Copyright © IFAC Power Plants and Power Systems Control, ELSEVIER

Seoul, Korea. 2003

IFAC
PUBLICATIONS
www.elsevier.com/locatelifac

COMPARISON OF SMALL SIGNAL STABILITY ANALYSIS METHODS

IN COMPLEX SYSTEMS WITH SWITCHING ELEMENT

Deok Young Kim *

A. P. Sakis Meliopoulos **

* School ofElectronic & Information Eng., Kunsan National University, Korea

** School ofElectrical and Computer Eng., Georgia Institute ofTechnology, U.S.A.

Abstract: A new small signal stability analysis method for eigenvalue analysis is
presented. This method utilizes the Resistive Companion Form (RCF) for the
computation of the transition matrix over a specified time interval that corresponds to one
cycle operation of the system. This method is applicable to any system with or without
switching elements. An illustrative example of the method is presented and the
eigenvalues are compared with those of the conventional state space method (analog) to
show the exactness of the proposed eigenvalue analysis method. Also, the variations of
oscillation modes which were caused by the switching operation can be analysed exactly
using this method. Copyright © 2003 IFAC

Keywords: Small Signal Stability Analysis, Eigenvalue, Switching Element

1. INTRODUCTION integral equations. These equations can be arranged in

the following general form:
FACTS technology utilizes high power electronic
devices to enhance controllability of electric power
flow in modem power system. This has the potential
to increase power transfer capability and provide
[~t)] = [iJ (~(t),y(t), V(t),y(t),U(t»] (1)
economic loading of existing transmission facilities. f2 (v(t),y(t), v(t),y(t),u(t»
These attractive features of FACTS elements also
where,
cause unwanted effects which are mainly waveform
i(t) : vector of terminal currents,
distortion and additional oscillation modes. These
effects are generated by the switching operation of v(t) : vector of terminal voltages,
power electronic devices. y(t) : vector of device internal state variables,
u(t) : vector of independent controls
In conventional small signal stability analysis, system
is assumed to be invariant and the state space
equations are used to calculate the eigenvalues of This form includes two sets of equations, which are
state matrix. However, when a system contains named external equations and internal equations
switching elements such as FACTS devices, it respectively. The terminal currents appear only in the
becomes non-continuous system. In this case, a external equations and the device state variables
mathematically rigorous approach to system small consist of two sets: external states(Le. v(t» and
signal stability analysis is by means of eigenvalue internal states(i.e. y(t».
analysis of the system periodic transition matrix
based on discrete system analysis method. In this An example of above modelling is a switching device
paper, Resistive Companion Forrn(RCF) method is represented with linear elements. Between switchings,
used to analyse small signal stability of a non- the model is described with a linear differential
continuous system including switching elements. To equation of the form:
demonstrate the relative merits of the proposed
method, a comparison of the conventional state space
method and RCF method is presented for application
systems with and without switching elements.
i(t)] =
[o
[All
A2l A22 y(t)
[B
A12 ][V(t)] + lI
B 2l
B12 ] !!...[V(t)] (2)
B 22 dt y(t)

Equations (1) are integrated using a suitable

2. RESISTIVE COMPANION FORM (RCF)
METHOD
For small signal stability analysis, any power system
device is described with a set of algebraic differential
numerical integration method such as trapezoidal
method. Assuming an integration time step h, the
result of the integration is manipulated to be in the
following form:

703
 The location of an eigenvalue of the transitioin matrix
i(t)] = [1';1 1';2 ][V(t)] _[~I ~2 ][V(t - h)] indicates the nature of the mode. In order to interpret
the eigenvalues in terms of modal damping factors
[o Y21 Y22 yet) P21 P22 yet - h) and natural frequencies, we can use the eigenvalue
_[CII C1z!i(/-h)] (3) mapping between the transition matrix eigenvalue and
state space eigenvalue. It is known that:
CZ1 Cn 0

To consider the connectivity constraints among the

1
Ad = e .l. T = eel"
<
-aT -
RT
= re J'(J
(8)
devices of the system, Kirchoffs current law is
applied to each node of the system. Application of
where, Ad and Ac are the eigenvalues of the
KCL at each node will result in elimination of all
device terminal currents. The overall network transition matrix(i.e. discrete system) and state space
equation has the form: matrix(i.e. continuous system) respectively, T is the
60Hz period, and A.c = -a + j P.

