Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

IEEE Transactions on Power Systems, Vol. 9, No.

2, May 1994

965

A POWER SENSITIVITY MODEL FOR ELECTROMECHANICAL OSCILLATION STUDIES

V. F. da Costa
UNICAMP-Campinas-Brazil
FEE - C.P. 6101

S. M. Deckmann

Abstract: This paper describes the derivation of a power sensitivity model for dynamic studies of power systems,
subjected to normal operation disturbances. The need of an
infinite bus representation is avoided with the linearized
nodal power balance approach. This permits the model to be
easily extended to any number of network buses.
In the linearized form, the resulting Power Sensitivity Model (PSM), presents some interesting features, such as
decoupled modelling and time scale decomposition properties. For presentation reasons, the PSM is first derived for
a single generator connected to an infinite bus. Its performance is then compared with the classical HeffronPhillips Model (HPM), as described by de Mello and Concordia. At last, it is extended for multinodal networks.

built into the derivation of coefficients K1 to

K6.

A T m $ / M; . s+D
F ] F d
K2
K3
-

AEq

1+ s K3Td 0

Ke
AEFD m

AVt

(K61

AVREF

Fig.la - De Mello & Concordias version of the HPM

Keywords: dynamic analysis; low- frequency; power sensitivity

vtLe

VOLO

INTRODUCTION
The study of low-frequency oscillations, which may
arise during normal operation, assumes capital importance
in power system control and system dynamic security analysis. It is well known that these kinds of oscillations are
mainly associated with electromechanical torque imbalances
at the synchronous generators, and produce tie-line oscillatory power exchanges. The major problem is that these low
frequency modes (in the range of 0.2 to 2Hz) are poorly
damped by natural means.
The possibility of using the voltage control loops, in
order to introduce artificial damping of the rotor speed
oscillations, was initially devised by Dandeno [11 and by
Schleif [21, in the early sixties. Effective results, obtained at field tests, led to the fast development of several types of Power System Stabilizers (PSS), which gained
great importance for power pool security during the last
two decades [3,5,6,7,101.
A simple linearized model, representing a synchronous
generator, connected to an infinite bus through an external
impedance, was described by Heffron and Phillips [41 in the
early fifties. The suitability of their model for the analysis of low-frequency oscillations was only fully recognized after the publication of the study made by de Mello and
Concordia [31 in the late sixties, where they explored the
important decomposition into the orthogonal synchronizing
( s A6) and damping ( s do) torque components.
The apparent simplicity and conciseness of their version of the Heffron-Phillips model (HPM), shown in Figla,
indeed hides a careful derivation
of each parameter, according to
Ref.[31. So it should not be surprising that
quite confident results were reported with this model,
specially under the presumed conditions.
The very satisfactory performance of the HPM has motivated others to seek a generalized multimachine model for
the analysis of low frequency oscillations in large interconnected networks [6,7]. However, there is an intrinsic
difficulty in simply extend the model, because the external
impedance connecting the generator to the infinite bus is

93 SM 516-5 PWRS A paper recommended and approved


by t h e IEEE Power System Engineering Committee of
t h e I E E E Power Engineering S o c i e t y f o r p r e s e n t a t i o n
a t t h e IEEE/PES 1993 Summer Meeting, Vancouver, B.C.,
Canada, J u l y 1 8 - 2 2 , 1 9 9 3 . Manuscript submitted Aug.
28, 1 9 9 2 ; made a v a i l a b l e f o r p r i n t i n g A p r i l 1 2 , 1993

PRINTED I N USA

Fig.lb

Generator connected to infinite bus

This paper describes a linear decomposition technique


[11,121, which overcomes those difficulties, and which can
easily be extended to large networks, including different
machine models, load characteristics and
allows a direct
representation of all control loops. The resulting model
presents several useful features, which make it an efficient simulation approach for dynamic studies of mid- and
long term responses by considering small impacts on normal
system operation. This has been an area of intensive research for many years [5,6,7,8,9,10,141.
THE NODAL POWER BALANCE
The basic assumption underlying the derivation of the
proposed model is the general principle that the active and
reactive power balances must be satisfied continously for
any bus during the dynamic process. As a consequence, the
local operating characteristics may be estimated by means
of Taylor decomposition in terms of the linear sensitivities between the variable changes. These characteristics
are used as the input/output relations for the disturbance
propagation inside the network and through control loops.
The resulting linearized model will be referred to as
Power Sensitivity Model (PSM). In order to show the basic
derivation steps and to contrast it with the HeffronPhillips Model (HPM), we will first derive it for the single machine connected to an infinite bus. Some common tests
are shown, in order to compare their relative performances;
thereafter, the extension to the multinodal PSM will be
presented. Also the inclusion of load characteristics, PSS
actions, and AFC control loops, will be briefly discussed.
THE POWER SENSITIVITY MODEL (PSM)
Consider a salient pole generator, connected to the
infinite bus, as shown in Fig. lb. The terminal power balance, to be satisfied at any desired time instant, may be
expressed by the equation pair (1) and (2):
Pg
Qg

