f-
<1
~ I8
2 4 6 8 10 12 14 16 18
Supplying the usually employed numerical value of MECHANICAL STARTING TIME, Tm -SEC.
The temporary droop type of governor has been so Assessment of the current need for governing function
predominant for controlling the hydraul ic turbines of as such is not complete until it is considered that the
power generating units in this country for the past initial reduction of speed deviations which
several decades that it is used here as the basis for accompanies the interconnection of systems does not
comparison and it is referred to as the "conventional" directly alleviate the need for governing. It tends
governor. A few electrohydraulic governors of various mainly to transfer part of the evidences from local
configurations and with improved characteristics are speed deviations to remote tieline load deviations. The
now beginning to appear? However, they are still in imbalances of load and generation which were formerly
the minority and none so far have incorporated all of reflected entirely by speed deviations, after extension
the features this study has indicated to be both of system interconnections become reflected partly by
desirable and practical. speed deviations and partly by deviations of load on
the interconnections. The relatively wide deviations of
The temporary droop type of governor was a model of load on the longer system interconnections is now a
simplicity. It was well suited to the familiar phenomenon to system operators.4
mechanical-hydraulic means of accomplishing its
functions and it thereby established a predominantly Obviously the system control problem could be
satisfactory reputation of reliability which has served significantly relieved by not only (1) improving the
to excuse it from much critical study heretofore. governor's speed of response to the area load control
Simultaneously, the results of growth and signal as previously discussed, but by (2) improving the
interconnection of power systems have tended to governor's sensitivity and speed of response to the local
obscure the limitations of this type of governor in speed deviations so as to handle promptly and directly
keeping up with the power systems present needs. as much as possible of that component of the control
3
function, thus freeing capability of the area load The servomotor system is not in reality a perfect
control for better handling of its share of control. integrator as implied by the simplified representation
1IT 7S, It is in fact, besides an integrator, a cascade of
The respective shares of the control function between delays including mainly two for the valve system and
the speed governors and the area load control are one for the main servomotor itself. These incidental
established by the coordination of area frequency bias delays may have as much degrading influence on
and speed droop. Hence, there is no conflict in overall performance as the water starting time.
restoring the governor's sensitivity to speed deviations
to the same order as that of the area load control. An additional limitation of this form of governor is its
Instead, the coordination is improved. However, this inability to compensate for the incidental delays and
cooperative overlap in the hierarchy of system control nonl inearity of the servomotor system. With those
should be accompanied by separation of the response delays in a simple cascade, performance of a very large
times of the two layers of control. The governor should unit such as for Grand Coulee Third Powerplant, would
respond preferentially to speed deviation. Its response be exceedingly poor.
to the load control signal should not be so fast as to
induce a speed deviation or to supersede correction of Double Derivative Governor
speed deviation.
A superior form of governor developed in associated
Temporary Droop Governor studies, particularly to accomplish better system load
control without conflicting with stable speed control,
The temporary droop type of governor for hydraul ic is the "double derivative" governor shown in Figure 3.
turbine control is shown schematically in Figure 2. It This is a special form of proportional-integral-derivative
has the virtue of simplicity but that simplicity also governor. Its superior performance is evident in Figure
limits its capability. As revealed by its transfer 4 (C and D) compared with performance of the
function, also shown in Figure 2, its dash pot provides temporary droop governor shown in Figure 4 (A and
only one term (1 + T rS), to compensate for the lags in B).
the entire control loops, of which the water starting
time is only one.
Speed
N
1st Deriv.
Speed
S
N KI
Pilot Servomotor I+T2S
Speed Gate
ref.a t - I Signal
G1 2nd Deriv.
Load T7S
Signal ~
Dash ot I+T4S
8 Tr S
Pilot Servomotor
I tTr S Gate
I Signal
T7S GI
llGI ItTrS
Tr S2
=
II N CTt (T7 t CT Tr t 8 Tr) StT7 ~GI_I+(T2+T4tKdSt(T2T4tKIT4tK2)S2
llN - (ItT2S)(ltT4S)(CTtT7S)
Figure 2. Block diagram and transfer' function of
temporary droop governor for hydraulic turbines. (See Figure 3. Block diagram and transfer function of "double
glossary for symbols.) derivative" governor for hydraulic turbines.
