Documentos de Académico
Documentos de Profesional
Documentos de Cultura
NINE
THE ENERGY SYSTEM IN STEADY STATETHE CONTROL PROBLEM
"./
I
j
I
The two oreceding chapters were devoted to the problems associated with the selection ~f a normal operating state for a power system. In this chapter we concern ourselves with the problem of keeping the system in this state by means of continuous automatic closed-loop control. As the demand deviates from its normal value with an unpredictable small amount the state of the system will change. The automatic control system must detect these changes, and initiate in "real time" a set of countercontrol actions which will eliminate as quickly and effectively as possible the state deviations.
.\
-r:
load-frequency control (ALFC) loop regulates the me awatt (speed) of the generator .The loop is not . I g . the c f h AVR . a Sing e one as 111 e case 0 t e . A relatively fast primary loop responds t f . '. . 0 w hi h as we h ave noted, IS an indirect measure of me awat a requency signal IC . speed governor and the control valves the steam (or hvd g) ft t ~alance. Via t.he - ro aw IS regulated With The automatic
~--------_/'-------------.
rTie~1ine
ALFC loops
IIVlllo()p
VOllil!JO
unOI
....
:I:
-c "
mixer
>
Z
....
o " o
c:
~ o z
\dP,
To network\ Figure 9-1 The automatic load-frequency and voltage regulator control loops of a synchronous generator.
-..;-==
...
___
THE
EN ERGY
C01?TROL
PROnLt.~ . .\ ~\ ~
~
'd
fJ
.f'l
. . the me awatt output to relatively fast load fluctuations. -, d.., the intent of matchmg t~ t take place in one to several seconds. By thus. \\ By" fast" we :nea.n changes t~ b lance this primary loop performs indirectly a ... ., tending to mamtain a megawa a ., . \ coarse speed or fre~Uen~y contra~~tains the fine adjustment of the [requency; and A slower secon ary o~p ~ . har with other pool also by "reset" action mamtams proper megawatt mterc ange b members. This -Ioop is insensi ti ve to rapid load and frequ~ncy cha~ges ut focuses instead on driftlike changes which take place over pen ods of rmnutes.
t:
'\ \
a..
9':'1-1 Cross-Coupling
!
I.
The AVR and ALFC loops are not in the truest sense noninteracting; crosscoupling does exist and can sometimes be troublesome. There is little if any coupling from the ALfC loop to the AVR lo,?p, but interactio~ exists in. the opposite direction. We understand this readily by realizing that control actions in the AVR loop affect the magnitude of the genJrator emf E. As the internal emf determines the magnitude of the real power [Eq. 4-22) it is clear that changes in the AVR loop must-be felt in the ALFC loop. I However, the AVR loop is much faster than the ALF9 loob and there is therefore a tendency for the AVR dynamics to settle down before they can make themselves felt in the slower load-frequency conttol channel. In thejdiscussions to follow we shall first study the two loops independently starting with the A VR loop. Later we shall discuss the effects of cross-coupling.
61
normal operating
';j (:l.'
v;- c:
~
::;. tI:l c
(\) '< ~
.':;;!;;
~ on '
o ....
.-. O
=.
.JI1~
eLecT
AN INTRODUCTION I
'I
;-/
/
(AVR)
I
'
The exciter is the main component in the AVR loop. It delivers the de power to the generator field. It must have adequate power capacity (in the low megawatt range for large generators) and sufficient speed of response (rise times less than 0.1 s~conds). The bas.lc role of the AVR is to .provide constancy of the generator terminal voltage during normal small and slow changes in the load. However, it is com~on practice to design tile exciter with enough margin to give powerful boosts in the excitation level also during emergency situations.
i
,I:
-,' .,j
9-~-:n. Exciter
Types
I.
There exists a variety of exciter types in use, the more important ones described in Refs. 1 arid 2. . , In older' power plants, the exciter consisted of a de generator driven by the main generator shaft. This arrangement required the transfer of the de power to the generator field via slip rings and brushes. Modern exciters tend to be of either brushless or static design. A typical .brushless AVR loop is shown in Fig. 9-2, where the exciter consists of an "inverted" three-phase synchronous generator. The latter has its three-phase armature on the rbtor and its field (Ill the stator. Its ac armature voltage is rectified in diodes mounted on the rotating s.aaft, and then fed directly into the main generator field. This deisign obviously eliminates the need for slip rings and brushes.
I !
J~
I
(
j'
I
Comparator
,...---......,
Amplifier
~,-~
Wlrel
+0
~ ~r-J=hllirt sit:>
e
. '_ _
.
+ I
I G
8
~
VR
Exciter
Rectifier
Synchronous generator
8
I I I I I I I
Il
L.r-._j
I Gf I
II II
i
J
I
Rotating components .
-'-_I
Stability compensators'
WI
~_---:--<l------=:~IL-~
IF'~e 5Jl-2 Brushless A VR loop.
/1
' ,r~\
THE
ENERGY
SYSTEM IN STEADY
STAlE-THE
CONTROL
PROULEM
303
i , I
,t :;,
In a static AVR-Ioop the excitation power is obtained directly from the generator terminals or from the station service bus. The ac power is rectified in lhyristor bridges and fed into the main generator field via slip rings. Static exciters are very fast and contribute to improved "transient stability" (Chap. 12).
9-2-2 Exciter
Modeling
I.
~ !
I I
Here follows first a brief description of the AVR loop depicted in Fig. 9-2. (The role of the" stability compensators" will be discussed later.) Assume that for some reason the terminal voltage I VI would decrease. This immediately results in an increased "error voltage" e which, in turn, causes increased values of VR' i., vj',t and if' The d axis generator flux increases as a result of the boost in if' thus raising the magnitude of the internal generator emf E and terminal voltage V. . Mathematical modeling of the exciter and its controls follows. If for the moment we disregard the stability compensator (shown dashed in the figure) we have for the comparator and amplifier respectively
\.
I
61 Vlref and
~I VI
~e
(9-'1 )
yields
:r.
I~
"
G
A
~ ~VR(S) 4e(s)
_ - KA
(9-2)
where G A is the amplifier transfer function. The last equation implies instantaneous amplifier response. In reality the amplifier will have a delay that can be represented by a time constant T and its transfer function will then be of the form A ,
(9-3 )
If R, and L, represent' respectively the resistance and inductance exciter field we have for voltage equilibrium in the latter of the
(9-4 )
t IEEE (see Ref 1) uses the symbol E'd for the field voltage We d~part reasons: .
(a) Capital letter symbols are reserved for phaso . rs, (b) S b 1M" .' ym 0 e IS reserved for emf and/or "error volt age. .
from thi
IS
S\'STc~IS
THI;llRY:
AN INTIWDUCTION
. . d es K armature Measured across the rnain field the exciter pro u~ )'1 ampere of field current i., that is, we have the proportlona ity t:"vf = K, t:"i, Upon Laplace transformation of the 1;5.t two equations
volts per
(9-5)
and elimination of
(9-6)
I.
I
I
From these equations and using accepted block symbols we assemble directly the transfer function model shown in Fig. 9-3. The time constants TA and T, have values in the ranges 0.02-0.10 and 0.5-1.0 seconds respectively. iNote: Sometimes the exciter block representation is given as
VF
VR
+ STE
We have chosen the form shown in Fig. 9-3 as it conforms with the general con trol system defini tions for" gain" and" time constant.") 9-2-3 Generator Modeling
We need to "close the loop" in Fig. 9-3 by establishing the" missing dynamic link" between the field voltageu. and the generator terminal voltage I VI. As the terminal voltage equals the internal emf minus the voltage drop across the
Comparator Amplifier Exciter
"I ~I
'1-
\'"
v'
':IV)
'VI
VI
'r
\
\
'~'''' o
\
\ THE ENERGY SYSTEMIN STEADY STATE-THE CONTROL
--:J.
'~o....'.i
.-~
'Po
~ 0
""ell"_" .....~;>,_
Amplifier
Exciter field
Generator -----, 1
field LlIV I
<, ~
KF
+ sTdo
la)
v
u
lb)
r
j~
LlWI
->
(c)
Figure 9-4 Closing the AVR loop; (a) i~dividual bloc~ representation; (e) closed loop model.
(b) condensed
model;
internal impedance it is clear that the relationship between V f and IV I depends on the generator loading. The simplest possible relationship exists at low orzero loading in which case V equals approxir:tately the internal emf E. Kirbhhoff's voltage law applied to the field winding yields in this case: I
I1vf
= Rf
I1lf
Lff
-d (111f) = -L - Rf I1E t W fa
d..Ji
[ + Lff
dt
(I1E)
I
(9-7)
d]
The last step follows from Eq. (4-5). . I After Laplace transformation. the last equation yields the field transfer ratio
I1E(s) ~ I1vAs) =
111 VI
(s) l1iJAs)
1 + s~o
(9-B)
where K and
.
A
F -
wLfo_
-fiR~
.
to)
T;,. was
We 'can now complete the AVR loop as .hown Fig. 9-4a. In tws easy-toin
the final form shewn' in Fig. 9-4c.
.,
t:
;0
_J!.1;ft~~Fjw.--;_"2'i:'102:.";:.'ii:-:;,j'S{"_"O=/J.::.~j;_T:'JJ.i5.,?J;;iS!f!./!i~;it!i!!fif![;!.{fi!!ifliff."
j/
1~
//
'"
;u
.~
'V
:
/
ELECTRIC
ENERGY
SYSTE~tS
THEOR
v:
AN INTRODUCTION
// ~
_ Tl]c oven-loop
t1'(II.J.~,/C,.[unction
(;1 s) equals
G(s)
. where the open-loop
K = _._ -~~--:::-:-:-
(1
-l-
sT,.j)(l
by
+ s:z;,)(1 + sT~o)
__
(9-9)
gain K is defined
1~ KKK ,.j
(9-10)
1. regulate the terminal voltage I V I to within required static accuracy limit 2. have sufficient speed of response 3. be stable
The sranc accuracy requireme-nt can be stated thusly: For a constant (subscript 0) reference input the likewise constant error l1eo must be less than some specified percentage p of the reference. We thus can write the static accuracy specification as follows: l1eo
'
= I1jV Irer.
I) -
I1jV
(9-11)
For a consfant input the transfer function is obtained by setting s Fig. 9-4c we thus obtain . l1eo f I1jV
= O. From
Irer. 0 -
G~}
1 - 1 + 0(0)
Clearly, the. static error decreases with increased loop gain. Substitution of (9~12) into (9-111) yields the minimum gain needed to obtain a specified accuracy: K> 100 - 1 p
6.jVlrcr.
(9-12)
(9-13)
For example, if: we specify that the static error should be less than one percent the open-loop gain must exceed 99.
tilt!
AVR Loop
:;y:;tCI11S12
t.1V1(t):o= 2.1-'{t.I~o((s).G(~)(1'1} .