- Pe = 0
- Qe = 0

(1)
(2)

where the subscripts g and e refer to generated and external powers, respectively.
Considering the transient conditions for this power
balance, and ignoring the internal and external losses in
order to simplify the derivation, these powers will be expressed as functions of the state variable pairs (E,&),
{vt,el and { V O , ~ )carrying
,
implicitly the time dependence:

0885-8950/94/$04.00 0 1993 IEEE

966
(3)
(4)

In order to solve (9) and (lo), we must aggregate independent equations for variables A6 and AEq'.
The additional equations will be the functions governing the rotor swing and generator flux balance, which are
represented in frequency domain, respectively, by:
Ad =

Xd'
-

(6)

where each variable can easily be interpreted by means of


the network shown in Fig. 1b.
LINEARIZATION
Applying Taylor linear expansion to equations (3 - 6)
for an initial steady state, the power balance equations
(1) and (2) may be expressed as the incremental balance
equations (7) and ( 8 ) :

[ APm - APg ]

s.(Ms + D)

AEq'=

Xd

1+

, Xd,

.TdO-

(,,,,

+ Kv.AVt

(13)

- KA.A(cS-I))

(14)

Xd

The flux balance equation (14) is derived in Appendix


11. The time solution of (13) and (14) requires a step by
step integration scheme, with the following inputs:
APg = Alg.A(6-8) + A2g.AEq' + A3g.AVt
AEFD = &e(AVREF

- AVt)

(15)
(16)

and:
A~~.A(~-B)+A~~.AE~'+A~~.AV~-A~~.A~-A~~.AV~-A~~.AVO=O
(7)
APm = mechanical input changes (AFC actions)
R ~ ~ . A ( ~ - B ) + R ~ ~ . A E ~ ' + R ~ ~ . A V ~ - R ~ ~ . A ~ - R ~ ~ . (A8V) ~ - R ~ ~ . A V O = O
AVREF = reference voltage changes (PSS actions)
where the coefficients IAg,Ae) and IRg,Re) represent the
local sensitivities of the Active and Reactive power functions (3 - 6 ) relative to the state variables. The relevant
coefficient expressions may be found in Appendix 1.
It should be noted that the reactive power equations
were first normalized (divided) by the local voltage level,
in order to favour the linearization process relative to
voltages. The incremental equations (7) and (8) represent
the intantaneous power balance that must be satisfied at
any particular time instant, after an imbalance or perturJation is imposed on the original equilibrium state.
Once there are no voltage changes at the infinite bus
(AVo=O), equations (7) and ( 8 ) will reduce to only two
state variable pairs: (AEq',A6), (AVt,AB). The terminal
variables IAVt,Ae) represent the fast algebraic interface
between the machine and the external network. The machine
internal variables IAEq',As) represent the slow varying
states, which are governed by the dynamic processes of flux
changing and rotor angle swings, respectively.
Each state variable pair presents an incremental orthogonality property in this complex polar form, as indicated
in Fig.2.

Equation (15) corresponds to the electric power imposed by the network on the swing equation (13), while (16)
represents a first order static excitation control system,
similar to that used for HPM analysis in Ref. [31.
In the flux balance equation (14) the effects of armature reactions are associated with both voltage magnitude
and angle difference changes, and are explicitly accounted
through coefficients KV and KA, in contrast to the HF'M,
which only presents the slow varying term K4.A6 (see further comments in Appendix 11). As will be shown, this
enables
direct simulation of the machine's response to
external disturbances, imposed from the network.
The mechanical power (APm) and reference voltage
(AVREF) adjustments will be considered as fixed independent
inputs. They are the natural entries of the AFC (Automatic
Frequency Control) and the PSS (Power System Stabilizer)
control actions, to be discussed later.
BLOCK DIAGRAM OF THE PSM
The complete PSM model, represented by equations (9) to
(16), may be interpreted by means of the block diagram at
Fig. 3. This diagram should be compared with the HPM of
Fig. l a . Both models represent a single generator connected
to an infinite bus as in Fig.lb.

