167
Working Group on Prime Mover and Energy Supply Models for System Dynamic Performance Studies
ABSTRA~
A working group on Prime Mover and Energy Supply Models for System Dynamic Performance Studies underthe IEEE System Dynamic Performance Subcommittee was established in 1987 to collect technical informationon dynamic characteristicsof prime mover and energy supply systems that can affect power system performance durlng and following disturbances such as faults, loss of generation or loads and system separation. A principal objective of the working group is to develop prime mover/energy supply models for use in power system simulation programs. The last endeavor in this area was the I E E E Committee Report (Ref. 1) in 1973 entitled "Dynamic Models for Steam and Hydro Turbines'.
KEY W O R D S
Other developments requiring new models include the widespread use of electrichydraulic speed control both in new construction and in modernization of older power plants. It is better to use models describing the actual equipment rather than make approximations to fit existing mechanical governor models. The tremendous increase in computer power eliminates the need for less detailed models. This report recommends hydraulic models suitable for a relatively wide range of studies. The two main sections of the report provide models for 1) prime movers including water supply conduit and 2) prime mover speed controls. The section on prime mover models includes both linear and nonlinear controls. Nonlinear models are required where speed and power changes are large, such as in islanding. load rejection. and system restoration studies. The block diagram of Figure 1 shows the basic elements of a hydro turbine within the power system environment. Excluded from the scope of this report are models for aeneration load control and electrical load dynamics.
Hydraulic Turbine Dynamic Models, HydreTurbine Controls. Power System Dynamic Petformonce.
1 .O INTRODUCTION
mechanlcal torque
The 1973 IEEE Committee report 'Dynamic Models for Steam and Hydro Turbines in Power System Studies' (1) has been used widely. Even in 1973. however, it was realized that more work needed to be done. Since then, modeling requirements have increased greatly and more detailed models have been implemented in the advanced simulation programs. The older models were considered adequate for typical first swing stability simulations common in the early 70s. Nowadays, models for the following types of studies are also needed: Longer transient stability program simulation up to and beyond ten seconds are now routine. In this longer time frame, prime mover and prime mover action can affect results. Low frequency oscillations.
Dynamics
Aaaigned unlt generation
Dynamics
Rotor Dynamic8
electrical torque
preaaure, h
flow. q
Automatic Oeneratlon
Load Dynamlcs
Figure 1. Islanding and isolated system operation. System restoration following a breakup. Load rejection. Load acceptance.
2.1
Functional Block Diagram Showing Relationship of Hydro Prime Mover System and Controls to Complete System
2.0 MODELING OF TURBINE CONDUIT DYNAMICS
Wcterhammer dynamics. Pump storage generation with complex hydraulic structures. Paper preparation was coordinated by F. P. de Mello (Chairman) and R. J. Koessler with contributionsfrom J. Agee. P. M. Anderson, J. H. Doudna, J. H. Fish 111, P. A. L. Hamm, P. Kundur. D. C. Lee, G. J. Rogers and C. Taylor.
91 4622 K:RS A p a p e r recoinnencied and a!l:xoveti by t h e I% .E Power Systeil I,ngineerin: Committee of Tower 2n:;ineering i o c i e t I f o r 2 r s s e n t a t i o n a t t h e IKEkl/P?S 1371 Su.xier : . e e t i n g , S a l Die?o, C a i i f o r n i a , J u l y 26  Auzus?. 1, l.;sl. .Ixn.iscri?t c u b i i i t t e d Xebruarv 1. 1941 innde a v a i l a b l e l o r p r i n t i n 1 June 19, 1971.
" I
The block diagram in Figure 2 represents the dynamic characteristics of a simple hydraulic turbine, with a pehstock, unrestricted head and tail race, and with either a very large or no surge tank. The penstock is modeled assuming an Incompressiblefluid and a rigid conduit of length Land crosssectionA. Penstock head losses hl are proportionalto flow squared and f is the head loss coefficient, P usually ignored. From the laws of momentum, the rate of change of flow in the conduit is
_ dq  (6 F i  4 ) g A/L
dt
where:
(1)
UYI
168
turbine flow rate, m 3 ,, / 2 Penstock area, m Penstock length, m 2 is the acceleration due to gravity, m/sec
is the static head of water column, m
is the head at the turbine admission, m
produces mechanical power. There is also a speed deviation damping effect which is a function of gate opening (6). Per unit turbine power, Pm, on generator MVA base is thus expressed as:
where qnl is the per unit noload flow, accounting for turbine fixed power losses. A+ is a proportionality factor and i s assumed constant. It is calculated using turbine M W rating and generator MVA base.