THE
ENERGY
SYSTEM
IN STEADY
STATI'.--TIIE
CUNTROL
I'IWIlLEM
3{)7
Mathematically, the- response depends upon the eigenvalues poles, which are obtained from the characteristic equation
or closcd-toon (9-15)
G(s)
+I=
In the AVR loop (Fig. 9-4) the open-loop transfer function G(s) is of third order and we thus obtain three eigenvalues, Sl' S2. and S). If these are distinct and real the transient response components are of the form
(9-10)
a jw.
If two of the eigenvalues. for example S2 and then the transient component will contain Ale'" sin (wt
SJ.
form
+ [3)
For the A VR loop to be stable the transient components must vanish with time. We thus must require that all three eigenvalues be located in the left-hand s . plane. For the loop to possess good tracking ability the transients must not only vanish but they must do so speedily. As the real parts of the eigenvalues determine the rapidity of the exponential decay, a high-speed loop must possess eigenvalues located well to the left in the s plane. The amplitude factors AI' A2, and AJ in formulas (9-16) and (9-17) express the relative size of the transient terms. If one or several of the terms are relatively large then the corresponding eigenvalues are said to be dominant. Generally the closer the eigenvalue is 'located to the jw axis the more dominant it becomes,
I
"
I I
I
By incre.as!ng the loop gai.n th~ eigenvalue 52 travels to the left and the loop response ,quIckens .. At a certain gam setting-the eigenvalues S3 arid S2 .. collide," Further increase III the loop gain results in S3 and S2 becoming conjugate
:'::'."
INTRODUCTI0N t. .
1'(')
'"
<(' c
,/
. \ compl:Xl s plane
_j
<,
'Critical gain crossing
a
_
.,_ ~~I
..J,C:-:
__
...;_''!..
.!J.~-*------l!H)--f--{}-X;:-t--------Pole at 1 s = -TA
Pole at
s = --
1 Te
complex. This dominant eigenvalue pair (b) makes the loop oscillatory, with poor damping. If the gain is increased further the eigenvalues wander into the right-hand s plane (c). The loop now becomes unstable. 9-2-7 Stability Compensation
High loop gain is needed for static accuracy but this causes undesirable dynamic response, possibly instability. By adding series and/or feedback stability compensation (Fig. 9-2) to the AVR lo"op, this conflicting situation can be resolved. As the stability problems emanate from the three cascaded time constants the compensation networks typically will contain some form of phase advancement. Consider for example the addition of a series phase lead compensator, having the transfer function G,=l+sYc The open-loop transfer function will now contain a zero
.
G(s)
(9-18 )
The added network will not affect the static loop gain K, and thus the static accuracy. The dynamic response. characteristics will change to the better. Consider for example the case when we would tune the compensator time constant
",.. !.
CONIROL
T
e'.
The oDen-loop
transfer
function
thea equal
K G(s)
1(9-19)
= (1 + sT )(l
A
+ sTdo)
sated system are depicted in The root loci (there are only two) of the con:pen 1 el'genvalues the dornin. )U It negative rea in '. Fig. 9-6. Low loop gam (a sn resu. s term Increasing loop gam (b) ant one of which, S2' yields a sluggIsh respo.nse f the oscillatory term will, results in oscillatory response. The ~ampmg h in the uncompensated ., . n a- was t e case 1 however, not decrease with mcreasmg gal :> system. . ..' k ensation In reality a combination of senes and feedbac cornp (Compare Prob. 9-5.)
9-2-8 Effects of Generator
is employed.
Loading
The above analysis of the AVR loop was based upon the assumption (Eq. 9-8) that E :::: I V I which holds true only for a lightly loaded generator. HO\;r. does increased loading affect the above analysis? It should be noted that the ability of the AVR loop to maintain a well regulated terminal voltage under widely shifting load conditions ("robustness") is an absolute prerequisite. . ' Added load does not change the basic features of the A VR loop as depicted in Fig. 9-4 . It will, however, affect the values of both the gain factor KF and the field time constant. High loading will make the generator work at higher magnetic saturation levels. This means smaller changes in I E I for incremental increases in if' translating into a reduction of KF The field time constant will "likewise decrease as generator loading means closing the armature current paths.
(namic
ompen:d. rstants
'b
t
1 a - Tdo
vancensator,
1
TA
0
~~~------~----~D-~~4----
--------~
t
loop.
~-'
:::J
()
0
;t!
0
I'"
f!>
j1
s
ENERGY SYSTEMS THEORY: AN INTRODUCTION
...,.
~
~. ~O
o " :::J?
ELECTRIC
/7
r
~1
~.
This circumstance permits tlu: formation of transien t " stator current the '. f hi h . ld S, e .exIstence 0 W IC yie s a lower effective" field inductance. The rotor-field time cons tan t, load can be sh a wn to 'equal
U
wyer
I
d, load -
X'
X + X exl . . Tdo +
I
ext
reactance
i~ .
" ., ~j,
11;
I
I
'I'
&\
i~~\
:.i
: }I
~!:
I:
~-,
I'.
~\
;'1 ~
'I l( .
~
The basic role of ALFCt is to maintain desired megawatt output of a generator unit and assist in controlling the frequency of the larger interconnection. The ALFC also Helps to keep the net interchange of power between pool members at p~ed~termin9d. values. Con~~\ should. be appl~ed in such a fashion that high!y differing response characteristics of units of vanous types (hydro, nuclear, fossil, I I. . etc.) are recognized. Also, unnecessary power output changes should be kept to a minimum in-lorder to reduce wear of control valves. The AL~C loop will maintai~ con~ol only during nor~al (small and slow) changes in load and frequency. I~ IS typically unable to provide adequate control during emergency Situations, when large megawatt imbalances occur. Then more drastic" emergency controls" must be applied {Chap. 12). In the present section we shall first study ALFC as it applies to a single generator supplying power to a local service area. We shall later extend the study to embrace several generators allpart of a single" control area."
\11 I,:'
?iJ,l
'fi'.1fi
Ii! \.,
I
~1{l'~ \
'
'I
;
~
~
I !
The term" Automatic Generation Control ~ (AGe) is also commonly used. terminology used agrees wit~ the standard 'accepted by the industry. AIl systc course. not identical. We chose a representative example for demonstration. ms arc,
:t The
f
0
~'J;,)",;?-.=~~. .. .:=_. . ,
.. ... -
CONTROL
PROn'LE,\.1 311
Steam
.IPv
Speed
governor
Hydraulic
amplifier
megawatt increment 6oPv. This flow increase translates into a turbine power increment 60PT in the turbine (not shown in the figure). Very large mechanical forces are needed to position the main valve (or gate) against the high steam (or water) pressure, and these forces are obtained via several stages of hydraulic amplifiers. In our simplified version we show only one stage. 'rhe input to this amplifier is the position XD of the pilot valve. The output is the position XE of the main piston. Because the high-pressure hydraulic fluid exerts only a slight differential force on the pilot valve, the force amplification isvery great. The position of the pilot valve can be affected via the linkage system in three ways: 1. Directly, by the speed changer. A small downward movement of the linkage point A corresponds to an increase 6.P ref in the reference power setting. ~. Indirectly, via feedback, due to position changes of the main piston. 3. Indirectly, via feedback. due to position changes of linkage point B resulting from speed changes. . It should prove a useful' exercise for the reader to find, qualitatioely, the workings of the mechanism. For example, give a "raise" command to the speed changer and prove that this indeed results in an increase in turbine outpwt. Prove also ..that a sm~""ci r"" U/;1I o-;,,~ th~ ~~~~ ~I'r __ n
AN INTRODUCTlON .
of the mecha~is_m'
. ints
.Il<lthe analysis to follow, incremental movements .of the five linkage po II .. , E in Fig. 9-7 are of particular interest. In reality these movements are
measured in millimeters but in cur analysis we shall rather express them as power increments expressed in megawatts or per-unit megawatts as the case may be. The movements are assumed positive in the directions of the arrows. The .. h (1. The governor output command 6.Py is measured by the positron c ange Xc' governor has two inputs:
I. Changes
2. Changes
6.P ref in the reference power setting tl/ in the speed of frequency of the generator,
as measured
by 6.xB
An increase in tlPg results from an increase in 6.Pref and a decrease .in 6.f.We thus can write for small increments 1 Mg = sr.; - Ii 6./ MW (9-21) The constant R has dimension hertz per megawatt, and is referred to as or droop. (For numerical values see Example 9-2 below.) Laplace transformation of Eq. (9-21) yields 6.Pg(s) = Using well-known as shown in Fig. 9-8.
regulation
6.Pref(s) -
Ii 6./(s)
(9~22)
the governor
block diagram
9-3-2 Hydraulic
Valve
Actuator
The input position 6.xD of the valve actuator increases as a result of an increased command 6.Pg but decreases due to increased valve output, 6.Pv. Equal increases in both 6.Pg and 6.Pv should result in 6.xD = O. We can thus write
tuD =
6.Pg
6.Pv
MW
. (9-23)
For small changes IlxD the oil flow into the hydraulic motor is proportional to position 6.xD of the pilot valve. Thus we obtain the following relationship for the position of the main piston: 6.Pv = kH
f 6.x
dt
(9-24 )
The positive constant kH depends upon orifice and cylinder geometries and ft:blid pressure. Upon Laplace t.ransformation of the last two. equations and upon eliminalioF} of 6.XD w~ obtain the actuator transfer function
'-'--,.
(2~.y!'~:;-a
__," '---~~--__..'------~~-----~
"
::t;;;'~jJ Z~~==-e5 -t'ci-~ arouzri Q.! s, 1t.~ b;'-~:-2.'l:t1:-{zl-;e 2.....~to.:- j;_as ceen represented G;![s):n :~~ 9-8.
by me transfer- function
fu
,!:-07~l';:.r '7 k~ P
';n!~c:e
(9-25) b. AF~' D "''-:;: ~ye Gf:T.--eo Gy for the most common turbine types. A b 'consisting of a .~~ mre cc;t~.?iPt, i.e...
~.t<tailil.e;flt;r;-Be~?at S1P..am 3J'POirE has the simples, transfer function,
iI
~bEe1'R'I~
ENERGY
SYSiFEMS TI-!EORY:
AN INTRODUCTION'
!D,e S'al,d t'h"ll[ the tU,rhine response is slow with response times ,,' nd se(\:Qn 's. .
"
lIil
. o'tm~r,ltlrbine .
t~ansfer functions
The, gen,er.atoi". P?wet increment I:lP G depends entirely upon the changes C1Pv the l?ad Pj;) being fed from the generator, Via the transient mechanism dis~ussed 10 ~ec. 4-13 the genei'ato~ always adjusts its ,output so as to meet tne e.man?, c~anges. !1P D '. These adjustments are essentially "instantaneous," eertainly m c9mpanson With the slow changes in Pr, and we can therefore set
!1PG D = !1P
(9-28)
the turbine-genera-
In vie~ of Eqs. (9-26) al1ci (9-28) we can thus symbolize tor dynamic response as shown in Fig. 9-8.