Fig.2

Voltage and angle orthogonal increments

This orthogonality has implicitly been explored before


in the derivation of the Fast Decoupled Load Flow [131, and
is used here to decompose the power balance equations (7)
and ( 8 ) in order to obtain the network solution for angle
and voltage changes. This is achieved by rearranging the
active power equation (7) in order to separate all Ae terms
and the reactive power equation ( 8 ) to separate the AVt
terms. After this arrangement, we obtain (for AVo=O):
Alg.A6 + A2g.AEq' + (A3g-A2e).AVt = (Alp + Ale).AlB
Rlg.(A6-A8) + R2g.AEq'

Rle.AB

= (R2e - R3g).AVt

(9)
(10)

The LHS of equations (9), (10) is interpreted as ter-

minal active and reactive power requirements to be satisfi-

~~

actiue +-L
aeactiue

ed at any time instant, and will be expressed as:


Fig.3 -The Power Sensitivity Model for system of Fig.lb
AP

A(%)&
Vt

Alg.A8 + A2g.AEq' + (A3g-A2e).AVt

(11)

Rlg.(Aa-Ae) + R2g.AEq' - Rle.AB

(12)

Tables I, I1 and 111 present data and parameters of


both models, obtained from the same initial conditions t31.

967
TABLE I

after each of the four subsystems is solved once.


Each time step solution reflects the exact matching of
the coordination equation pair (7) and (81, relative to the
original disturbance. Thus after n time steps (nAt) the
resulting main block outputs are interpreted as follows:

Basic Common Data

TABLE II - HPM Parameters


K 2 I 1/K3( K 4 I
K5 I K6
0 . 9 5 1 1 . 0 7 1 2.781 1 . 3 7 1 0.05441 0 . 4 3 9
K1

A6(nAt), A9(nAt)

TABLE 111 - PSM Sensitivities

I
I

represent the angular displacements


of internal and terminal
voltages
from the -6
angular position.

AEq(nAt),AV(nAt)- represent the internal and terminal


voltage magnitude deviations from
initiae voltage values.

Alg I A29 I A39 IAlel A2e 1 Rlg I R2g 1 R3g


1.5751 1.916 1-0.701 I 2.510.4921 0.7011 2.469 1-2.192
R l e I R2e
0.4921 2 . 5

Kv I K A
3.161 2 . 4 5

Comparing Fig.la and Fig.3 we notice that both models


present some similar blocks. However, there are important
differences to be considered. At first, two important decomposition features of the PSM should be pointed out:
The Active-Reactive Power Decomposition

This orthogonal decomposition becomes apparent by


noting the variable exchange between the left and the right
subsystems, as delimited by the vertical interface line in
Fig.3. While the left side a d h e power model produces
angle corrections (A6,A9), obtained from active power requirements, the right side tmdime model responds with the
quadrature voltage magnitude corrections (AEq,AVt), resulting from reactive balance requirements. As a matter of
fact, an alternating decoupled solution scheme for (AP-Ae)
and (AQ/V-AV) subsystems is preserved, not only for the
static, but also for the dynamic power matching, along with
the time simulation.
The Slow-Fast Time Frame Decompositton

This means that the time evolution of each bus angle


reproduces the local angular oscillation mode, relative to
the initial
position, and that the angle difference
[Aek(t)-Ael(t)] represents the oscillation mode between
buses k and I. Similarly, we conclude that the changes of
each voltage magnitude reproduce the amplitude modulation
resulting from the initial perturbation.
The automatic decomposition into the local PowerAngle/Voltage oscillation modes, produced by the PSM, is of
prime importance in power system stabilizing studies, simplifying the application of the synchronizing-damping torque concepts. These concepts may be naturally extended to
multimachine, multibus networks, as shown later.
COMPARATIVE PERFORMANCE

In order to show their relative performance, the PSM


and the HPM were tested for the same basic conditions. The
Fig.5
test parameters are given in Tables I, I1 and 111.
shows the curves obtained for a APm = -10% step.
o.m.,
w.r&

A time frame decomposition of the model becomes apparent from the variable exchange between upper and lower
subsystems, as indicated by the horizontal interface line
In Fig.3. While the lower c@ehaic subsystem represents
the east network state variables (AB,AVt), the upper
&edial subsystem produces the &up
internal machine state variable corrections (A6,AEq). This time frame decomposition is very convenient for the step solution of dynamic
problems in power systems, as pointed out in Ref.[8,91.