'
(6)
Figure 2. NonLinear Model of Turbine  NonElastic Water Column Expressed in per unit this relation becomes
where hr is the per unit head at the turbine at rated flow and qr is the per unit flow at rated load. it should be noted that the per unit gate would generally be less than unity at rated load. , defined by Equation 6 converts the gate The parameter A opening to per unit turdine power on the voltampere base of the generator and takes into account the turbine gain. It should be noted. however, that in some stability programs, A+ is used to convert the actual gate position to the effective gate position, i.e. A, = 1 /(G  G ) as described in (6). A separate factor is then used to convetthe "p',wer from the turbine rated power base to that of the generator voltampere base.
2.2 Linear Models
Neglecting friction losses in the penstock, a small perturbation analysis of the relationships in Figure 2 yields the black diagram of Figure 3. where h and hl are the head at the turbine, and head loss respectively in per unit, with hbase defined as the static head of the water column above the turbine. Tw, called water time constant or water starting time, is defined as:
qbose is chosen as the turbine flow rate with gates fully open (Gate posltion G = 1) and head at the turbine equal to hbase. It should be noted that the choice of base quantities is arbitrary. The system of base quantities defined above has the following advantages: Base head (hbase) is easily identied as the total available static head (i.e. lake head minus tailrace head). Base gate is easily understood as the maximum gate opening. Having established base head and base gate position, the turbine characteristics define base flow through the relationship: q = f (gate. head) The per unit flow rate through the turbine is given by:
7AW
Figure 3. Linearized Model of Turbine  NonElastic Water Column From this figure, the change in mechanical power output can be expressed as:
AP,,,
where Go
Ti
4(1  T,s)AG
(1
+
TA
 DG0Ao
(7)
= =
=
=
per unit gate opening qt operating point (qoqn$TW G0Tw/2 per unit steady state flow rate at operating point
In an ideal turbine, mechanical power is equal to flow times head with appropriate conversion factors. The fact that the turbine is not 100% efficient is taken into account by subtracting the no load flow from the actual flow giving the difference as the effective flow which, multiplied by head,
T2
90
Note that Go = qo
169
With the damping term neglected, equation 7 i s similar to the commonly used classical penstock/turbinelinear transfer function
=
(10)
AP,
AG
 GoTwS XA, G T s 1 + A
1
c
(8)
T,=L/a
where GOTw is an approximationto the effective water starting time for small perturbations around the operating point. Other, more elaborate linear models have been proposed (2.3A.5). They require more detailed turbine data. Linear models are useful for studies of control system tuning using linear analysis tools (frequencyresponse, eigenvalueetc.). Their use in time domain simulations should be discouraged since in addition to being limitedto small perturbations, they do not offer any computational simplicity relative to the nonlinear model. where:
(1 1)
PK
f
E
= =
=
= = = =
'base hbase = 9 =
pg (1/K + D/fE) density of water bulk modulus of water internal penstock diameter wall thickness of penstock Young's modulus of pipe wall material base flow base head acceleration due to gravity length of penstock wave velocity
L
a
=E
Tw =
q d
4umag
rL
T ,
= ZoTe
Figure 4.
This block diagram incorporates the traveling wave transfer function between head and flow rate:
Typical values for wave velocity are in the range of 1000 to 1200 m/sec. In the block diagram of Figure 4 the effect of friction head loss in the penstock is shown proportional to flow squared. An alternative numericalmethodof time simulation of traveling wave effects is the method of characteristics solution, detailed in reference 9. An example of this solution technique i s given in Section 4.4. The dynamics of turbine power are an almost instantaneous function of head across the turbine and gate or nonle opening including deflector effects where applicable. The head across the turbine is a function of the hydraulic characteristics upstream of the turbine and also downstream in cases where the flow in the draft tube and/or tailrace i s constrained as in the case of Francis or Kaplan Turbines. In the case of Pelton (impulse) turbines, the downstream pressure i s atmospheric hence the hydraulic effects are only from the conduits between the reservoir and turbine. This modulor separation of effects is shown in Figure 1 with the distinct blocks labeled "Turbine Dynamics" and "Conduit Dynamics". The model of the combined system. turbine and conduit. is shown in Figures 2 and 3 for the simple penstock/turbine system without elasticity effects and in Figure 4 considering elasticity effects. Particular hydraulic conduit arrangements may require special modeling in cases such as constrained or vented tunnels. individual penstocks fanning out from a common pressure shaft etc. The basic models for conduits can be put together to describe the specific arrangement, much as the basic models of electric components are used to describe specific networks. Examples of models for more complex hydraulic systems are given in Sections 2.4 to 2.6.