4
5A and 58. In each case the speed deviation is lim ited
05 20 to about the same magnitude but other characteristics
are different.
fIN B
t;P In Figure 5A the servomotor integration time was quite
short, 1.5 seconds. With this short integration time
only modest amounts of th'e derivative terms are
necessary for stability. Response of the gate system is
00 10 20 SEC00 20 40 SEC
Iively with appreciable overtravel for a short period to
quickly accelerate the inertia back to normal speed.
0.5 20 Corresponding response of the gates to a load control
signal is shown by the bottom curve in Figure 5A to be
quite prompt, with a time constant of 30 seconds. This
M l'.G
l'.P lIT set of adjustments would be proper for normal
interconnected power system where activity for speed
control is modest but where good response to load
00 0 I
control is needed.
10 20 SEC. 0 40 SEC.
10+---
L
t75l'.G
z: l'.L
10
It is an interesting property of this and other
~
high-order derivative governors to be described later
that there is not just one optimum adjustment but a ~O0
<:)
20 40 SEC 0
o~ 20 40 SEe.
family of optimum adjustments. For each integration
time chosen for the pilot servomotor there is a A B
different optimum combination of derivative terms.
Over a wide range of these optimum adjustments,
Figure 5. Performance of derivative governor of Figure 3
control of the speed transient is essentially the same;
with short and long integration times.
however, movement of the turbine gates takes on
A. Short integration time: T7 = 1.5, K1 = 5.3, K2=7.4,
different characteristics especially the response to a
a=0.05, Tg=T7/a=30.
load control signal. It is this flexibility that makes this
B. Long integration time: T 7 = 4, K1 = 10, K2 = 15, a=
type of governor superior for reconciling the otherwise
0.05, Tg =T 7/a=80.
confl icting needs of (1) speed control without sacrifice
of stability, and (2) rapid response to system load
control without conflict with speed control. For In Figure 58 the servomotor integration time is much
illustration, performance of the governor with two longer, 4 seconds. With this long integration time the
widely differing sets of adjustment is shown in Figures derivatives must be much stronger to make the
5
governor active enough for good stability. Response of governor, but response time for this path may be
the gate system is conservative. Note that overtravel of modest. For the derivative paths, a fast response signal
the gates is quite modest but this overtravel is sustained is needed but since any long-term drift is not
for a longer period while the inertia is gradually transmitted through these stages the drift
accelerated back to normal speed. The magnitude of characteristics of this sensor are not critical.
speed deviation is not greater but the recovery takes
longer. This set of adjustments was chosen to illustrate Finally, the derivative configuration allows freedom to
conservation of servomotor energy or activity. It would insert the load control signal after the derivative stages.
be applicable on a widely fluctuating load where there This retains for the speed control channel the
might be a need to minimize burden on the governor compensation specifically tailored for best speed
oil supply and where load control would not be control but does not impose this compensation on the
needed. As shown by the bottom curve in Figure 58, load control which should not be allowed to conflict.
response to a load control signal would be exceedingly
sluggish with a time constant of 80 seconds. Few Without this feature of detouring the governor
applications exist anymore which would be benefited compensation with the load control signal, conflict
by the characteristics in Figure 58 but the comparison would come from the practice of proportioning the
is revealing. load control function by breaking up this signal into a
series of intermittent, arbitrarily timed pulses to the
Advantages of this double derivative governor stem governor speed level mechanisms. Worse still, it has
both from its higher order of compensation and from been common practice to employ one such controller
its particular form. The first derivative of speed is to generate control pulses for all units in a plant and
sufficient to accomplish the equivalent function of the sometimes more. Sampling rates for generation of the
temporary droop feedback in the more conventional control pulses may range from 4 seconds to 5
governor, that is, each produces a single lead term of minutes.s Augmentation of the fronts of these
the form, (1 + TS), in the numerator of the transfer arbitrarily timed pulses by the speed control
function. This operates to compensate a dominant lag compensation not only accomplishes nothing beneficial
in the control loop, primarily that of the penstock for system control but actually contributes to random
water column. A second derivative of speed used in this swings or system "noise." This influence is easily
configuration provides an additional lead term for relieved by inserting the load control signal after the
compensating other lag in the control loop from the speed control compensation. This achieves smoother
servosystem. but nevertheless rapid control.