AE;I

wo.m
a)

HPM

tine

THE SOLUTION SCHEME

m.m

( 5 4

APm=-10%
.

Ke=25
.

Considering both decomposition interfaces, we can view


the PSM as four coupled subsystems, as shown in Fig.4:

APac - A6

I-

AP

slow
fast
A9

active

*.mal

reactive

Fig.4 - The PSM Main Subsystems

APac = APm

APg

(17)

- KA.A(6-8)

.
b)

The upper blocks net inputs represent the accelerating


power APac, and the exciter control combined with the armature reaction AEQ,, as represented in Fig.3 :
AEQ,= AEFD + KV.AVt

o.m

(18)

In order to reduce memory requirements (not critical


in this case), a decoupled solution scheme will be given
preference. With decoupled processing it is easier to include local control non-linearities, such as limits and
function rate changes, without destroying the nodal matching properties of the model.
The general decoupled simulation scheme consists of
performing a complete solution cycle for every time step,
in order to obtain the input-output response of each subsystem, solving first the faster (algebraic) subsystems and
proceeding in a block sequence through the slower (differential) subsystems. One cycle or time step is completed,

PSM

J.

m.m

tlrr bsc)

APm=-10%

Ke=25

Fig.5 - Comparative performance of HPM and PSM models


From these curves, it becomes apparent that both models present quite similar responses. In this case, there
is enough damping to reduce the rotor oscillation by natural means. These results are indeed expected, once the
models represent exactly the same linearized system, using
different approaches. While HPM represents the generator
internal variables as viewed from the infinite bus, the PSM
represents the same problem as viewed from the terminal
bus. In this way the sensitivity terms of the latter separate the generator internal and external contributions to
the terminal power balance. This explains also the explicit
presence of both armature reaction terms KVAVt and
term K4A6 of the HPM.
KA(AS-AO), in contrast to the
As noted in Ref.[3], the critical parameter, responsible for positive damping of the HPM, is KS. A negative KS
value produces positive A6 feedback, through the exciter
control loop. It can be verified that the corresponding

968
critical parameter of the PSM is Rlg. A negative Rlg
also produces a positive (A6-Ae) feedback through the
ter control loop. Thus, for stable operation we must
Rlg>O, which leads to the following condition (derived
equation 1.4 of Appendix I):

value
excihave
from

(19)
Equation (19) shows that the dynamic stability margin
decreases with generator loading and with field voltage
rising. This becomes more critical for machines with high
Xd/Xq ratios (low short-circuit ratios).
Fig.6 shows the response of both models for the same
unstable condition, as proposed in Ref I31 (Pg=l.O pu and
Qg=-0,5 p). For this case, we obtain negative values for
both critical parameters (K5=-0.255 and Rlg=-0.64).

In the first test (7a), a pure inductive load increase


of AQ=5% is imposed on the terminal bus, and in the second
(7b), a combined active and reactive load increase of
AP=lO% and A@-% is imposed simultaneously on the terminal
bus. For both cases Ke=25. All important variables can be
directly displayed for analysis. These simple tests are
difficult to simulate with the HPM, because the terminal
bus has been eliminated during the parameter derivation.
In both tests the fast-slow time decomposition becomes
evident, once the network state variables (AVt,AB) change
abruptly after the load disturbance, in contrast to the
slower changing of the machine state variables (AEq,A6).
These tests indicate that PSM can reproduce with advantages the dynamic behavior of a single generator connected to an infinite bus.
THE PSM FOR MULTIMACHINE SYSTEMS
In order to extend the PSM to multibus systems, it is
necessary only to write down the nodal power balance equations (9) and (10) for each network bus. Let us focus on a
general bus k , connected with buses i and j, as shown in
Fig.8:
pkj ,Qkj

Pki,Qki

o.m
a)

tlR

HPM

m.m

kR2)

APm=-10%

Kez12.5

Fig.8 - General bus k power balance


The incremental power balance at bus
APgk

APLk

- 1APkl

may be written as:

= 0

(20)

IEK
A(%.!!) - A(%k)

4.ml

om
b)

PSM

tlRe (sec)

APm=-10%

Vk

s.m

Ke=12.5

Fig.6 - Comparative responses for unstable condition


These curves show that both models present the same
response under unstable conditions, if submitted to a mechanical power step.
For this simple example it is also possible to simulate a number of tests for which the HPM is inadequate. Fig.7
presents two of these cases.