2.4
also written as
Z,tan h(Tes)
(9b)
In hydro plants with long supplyconduits.it is common practice to use a surge tank. The purpose of the surge tank is to provide some
170
hydraulic isolationof the turbine from the head deviations generated by transients in the conduit. Many surge tanks also include an orifice which dissipates the energy of hydraulic oscillations and produces damping. The hydraulic model shown in Figure 5 includes representationof penstock dynamics surge chamber dynamics tunnel dynamics penstock, tunnel and surge chamber orifice loses
I I
css
q"L
AW
Head at turbine(\
d
A
I
I '
Figure 6.
2.6
NonLinear Model of Turbine With Surge Tank Effects and Traveling Wave Effects in Penstock
Nonlinear Model d Multiple Penstocks and Turbines Supplied from Common Tunnel. Inelastic Water Columns
I
Surge Tank Level (head)
Figure 5.
NonLinear Model of Turbine Including Surge Tank Effects  NonElastic Water Column
Flow base, head base and water time constants are determined as in 2.1. C,, storage constant of surge tank, is defined as:
Figure 7 shows a configurationwhere a pressure shaft or tunnel brings water to a manifold from which penstocks fan out to several turbines. The coupling effect of head variations at the manifold is illustrated in the model of Figure 8 for the case of three turbines and Tw2 and Tw3 their penstocks with water starting times of Twl, respectively and a tunnel water starting time of Tw. The model of Figure 8 is derived from the basic momentum equations for each conduit and eliminating the variable head at the manifold through use of the continuity equation forcing the flow in the upper tunnel to be equal to the sum of the flows in the penstocks.
Penstocks
Tunnel
Turbines
c,
where
= As*hme
~
secs
2
(14)
%lSE
Upper and lower penstock head loses are proportionalto flow rate squared through loss coefficients f p l and Head I o S S e S in the orifice'to the surge tan are proportional to the coefficient fo times flow rate times absolute value of flow rate to maintain direction of head loss. The same applies to head loss in the upper penstock where flow can reverse. The head across the lower penstock is defined by the level of the surge tank, which can undergo low frequency oscillations (in the order of .01 Hz) between surge tank and reservoir. The inclusion of surge tank effects is warranted in cases where dynamic performance is being simulated over many seconds to minutes.
Figure 7.
01
Q2
2.5
NonLinear Model Including Surae Tank Effects. Elastic Water Column in Penstock
G3The equivalent relf and mutual rtarting timer, T1, , T12 , T13 , etc. are derived from the solution of h 1 , h 2 and h 3 as function of ql , Q2, 43 They are basically the terms in the inverre of the matrlx below.
in cases where traveling wave effects in the penstock are important the model of Figure 5 is modified to that of Figure 6. Here the upper penstock or tunnel is considered inelastic because the dynamic effects contributed by that system and surge chamber involve low frequency effects, while the high frequency response components are contributed by the lower penstock which is subject to abrupt gate or flow area changes. The difference between the model in Figure 6 and that in Figure 4 i s that the head acting on the lower penstock is the surge tank level rather than the constant reservoir elevation taken as 1 pu.
Figure 8.
171
2.7
NonLinear Model of Multiple P w t o c k s and TurbiheS SuPDhd From Common Tunnel. Elastic Water Columns in Penstocks and Tunnel
proportional control gain would be limited to about 3 per unit for acceptable stabilfty which would imply an unacceptably high regulation of 33%.
Figure9 shows the model accounting for traveling wave effects in the penstocks and tunnel.
ia
1%
1
202 Surge impedance of indlvidual penstocks 2 0 7 Surge Impedance of tunnel Te,, T 9 Te? Travel tlme of lndlvidual penstock8 T y Travel time In tunnel
I
G3=
Figure 9.
Model for Configuration of Figure 7  Including Traveling Wave Effects in Penstocks and Tunnel
n . .
This model incorporatesthe single penstock model of Figure 4 and introduces the effect of the tunnel b y using the same form of transfer function between downstream head and flow, which. for the tunnel is the sum of flows in the penstocks. Whereas the algebraic loop between flow and head of the simple penstock can be solved in closed form, these loops in Figure 9 are best solved by iteration.