The tandem configuration for combining the speed Equivalent High-Order Governors
signal, its first derivative and its second yields
additional advantages, both technical and practical. Extension of the study to identify some other governor
This combination has flexibility such that the configurations capable of the desired performance
compensating terms may be complex, that is, the yielded the configuration shown in Figure 6. This
quadratic expression in the numerator of the transfer governor employs a cascade of lead-lag stages for the
function may have complex roots. The significance of compensation instead of the tandem path method of
this is that still faster rise time of the control signal is Figure 3. Three lead-lag stages are necessary to
possible so that performance can be approximately accomplish performance equivalent to the tandem path
equivalent to third order compensation, but without method of Figure 3 but technical performance is good
the extra circuitry. and considerable flexibility of adjustment is available.
This configuration would be confined to only one
A practical advantage of this tandem configuration is speed signal input for all functions but whether that
that the derivative paths do not transmit any constitutes a limitation depends upon the quality of
fundamental component. Therefore any drift of speed sensor employed. The cascaded stages would
operational amplifiers used to perform these functions tend to amplify any drift problem but this should be
does not accumulate in a cascade product. Hence, drift regarded merely as imposing stricter drift specifications
is minimized. within the individual stages.
Another peculiar flexiblity of this tandem arrangement While it might appear that only two lead-lag stages
is the ability to use separate speed sensors for the should be equivalent to the tandem path method with
fundamental and for the derivative paths, if desired, to
first and second derivatives, that is not the case. There
better utilize characteristics of the respective sensors.
For the fundamental path, stability of the sensor is is no cross coupling between successive lead-lag stages
important as this controls the long-term stability of the
6
Cascade Compensator
Speed
(I+T,5)(I+T35)(1 +T55) 5 eed N
( I + T2 5) ( I + T4 5){ I + T 65) N
Speed Gate
Pilot 5ervomotor Gate ref. a +
-
Signal
I 5i nal Lood G1
GI Signal -
T75
Droop
(J
to generate the more rapid rise time reflected by feedback loop. The significance of this configuration is
complex roots in the tandem compensation. Only that it represents upgrading of the temporary
terms with real roots are produced by the lead-lag droop-type governor system, with two additional
stages, consequently one more stage is required to lead-lag stages of compensation, to accomplish
generate equivalent performance in the signal shaping performance equivalent to the other high-order
(electronic) circuitry. governor systems described. A temporary droop
governor improved with one lead-lag function has been
The computer simulations show that by permitting described by Schiott.6 Although the block diagram of
underdamping of the succeeding servocontrol loop, Figure 7 shows the third compensating term
seemingly equivalent performance would be obtained accomplished with a lead-lag, it could also be
with only two lead-lag stages. However, this should not accomplished with the temporary droop feedback.
be accepted as truly equivalent. The servocontrol Both methods reduce to the same mathematical
system actually contains many nonlinearities such as expression.
valve ports, friction in the gate mechanism, and
compressibility of the unavoidably somewhat aerated This configuration somewhat reduces freedom to
servomotor oil system. It is subject to variables such as bypass the load control signal around the
the oil pressure and hydraulic thrust upon the turbine com pensation designed for speed control. Th is
gates at different gate openings. Thus, considering that confinement is minor if the last stage is used for one of
the servocontrol loop is a high energy loop of massive the shorter time constants of the compensation. When
moving parts with many nonlinearities and variables, it this is done performance is essentially the same as
is clearly preferable that this loop of the overall control Figure 4 (C and 0) and it is therefore not shown
system be kept well damped and the required rise time separately.