R !1Jo
(9-29)
Case A The generator is synchronized to a network of. very large size, so large in fact, that its frequency will be essentially independent of any changes in the power output of this individual generator (" infinite" network). Since we thus have '
6.1=0
'we obtain from Eq. (9-29)
!J.PT,o = !J.Pref, 0
(~-30)
For a generator operated at a lforced) constant speed, we thus have a direct proportionality between the 'turbine power and reference power setting.
Example 9-1 A lOO-MW generator is operated onto an .. infinite" network. Hew would YO\l make this generator increase its turbine power by 5 MW?
SeW'IJ:0N
'j,
Case B Now we consider the network "finite," i.e., its frequeney is variable. W'e do, however, keep bae speed-changer setting constant, that is, Mrcf = O. FFOIn Bq. (9-19) we obtai;l'l , 1 .
0
:;
(9-31)
.. '.:.
CONTROL
PROULHI
315
For a COl1stant setting of the speed chanqer the static increase ;11 turbine !'(JI\'('I' output is directiy proportional to the static [requency drop. This result points out the physical significance of the feedback parameter R. We remember that the physical unit for R is hertz per megawatt. III practice. the power is measured in per units, and in this case the unit for R will be hertz per unit megawatt. If the frequency drop is likewise measured in. per units of normal frequency (= 60 Hz), and if the power is measured in per-unit megawatts. then the unit for R will be in per units also.
Example 9-2 Consider the tOO-MW generator in (he previous example II ha, a re!?lIIalit)11 parameter R of four percent (or 0.04 pu). By how much will the turbine power increase if the frequency drops by 0.1 Hz with the reference unchanged?
SOLUTION
A 0.04-pu regulation parameter means thai the turbine power will increase I I'll. or tOO MW for a 0.04-pu, or 2.4-Hz drop in frequency. Thus we have 2.4 R = 100 = 0.024 HzjMW
According experience
to Eq. (9-3l) for a frequency change of 6.[= -0.1 a static change of I 6.Pr.o=-0.024 x(-0.1}=4.17MW
Case C In the general case, changes may occur in both reference setting and frequency in J.,hich case the more general relationship (9-29) applies. In a frequency-generation graph Eq. (9-29) represents a family of sloping lines as depicted in Fig. 9-9, each line corresponding to a specific reference power setting.
i
Speed changer set to give rated frequency at 100 percent Of rated output
---I>
98
96,
94.
Percent of rated output Speed changer set. to give rated frequency at 50 percent of rated output response of speed governor (Graphs correspond to R = 0.04 pu.)
___
--:'
~ ...:t>::I._~_"'" .---:-
-----.-
.11'0
El.ECTRIIC
EN:ERGY
SYSTEMS
THEO
RY'
AN INTRODL;CTION" .
.. MW enerator in the previ0us example. If the frequency Example 9-3 Consider again the 100g . h ed by how much should the drops by 0.1 Hz but the turbine power must remain unc ang , referense
SOLUTIO:--:
setting
be changed?
T.
As t>.P
=.!.. R
t>.f,
0
= 0.024 -'--
-4.17
MW
Rcsul:
We
mUSI
command
"Iower"
by 4.17 MW.
Th err ra t'Ings are 50 and . Example 9-4 Two generators arc supplying power to a system. , 500 M W respectively, The frequency is 60 Hz and each generato( is hulf-Ioaded. The system load increases by 110 MW and as a result the frequency drops to 59.5 Hz. What. must rhe individual regulations be if the two generators should increase their turbine powers In proportion to their ratings?
SOLUTION
The
IWO generators
should
pick
up 10 and
100 MW
respectively.
From
formula
their R values:
= - --
-0.5 10
= 0.05
Hz/MW
(smaller
unit)
-O.S R = - -- =
100
(larger
unit)
If we express
the regulation
in per-unit
per per-unit
megawatt,
This result thus teaches us that generators working in parallel on the sa~e network ought to have the same regulation (expressed in per unit of their own rating) in order to share load chanqes in proportion to size.
fTIega watt-seconds. 2, By' connecting additional load objects to the system the load demand increases by !::"P D which we shall refer to as "new'; load. (If load objects are discennected thea tJ.PD < 0, that is, the "new" load is negative.] The generator immediately increases its output !::"'Pa to match the..new load, that is, !!.PG = !::"PD,
~.
:co ~,
'~
~,I..; L PROl\L~ ~
~
.
k< zs:
.;:;
STATE-THE
CONTRO
.
n"9-
o~ \;\
3, "There will now be a power imbalance ill the ~rea that equals 6.PT - ~t ~ megawatts. As a result the speed or frequency will change. This change will b (ilssumed uniform throughout the area. As the kinetic energy is proportional t~ , the square of the speed we can thus write for the area kinetic energy Wkin =
~in
( JO J 2)
MWs
(9-32) Example
4. The" old" area load has' a frequency dependency 3-6) that we can lump into one area para neter IvI:W/Hz
ap olaJ (compare
(9-33 )
Area power balance requires that the increase in turbine power equals the, sum of "old ,; and "new" load changes plus the: rate of change of kinetic energy. We can thus express this area power balance as follows: (9-34 )
As
f=fO
and as I1f is small relative to
~in
+ /)"f
write Eq. (9-32)
r we can
~jn
~jn
fO fO I1f)2 = +
;::::; ~In
f 1(+: 2 6. ) fO
(9-35)
equation
(9-36)
by the generator
~jn
Pr
the form 2H d
MWs/MW
(or s)
(9-37)
l '.:'
I1PT -
sr; =
fO dt (~f)
+D
~f
puMW
The ~P's are mow measured megawatt per hertz. The H para~eter
that
it
is essentially
iadepen.
,
pu M 'The eq1uation is now written in terms of AI/ any ana ysts prefer this.] Laplace transformation of Eq. (9-3&) yields
MW
(9-39)
.
I",
APr(s) - .1PD{s)
which equation
. I
= =Os
?H
lJ.f{s)
+D
lJ.f(s)
(9-40)
I;, K
p
IOD
D
2H
s
Hz/pu MW
(9-43)
(9-44)
~ _!_
Equation (9-41) represents the" missing' link " in the control loop of Fig. 9-8. By adding a summing ju~ction. and a transfer block we close the loop as depicted in Fig. 9-10. (Disregard at this time the dashed portion of the diagram marked" secondary loop.")
r-------------r------------------.
I
!I
I I
I
I'
I I I ... ----.,LlPr.'(s)_ I. 'K, 1_+ ..... --1. -~I
S L.._-' __ ~
df.(s)
I
/1 '
~,:',
319
'-40)
.' lor feed ing lion of a single genera Our above theory was based upon the ~ssur:nP I the striClest sense the theory power to a local service area-a rare sltuatl~n. a~ arallel-working gcnerato~S does not apply to the more common case 0 m YtPms normally control their st power sys e I loops serving a larger area. However, mo . lier the individual contra . . . Imp ortaIH---t he generators . Unison. F or r easons mentiOned ear I d his is qUite . In have the same regulation parameters. If also-an t 11 onse characteristics then individual generator turbines tend to ha.ve t~e r;:gresent the whole system it is possible to let the control loop In Fig .. which then would be referred to as a control area.
s;~~
r
Example following 9-5 Determine data: Total Normal Inertia rated area capacity load the primary ALFC loop parameters or
a conlrol
area
having
the
1-41)
operating constant
P~
Regulation
R
that
= 2.40 Hz/pu
MW
(all area
generalClr' aning
)-42)
We shall load would assume the load-frequency for onc percent yields D: dependency frequency increase one percent
IS
. r Inear,
me
that
the"
old ..
increase.
i-43)
SOLUTION
The latter
load assumption
9-44}
or in per units of area capacity
iJP~ D =- = -
of
10. = 0.60
16.67 MW/H
~.9-8. )p as .gram
We then get from
pu MW/Hz
= 20 s
= 120 Hz/pu MW
ALFC
Loop
Having closed the primary ALFC loop we wish now to-study both its static and dynamic features. In this section we shall limit attention to the static characteristics. One of the basic objectives' of the loop is to maintain constant frequency in spite of changing loads. How accurately does the loop maintain the frequency? The primary ALFC loop in Fig. 9-10 has one output 6.J and two inputs 6.P rer and IlP D' From the block diagram we obtain by inspection'
{[6.P
rer -
N ]GHGr -
MD}G
=N
(9-45 )
Fcii.r a
(:,OI1SlUllt
.r ' relenmce
inp u t
6./(s) = - 1 + (I/R)GpGHGT
For
(9-46)
a step load
change of constant
magnitude
6.Pv = M we have
M s
0 bta tarn
we rea dil Y I
[s 6./(s)] =
the
Kp
1 + Kp/R
M_ _
I -
+ l/R
response
Hz
(9-47)
so-called
area frequency
charactertistic
D.
1 R
pu MW/Hz
M f3
(9-48)
6.fo =
Hz
example
(9-49)
follow-
Example 9-6 Find the static frequency drop for the 2-GW system iri previous ing a one percent load increase. that is. 6.PD = M = 20 MW = 0.01 pu MW.
SoLUTION
data I
X
we get
f3 = 8.33
.'. or 0.04 percent Example governor
50LtJT'I0N
10-3 + -
2.40
= =
0.425 pu MW/Hz
6./0
= -
0.01 0.425
-0.0235
Hz
of normal
9-7 What would the frequency drop in the previous loop were nonexistent or open? Opening the loop is tantamount to setting
P=
.and
D = 8.33 x 10-
6.( = -~
vo or fw.o: percent .: of normal value.
0.00833 -
~-
1.20, Hz
'.'
.
"
=:
.~
THE ENERGY
SYSTEM IN STEADY
STATE-THE
CONTROL
PROfiLGM
321
ic
7)
til-
48)
.s.:
!::,.f(s) ;:::; _
1 +--p-
1 + sTp 1 K
M s
R. 1 + a,
49)
tow-
=-MR:K~p(~_t::,.j(s) ;:::;-0.0235(~ s
s+~
. p
Rl+Kp)
(9-50)
+ 2.55.