3.0 HYDRO TURBINE CONTROLS
K = 2
0.01
Oel
rad/sec
'
10
Figure 11.
Hydro turbines, because of their initial inverse response characteristics of power to gate changes, require provision of transient droop features in the speed controls for stable control performance. The term 'transient droop' implies that, for fast deviations in frequency, the governor exhibits high regulation (low gain) while for slow changes and in the steady state the governor exhibits the normal low regulation. (high gain). From a linear control analysis point of view. the case of a hydro turbine generator supplying an isolated load can be represented by the block diagram of Figure 10.
Qate
K = 3 K = 4
WCROSS WCROSS
= 0.28 radlsec
=
WCROSS
= 0.71
45.8 deg
2 2 . 2 deg
0.0
deg
Pmech
ACC.
Acc.
Syrtem
Pelec
Figure 10.
Linear Model of Hydro Turbine and Speed Controls Supplying Isolated Load
100
1?,I
,
.oo30 3.0000
5.0000
7.0000
9~oooo
TI M E
11 .WO
3.000 15.noo
,ooO 19.000
Conventionalfrequency response and Bode plot analysisof this control system shows that a pure Droportionalcontroller would have to be tuned with a very low gain for acceptable stability yielding a very poor (high) regulation. This is evident from Figure 11 showing the open loop asymptotic gain and phase angle plots and in Figure 12 showing the response to a step change in electrical load for different values of proportional gain K. This example using a water starting time Tw of 2 sec and inertia constant H of 4 sec, shows that a
Figure 12.
Response of Mechanical Power for a Step Change in Electrical Power in System of Figure 1 0 with ProportionalGovernor
172
Transient gain reduction is thus necessary to provide acceptable steady state regulation with adequate stability.
3.1
Figure 13 shows the model block diagram of a typical governor in which the turbine gate is controlled by a two stage hydraulic position selvo. The physical meaning of the parameters used in the model is as follows:
T T ?I RP TR
Pilot valve and servo motor time constant ~ervogain Main servo time constant Permanent droop Transient droop Reset time or dashpot time constant
'
U
Rp
I
0.01
radlsec
10
Bode Diagram of Governor in Figure 13, Forward and Inverse Feedback Function
The permanent droop determines the speed regulation under steady state conditions. It is defined as the speed drop in percent or per unit required to drive the gate from minimum to maximum opening without change in speed reference. As noted in Section 3.0. due to the peculiar dynamic characteristicsof the hydraulic turbine. it is necessary to increase the regulation under fast transient conditions in order to achieve stable speed control. This is achieved by the paralleltransient droop branch with washout time constant TR. Because of the choice of the per unit system, with maximum gate opening defined as unity, the speed limits must be defined, for consistency, as fractions of the maximum gate opening per second. The Bode diagram in Figure 14 gives an asymptotic plot of the inverse of the feedback path l/hl i.e.
Hence, the closed loop response may be approximated by plotting both g1 and l/hl and choosing the lowest of both gain responses at any frequency as an approximation to the closed loop response at that frequency. Referring to Figure 14, the speedregulating control loop will "see' the governor as having a gain of 1/Rp in the steady state, and a 'transient' gain of l/Ri for phenomena above the l/TR frequency range. An equivalent time constant of 1/(QRt)sec will result from the second intersection of the g1 and l/hl traces. The speed regulating loop will have acceptable stability if: a) The transient gain, (l/Rt) does not exceed
 i 1.5
H 
Rt
Tw
b) Crossover frequency, Wc, approximately equal to 1/(2HRt), occurs somewhere in the region between l/TR and QRt. Th's reduces phase lag contributions from the governor. Several authors have proposed relations for temporary droop.
Rt = TdH*(1.15  ( l , 1)*0.075]
All three will result in crossover frequencies that are close to 1/2Tw,
and, therefore, satisfy condition (a). Regarding the dashpot time constant, Ref 11 suggests a value of 4 times Tw, Ref 12 proposed a TR equal to five times Tw and Ref 7 proposes
173
TR = Tw*(5  (Tw
 1)*0.5]
(20)
For crossover frequencies in the order of 1/2Tw, l / T R values in the order of 1/5Tw will minimize 'low frequency" phase lag contributions from the governor. Reference 7 suggests large values of 2, the servosystem gain, to attain improved performance characteristics. The typical maximum values of 5 to 10 reported in that Same paper, will minimize the 'high frequency' phase lag contributions from the gavemor.