be generated in the compensating network where
computer-grade electronic circuitry can be employed. A parallel configuration of proportional-integral-
derivative governor, sometimes referred to as an
Another governor configuration capable of industrial controller, is shown in Figure 8. This
performance equivalent to that of the tandem or configuration of governor can yield the same transfer
"double derivative" system is shown in Figure 7. This function as the "double derivative" governor of Figure
is functionally a minor variation of Figure 6 in which 3 with certain modifications. For best performance the
one of the compensating terms is enclosed along with speed droop feedback should be returned to the
the pilot servomotor within the permanent droop integral stage only as shown by the sol id Iine rather
7
Ref
+ 5peed
Droop feedback
A
N to ref. junctiun
Proportional
~ ..,../-
\~ Droop feedback
l'.N
\ /( to integral stage
l'.P
/ ,
2
Load signal to B
LlG' 1+ {T4 + K, T7)5 + (K1 T4 T7 + K2 T7) 52 ref. junction
LlN = (I + T4 5) (cr + T7 5 + T2 T7 52) I
I
\
Figure 8. Parallel configuration of \
pro po rtional-integral-derivative governor. Conventional l'.G
\
arrangement is shown by dotted lines. Improved l'.L \ Load signal to
arrangement is shown by solid lines.
/
/- integral stage
than summed with the speed signal for all three terms
as shown by the dotted Iine. The difference in ,
response to the load control signals through the proportional-integral-derivative governor shown in Figure 8
proportional and derivative terms whose function is with conventional and improved arrangement.
correct only for speed control. The difference in A. Response of speed to load change.
performance is shown in Figure 9B. B. Response of gates to load control signal.
8
Refinement of Servomotor Control of 0.2 second. Extrapolating from a servomotor lag
term of 0.12 second measured on the existing 108-mw
The best servocontrol heretofore ava ilable has been Coulee units, a servomotor lag time constant in the
accomplished by enclosing the pilot valve, main valve, order of 0.5 second for the 600-mw units might be
and servomotor in a closed loop as shown in Figure 10 expected. With these values a simple feedback loop as
shown in Figure 11A would be restricted to 0.75 per
unit gain and response would require over 2 seconds.
G
A
W
I-
(f)
Valve s and Servomotor
Gate
K
Gate ::J I
Signa I Position
B
G1 ItTvl S)(I+Tv2S)(ltTs S)S G ~ G
(f)G' B
~ 0
Position feed back :!:
w
Valves and Servomotor I-
Gate
i(tJ
K Posit ion
c I tTvlS)(ltT V2S)(I+ Ts S) S G
G
Damping terms
u..
0
(f)
Figure 10. 8ervosystem representation. I.LJ
dr
(f)
A. Component functions. z:
0
B. Combined functions of components. a..
(f) D
C. Closed loop. I.LJ
a:: I I I
00 2 4 6 SEC.
and driving that loop by a pilot servomotor in which
the governor's integration function is performed with Figure 11. Responses of closed loop servosystem of Figure
higher accuracy. This simple feedback loop around the 10C to step signals within proportional range of valves, T
; 0.2, T 2; 0.2, Ts; 0.5. v 1
main servomotor system is quite hel pful but v
performance is still limited because of the presence of A. Gain K ; 0.75, no damping.
three time constants in the loop which requires that B. Gain K ; 3.3, no damping.
gain be kept low to preserve stability. Because of the C. Gain K ; 3.3, velocity damping; 0.578.
low gain, linearization and the reduction of response D. Gain K ; 10, velocity and acceleration damping;
time by this technique are still modest. 0.418 + 0.08582((1 + 0.18).
If the gain is increased to 3.3 per unit, the response
Exploration of additional measures to improve becomes underdamped and approaching instability as
response has included valve ports, oil viscosity, barriers shown in Figure 11 B. By the addition of velocity
to control oil aeration, and phase correction or feedback the response becomes su itably damped, as
damping terms in the feedback loop. The latter showed shown in Figure 11 C. Still further improvement of
such promise as to supersede all the other relief response and linearity would become possible through
measures considered. It permits increasing the servo the addition of some acceleration feedback to the
loop gain by a factor approaching 10 with good damping signal. Gain could then be increased to 10 for
stability. This accomplishes both linearization and the response shown in Figure 11 D. Ideal stability at
reduction of response time. high gain would result ,from the combined use of
velocity and acceleration to supplement the
For units as large as those for Grand Coulee Third displacement feedback, but some practical aspects
Powerplant the valve delays might be held to the order warrant consideration.