Hz (9-51 )
1_)
The approximate time response is therefore purely exponential. t::,.j(t);:::; -0.0235(1 - e-2.S5,)
peed-
Figure 9-11 shows an analog computer recording. of this response. For comparison, we also simulated the loop response with the inclusion of. the time constants TH and TT' We make the following observations in regard to our results: .' 1. The overall closed-loop system time constant iequals only .1/2.55 = 0.393 s, which is a considerable reduction from the value T" = 20 s, characterizing the " plant" itself. This speedup is a result of the feedback arrangement of the speed governor. Note that the system can be made-still faster by reducing R, that is, by increasing the static loop gain. 2. Reduction of R aliso reduces the static frequency error. 3. If we performed the above analysis by not disregarding the turbine response, then the response would net be purely exponential as above. :I.o,'Fig. 9-1i ,we
show the difference. Note that the added delays cause a larger transient frequency dip. Why? 4. The speed governor gives a reasonable performance with a static frequency . drop of only 2.4 Hz between zero and full load and settling time of the order of 3 s (as depicted ill Fig. 9-11). However, with the extremely severe restrictions we in reality impose on frequency constancy (see Sec. 9-3-10), the results are, in fact, entirely u.nac.ceptable. We must do much better. 5. The foregoing analysis may not have given the reader a full physical understanding of the load frequency mechanism of the' single-area system. Since . such an understanding is invaluable to appreciate fully so-called bias control in multiarea systems (see Sec. 9-3-10), we shall attempt now to shed some light on the physical mechanism.
of Results
When the load suddenly increased by 1 percent (= 20 MW), where did this power come from? Certainly, it must hav~ come from somewhere, as can be certified by the customer who threw the switch and expected and got instantaneously the demanded 20 MW. """'-.'. In the milliseconds following the closure of the switch, the frequency has not changed a measurable amount, and therefore no power increase has had time to develop in the turbine (where the steam valve has not yet moved).
.. . In those first instants the total additionalload-demand 20 MW is obtainedfrom the stored kinetic energy, which therefore will decrease at an initial rate of 20 M W. The kinetic energy is released. by speed reduction. Since the speed is dropping (and from. Eq. (9-51) .we note that the initial deceleration is 0.0235 x 2.55 = Q.06 Hzjs), the steam valve is opened up; due to the mechanism' described ea~lier and thllis meains increased generator output. Also, and this, is important to real'iz-e' ,
THE ENERGY
SYSTEM IN STl:,\I)Y
Sl'ATIi,-THll
CONTIlOL
PROBLEM
323
since the speed is now dropping, the, "old" 10tHI llOOO rvlW in this case) decre~ses at the rate of D = 1000/60 = 16,67 rvtW/Hz, SInce the appearance 011 the scene of-thts I'c/t;oscd" power //IeallS that less power l1eeds to be qenerated. we carr in effect consider itto /)(1 (l direct cOlllributioll to the new load demand, ' - In conclusion,' as the speed is dropping, the demand increase of 20 MW is thus made up three components:
H
or
system
machines
3. "Released"
Initially, the last two components are zero, but as the speed is dropping, they will account for an increased contribution. Consequenlly. the kinetic energy will lie consumed at a decreased rate , and this is confirmed in Fig, 9-1 L which shows that the deceleration decreases as time goes on, Eventually (theoretically, after infinite time), the speed will level off at a new constant lower value. At this time the kinetic energy is constant (at a lower value), and the 20-MW load increase is therefore made up of the last two components only. It is interesting to see in what ratio the last two components contribute. The static speed drop being 0.02.3.5Hz, we can compute these contributions as follows: 1. The generator regulation is 4.00 percent, or 2.<.l.()Hz/pu MW,'and in generation .after the new steady state is reached will be 0~~5 the increase
= 0.0098 =
pu MW
0.0098 x 2000
19.6 MW
s
e
1-
2. Since the" old" load decreases at the rate of 16.67 MW 1Hz, the speed drop of 0'.0235 Hz will," release"
~t
LO
m'l
IV, :ng
n:
=
er, ze,
324
ELECT~IC
ENERGY
SYSTEMS THEORY;
AN INTRODUCTION
the stringency
the sophistication
of the
proposed control method. A Suggested control system specifications We are presently discussing ~ singble l .. .f . . r h a system are considera y area system, and the con trol spec! catI~ns lor sue simpler than those imposed upon a multiple-area system (power pool). Here follow some realistic specifications:
1. The control loop must be characterized by a sufficient degree of sta~ilitYT' hi 2. Following a step load change, the frequency error should retur~ to ,zero. IS is referred to as isochronous control. The magnitude of the transient frequency deviation should be minimized. (This magnitude depends, of course, upon the magnitude of the load change.) ,~ 3. The integral of the frequency error should be minimized. 1 4. The individual generators of the control area should divide the total load for optimum economy (Chap. 8).
Let us comment
1. Stability is always a problem in closed-loop control. The tighter the error specifications, the greater the risk-that the proposed loop win turn unstable. 2. The need for frequency constancy was discussed in Chap. 7. Isochronous (iso = equal, chronos = time) control gives assurance that the synchronous clocks run on time, but not without error (see point 3). 3. Isochronous control guarantees that the static frequency error following a step load change will vanish. No control system can, however, eliminate the transient frequency error (see Fig. 9-12 for a demonstration of this point.) The time error of synchronous clocks is proportional to the integral of the frequency error. This integral has the dimension of" cycles." If we divide the integral by I" we obtain an expression for time error. !:it expressed in seconds, i.e., 1 r s !:it = fO !:if dt (9-52)
fo
The particular control strategy we shall propose below does not have the ability to reduce the error integral automatically to arbitrarily small values. During the periods of heavy load, the system frequency will have a tendency to be below 60.00 Hz, during periods of the light load above. For example, ass~me that the frequency averages 59.99 Hz during a prolonged heavy-load period. In one hour we would now accumulate a time error of !:it
= 60 fo
3600
36
The customary way to handle such a slowly accumulating the U nited States) is as follows. .
1",
o ~ ~
1"l-1E ENERGY SYSTEM. IN STEJ.!}Y STATE-THE CONTROL
.).
ph '':_".
(... B~ O
f (t),
H:j
4 -
-0.Q1
re~ponse of the ALFC loop including the reset action of the secondary
loop.
The error is measured by comparing the "system time" with that of the National Bureau of Standards. When an error of 3 s has accumulated (and, as exemplified, this may normally take many hours), the speed changer is offset intentionally an amount of, say,.O.02 Hz for a sufficiently long time for the time error to reduce to zero. In the case of a power pool, one usually designates a pool member to keep track of the time error and inform the other pool members regularly of the need to reduce' the error by a unison effort. 4. The first three specifications' are taken care of by a control system (see below) with a response time of the order of a few seconds or half a minute. When these three control requirements are met, one attends to the fourth economic 'requirement. This is usually being done by a slower economic dispatch con-' trol scheme, having response times of the order of a minute or longer. . (Sec. 9-3-11.)
B. Integral control By using the control strategy shown dashed in Fig. ~-10, we obtain an overall system that will meet performance specifications 1 to 2 above. We have.added to the primary ALFC loop.in Fig. 9~1O.s.()~calledintegral control': i.e., we let the speed changer be commanded by a signal obtained by .first. amplifying and then integrating the frequeacy error; i.e.,
-
'.
tlPref ~
.,
K[ J t:.fdt
(9 53)
0
.:.~~
~ :?
$
.>..~
/
.'>j
t!7
~
erNfiJ,(JY
s8 ff
, ,'the
[flni,t or K_I is "per-unit megawatt per hertz and second." f Fef exaI?ple, If t~e frequency drops by 1Hz (t:.J == -1 then . , salIs fer an Increase In power with the "call" . .) the Integrator -K[ pu MW/s. Note the negilti~e polarit of cthe .ll1creasmg at the init!al rate. of must be chosen so as to cause " iti ~ Integral controller. This polarity or "decrease," com~a~d ~ha P?SI ive re~uency er~or to give rise to a negative, area control error (ACE) '. e signal fed Into the Integrator is referred to as , I.e., , ' ACE ~
C
gain,
i.e. if
I dI:nhtcgral control will give rise to zero static frequency error following a step oa c ange, for the following physical reason. As long as an error remains, the integrator output will increase, causing the s~ed ch~nger to move. The integrator output, and thus the speed-changer position, attains a constant value Dilly when the frequency error has been reduced to
zero. The gain constant KI controls the rate of integration, and thus the speed of response of the loop. The integration is mostly performed in electronic integrators of the same type as found in analog computers. C Analysis of Loop Respons-e Here follows an analysis of'the proposed system, subject to a step load change. To avoid cumbersome numerical analysis, we shall as before neglect the turbine dynamics. In addition, we also make the assumption that the speed-changer action is instantaneous. This is not perfectly correct, since the device is electromechanical and will therefore have a nonzero response time. These approximations will make possible relatively simple analysis without distorting the essential features of the response. It is also worth mentioning that the errors we thus introduce into our analysis affect only the transient, not the static, response, (This was already demonstrated in Fig. 9-11.), From Eq, (9-53). we get, upon Laplace. transformation,
n
then'
where
t
corre
,
llP,ef(s)
K
= - _!_ 6./(s)
we il
(9-55)
SeC0,!
simu
After making use of the block diagram in Fig. 9-10 and Eq, (9-55) we get, after
some algebra,
cup", Figtl
(9~56) We obtain the time response !lJ(t) upon inverse transformation of this ex,pFessien. Since the response depends upon the poles of Eq. (9-56) we must first , turn 011f' attention to the second-order-denominator polynomial, whJ.s:hcan be w;nlttca
l.U 01
CI U: Fe
i +, Kpl'R
T"
+'K/K~
Tp'
,=
(s + 1 + Kp/R)'2
27;,
+ K/ Kp _
Tp
(1 + Kp/R)
2Tp'
2. PI. fe
s.
(9-57)
CONTROl. rRonLE~1
327
)r
)f ~y is ~ )
p
gam K/.
. Clearly, the nature of the poles depends upon the magni tude of the in tcgr a I
If
i.e., if
1 K/ > 4~Kp then we can write the denominator (s
2
+ RP
K)2
2
K,.
(9-58)
eril
polynomial
in the form
+ Q:)2 + w
where Ct. and w are both positive and real. The expression ~f t,f(s) now has a conjugate-complex pole pair in the stable s plane, and the time response t,f(t} therefbre contains damped oscillatory terms of the types e-ar sin wt and e-r cos
(VI
(" subcritical " gain setting) then we can write the polynomial in the form
(9-59)'
(s
+ P,)(s + /32)
where /31 and P2 are both positive and real. Equation (9-56) now has a real pole pair in the stable s plane, and the corresponding nonoscillatory time response t,f(t) contains terms of the types and In either of the above two cases, t,f(t) will thus approach zero, proving tha t we indeed have bolh stable and isochronous frequency control. Our first and second system requirements are thus met. In Fig. 9-12 we depict the actual simulated time responses for different values of the gain setting K/. The family of curves in Fig. 9-12a neglect. turbine dynamics, i.e., the case we just analyzed. Figure 9-12b includes the effect of turbine and hydraulic motors. We make the following general comments about integral type control: 1. If we use subcritical gain settings we obtain sluggish nonoscillatory response of the control loop, This means that the integral of 6.f(t), and thus the time error, will be relatively large. In a practical situation thi~ setting is most often used. The advantage is that the generator now will not unnecessarily" chase ,. rapid load fluctuations, causing equipment wear. 2. A careful study of the response graphs in Fig. 9-12 r~veals the following features: As the sudden 'load increase sets in, the frequency starts falling off at tile same exponential rate as for the system of Fig. 9-11. During these first instants
3./.X EI.ECTRIC
runon Y:
AN II"TI((lDl)C'I'lON
has not yet had time to IJO j:nto t~Cli()ll,. an,d the :r;l:.m the 1)!'!'I'II'll'y (\1 J..... loop which VIC (!Jf;r;WWC( PIC1':'(, viously in great detail. After a certain lime (lh,c short~r the _til: h,I;~)_1t;: integral gain Kil. the integral controller comes Into acuon and eventually 11ft.,
rhc inicnral controller . rcspOflse ~. IS d elermtne d by
u ,
y/(J~,
.