K=(i, jl

Vk

A(-)=
Qkl
icK

(21)

Vk

Now the load increments (APLk, AQLk) are also included, and will be assumed to change with local voltage and
frequency.
Proceeding as before, we can express each power increment in terms of the sensitivity coefficients, calculated
for the basic steady state, and obtain a rearranged equation pair, similar to (9) and (10). For the selected bus k
the resulting equations will be:
Algk.A6k+A2gk.AEqk+A3gk.AVk-C(A2kl.AVk+A3kl.AVI)-APLk=
=Algk.ABk +CAlkl.(AJk-ABI)

(22)

R l g k . ( A 6 k - A e k ) + R 2 g k . A ~ k - C R l k l . ( A e k - A e l ) - A ( ~Vk
k) =

=-R3gk.AVk +pZkl.AVk -CR3kl.AVI

where

includes all buses

(23)

connected to bus k (IEK).

represent the bus mismatchThe LHS of (22) and (23)


APk and AQk/Vk, including the varying load effects and
the cross coupling terms (AP-AV and AQ/V-Ae).
For a network consisting of n buses ( k = l to n), we obtain two n-dimension decoupled algebraic systems for equations (22) and (23), which we represent in matrix form as:
es

4.m

m.m

tim (sec)

a) AQ=5% load increase

o.om/

AP1

A11
,

APk
APn
AQ1/V1
AQk/Vk
b ) AP=10% and AQ=5% load

...

Ak1

An1

. ..
Akk

. ..

Aln

..,.

Ann]

.?

AQn/Vn

Rnl

...

.,

-RI1
...
.. .._.
= Rk1
Rkk

increase

Fig.7 - Special tests with the PSM

Akn

Rln

AV1

Rkn

. AVk

Rnn

AVn

(25)

969

AEq k

I
act ise

i
I

laad

madet.

Fig.9 - Multibus PSM network extension


Both power sensitivity matrices [A] and [RI correspond
to the Jacobian principal submatrices, considering all network buses. They present a [Ybusl sparse structure. Each
generator contributes to the respective network bus with a
"shunt" coefficient (Alg from eq.22 and -R3g from eq. 23),
thus providing a well conditioned decoupled solution for
[AP-Ael and [AQ/V-AVl subproblems. Load characteristics may
also contribute as local shunts, especially if they are
voltage sensitive.
The matrix relations (24) and (25) correspond to the
fast acting network extension of the single bus mismatch
equations (9) and (10). The slow acting differential subsystem of each generator can be aggregated as external
control loops, in a way similar to that shown in Fig.3.
Fig.9 shows the n bus PSM diagram, considering, for
simplicity reasons, only the generator connected to bus k.
The main difference between Fig.3 and Fig.9 relies on the
matrix representation of the nodal power balance, with the
inclusion of load effects.
One important aspect to be pointed out is that we have
not specified the reference to any particular infinite bus.
Due to the local expansion, each variable will change with
respect to its own reference, represented by the original
steady state value, used for the sensitivity calculations.
In this way, the PSM balance equations can be extended to
any number of buses, preserving the basic decoupling features of the single bus case, as discussed before. The solution cycle includes now the calculation of mismatch vectors
[AP] and [A(Q/V)], containing the LHS of eq. (22) and (231,
for all network buses (k=1 to n).
The prior decomposition diagram of Fig.4 applies also
to the extended network model, and a single solution cycle
corresponds, as before, to one solution of each subsystem.
Due to the linearized formulation of the power balance
equations, no model interface error will appear. Thus, all
the advantages of a decoupled solution may be fully explored in this extended approach as well.
LOAD REPRESENTATION
It is possible to include directly the changes of the
loads, as imposed by the network state variables or by external load demand changes. Returning to the generic bus k,
a change of the active and reactive loads at this bus can
be functionaly related with the local voltage magnitude and
frequency changes, in linear form represented as:
APLk = Dlk.Afk + D2k.AVk + APEk