9
It would be essential that the hydraulically operated
main servosystem, the power "muscle" part of the 0.5
governor, be sufficiently invulnerable to failure of any
electrical or electronic component or power supply as
to remain safely controllable. But mechanical-hydraulic
means of generating a velocity signal are somewhat a.. .25 A
awkward, and for generation of an acceleration signal ~ /-"""'"
are still more difficult. The most accurate and :z (
convenient means of generating these signals is <J
electrical or electronic, but these signals, from or
through a separate power source, could conceivably be
lost while the main hydraulic power source remains. If I--
Z
this were to happen, some degradation of performance w.25 B
could be accepted provided safe control could never be (/)
lost. z
<t
a:::
Thus, unless unusual independence or fail safe I--
interlocking could be devised, it would be prudent to a
confine gain in the main servo loop to a value just w
w
below the point of instability without the stabilizing a..
(/)
terms of the feedback. Then if the stabilizing terms .25 c
were accidentally lost, servomotor response might
become underdamped as in Figure 11 B but oscillation
would subside rather than increase.
10
:)
N
a.. ,
~ in; '
,n, ::>I
WC\J
f1n :
. .
0 .
~n n n A ~1
A
~:WJvvvv~ ~Tj
!
T
Figure 13. Field test of closed loop servosystem on Unit
G-4 in existing Grand Coulee Powerplant. Response to 0.02
per unit step signal with loop gain of 7. Figure 14. Field test of governor systems on Unit G-4 in
existing Grand Coulee Powerplant. Speed transient response
A. Position feedback only, no damping.
B. With velocity feedback damping. to 0.03 per unit load increment by simulated isolation.
A. Mechanical temporary droop-type governor.
B. Electronic derivative governor of Figure 3 with
closed loop servo system response as in Figure 13B.
difference in usable gain is because the time constants
in the existing units are not as long as those expected
in future larger units. When velocity feedback damping stable response to the small deviations of system
is added, the response becomes nearly ideal as shown in frequency and to system load control signals in normal
Figure 138. The result compares well with the study of operation is only part of the governing requirement. Of
Figure 11 C. While the improvement in response time is considerable importance also is the performance under
evident, the reduction of nonlinearity can be inferred the extreme conditions of startup, and various degrees
from the reciprocal of gain improvement; that is, to of load rejection ranging up to rejection of full load.
one-seventh. Performance under these conditions can be classified as
large signal performance. It is characterized by certain
Performance of the conventional mechanical unavoidable nonlinearities such as gate limits and valve
temporary droop type of governor is shown in the field
limits. It was considered appropriate to examine
test result of Figure 14A while Figure 148 shows the performance under these conditions to establish
field test performance of the more refined type of dynamic ranges required for the respective signals and
governor of Figure 3 and with a high gain, damped to choose simple but adequate auxiliary control
servo loop with characteristics shown in Figure 138. features such as for automatic startup.
These field test results compare well with the study
results of Figure 4. A minor difference is that the Startu p
mechanical temporary droop governor performance
shown in Figure 14A does not have benefit of a Well controlled startup performance is especially
closed-loop servosystem, while benefit of that feature important for peaking plants where the units may be
is represented in the study shown in Figure 4A. With started and stopped daily. Computed performance
allowance for that difference the field test results are during startup in which the gates are allowed to open
considered a close verification of the improved to 40 percent is shown in Figure 13. This was
performance to be expected from the refinements computed for a derivative governor of the type shown
indicated by these studies. in Figure 3, servocontrol system of Figure 11 C, and
basic parameters as for Figure 4 (C and D).