.'
.. ,It
3. The reader must realize that in order to keep the analysis (jf thCnyW tc1h"ra1J J controller simple. we have made several simplifying assumptions. v: C 13 a
summarize these: "y n. The hydraulic and turbine dynamics were neg lcctc d (Iiu l thie v/cte included in the simulated graphs in Fig. 9-12/J). 1>. The speed-changer response was assu rued ins tan tuneous. c. All nonlinearities in the equipment, such as dead zone, ctc., have been neglected. d. It has been tacitly assumed that the turbine can change its torque as fast as It is commanded by the speed changer to do so. In reality, there is a practical limit to the rate, expressed in megawatts per second, at which a steam generator can pick up load. We have neglected this rate limitation during the few seconds we are considering. e. We have assumed that the ACE is available as a continuous signal. In reality, the measurement of the frequency deviation t..j takes place discontinuously in sampled-data fashion. If the sampling rate is relatively high (compared with the fastest changes in the response of Fig. 9-12), then the above analysis gives good results. 9-3-11 Economic Dispatch Control
The integral controller just described results in a system that meets the first three of the specified control requirements. The fourth requirement, i.e., that pertaining to economic dispatch, can be met only by application of the optimal dispatch eql:lations,\ODE's (Eq. 8-15). .' The primary ALFC loop makes the initial coarse readjustment of frequency. By its actions the various generators in the control area track the shifting load and share it in proportion to their size. The speed of response is limited only by the natural time lags of the turbine and the system itself. Depending upon turbine type the primary loop responds in 2 to 20 seconds, typically. The secondary ALFC loop takes over the fine adjustment of the frequency by resetting, through integral action, the frequency error to zero ..This loop is considerably slower and goes into action only when the primary loop has done its job. Response time may be, of the order of one minute. B-oth of the above loops base their simple control decisions upon the frequency.error that can be measured locally at the power stations. They can thus be implemented locally in the plants. Economic dispatch control can be viewed as an additional tertiary control 10OY.p. As the control decisions in this loop are based upon the solutions of the OPE's it is necessary to incorporate a digital computer as part of this control )'c:m,p. ypiea,Hy this computer is located in an .. energy control center T which is
II
\
ier
f,'
sr$
.:l'"'....
'V
'Q
.W]
,~ s# /\~
~
~;,:,.q_j$~
!::. 'J)"
o ...
<,
~CtJ'tJOo~~'?
q;t;
fts
re
ttr linked to the various power plants via communication channels (micro~,~~ & ::S;s c.telephone, etc). Periodically, e.g., every five minutes, the com~uter is providtf ~ f:1'-l ~ ~"q;' with the megawatt settings in the power plants, These settings are compare\~ W ~ ~,.'f ~ with the optimal settings derived from a run of the type of "optimal dispatch\~ S:"," ~ ,'tJ <$" I ~ \",. '- .. program" discussed in Chap. 8,.lf the ~ctual settings are ofT ~i:O~ theoptimal \ R' ~ ~ ~ values the computer sends back instructions to the plants to readjust the mega~ {J.g ~t::;- !..'" . .. ,~i/r" ~ 0 ~ watt outputs accordingly. This readjustment takes ... Iace, 0 fl. p course, via th e spee d \ 1f ~. ~ changer. ' '~if. ~ & ~ ~ ~
STATE-THE
CONTROL PROBt
~ ~ :" ~
I t': $ (;~ .
G.&
(
J5 ~~~
O~ (
f) .
,,0 ~
:n
is
'a
SYSTEMS
a on
n
1-
h e
From a practical point of view, the problems of frequency control of interconnected areas, or power pools, are more important than those of isolated areas. Practically, all .power systems today are tie-l together with neighboring areas, and the problem of load-frequency control becomes a joint undertaking. Closely associated is the problem of controlling the power flows on the interties. Many advantages can be derived from 1'001 operation, and they can all be summarized in two words, mutual assistance We should at the outset state the following operating principles, which are basic:
1. Under normal operating conditions each pool member or control area should
strive to carry its own load, except such scheduled portions of the other members' loads as have been mutually agreed upon. 2. Each control area must agree 'upon adopting regulating and control strategies and equipment that are mutually beneficial under both normal and abnormal situations. The advantages of belonging tl) a pool are particularly evident under emergency conditions. Let us consider some of the advantages referred to in point 2. First, consider the effect of size. We have: seen in the preceding examples in this chapter how the frequency, following.a sudden load change, will sag as a result of the fact that, during the first moments following the load increase, the needed energy is being "borrowed" from the kinetic energy of the system. Clearly, the larger the system, the more kinetic energy it possesses, and therefore the. more energy can be temporarily borrowed for a 'given speed drop. For example, the large interconnected group that covers the central and eastern parts of the United States and some Canadian provinces can absorb a sudden load change of about 3000 MW and experience a frequency deviation of less than 0.1 Hz. A small system of, say, 1000-MW capacity, if it suddenly loses 300-MW capacity (which represents 30 percent), .and if operating alone, may be 1R dire trouble. Its frequency will undergo an extensive drop, and chances are that the transient power angle swings that would ensue (Chap. 12) W0u!cl tear the whele system apart, Fesulting in a complete blackout. . .
~. ~ 'if ',...
c;
? ()
~
~
0 ..-l
'\
/'" -
i'1.lea""'L;.,--,-_.
-
__ --,
~
-uu
mrroduction
_--=-- __ '
~~
'"
~~~<fl()<;{p
'" (.>
Jf ~ ~
--z_
; ~ -<"'~Q.-~~
'2..---'v-.... ~
." 0
r" /
I-
w/ ,>
,..-
o,6c1f1.
t$)pi'.
/ ..
ic N
RGY SYSTEMS
Ti-IEORY;
AN
INTRODUCTION
I same system
's. ~ ~ ~ g ~ ~ /
;. ~/~ ~ ~ ~.,~/
were part of a pool of say 100 000 MW . f~lure. would repres~nt ~nly ~ 03 perce catpalcltY'Tthe (l ~ .... ?' ~ ~./ J1lf: Id b" d '1 d. ' n oss. he <fl ~ d .... ("> C<;/JP ency WOU e save, an support power to the extent 0[300 MW . --0 00 ';.> o o, frcqu 1fl . he cri '. would %.. ~ e-?' jflS!antaneous y. 0V: Into t e crippled area via the tie-lines to carry its load until normal generation IS restored.
l('l~J-MW- eneration g
System size also reduces the need for reserve power among' the pool members. A system that op~r~tes alone as a "single area" must provide power to handle not only the anticipated peaks, but also the sudden requirements due to equipment failures. The reserves are classified on the basis of their readiness. " Spinning" reserves can be called upon for instantaneous (at least as fast as the controllers can react) assistance. They are made up of fully operating but only partially loaded units. Hydroelectric-, diesel-, and gas-turbine-operated units can be called upon at a Jew minutes' notice. Steam reserves range all the-way from "low bank" (i.e., idling at pressure and heat levels below that required for service) to "cold .reserves " (operative, but not in operation): . Since peak demands occur at various hours of th~ day In vanous areas, the ratio between peak and average load for a large pool IS smaller than that of the individual systems. It is obvious, therefore, that all pool members can benefit from a reduced need of reserve capacity by a scheduled arrangement of energy interchange.
the Tne-JLiII,e .
.'
THE ENERG
CO,NTReJL
PReJB,LEM
'3;31 ,
Wl'l~,ne (;/1
, 1 The order 0 . and 02 are the angles of end voltages J/1 an d V2 respective y, direction "in ()'fr tlii'e sum scripts indicates that the tie-line power is defined positive .. ...
so
the amoant
In
, " [he angles 01 and O2 the tie -1'Ine , ower changes with p. (c5? _ (5~)(tlc51 -l'lc)l). MW ,(9-61)
6P!2
machines Analogous to the concept of "electric , su 'rr "f 0 synchronous of a line nes,s, ffici t " (compare Eq. (4-33)) we now de~ne the "synchronizing coe cien
-r<l_A 1 - "'"
-'---'-':::-::-\)(
1V? 11111
cos (c5? _ t5 0) 2
MW/rad
(9-62)
The frequency
formula
deviation
61=211:
or inversely
1 dId -W+6c5)=dt
211:
r
Hz
(9-64)
110
= 2n: J 61de
rad
By expressing tie-line power deviations in terms of Co.! -rather than Co.fJ we thus get
MW
Laplaee transformation of the,l~st formula yields
211:YO .
(9-66)
APl~(S)
'<i!lia:gram in Fi:g, 9-13,
p.Gwe-r
j,t\}
,~~ I
= :- Mi"u
.MW
If
I!v !
,.:IP1?
representation
or
We shall find it convenient when we work with multiarea systems always to consider the tie-line power positive in the direction out from the area in question. We remember that the sum of the right-hand power terms in Eq, (9-40) is represented by the right summing junction of Fig. 9-10. If we thus observe the above tie-line sign rule we can interconnect two single-area diagrams by means of the tie-line symbol in Fig. 9:13. The end result is the block diagram depicted in Fig. 9-14. In reading this diagram the following clarifying poin ts are in order:
.:1f 1 (s)
r - -. I
- - -----~------------------.
+--
r....Ll
18; 1
L~J
/'L')---+
\
+i ..J..
r--l
K/,
i
+
rof.'
+1
L1P'2(Slj
cr.,---,--l>-LS--l>-
__
.J
II P
I
I I
ACE,
= LlP'2
+ 8,.1(,
.dP,2Is)
04-
\.
~----------------------I
r..Ll
I -/ I
L.,-J
.IP
-
() 1 I
21 S ~ 1
ACE2
+J..