3 . 2
There are cases where specific governors require more complex representation than in Figure 13. The differences may be due to added time constants in hardware and also where derivative action is included. Figure 1 5 shows an example of more complex representations.
Saeed _r_Reference
*
U
.o 1
.1
radhec
10
Servo
Neglecting the pilot servo and additional servo dynamics in Figure 15, shown in the Bode diagram of Figure 16 are the inverse of the intemal loop feedback path l/h i e 1/Rp. and the forward gain g , i.e. KP + Kl/s. When comparing tLe resulting frequency response characteristics with those of Figure 14 it is apparent that both governors achieve the same objective, i.e. transient droop increase. Tuning objectives are identical: Transient droop Rt is given by l/KP. KP/KI is equivalent to the dashpot time constant T , and care must be taken that crossover does not occur at frequenfies that are close to the inverse of the smaller servomotor time constants.
3.2.2 PID Governor
10
The purpose of the derivative is to extend the crossover frequency beyond the constraints imposed on PI governors. Figure 17 shows the governor loop frequency response when the PID governor is tuned according to the authors of Reference 10.
Rt = 1/KP = 0.625Tw/H
'
(21)
0.1
0.01
radlsec
'
10
Transient gain (1/Rt) has been increased by 60% over normal PI values. This results in roughly the same increase in crossover frequency. and thereby, in governor response speed. The detrimental effects on stability are averted by the phase lead effects resulting from derivative action. There is a risk, however, that the rise in magnitude due to the derivative action, compounded with that resulting from the hydraulic system. may result in a second crossover at higher frequencies. Due to the high phase lags at these frequencies, a second crossover will certainly result in governor loop instability. This is the reason for the minimum limit imposed on the value of KP/KD.
3.3
The governor model described in Figure 18 has modeling capabilities not frequently found in typical hydro plant models. Its features may be critical for the correct simulation of partial or toto1 load rejections:
114
Buffering of gate closure may produce a reduction in overpressures under load rejection. It also reduces impact loadings on the gate linkage and limits the magnitude of the pressure pulsations while the gates are fully closed during the decay of load rejection overspeed. Jet Deflector MXJDOR
defined in the Appendix serves to illustrate various aspects dynamic performance and simulation approaches.
4.1
Of
Govemor L o o p Stabili
Speed
Traveling wave effects may become a significant factor when analyzing governor loop stabili. F$ure 19 shows mognitude and phase lags for the classical linear hydro turbine model (Eq. 8. G =1, Tw=2 sec) and for the same ideal modelwhen travelingwave e d c t s are considered (Te=l sec) (14). The larger magnitudes in the traveling wave model will result in higher crossover frequencies. Higher crossover frequencies, compounded with larger phase lags, result in smaller phase margins,and therefore lessstable performance than when assuming a lumpedparametermodel.
%e
Ref
MXQTCR or
Opening Relief Valve Permanent Droop Temporary Droop Dashpot Time Constant Pilot Valve Tlme Constant Qate Servo Tlme Constant Maxlmum Gate Openlng Rate Maximum Gate Closing Rate Maxlmum Buffered Gate Openlng Rate Maximum Buffered Gate Closing Rate Maximum Gate Limit Minimum Gate Llmlt Relief Valve Closing Rate Maxlmum Rellef Valve Llmlt Maxlmum Jet Deflector Openlng Rate Maxlmum Jet Deflector Cloalng Rate Enhanced Governor Model Used In Load Rejection Studies Figure 19. In some installations a relief valve i s attached to the turbine s operated directly from casing providing a bypass for the flow. It i the governor or the gate mechanism of the turbine. The amount of water bypassed i s sufficient to keep the total discharge through the penstock fairly constant, hence controlling pressure rise. Turbine flow, as used for turbine power calculations, is determined in this case as:
AP /AG Function  Lumped Parameter and Traveling
.y
..
rR
1.0
I
o (rad/.sec)
1.o
2.0
MXGTOR MXGTCR MXBGOR MXBGCR GMAX GMlN RVLVCR RVLMAX MXJDOR MXJDCR Figure 18.
On.