Performance is seen to be quite satisfactory without
SEVERE CONDITION PERFORMANCE auxiliary control, overshoot of speed being only 2
percent and reverting to stable speed for synchronizing
The conditions under which a governor must function in about 50 seconds.
encompass a rather wide range. Rapid, linear, and
11
rapidly. The first derivative of speed becomes slightly
1.0 IT 1 negative for a few seconds after maximum speed
Z
0
2 % over shoot corresponding to the slight downward slope of the
I-
(f) .8 Speed
I speed curve until rated value is reached.
0
a... The second derivative signal corresponds to curvature
I- of the speed trace and it shows a modest positive value
w -
I-Z .6 at breakaway as the speed bEmds upward. A modest
<:[=>
<..:)Q:: negative value is displayed between 32 and 40 seconds
0 w
za... as the speed bends downward. For each derivative
<:[
.4 signal a dynamic range of 0.2 per unit of the rated
0 speed signal is more than enough for this set of
w
w conditions.
a...
(f) .2
A more cautious startup operation is shown in Figure
16 where the gates are allowed to open only to 20
0 percent of full opening. As would be expected,
w ~ +.16 acceleration is more gradual and the overshoot of speed
>z of 1 percent is half that developed when starting with
f==> 40 percent of full gate opening.
<:[Q:: 0
2:w
~01 a... +12
10
o~ 0
~ ~
a...<..:)
-08 z
0
(f)
(-f) f- .8
(f)
0 20 40 60 80 a
a...
TIME - SECONDS w-
f-
.6
f-Z
<:[=>
Figure 15. Startup with gate limit at 0.4 per unit. <..:)Q::
ow
za... 4 H-1
<:[
The startup in this study was initiated by allowing the 0
I
gates to open at their maximum rate until held by the w
w .2
gate limit at 40 percent of full opening. The maximum a...
(f)
rate of gate movement is 8 seconds for full travel or 3.2
seconds for travel to the 40 percent position. The rate
is limited by the valve stops. After breakaway, speed 0
w ~ +16
rises smoothly until at 84 percent normal speed the >2:
speed signal plus acceleration terms start the gates f==>
<:[~
rapidly toward the closed position. Finally, the gates 0
:=:w
reopen to about 8 percent to maintain normal speed. ~a... +.12
01
(f)
of speed
This computed sequence allows a number of interesting 0--1 0
observations. A speed switch set at 90 percent of ~a...<..:)~ +08
normal speed could safely be used to release the 40 (f)(i)
percent gate limit used for the starting sequence. At
0 20 40 60 80 100
that point the gates are already under control of the
governor speed terms. The first derivative of speed, TIME-SECONDS
which is acceleration, becomes uniform at about 0.16
per unit soon after breakaway and remains uniform
Figure 16. Startup with gate limit at 0.2 per unit.
corresponding to the uniform slope of the speed curve
until at 32 seconds the sum of speed (0.84 per unit)
and its first derivative (0.16 per unit) exceeds the From a governing standpoint, this cautious startup
reference (1.0 per unit) and the gates start closing procedure may be seen to yield more constraints than
12
benefits. The time for speed to become stable at set I.
value is 100 seconds or twice as long as the time t-
required when starting with 40 percent gate, yet the z
reduction of overshoot is of small consequence since a ~1.4
et:
larger value could easily be tolerated. Furthermore, the lLJ
a...
speed signals do not assume control of the gates until
the speed reaches 94 percent of normal. If the 20 ~
lLJ
1.2
percent gate limit for starting is to be released by a a...
(f)
speed switch, its setting must not be less than 95
percent of normal, a more exacting setting than that t-
Z
required when starting with 40 percent of full gate ~
opening. et:
~ .4 .8
(f)
From these computations it is evident that governing -I t-
<[ -
needs impose no restriction against the more rapid z Z
19.2 ~.6 1st Derivative of speed
starting proced'ure. Hence, final choice of procedure (f)
can be made on the basis of mechanical, structural, or
et:
lLJ i I I
hydraulic conditions.
lLJ
>
a... - 2nd Derivative of speed
~0~.4
> <[
A few comment~ concerning parameters represented in a: 19
lLJ
these computations are in order here. The governor a -.2 .2
characteristics, water hammer, and mechanical inertia a
lLJ Gate
were all adequately represented in these runs by analog lLJ
a...
computer. The turbine torque versus speed (f) 0
chara{;teristic was represented as constant whereas in 0 20 40 60 80 roo
reality the torque of a Francis-type turbine is higher at
reduced speed. More rigorous representation would TIME - SECONDS
show the speed to rise a little more steeply at first,
diminishing to the slopes shown near rated speed. Figure 17. Computed dynamic performance for rejection
However, since the correct value is represented for
of rated load.
rated speed, the dynamic behavior near rated speed is
accurate and this is the region of prinicpal interest
here.