/"",'\~__j'
LlP21
K/2 ~
82M2
rof.2
r---,'
Ll"P
i
\ ... ,-. ...:..:...___J _
"'" L.. -I
+1 r..L-, I 82 I
LTJ
t ---
----O--------
t
THE ENERGY SYSTEM IN STEADY STATE-THE
\u~
..... ....-~
0~
~ ,
CONTROL PROBu:.~~'
i \. ~ ~ \1)
'&...,~ (5
'!:J'<.J (
portions
should
be
disregarded
for
(~~
~.
r!
i$
(A~
V ~O
....,~
iP;' ~
,,0 '0.-";
eo ;S>"'<J~
'2. We r~rne~ber that .the powers i~ the single-area diagram (Fig. 9-1~) welrf~vc.,.l~f:'~O expressed in per unit of area rating. The parameters R, ~, arid H, likewise -.,Vl s; ~ were based on the same base power. When two or several areas, generally of ;It" different ratings, are involved, we must refer all powers and/parameters to the ~ one chosen base power. I 3. The added blocks of transfer function -1 follow from Eq. {9-68}.
Before we proceed with an analysis of the two-area system" we shall present, without proof, a mechanical system that is a perfect analog of our electrical two-area system. Consider the train in Fig. 9-15. If this traiq is traveling at a constant speed VO and ,suddenly is subjec t to incremental load changes, the velocity will change. The two car-assemblies will experience the velocity perturbations t..Vl and t..V2, respectively, and the t.ie-spring power will change with the amount t..P'ic' By writing the incremental dynamic equations for this system, one can readily confirm that they are of the identical form as our electrical system equations. This means that the systems are analogs of each other. The following variables correspond to each other :
t..Vl ~
!jfl
The train analog can thus be used as an aid in achieving the behavior of the electric system.
,
l
!
~.
Analog of area 1
... ,::: .:if(.,':;
Analog of area 2
~~.;"'
.
~.
~
. .
\
I . . :,;,: ~
";~:::;o.
~(~~~.~~~~~
2.
[.
~~ ~"~
.,0+.::1".1 .
~~~
!:
"O"'~U2 ---
"
"i 'Ii
of two-area system.
',-.'0
iB _
~~
al~'
o - 9 . ~
~ ~
j;' -I:v"/
6NKGY SYSTEMS THE9 RY: /'' AN
J~ "....p
~ ~
g:::,
/
~.It':[(._'
::/<
C"JQ
/"
INTRODUCTION
'~~'
,(1)
~. ~
(//~~;e
assl:H11C
increased
by the constant
@ '::;~JcreIl1ental steps
/,(' Il
tl.PD1 = MI and 6PD2 = M2 We shall presently limit our ysis to finding the static changes that result in frequency and tie-line power. rlaJ
case is
,/
Let us call those changest 6Jo and 6P 12. 0 respectively. Since the incremental increase in turbine dynamics in this static determined by the static loop gainsj we obtain from Fig. 9-14
(9-69)
1
DPn.o= By adding the powers -
R2
t.Jo
we obtain
at the summing
junctions
(9-70) 1 - R ~fo - Mz = Dz ~fo z We solve for ~fo and ~Pu. () and obtain !'J.fo = _ Ml ~PI2.0
+ Ml 132M
I
131 + 132
131M
!'J.P12.o=-!'J.Pu.o=
where, in analogy
2 -
Hz (9-71 : pu MW
/31+/32
of each area:
1 RI 1 (9-72)
D2 + Rl
I
I
I
I I
l I
/31=fJ2=/3
.t In ~tead~"~tate the frequency dr~'ps in the two areas will be equal (compare graphs in Fig. 9-1'6'). ,~''Obtained letting s .... o.
," "
Ib~
I,
'!
THE ENERGY
CONTROL
PROULDI
335
We then gel
No=
6.Pt2.0
=
Hz -6.P210
M2-M1
=
(9-73 )
pu MW
Hz
(9-74)
pu MW These two last equations operation: tell us, in a nutshell, the advantages of pool
1. Fifty percent of the added load in area 2 will be supplied by area 1 via the tie-line. 2. The frequency drop will be only half that which would be experienced if the areas were operating alone (compare Eq. (9-49)).
Example 9-$ A 2-GW control area (I) is interconnected with a 10-GW area (2). The 2-GW area has the system parameters given in Example 9-5, i.e.. R
2~40 Hz/pu MW
X
D = 8.33
10-1 pu MW/Hz
Area 2 has the same parameters, bur in rerms of the 1O-G W base. A 20-MW load increase takes place in area 1. Find static frequency power change .. drop and tic-line
SoLUTION
I
I I
I
We will choose the generator capacity of area I, 2000 MW, as our power base. In EJtample 9-6 we had already computed PI = 0.425 Hz/pu MW in terms of this base. p, has the same numerical value based on I()"GW. Based on 2-GW it must be five times larger numerically, i.e. p, = SPI ;". 2.125 Hz/pu MW Also 20 MI = 2000 =O.QI pu MW
Equations
No
tlP
12.0
= 0.00392
= -0.00833
Hz
I I
I
om
pu MW (or -16.7 MW)
. Note that the frequency drop is now only one sixth of that experienced by area I running alone (co~pare Example 9-6). Note also that this "frequency suppcrt " is accemplished by an added delivery of 16.7 MW from the larger area.
........ v
LI.I.\..
I r..1\..
r.."'CIU .... 1 1
.y't~Il:MS
!"I-IEORY: AN lNTRODUCTION
9-4-6 Dynamic
Response of Two-Area
System
model that we have used, the two-area system be meaningless, in our text, to attempt a direct analytic approach for finding the dynamic response of :he system. We perform instead an approximate analysis based upon the following assumptions:
I, Consider the case of two equal areas, } Consider the turbine controller fast relative to the inertia part of the systems. i.e .. we set Gil = GT = I. 3. Neglect the system damping, This means that we assume the load not (0 vary wi til frequency; i.e.. we set D 1 = D 2 = 0, This means, in accordance wi th Eqs. (9-42) to (9-44), that G pi (s) = G p2(S) -+ fO js2H.
Under these highly simplifying assumptions we can readily derive the following expression for the tie-line power from Fig. 9-14
nfoyo llPi2(S)
This expression 1. The denominator
=~
S2
llPD2(S)
facts:
llPD1(S)
+ (f0j2RH)s
+ 2nfoYOjH
(9-75)
+ 20:s + w2 = (s'+
0:)2
+ w2
0:2
where 0: and w2 are both positive; we know that the system is stabieresv: damped. 2, Following a disturbance, the system will oscillate at the damped angular frequency (9-76) 3. The system damping is strongly, dependent upon the 0: parameter. Since fO and H are essentially constants, the damping will be a function of the R parameters, Low R values will give strong damping; high R values, weak damping, The system will perform undamped oscillations of frequency Wo = w if R = co, tha t is, if the speed governor is nonexistent. 4, The fact that the system _is inherently oscillatory could have been immediately predicted from the mechanical analogy in Fig. 9-15.
Example 9-9 Consider two equal areas, each having the parameters
R = 3,0 Hzjpu MW
If =
I"
5 s
= 60 Hz
THE
ENERGY
SYSTEM
IN STEAfilY
Sl'ATE~TME
:,
"\
I'"
TH1 . :~ ,
I
TH2
= 0
R1
VI
R2
= 2.4 Hz/pu MW
Tn = Tn
Figure 9-16 Dynamic response of two-area system sublect to a step-load increase in area
I
2.
r/s
Assume the tie-line has a capacity of 0.1 pu (10 p er cent of area capacity) and is operating at a power angle of 45. From Eq. (9-62) we thus have T" The oscillating frequency is thus
COo
= 0.1 cos 45
(.0707
= .j75.3
rls
The graphs in Fig. 9-16 depict the tie-line power a id frequency oscillations (in both areas) for a step load increase in area 2. (The dashed pc rtions of the graphs are explained in Sec. 9-4-8.)
i,
iIi
';;j
It
tTl
.,j ;0'
r"e"quehCY reset and area 2 take care of the tie-line power. We would thus a-rr~nge for the following area control errors: .
U'
ACE1
Afl
(9-77)
h These AC:E's would be: fed via slow integrators on to the respective speed c angers. This arrangement would work-but not too good. Actually in the early days of pool opers.tion one area was designated to reset the system' frequenc,~ and the others would be responsible for zeroing their own" net interchanges (see also Sec. 9.4-9). The problem with this arrangement proved to be tha.t the central frequency controlling station tended to regulate for everybody trying to absorb everybod y else's errors and offsets. As a result it would swing wildly ,between its generating limits. ' As a result of the original work by Cohn" a control standard has developed that has been adopted by 100st operating systems. The control strategy is termed "tie-line bias coritrol " and is based upon the principle that all operating pool members must contribute jr.eir share to frequency control in addition Cotakinq care of their own net interchanq.:
9-4-8 Tie-Line Bias Con trol of Two-Area System In applying this reset control method to our two-area system we would ,add the dashed loops shown in Fi:~. 9-14. Th~ c~ntrol error for each area consists of a linear combination of frequency and tie-line error:
A CE ACE2
I ~ ~
ilP i2 ilP21
+ B 1 6fl + B26f2
(9-78)
The speed-changer
commands ilP,,,r,
1
= -K/1
= -KI2
I
,
"
"
ilP,ur,
I (ilP f (LlP
12
+ BINI) + B2Llf2)
dt (9-79) dt
21
' nts K and tc., are integrator gains, and the.con~tants s, and B2 are Th e cons a." r1 . '. b . 1d d . thefreql ency bias parameters. The mmu~ SIgnS must e inc u e: sl~ce.each area should ~nc~ease its. generation if either ItS frequency error or ItS tie-line power increment IS negative.
It
iI
1,
I,
.. <>
"
t ':;
I I I
THE ENERGY
SYSTEM IN STEAI)Y
STATE--THE
CONTROl.
!"WIlLEM
),39
. I . id changer commands all equilibrium exists, can be achieved only alter I ic spec - b: I . te zrands in . evidentI y rcqurrcscs uia- ot 1 111 . . thai eg ,. have reached COl1stanl values. But this Eq. (9-79) be zero; i.e., 6P'2.0 6P21.0
+ B, + B2
(9-XO)
6PzI.o
= II
(Y-X I)
e
y g d d
:JI
-e
le
'allies l nfact, 0/1(:' oflhe Note that this result is independent of the B ,(Ill d B 2 (,;. II t b e zero, and lVe still have a gllarantee 10 bias parameters (but 110t both ) can . Eq. (9-8I~ is satisfied. t has been The question what "best" value to choose for the B para me ers d hotly debated. Cohn has shown that choosing B = f3 (i.e., the A ~R C) pro uc~s . t d system satisfactory over-all performance of the mterconnec eo svsvern. The Irltegralor gamI constants K" and KI2 are not critical-but they must be chosen 'small :1~OUg 1 not to stimulate the area generators to "chase" load offsets of short duration. The actual effect on the frequency and tie-line power graphs of the. added tie-line bias control is shown in Fig. 9-16 (dashed parts of graphs). Following the immediate excursions which are entirely determined by the pnmary sp~edgovernor loops of each area, the secondary integrator loops of each area go mto action and reset both the frequency and tie-line power back to original values.