(26)

AQ* Vk = D3k.Afk + D4k.AVk + AQEkVk

(27)

where the load sensitivities (Dl,D2,D3,D4) can be derived


from typical load characteristics, or may be obtained from
specific load monitoring and data aquisition systems. If
the relevant load has a controlling loop, such as industrial motors, transformer automatic tap-changers or reactive
compensators, then the controlling transfer function may
also be explicitly represented as an input-output interacting process between the network and the load variables.
The frequency change (Afk) can be derived from local
angle changes (d(AOk)/dt). However, the main regulating
effect of the loads will be retained, if Afk is simply substituted by the mean frequency changes, calculated from the
weighted generator frequency deviations as a function o f
'Afi
i = 1,2 ... ngen
Afk =(28)
XMi

The inclusion of load characteristics into the


power balance, may be represented as shown in Fig.10:
APL k
AQLk

w e
-Afk

Active load effect

%+

dvk

APE k
a)

nodal

b)

AVk
QEk
AVk
Reactive load effect

Fig.10 - Load contributions to bus mismatches


Fig.11 shows some curves obtained with the 9-bus test
system, obtained from Ref .[141. These results correspond to
AP~=108and AQE=5% at bus 8, with Dl=lpu, D2=.5pu, D3=0 and
D4=2pu for lpu loads. Only machine 2 has an excitation
control (IEEE type 1). Machines 1 and 3 were assumed with
constant Eq'. Fig.lla shows the time evolution of the original angle differences between each generator and the
disturbing bus 8. Fig.llb shows the time evolution of
AEFDZ, AEq2' and AV8. These curves serve only to illustrate
potential applications of the PSM, especially for low frequency oscillation studies.
THE PSS REPRESENTATION
Linear dynamic models are very useful for parameter
adjusting and the optimization of control loops performance.
This is specifically the case for PSS (Power System Stabilizer) adjustment. The PSS action is usually imposed on
voltage control loops. At synchronous machines, this can be
realized via the excitation loop, using adequate phase
compensation circuits. More recently, control actions have
been introduced by Static Switched Compensators (SSC) l1.51,
which can be located at the most sensitive buses.

970

a)

an infinite bus, allows a direct comparison with the widely


used Heffron-Phillips versions.
Numerical results show the advantages of using the new
method, particularly in extending the dynamic studies to
networks, for which the original topology preservation may
be of interest, in order to reproduce the low frequency
electromechanical oscillation modes over the system.
It is also shown that the proposed model presents very
attractive decomposition features, which can be effectively
explored for fast dynamic simulation purposes. The nodal
power sensitivity approach is specially suited for decomposition into the basic synchronizing and damping torques of
the oscillation modes. Furthermore, it naturaly overcomes
the infinite bus barrier,which has made difficult to extend
the Heffron-Phillips model for multibus networks, thus
limiting the study of damping and synchronizing effects of
several distributed controlling devices upon the dynamic
behaviour of power systems. It is expected the PSM will
become a useful analytical tool for these studies.

Generator angular displacements

:mm

REFERENCES

o.mml ~

0.m

b)

Fig.11

tlR

(5R)

ism

Voltage changes after disturbance

- PSM

for the 9-bus system. (APs=lO% A@=%)

The PSM formulation enables us to verify the performance of stabilizing devices directly connected to selected
locations, and thus verify the damping effect in the presence of multimachine oscillations and faced with the load
characteristics. The detailed network representation also
permits to monitor any desired tie-line oscillation, as
well as the local electromechanical and exciter modes, as
affected by the generator control loops. This topic deserves a specific treatment, and therefore is out of the scope
of this paper.
THE AFC CONTROL ACTIONS
Another special feature of the PSM is the abillty to
represent the detailed dynamics of the AFC. The mechanical
power control loops (APm) affect the low-frequency oscillation modes, thus influencing the PSS actions. The primary
control is a local speed feedback loop, while the suplementary control combines frequency and inter-area power flow
deviations. This mixed signal is usually synthesized in a
remote Control Center and then transmitted to the individual power plants for the corrective actions.
For the complete AFC simulation in a multimachine
power pool, it is necessary to monitor all tie-line power
flows between the interconnected areas. With the PSM, this
information may be obtained directly. Once the network
topology is preserved, the power change at a generic tieline k-l can be calculated from the available terminal
state variable changes:
APkl= Alkl.(ABk-ABI) + A2kl.AVk + A3kl.AVl

P.L.Dandeno, A.N.Karas, K.R.McClymont, W.Watson, "Effect of High-speed Rectifier Excitation Systems on


Generator Stability Limits", IEEE Trans. PAS, Vol. 87,
pp. 190-201, Jan. 1968.