13
recovery is shown to be smooth. The gates reopen to a
~~
~1 1
position to maintain speed at no-load and the speed
presently becomes stable at the set value.
~
(J)
10
.
2hJ -~
,
II
I
_-: I
.
I
m
is 8 l I
A partial load rejection can be considered a more S; I '
severe condition from a system load standpoint than a 0Q . 5-- ~
0:: W
Figure 18A shows performance to be expected W
Q >
-J
following rejection of the maximum amount of load <!
>
>0
for which the governor performance is reasonably 0::
linear. This is a rejection of 30 percent of rated load. It W => +1
0 Q
14
adjustments which yield best performance under
loaded conditions would yield performance which
could be appreciably improved upon for synchronizing.
A
Performance computed for the Coulee Third
Power plant units under typical synchronizing
conditions is shown in Figure 19. With governor 0
parameters set for loaded operation, the response of w
w B
speed to a speed change signal is shown by Curve A to CL
be quite stable but to require about 30 seconds to <f)
reach the final value. Appreciably better performance w
for sychronizing can be achieved by the readjustments >
yielding Curve B which reaches a steady value in 9 I-
<I:
c
seconds. Still more rapid arrival at ultimate speed is -.J
potentially possible with the more extensive W
readjustments yielding Curve C which arrives at the a:
final steady value in 5 seconds. For comparison, the
adjustments for loaded operation and for better
synchronizing, Curves Band C are tabulated following.
T7
K1 K2 pilot servo
1st deriv. 2nd deriv. integra- 0 10 20 30 40
coefficient coefficient tion time
TIME - SECONDS
Loaded oper-
ation (A) 5.2 7.4 1.5 sec- Figure 19. Response of speed to speed change signal under
onds conditions for synchronizing.
Synchroniz- A. With governor adjustments for loaded operation,
ing (B) 3.5 1.5 0.5 K1 = 5.2, K2 = 7.4, T7 = 1.5 sec.
Synchroniz- B. With governor parameters readjusted for
ing (C) 2.5 1.0 0.25 synchronizing, K1 = 3.5, K2 = 1.5, T 7 = 0.5 sec.
C. With governor parameters readjusted for most rapid
In practice, some influence from gate friction and synchronizing, K1 = 2.5, K2 = 1.0, T 7 = 0.25 sec.
minute amounts of backlash in the gate mechanism
may somewhat limit ability to accomplish ideal control
at the maximum rate. Adjustments in the vicinity of
Curve B are considered to be reasonable from a 3. Governing of large hydraulic turbines can be
practical standpoint but the potentially useful range of substantially improved by the addition of damping to
adjustment is nevertheless indicated by the tabulation. the servomotor control loop so that its gain can be
increased for faster response and better linearity.
CONCLUSIONS
APPLICABILITY
1. Refinements of governing systems detailed herein
will benefit speed control, area load control, and the Results of this investigation are considered applicable
coordination of these two functions. to hydrogenerating installations in general. Although
the study was undertaken for benefit of the larger
2. The refined governing systems can accomplish the installations, expense of incorporating the refinements
desired control with less confinement from the basic would be small.
parameters of water starting time and mechanical
inertia that would be imposed by the heretofore
conventional governing systems. Some guide for
proportioning of parameters has been offered.
15
REFERENCES GLOSSAR Y OF SYMBOLS
16
For speed no load Tw = 1/10 normal = .183 For velocity damping only
*
I
18
.560
9
KI L
FS3
A
PI LOT SERVO SERVO LOOP TURBINE
I
1/10
~
-.J .26P L
8 A
~ + 10
A4~ - - - FSI
1/10 R
AI8 TEMP. DROOP GOv.