8)
Systems
19)
ire 'ea
fer
:1
In reality a control area is interconnected not with one tie-line to one neighboring area but with several tie-lines to neighboring control areas, all part of the overall power pool. Consider the ith control area. Its net interchange equals the sum of the megawatts on all m outgoing tie-lines. As the area control error ACE; ought to be reflective of the total exchange of power it should thus be chosen of the form ACE, =
I !
I ilPll + BI 6;;
j=1
(9-82)
'ia-
sch
Typically, the reset control is implemented by sampled-data techniques. At sampling intervals of, say, one second, all tie-line power data are fed into the central energy control center where they arc added and compared with predetermined contracted interchange megawatts. In this way is obtained the sum-error of Eq. (9-82). This error is added to the biased frequency error and the ACE results. The ACE is communicated with all area generators' that are participating in the secondary ALFC If optimum dispatch is employed, a tertiary slower "OD loop" is added of the type discussed in Sec. 9-3-11.
34(}
FIICTRW
AN INTRODUCTION
',i
9-5 .. STEADY-STATE"
INSTABILITIES
The control methods outlined above-or variants thereof-are incorporated as standard in most of the interconnected power systems around the world. The operating experience has been good-exceptionally good in view. of the vast geographical spread and different generator mix characterizing many of the svstcms. . However. on occasion. u power system may experience stability problems usuull y in the f0f111of self-excited low-frequency oscillations. Power engineers arc rcfcrr ing to these as "steady-state" instabilities as they occur, mostly unexpected, as the system is running in what appears as a smooth steady state. ligure 9-17 depicts ,: graph of the frequency of the peninsular Florida grid. It is operating normally at 60 Hz with a slight unavoidable" noise" (which is always characteristic of the ALFC loop action) when, gradually, an oscillation starts to build of frequency typically less than 1 Hz. The oscillation is detected not only in the system frequency but also in the tie-line which will experience pow.er "sloshings " of increasing magnitude. This type of instability is actually caused by cross-coupling between the AVR and ALFC loops. The phenomenon has been detected in many parts of the world. It is directly associated with the so-called natural oscillatory modes of the network and it is therefore fitting that we start this section with a discussion of the latter.
:j
li
fi .~
f.
Hzj
I
Normal operation
Oscillatory
buildup
60.2
60.1
600~
20
I 30 in inter-connected
I 4.0
50
60
oscillations
power system.
THE ENERGY
SYSTEM IN STEADY
ST,\TI:-THI!
CO~TI1.0L
PROBLE~\
"-
,
\
I
I
I /
, , ,
----
5
,te) . I '11 tory system' (a) identical masses andshaft Figure 9-1S Torslona OSCI a . mode; (c) higher natural mode. elasticities;
rrr.~ ..
,;' ~
..
'~.?te'flil IS ~~e
.<,
so
1,
a o1
!il:J!JQr}
release:
position setti 0 mgs. For exa"'" I 'r'" = - f>3 and f>o == 0, '. . "'~~' I InItially the 2 , It WIll oscIlJat In the lower mode
"
"
.9... 5
i:
2. By some oscillatory Source that is " '" will arise if the SOUlce frequency pu~pmg. the system. Natural Qscillatiofls ~'. Sudden major fault ..induced to ,col~clbels with one of the natural frequl'Icies rque irn a ances. . . The natural oscillations can tak Rig. 9-18, or around a reference l e place around a fixed reference, as in case is analogous to what h sy~tem rotating at a constant speed. The latter here the enerator+ r a~pens a power system. The individual masses are "shaft" g h I ('t.or~ (I~c)udmg the fixed turbine rotors). The elastic s zle t e transnussior, lines of the grid. The position coordinates are the power ang es 15 [compare Example 9-9). If the system conla ins N generators then there exists, in the strictest sense, N - 1 natur~l modes. However, many subgroups of generators, particularly those belon~m~ ~o th.I,~ same power stations, are swinging as coherent groups ra ther than individually. Their close proximity makes their connecting "electric shafts" very stiff. These natural frequencies typically lie in the range 0.1-5 Hz. When one of them IS b eing excited the power system is said to perform intermachine oscillations. Every individual generator rotor including its directly coupled turbine rotors ,:and excitor rotor, if present) participate in the oscillations as one unit. The o scillations shown in Fig. 9-17 are of this type. For a turbogenera tor unit of very large rating the shaft is very long and generator rotor, exciter rotor, and the often three to four different turbine rotorst constitute a rotational system which by itself possesses maybe half a dozen natural modes. 'I he eigenfrequencies of these modes typically lie in the freq)J~ncy range .20-50 Hz. ?ne of these mo~es can on occ~sio~ ~lso b~ excited, partidul'arly when the.f.:lfc~nc network con-tams very ~ong lines WIth series capacitors! Such an instalinitt:1 IS referred to as subharmonic resonance. is important to realize that both the low-frequency" interrnachine " and the high-frequency "subharm.oni~" oscillations ~re po~rly .dampe~. This means that a\ relatively weak" pumpmg source can excite oscillations which may grew . to amplitudes of destructive magnitudes causing the" elastic shafts" to break . . .tnl the" subharmonic" phenomenon the elastic shaft consists of the genera.1>0r'shrf,t itself. In a elasslc accident this shaft broke causing lOO-millioN-dollar
Qp
It i1
SySI
sels gen
Thi
I?
ben
the det
J
j
ger pel
(E!
~ ~
foli 1. 2.
p~
be
Itl
.~
~..
da'ma9~' ' Itl:\the "iNtefmachine" phenomenon th~ "~lasti~ shafts". are the transmission lk.J.es of,the ~itl. They "~r.eak" when the tle~lme power swings become S0 large that '1'h6p.e:we.l1 ang.le 0 g'1e.s beyond. 90 electrical degrees cal!sing loss of synchro~""In,ism a~fd"island" feFIDit10n.:J; .
t, ,r.l.i:g!\, low" and' rntermec d I~t.! pressure sec t' rons.
~: WOfn, periinsill,ar' flor.ida e~pe.r.ienCes the s~ihgs. depicted in Fi.g. ~-17, and if the oscillation 'lfi.oi'I{luB 'Gannet DC steppe<\; the .e.nd.result usually IS a, ~reakup of-the tie-hiles .to -Georgia, The wJiele .~tlinSU'11 gnq ',wHl th"eFetip-en;Am isolated, One talks- In sueh cases .apont "'int~rar.ca" esciJ.Iat,ion's.
I."
i. .
", ,I,'
,
CONTROL PIWBLI;M
343
9-5-2
Operating s
Generator
It is instructive
e
c
y s c
to study the factors that determine the natural modes in a power system. Because the interrnachine oscillation is the most common one we shall select it for our study. As our" textbook case" we shall choose the simplest of all ge~er.ator configurations-a single machine operating onto an .. infinite" bus. This IS analogous to the case of two rotational masses (Fig. 9-18) one of which being of infinite size. With N = 2 the system possesses only nllC' natural mode. the characteristics of which we now explore. The single-generator case was extensively studied in Sec. 4-12 where we derived static characteristics. In Sec. 4-13 we derived formulas for the cransienr generator power which will now prove useful. In the study of .. area" frequency dynamics in Sec. 9-3-5 we derived the pertinent differential equation (Eq. 9-38) from a .power-balance" expression (Eq. 9-34). We use the same approach now but need to slightly modify it in the following regard: 1. We study the power balance of one rotating unit-not an entire "lumped" area. 2. As we shall later use our derived equations for large-scale dynamics (Chap. 12) we do not limit.our derivations to incremental excursions. As the differences between the turbine power PT and the generator bus power P G is used for increasing the kinetic energy of the unit we obtain the power balance equation .
PT
PG
di(Wkin)
MW
(9-83 )
that
Note that the last term in Eq, (9-34) is nonexistent here, as we assume no local load is served from the generator.r Following the same steps as in Sec. 9-3-5 we transform Eq. (9-83) Co PT-PG=-
2H df
fO
dt
pu MW
(9-84)
We shall find it more useful to let power angle, lJ, rather than frequency.j, be our independent variable. In view of Eq. (9-65) we thus get
pu MW
(9-85)
SM'
ELECTRIC
AN INTRODUCTION
This is the famous genera tor swing equation (GSE) that plays an extremely irnportant role in all types of studies of power-system dynamic phenomena, Ifwe want to use the GSE to study incremental dynamics around an operatingpoint (P~, Pg, bO) then we readily prove it to be of the form . H I1P T
-
.'0
I1P G = 1[/0
d2 dt2 (116)
(9-86)
SoLUTION
Seep
1 Find the steady-state operating point. This was done in Example 4-3 for the power level P~ = P~ = 0.667 pu, an excitation level EO = 1.22 pu and with the network voltage held at I I = 1.0 pu. It was found that the steady-state power angle equals
v:,
(9-88) or pu MW where S' is the" transient stiffness" of the generator. Compare component (Eq. 9-89) is referred to as synchronizing power. From Eq. (4-83) we obtain (9-89) also Eq. (4-33). The-power
dP ) 0 S = ( = 2.398 cos
dO;
o~-
=
Step 4 Write the incremental (9-86) yields
. (9-90)
of (9-89) into
t!.Pr = O. Substitution
(9-91)
or shorter
(992),
CONTROL
PROBLEM:'
where Step 5 Solve the GSE. The solution of the second-order of the form
linear, constant-parameter
differential
equation
(9-92) is
(9-94)
The integration constants the oscillations are started, In Our numerical case A and", depend upon the initial conditions, i.e., upon how
WI
J11
corresponding
to
a .. natural"
frequency
It is interesring to note that increased inertia lowers the natural frequency. Increased stiffness (caused by lower generator and line reactances) will increase the synchronizing power and thus the frequency.
Equation (9-94) implies that the inertial oscillations are undamped. In actuality, the oscillations in all probability would be damped and would ~ctually vanish after some time. The system evidently possesses positive damping.' This damping originates mainly in the rotor "damper" winding (Sec. 4-1). As the rotor performs incremental dynamics relative to the synchronously running "infinite" system its damper windings will no longe~ be stationary relative to the armature reaction flux wave. The relative rnotidn' will in fact induce currents in the damper bars and according to Lenz' lawl these 'current~ will have a direction such as to try to stop the relative motion. . The damper winding will in fact dissipate power and the dissipation rate is . ,. almost proportional to the relative velocity, d(~o)/dt. The presence of. the damper winding thus .adds a positive term on the right-hand sid~ of the power balance equation (9-85). The incremental form of the GSE will change from (9-B2) to
I
I
(9-95) The general solution of this equation is !iON = Ae-br where Typically,
COl
(9-96)
(9-n)
p b, and therefore
f
",
:l
;j
-<
e
ENERGY SYSTEI~'S THEORY: AN INTRODUCTION
!2
'l7
:>;j ~P
cr;</c
6riN <I-POIt
Stiffness
1. 1
di
3
Natural dampi~g
(tlbN) 1
S
66N
--l>
2. { 3. 1
---l>
iF
(dr)N) '-q-
9-5 1ft!