W 8 e Models These effects have usually a negligible impact on P I controller stability. They should not be neglected. however, when analyzing PID controller. This is illustrated in the example in Table 1, where travelingwave modeling i s shown to have a significant impact on PID controller stabili. It can be shown that the per unit error in the head/flow transfer function due to ignoring traveling wave effects is approximately equal to
(22)
In longpenstock impulse turbines, rapid reductions in water velocity are not allowed to avoid the pressure rise which would occur. T o minimize the speed rise following a sudden load rejection, a governorcontrolledjet deflector is sometimes placed betweenthe needle nozzle and the runner. The governor moves this deflector rapidly into the jet, removing part or all of the power input to the turbine. Turbine flow in this case i s calculated as: Turbine Flow = Openstock x Min(l.,DefPos./Gate Opn.)
error
. I
To' s2 
(24)
Their significance is therefore larger for long penstocks. Long penstocks result in larger water time constants and therefore lower governor response speed. The larger bandwidth in PID controllers is most attractive in such conditions. Traveling wave analysis thus becomes cfiical.
4.2
(23)
Linear vs. NonLinear Hydraulic Model (Inelastic Flow) The advantages of nonlinear versus linear models become
In Figure 18, position limits are shown on the controller. Limits could also be included on the jet and gate Servos.
175
apparent when both models are subjected to large excursions in turbine loading. Figure 20 shows the models' responses to a relatively small (0.01 pu) step in gate position. Figure 21 shows simulation results for a larger (0.2 pu) step. For comparison purposes, the governor representation was deactivated in both models and both noload flow and sDeed deviations were set to zero in the nonlinear model. Table 1. Governor Loop Stability for Typical Controller Tuning and Alternative Hydraulic Models
4.3
Surge tank effects should be included in dynamic analyses of hydro plants when the time range of interest is comparable to the surge tank natural period. For shorter time periods. the simpler shortterm model can be used.
(25)
1 :[ 1 1 1
Hydraulic Controller Frequency rad/sec Lumped Parameter Parameter Lumped Traveling Wave
r ?
rad/sec
Traveling Wave
0.54 rad/sec
C C
13.8 deg
MECHANICAL POWER
H=4sec.TW=2sec.Te= 1 sec PI Controller: KP = 2, KI = 0.25 PID Controller: KP = 3.2, KI = 0.48. K D = 2.13 The hydroulic system parameters were:
T ,
Go = 0.762 pu
The linear model fails to represent the increases in effective water time with changes in penstock flow.
I I I I I
e
i
MECHANICAL POUER
Figure 21.
Mechanical Power Response to .2 pu Step in Gate Position. Linear vs. Nonlinear Model
' z v
Figure 22 shows the result of simulating a 0.1 pu step load increase on on isolated hydro plant with and without surge tank effects. The surge tank natural period is 3 minutes. For the normal 3 to 5 sec transient stability range simulation results are almost identical. For longer simulation intervals. surge tank level starts falling, and mechanical power recovery lags behind that of the shortterm model, which assumes an infinite surge tank. Simulation of surge tank dynamics is necessary when the tank is small enough to be emptied by a large load increase (8). Longterm simulations are valuable in establishing acceptable operating procedures that avoid such catastrophic consequences. For plants with more complex layouts. 'high frequency" oscillations resulting from pendulum action between surge chambers and other hydraulic resonant modes may interfere with the governor's speedregulating loop. Dynamic simulations and frequency response analyses representing these 'longterm" effects are required tools in such types of analyses (81.
Figure 20.
Mechanical Power Response to 0.01 pu Step in Gate Position. Linear vs. NonLinear Model
176
4.4
where: a  condult wave velocity g  gravitational acceleration A crosssectional area 0  conduit siope
The results shown in Figure 22 were for the case of inelastic condults and incompressible fluid. Taking into account the effect of elasticity and compressibility leads to a traveling wave solution method described in Figure 23. This method of calculation of traveling wave effects is an alternative to implementation of the Figure 6 model with time delay of transport time simulation of the wave.
Ond On
equation Of motion:
SPEED DEYIATIffl
I
TIME
Figure 22.
Response of Mechanical Power and Speed to a 0.1 pu Load Increase With and Without Surge Tank Effects
Time
Tunnel Inlet Constraints
Figure 23.