(29)

Thus, the dynamic studies of the complete system can


be extended for long time intervals without carrying unnecessarily complex models. The time steps to be choosen
depend of the integration method used, but they are mainly
dictated by the time frame of the dynamic process to be
studied. For the present simulations, a trapezoidal rule
was chosen, with steps of .01 and .02 seconds.
CONCLUDING REMARKS
The main purpose of this paper was to describe the detailed derivation of a power sensitivity model, suited for
direct simulation and analysis of dynamic processes in
multimachine, multibus networks. The proposed model is
based on power sensitivities, used to establish the dynamic
power balance conditions at the network buses. It was derived from the basic concept of instantaneous nodal power
balance, with the preoccupation of devising a physical interpretation for every step of the derivation. Furthermore,
the block representation of a single machine, connected to

F.R.Schleif, H.D.Hunkins, G.E.Martin, E.E.Hattan,


citation Control to Improve Powerline Stability",
Trans. PAS, Vol. 87, pp. 1426-1434, June 1968.

"Ex-

F.P.de Mello, C.Concordia, "Concepts of Synchronous


Machine Stability as Affected by Excitation Control",
--IEEE Trans. PAS, Vol. 88, pp.316-329, April 1969.
W.G.Heffron, R.A.Phillips, "Effect of Modem Amplidyne
Voltage Regulators on Underexcited Operation of Large
Turbine Generators", AIEE Trans. PAS, Vo1.71, pp. 692619, 1952.

E.V.Larsen, D.A.Swann, "Applying Power System Stabilizers", IEEE Trans. PAS, Vol. 100, Part I I1 and Ill,
pp. 3017-3046, June 1981.
H.A.M.Moussa, Yao-nan Yu, "Dynamic Interaction of Multi Machine Power System and Excitation Control",
-Trans. PAS, Vo1.93,~~.
1150-1158, Jul/Aug. 1974.

M.Mobarak, D.Thome, E.Hill, "Optimization of Exciter


and Stabilizer Parameters of large Thermal Generating
Station", IEEE Trans. PAS, Vo1.97, pp. 1182-1193, Jul/
Aug 1978.
G.Gross, A.R.Bergen, "A Class of New Multistep Integration Algorithms for the Computation of Power System
Dynamical Response", IEEE Trans. PAS, Vo1.96, pp. 293306, Jan/Feb. 1971.

E.G.Cate,
K.Hemmaplardh,
J.W.Manke,
D.P.Gelopulos,
"Time Frame Notion and Time Response of the Models in
Transient, Mid-Term and Long-Term Stability Programs",
--lEEE Trans. PAS, Vo1.103, pp.143-151, Jan. 1984.
P.Kundur, M.Klein, G.J.Rogers, M.S.Zywno, "Application
of Power System Stabilizers for Enhancement of Overall
System stability", IEEE Trans. PWRS, Vo1.4, pp.614-626
1989.
S.M.Deckmann, V.F. da Costa, D.A.Alves, "Dynamic Simulation of Interconnected Power Systems", IFAC Sym. on
Plan. & Oper. of Electric Energy Systems, pp. 305-312,
Rio de Janeiro, Brazil, 1985.
V.F. da Costa, "A Power Sensitivity Model for LowFrequency Oscillations Analysis", Ph.D. Dissertation,
Unicamp Univ., Brazil, 1992, (in Portuguese).
B.Stott, O.Alsac, "Fast Decoupled Load
Trans. PAS, Vo1.93, pp. 859-869, 1974.

Flow",

EPRI EL-484 "Power System Dynamic Analysis", Phase-1.