I
110
TIME BASE
PLOTTER X L
-10 IV/IN=IOSEC/!N AN AI7~PLOTTER Y
2V/1 N=20 SEC/IN AG A37--o FS4 .5V/IN
R
6N
I
Ito
.5
/2 o-X 10
10 +/0
P43 A4 --@
(A36CXJ)
~'For velocity damping only
+10 N
Servo loop and turbine
(See Figure A , Appendix I,
....
Add IC to A37)
CD
0
Pilot servo position R
Full gate = -IOV
Zero gate = OV
Max. gate velocity
Normal speed = OV
:t1.25V= 8 sec. travel
Zero speed = -IOV
The following conversion fa::tors adopted by the Bureau of Reclamatiol1 are those published by the American Society for
Testing and Materials (ASTM Metric Practice Guide, E 380-68) except that additional factors (*) commonly used in
the Bureau have been added. Further discussion :>f definitions of quantities and units is given in the ASTM Metric
Pra::tice Guide.
The metric units and eonversion factors adopted by the ASTM are based on the "Intern'ltion'll 3-jstem of Units" (design'lted
SI for Systeme Intern'ltional d'Unites), fixed by the International Committee for Weights and Measures; this system is
also known as the Giorgi or MKSA (meter-kilogram (mass)-second-ampere) system. This system has been 'ldopted by
the Intern'ltional Organization for Standardization in ISO Recommendation R-31.
The metric technical unit of force is the kilogram~force; this is the force which, when applied to a body h'lving a
mass of 1 kg, gives it an acceleration of 9.80665 m/sec/sec, the standard acceleration of free fall toward the e'lrth's
center for sea level at 45 deg latitude. The metric unit of force in SI units is the nei.\Tton (N), which is defined as
that force which, .when applied to a bo&y having a mass of 1 kg, gives it an acceleration of 1 m/sec/sec. These units
must be distinguished from the (inconstant) local weight of a body having 'l mass of 1 kg; that is, the weight of a
body is that force with which a body is attracted to the earth and is equal to the mass of a body multiplied by the
acceleration due to gravity. However, because it is general practice to use "pound" rather th'ln the technically
correct term "pound-force, the term "kilogram" (or derived mass unit) has been used in this guide instead of "kilogram-
"
force" in expressing the conversion factors for forces. The newton unit of force will find increasing use, and is
essential jn SI units.
Where approximate or n:>minal English units 'lre used to express a value or range of values, the comrerted metric units
in parentheses are also approximate or nominal. Where precise English units are as ed, the converted metric units
are expressed as equally significant values.
Table I
Multiply By To obtain
LENGTH
ABSTRACT .
In considering appropriate parameters for large hydropower generating units, the choice of
control characteristics
design. Parameters
to satisfy power system needs strongly influences the economics of
fundamental to control characteristics which can be determined by the
. In con
control
design.
designers are mechanical inertia or flywheel effect and penstock time constant. To aid the
designer with a basis for the most economical combinations of parameters to satisfy the power . designE
designe
system needs, a study analyzing the requirements and their interrelation was made. Results of
system
the study are given. The influence of the governing system and refinements with several new ... the stu
....
features for bette"r speed control were investigated toward improved coordination between unit
feature
speed control and area load control. A guide for proportioning of parameters is proposed.
speed c
. . . . . .' . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
:
ABSTRACT
In considering appropriate parameters for large hydropower generating units, the choice of
control characteristics to satisfy power system needs strongly influences the economics of
design. Parameters fundamental to control characteristics which can be determined by the
designers are mechanical inertia or flywheel effect and penstock time constant. To aid the
designer with a basis for the most economical combinations of parameters to satisfy the power
system needs, a study analyzing the requirements and their interrelation was made. Results of
the study are given. The influence of the governing system and refinements with several new
features for better speed control were investigated toward improved coordination between unit
speed control and area load control. A guide for proportioning of parameters is proposed.