Flows and heads along the penstock and tunnel are analyred in terms of a continuity equation
Timespace lattices such as those shown in Figure 23 are defined for each of the conduits, and both equations are simuitaneously solved using the method of characteristics [9). The accuracy of the results is proportional to the number of segments into which the conduit is divided. Practical application of these models seems to suggest that a minimum of ten segments is required. Time and space increments are related by conduit wave velocity. Time increments must be equal to or multiples of the Simulation time step. The minimum size requirements on simulation time step may create additional computational burdens for large system simulations. Simultaneousconsiderationof two or more conduits, while using a unique simulation time step makes it impossible to T i an exact lattice on each conduit. Recognizing the problem uncertainties, particularly on conduit wave velocity, conduit lengths are adjusted to the nearest increment. The sine term in (26) recognizes pressure rises. and therefore, specific volume and flow reductions, resulting from reductions in elevation. This complicates the initializationprocess (flows along the same conduit are not equal in the steadystate), but has negligible effects in simulation results. Horieontal conduits may be assumed. The additional computational burden of programming and running a travelingwave model has to be weighed against the error caused by the use of an inelastic model. As previously mentioned, the perunit errors are proportional to the square of conduit travel time times the square of the main frequency of the dynamic phenomena. As shown in Section 4.1 this perunit difference will usually be negligible unless very long penstocks are studied or unless governor bandwidth has been expanded by derivative action. A critical case run under both assumptions assesses the difference. This is shown in Figure 24, where the hydro plant described in e = 0.42 s) is subjected to a 0.2 pu the Appendix cTw = 1.83 s. T increase in load under isolated conditions. Except for some transient high frequency effects, the difference between the elastic and inelastic solutions is negligible. There are times, however, when traveling wave analysis is essential. The analysis of overpressures and pressure pulsations due to total load rejection is generally carried out with this type of tool. A closed or almost closed gate gives rise to poorly attenuated traveling waves of pressure. Figure 25 shows a total gate closure simulation for the system described in the Appendix. For gate positions at or near total closure, the inelastic simulations of scroll case head and penstock flows are no longer applicable, and are replaced by an algebraic, steadystate solution of the penstock. Surge chamber levels and tunnel flows are not affected by these highfrequency effects. Frequency control is not affected either since turbine power is practically zero at these small gate openings. The effect of buffering the gate closure is shown in Figure 26 for the same total load rejection simulation as in Figure 25, but with a . 0 5 pu/sec. applied after gate maximum buffered closing rate of 4 opening is less than 0.15 pu. Overpressures and pressure pulsations ore significantly reduced. at the expense of a larger overspeed. Figure 27 shows the same total load rejection as in Figure 25, but including simulotion of a relief valve with a 0.01 closing rate. Both overpressures and overspeed are significantly improved by relief valve operation. Figure 28 simulates total load rejection including jet deflector action with a 0.5 closing rate. While gate closing rate has been reduced to a tenth of Rs value in Figure 25, reducing overpressures. the jet deflector manages to control speed.
177
It is recognized that specific applications may require the s deadbands. development of special models including effects such a hysteresis. etc. One of the objectives of the paper is to Present the basic physics Of hydraulic turbines and thekcontrok that, in the State of the afl, the development Of code for a particular model is routine once the physics are well defined.
5.1
TIME
igure 26.
Speed and Head Response to a Full Load Rejection With Gate Buffering
TIME
Figure 25.
Speed and Head Response to a Full Load Rejection With and Without Traveling Wave Effects
1.60
,lmm
:
TIME
I
13,ooo
'
15.u00
19.000
l7.000
3!0000
5!woo
7.0000 9 . ~ 0 0
l'.ooo
Figure27.
Speed, Head and Flow Response to a Full Load Rejection With Relief Valve Operotion
178 In studies of small isolated power systems, the governor and turbine characteristics play an important part in the response of the system frequency to disturbances. Here the action of the govemor speed regulation and the response of the turbine must be included f gate position and speed limits can be in the model. The effects o significant in such cases, and should thus correspond to those in service in the modeled plants. APPENDIX Sample System (Figures 5,6 and 18)
Rated
MVA:
Rated Turbine Power: Rated Turbine Flow: Rated Turbine Head: Gate Position at Rated Cond: NoLoad Flow, qNL:
138.9 m
0.90 pu.
4.3 m3/s
.
A S
L~
OPEN IN^
TURBINE HEAD
Permanent Droop. RP: Temporor) Droop, Rt: Dashpot Time Constant. TR:
0.05 pu.
0.45pu. 8. s
52
m +
. 2 9' ,     OfFLfcroR FIOU f 4 fi .. Q __    _ . 0 . 2 5 *c. OPEN IN^ PEED ~ ~ y l ~ r l o N *   . __ . b . ,/? _ _   f " _ y , : . " s
I m O
 4 7 7 0
L
0.02 s Pilot Time Constant, T : P 0.5 s Servo Time Constant, T : 0 Maximum Gate Opening Rate, MXGTOR: 0.1 pu./s Maximum Gate Closing Rate, MXGTCR: Maximum Gate Limit, GMAX: Minimum Gate Limit, GMIN. 0.1 pu./S
1. pu. 0. pu.