Boeing Computer Services, Final Report, July 1977.
T.J.E. Miller, "Reactive Power Control
Systems", John Wiley & Sons, 1982.

in

Electric

97 1
APPENDIX I

- Power Sensitivities

In a lossless salient pole synchronous generator the


transient power functions Pg and Qg of equations (3) and
( 5 ) present the following sensitivity coefficients:

This suggests that we can enlarge the disturbing range


with attainable precision in the decoupled solution scheme
with constant matrices, only by correcting the most critical coupling sensitivities for each time step, thus transfering compensated outputs for the next subsystem in the
solution
cycle. This on-line non-linear
compensation
feature would be lost in a simultaneous solution scheme.
APPENDIX 11

Armature Reaction Coefficients

For a lossless unsaturated, salient pole synchronous


generator with constant q-axis flux, the transient d-axis
flux balance may be expressed by the following voltage relation, in frequency domain:
I

TdO.s.Eq = EFD

- Eq -

(xd

Xd).Id

(11.1)

with
Id

Eq -Vt .cos(6-8)
Xd

(11.2)

Replacing ld in the first equation, we obtain:


The sensitivities to angle differences may be decomposed into local sensitivities, with opposite sign:

or, in a linear expanded form:


Xd
TdO.S.AEq = AEFD --,AEq
Xd
A non-salient pole machine can be represented as a
special case, in which Xq = Xd.
The power sensitivity coefficients of the external
lossless impedance in Fig.lb, are calculated from Pe and Q e
functions (4) and (61, resulting:
Ale=;

aa e

VtxHocos(e)

(1.9)

+ r&:d
-cos(60-00)] AVt-

-[y 1

Vto.sin(6o-Bo) A(6-8)

(11.4)

If we define the armature reaction coefficients as

xXd -xd
Kv s d , cos(ao-eo)
KA

(11.5)

Xd-Xd
= Vto.sin(60-80)
Xd

(11.6)

than we can write (11.4) as

[E,+s.TdO)AE;

= AEFD + Kv.AVt

- KA.A(6-8)

(11.7)

which is equivalent to equation (14) in the text.


It is worth noting that for
no-load conditions the
coefficient Kv is maximum, while KA is zero. For normal
load, both coefficients assume comparable positive values.
We see from (11.7) that the demagnetizing armature effects
are due both a terminal voltage drop [AVtl and an angle
difference increase [A(s-e)l. If we decompose the last term
of (11.7) into
Equations (1.9) to (1.14) also apply for a generic
line connecting buses k - I, by simply replacing variables
{Vt,VO,Bj for the corresponding {vk,vl,(ek-ei)). Again, the
angle sensitivities can be locally decomposed as:

-KA.A(6-8)

= - K A . A ~ + KA.Ae

(11.8)

it becomes clear that A6 and AB produce opposite magnetizing effects, and present different time scale responses.
While+AB produces a fast
magnetizing effect imposed from
the network, +A6 introduces a slow demagnetizing contribution,
associated with the rotor dynamics.
BIOGRAPHIES

Losses and shunt effects can also be introduced in the


above derivations, resulting in a more complete model.
Dependency o f State Variables

Most of the coefficients have a nonlinear dependency


terms (P-V and
Q-e) assume a sine functional dependency (like A2g,A3g,A2e
ASe,Rlg,Rle) and zero value for unloaded conditions. The
direct coupling terms (P-e and Q-V) are cosine functions
(Alg,Ale,RZg,R3e). and will be maximum for unloaded conditions. Only R2e assumes a constant value. This operating
point dependency limits the extension of the MSP (as much
as the HPM) to moderate disturbing dynamic studies.
However, it is worthwhile noting that the most statesensitive parameters
are the cross-coupling
coefficients
and the less sensitive are those parameters which form the
active and reactive matrices [AI and [RI in the multibus
network representation.
upon the operation state. The cross-coupling

Slgmar M . Deckmann received his B.S.(1973)


M.S.(1976) and Ph.D.(1980) degrees from
Unicamp Brazil, where he has been teaching
Electrical Engineering
since 1974. His
main research interest areas are power
system modelling,
analysis and control.
After his Ph.D. on External Static Equivalents he has concentrated his work on lowfrequency dynamic studies and flicker disturbance measurement and analysis.
V f v a l d o F . d a Costa received his B.S.
(1976), M.S.(1981), and concluded his Ph.D
at Unicamp (1992). He joined the Electrical Engineering Faculty in 1977. His main
research interests are transient and steady state stability of power systems. He is
currently working on low frequency oscillation and power system stabilizing analysis.

You might also like