9 AVB
10
pre I
Figure 9-1~ A linearized AVR coupl,ing effects.
I
dynamic
the
In conclusion:' Under the influence of the damper winding, the generator is '" oscillat6rilY 'stable," meaning that if subject to any small disturbance it will return to its stebdy-state equilibrium by a damped' oscillation. This i" natural" damping of the system is further enhanced by the resistances of the network, which we have neglected in our analysis. It is instructive to vishalize the rotor dynamics by means of the boldface portiont of the block dia~am in Fig. 9-19. The blocks 1 through 4, plus the summing junction model the basic undamped." inertia dynamics of Eq. (9-92). The damper loop (block 5) takes into effect the natural damping of the system, as expressed by Eq. (9-95).
~ Ii ~ \\ ~ ~ ~ r,
I~
tiv
an in~ en gi
~ ~)
Ii
\
Ll,
9Ir.
SI.
a:
Qi
, III
performing a
dynamic:;
around
some operating
point
Llla~:
1. The response cClIwil;tl! flat of fine radian natural frequency (1)1' as in the above case, but N - I natural Ircqucncien (/)1> .,,' (I)/:_!, 2. Under the influence or damper windiJl&l and line resistances all N - I natural modes arc positively damped. 3. The system would thus: at; "o;;cillat0rily stable."
9-5-5 Negative
Damping
If the" natural" tendency for the Interconnected machines is to be " oscilla~orily stable" how then can phenomena of the type depicted in Fig. 9-17 be explained? Clearly, our above analysis must have neglected some important system ~echanism that will be present in real life. To understand this let us return for a minute to Eq. (9-95). Assume tha.t somehow the damping coefficient b becomes negative. In that case the factor e"!" will qrow with time rather than decay, and the inertial oscillations will be of the type in Fig. 9-17. Furthermore, the system will self-excite. i.e., the oscillations will be initiated by any infinitesimal disturbance that is always present. The damper winding and line resistances will, of course, never yield a negative b. It is entirely possible, however, that signals emanating from other sources will enter the loop at the summlnq junction in Fig. ~-19. These signals may, like the damper signal, be proportional to d(Llojjdt. If they, like the latter, are positive they will further help damp the inertial oscillations. If, on the contrary, the v are negative they will reduce the natural damping, possibly to the paine of eliminating it altogether. . It. is instructive t(: sh?w ho"~ sue? signals may be generated and how they enter into the dynamic prcture. r or trns reason we will in the next few sectio s giv~ a qualitative, step-by-step presentation of these added effects. n
2:1
f
,f
IS
(9-98 )
This Yi.clded the C'xpre;wion (9-89). In reality the emf ' will chance and thi
r., IS
's
to
I I I"
1'\'\
l .. '
1''''\.1
., 1 .... 11:.'1:'1
IIII:UK'I
. ,\:"
I;'\;
11(.JLHJCTIO:"i
Thus
transient
power
output
(9-100)
or shorter
sr; =
as rhc new input via block 6. 9-5-7 Factors
Causing Changes
S' t:.b,.,. + K,
t:.E'
(9-101)
where K I has the unit" per-unit megawatt per per-unit kilovolt." The added term will clearly be represented in our block diagram
in Fig. 9-19
in
E'
There
I. Manipulation (via AVR) of the field voltage 2. Changes in relative rotor position b,.,. .
"r
(
An increase in vf will clearly tend to ill crease E'. Not so clearly, an increase ill rot or angle tends to decrease E'. The latter statement is, however, easily confirmed from Fig. 4-25. As 0 increases, the voltage X~ Id and thus I d will increase. But since Id is demagnetizing (Fig. 4-25b) the end result is a decrease in the emf. The relationship between increments of t:.oN, tlvf, and t:.E' will thus read t:.E' = K2tlvf - K3M,v
(9-102)
The transfer blocks 8 and 9 plus the summing juncton account for the effects stated in Eq. (9-102). Equation (9-102) expresses the steady-state changes in E resulting from the changes in t:.vf and t:.oN.', Dynamical/y, the change in E' can take place only after S f and t:.o,~. have overcome the field winding time constant, the value of which (under load) was given by formula (9-20). The added delay transfer block 7 takes this field delay into consideration.
2::2
2
;
9-5-8 Inclusion
of the A V 1{ Loop 4.
At this stage the increment t:.v f in Eq. (9-102) must be accounted for. Returning to the A VR block diagram in Fig. 9-4 we obtain AUf directly as the output from the exciter. This results in the addition of block 10 plus the third summing' junction to the growing diagram of Fig. 9-19. . We finally are left with only t:.1 VI, the incremental change in the generator terminal voltage. t:.1 V consists (see Fig. 4-25) of the two components t:. V and tl Vd The former increases as E' increases, the latter as {J increases. Thus we \ave the incremental relationship:
2::3
5.
(9-103 )
6.
----~
1
349 .
.
THE ENERGY SYSTEM IN STEADY STATE-THE CONTROL PROBLEM
l)
In Fig. 9-19, we symbolize this equation with the summing junction 2:...and the transfer blocks 11 and 12. The block diagram is now complete with alLsignal branches, except .6.1 V Iref' fully accounted for. 9-5-9 Discussion of Results
The boldface portion of the diagram in Fig. 9-19 is readily understood. Under the influence of the synchronizing power the ~enerator rotor when disturbed tends to perform undamped natural oscillations. much like the spring-mass system of an auto when hitting a bump. The damping power emanating from the damper winding serves as the system "shock absorber." Note that the synchronizing power is in negative phase with the swing amplitude .6.bN, whereas the damping power is 90 out of phase. What can we make out of the lower portion of the diagram in Fig. 919? Without getting involved _in analytical details we summarize these observations:
1. The swing amplitude /:1tJN is fed back via a dual-path system, i.e., directly via blocks 6, 7, and 9 and indirectly via the AVR. This feedback can result in
2.
3.
4.
5.
6.
" parasitic" power components in phase with either the synchronizing power or the damping power. The actual phase is determined by the frequencydependent blocks 7 and 10. DeMello and' Concordia analyzed the actual nature of these parasitic power components in a classic papers and contributed quantitative data. Andersson and Fouad" applied Routh's stability criterion to find conditions for stability. The resulting formulas are too complex tobe included here. The parameters S', K!, ... , [(5 are partial-derivative expressions (compare Eq. (9-100)) and are therefore depending UPOIl the loading. DeM~llo and Concordia made the following comment: "It is important-to recognize that ... the parameters change with loading. Since they change- in 'rather complex manner, it is difficult to reach general conclusions based on parameter values for one operating point only." I \ . In some systems the combination of long lines, fast AVR's, and heavy loading can result in a situation where the parasitic feedback power arinuls the. natural damping thus making the system "'oscillatorily Jnstable" as demonstrated in Fig. 9-17. It should be realized that Fig. 9-19 in its complexity models only a single generator operating onto an infinite network. The results obtained from this diagram either by analysis or simulation are useful but may nbt be directly applicable to the conditions existing in an interconnected netw9rk. A recent EPRI sponsored project? has resulted in a computer program for simulation of intermachine oscillations in large systems. I When .a system experiences self-excited oscillations addition of so-called
'.
.,
j'W: EI:;ECTRIC
ENERGY
SYSTI!MS.THEORV:
AN INTRODUCTION
"powt'!r system stabiliz ers " (PSS) will quench the oscillations.V" A PSS injects stabilizing "supp: ernentary " signals to the summing junction in rig. 9-19. These signah typically are derived from the generator shaft speed and shaped in special networks.!? See also Prob. 9-14.
L3
Th~.rJder must realize by now that design of generator controllers as we know them tbday has followed "classical" lines. Root-locus methods, block-diagram representation, intuition, trial and error, empiricism are but a few of the "classi~al" design tools'. The high degree of reliability and general excellent functioning of today's power systems is proof of the soundness of design. In view of this fact is there really any room for improvements? , Tpb phenomenon of intermachine oscillations discussed in the previous section ddmonstrates that our present power systems are not immune to instabilities. also accentuates an inherent weakness of classical design-its inability fully tO .anticipate and adequately to cope with all dynamic features in systems of I high dimensionality. ' A inodem gigawatt generator with its multistage reheat turbine, including its ALFCjand AVR controllers, is characterized byan impressive complexity. When all its nonnegligible dynamics are taken into account, including cross-coupling I I ' between control channels, the overall dynamic model may be of 20th order. Th~s "dimensionality barrier" can be nicely overcome by means of computer-aided optimal control design methods originated by Kalman.Uv'? Yu; Fosha, and Elgerd1314 ,made ear~y attempts to apply these methods to power systems. A recent surveyl~ classifies the many contributions that have later followed. "Optimal I 'conho1" covers today a widening spectrum of computer, oriented techniques of which the so-called optimum linear regulator (OLR) design has proven particularly useful. Compactly stated, the OLR design results in a controller that minimizes both transient variable excursions and control efforts. In practical power systems terms , tbis means optimally damped oscillations with minimum wear and tear of CORtrol valves. OLR design proceeds as follows:
Itl
Stefl 1 Casting, the system dynamic model in state variable form and introduction of appropriate central forces ' , ' , Step 2 Choosing: ai:! }ntegi'al-squared-error control index, the minimization of which is the control goal Step 3 Hnding, the structure of the optimal controller that wiU minimize the chosen control index " We clemonstr~te
by applying ',i,t to the siaglein Fig'. ~-lQ. We propose to fiAIil a eon;tF@i felice I\(
THE ENERGY
SYSTEM IN STEADY
STATE-THr:
CONTROL
I'ROIlI.EM
351
(depicted in Fig. 9-20) which in an "optima}" manner steers or controls the system. We define "optimal" later; presently we develop an appropriate model for the system.
,
v
" ii
rl
~
II
."f If .s n
~
t1
1+ STH
I
- _!_ Llf(S))
R
g LlPT(s) = -1 -Llf(s)
+ sTT
K
P
(9-104)
:r ;r
r-
= -1
+ sT P
LlPD(s))
.n
:h ~s
n-
+ TH -d
(IlPv) = u - -
III
(9-105)
+ TT dt (LlPT) = sr;
Llf
.0f
and
X3
forming the
ble
= [ ~: .
XI]
[LlPvl
~;T
(9-106)
!e.U
(9-107)