4 7 7
&(
I '
: ,
'
/I
/
,I
1 o
1 ' I '
.$
.' 3 %

_  . Lake Head,
Tail Head,
307.0 m 166.4 m 465m 15.2 m2 1100 m/s 0.0003042 3850 m 38.5 m2 1200 m/s
0.00101 12 m/(m3/s)2
Penstock Wave Velocity: Penstock Head Loss Coeff.. fpl: Tunnel Length: Tunnel Cross Section: Tunnel Wave Velocity: Tunnel Head Loss Coeff., fp2: Surge Chamber C. Section: SCh. Orifice Head Loss C.. fo: Turbine Damping: R E F E R E N C E S
1.
Figure 28.
Speed, Head and Flow Response to a Full load Rejectionwith Jet Deflector Operation
The turbine model of Section 2.1 coupled with a governor model chosen from that of Section 3.1 or Section 3.2 as appropriate is recommended for use in transient stability programs. A simple linear turbine model is not recommended since its parameterswould have to be adjusted as a function of operating conditions and the accuracy of representationwould be affected by the magnitude of the perturbations.
5.2 Small Signal Stabilii
In small signal stability studies, it is the effect of the governors and turbines on the damping of low frequency interarea modes which is of concern. These effects can be modeled adequately by linearizing the nonlinear turbine and governor models about the appropriate operating point. Fixed time constant lineaked models of the turbine without adjustment for operating point are not recommended. Linear models are also used for guidance in speed control tuning using linear control analysis techniques. The most criiical condition for such studies of governor adjustments would be with the unit supplying an isolated load at maximum output.
5.3
2.
3.
4.
J.L.Woodward,"HydraulicTurbine Transfer Function for Use in GoverningStudies",Proc. IEE, Vol. 1 15, pp. 424426, March 1968. J.R.Smith et al..'Assessment of Hydroturbine Models for Power System Studies", Proc. IEE, Vol. 130. Pt. C, No. 1, January 1983. P.W.Agnew,"TheGoverning of Francis Turbines'. Water Power, pp. 119127, April 1974. R.Oldenburger. J.Donelson, 'Dynamic Response of a Hydroelectric Plant", Trans. AIEE, Vol. 81, Pt. 111, pp. 403418, 1962. J.M.Undriil and J.L.Woodward, 'NonLinear Hydro Governing Model and Improved Calculation for Determining Temporary Droop'. I E E E Trans., Vol. PAS86, No. 4, pp. 443453, April 1967. P.L.Dandeno,P.Kundur and J.P.Bayne, 'Hydraulic Unit Dynamic Performance Under Normal and Islanding Conditions  Analysis and Validation", I E E E Trans., Vol. PAS97, No. 6, pp. 21342143. November/December 1978.
SrJecial ADdications
5.
In special circumstances, additional complexity must be included in the turbine and governor models to study the detailed response of the plant to disturbances, or to study the effect of the units on longterm dynamics. Other instances requiring additional complexity are studies of interactions between turbine hydraulic dynamics including draft tube pulsations and electromechanical power oscillations. In such cases, the model must correspond as closely as possible to the actual tubine and controls that exist at the plant. Detailed modeling of special controls, such as those discussed in Section 3.3, and the penstock,including traveling wave effects and surge tank dynamics, as in Sections 2.3 and 2.4, may be required.
6.
7.
179 8.
J.M.Undrilland W.Stmuss. 'Influence of hydro plont desiQnon regulatingand resewe responsecapacity'. IEEE Trans., Vd. PAS 74,PP. 11921200, July/August 1974. V.L.Streeter and E.B.Wylie,'Fluid Mechanics' (McGrwHill,New 9 7 5 ) . York, 1 Hydraulic Turbine Generating Unit Controlled by PID Governof, iEEE Trans., Vol. P A S 98 No 6.pp. 22942298, Nov/Dec 1979.
9.
11.
L. M. Hovey, 'Optimum Adjustment of Hydro Governors on Manitoba Hydro System'. AlEE Trans., Vol. 81, Part 111. pp. 581587. Dec 1962.
13.
L. K. Kirchmayer, 'Economic Control of InterconnectedSystems. Vol. 11, Chapter I., John Wiley and Sons Inc. 1959.
14. C. K. Sanathanan. 'Accurate Low Order Model for Hydraulic TurbinePenstock', I E E E Trans., Vol. EC2, No. 2. pp. 196200, June 1987.