Y,12 ][V(t)] _[P.II P.12 ][V(t - h)]

[°0] =[Y Y
slI

s21 YS22 yet) P.21 P.22 yet - h) The magnitude r is related to the damping factor a
and the angle (} is related to the angular frequency
_[Q,(/-h)] (4)
fJ . Therefore the eigenvalues of transition matrix
Q2(/-h)
has the effect of mapping those of state space matrix
or the equivalent: to unit circle. It implies that highly damped modes are
identified with eigenvalues near the center of the unit
circle, stable oscillatory modes are identified with
V(t)] =[Ys11 YsI2 ]+ [1'.11 PsI2 ][V(t - h)] eigenvalues within the unit circle and unstable modes
[yet) Y,21 Y,22 1'.21 1'.22 yet - h) are identified with eigenvalues outside the unit circle.

+ [YslI Y ]+[Qt(t-h)]
S•2
YsZ1 Ysn
where the superscript + indicates generalized invere
matrix. Note that the above equation represents the
state transition equation for the entire system from
time t-h to time t. The above linear equation form is
the resistive companion form which results from the
trapezoidal integration method. The transition matrix
IS:
(5) 3. APPLICATION EXAMPLES
Qz (I - h)
To compare the eigenvalues of the transition matrix
and the state space matrix, two complex systems with
and without switching elements are used.
3.1 Complex system without switching element.
The application system without switching element is
shown in Fig. 1. The parameters of the application
system are:

(6) R. =IO.qO],Rz =50.0[0],R) = 30.0[0],R. =20.0[0]

L. = 0.05[H],L 2 = O.I[H],L) = O.2[H],
Cl = 0.002[F],C2 = O.OOO5[F],V = IIO.O[V]
Eigenvalue analysis of the transition matrix provides
the small signal stability of the system.

In general, we are interested in the transition matrix

over at least one period of operation of the system.
The proposed method provides an algorithm for the
recursive computation of the transition matrix over a
desired time period and around the operating
conditions of the system. The entire transition matrix
over a desired time period can be done by sequential
substitution of the transition matrix state variables in Fig. 1. Complex system without switching element
each time step. The overall transition matrix has the
form:
State space method.
<1>T (t. ,/0 ) =<1>. (I. ,1._1)' <1>._. (I._I ,1._ 2 )", <1>z (tz ,t.) .<1>. (t. ,to)
The state space equations of Fig. 1 are:
(7)

where, <I> i (tpt l) means

i
_ the transition matrix of
specified time step.

704
 -RI -1 -1 Table 2 Comparison of eigenvalues, t= 0.0002 sec
0 0
LI LI L,
iLl
0
-R.
0 0 -
1 iLl J... V
L2 L2 L. Mode RCFMethod State Space Error
i L2 i L2
-R) 1 0 Method Ratio (%)
0 0 0 ill +
ill L) L) 0
Vet
.974697 .974681 .001654
Vet -1 -I 0 +j .051779 +j .051783
0 0 VC2
Cl CI R 2 CI R2 0 2 .974697 .974681 .001654
VC2
1 -1 -1 -I -I -j .051779 -j .051783
- -
C2 C2 C2 C2 R2 C2 R2 3 .963425 .963425 .000428
4 .994806 .994806 .000001
RCFmethod 5 .971352 .971352 .000206

From the circuit diagram, the state transition

equations are: Table 3 Comparison of eigenvalues, t= 0.0003 sec

L, + hR, h h
0 0
2 2 2 Mode RCFMethod State Space Error
-h -h
0 L, + hR. 0 0 iLl (I) TV(I) Method Ratio (%)
2 2 iL,(I)
L hR, -h 0
0 0 ,+ - 0 iL,(I) + 0 .961297 .961245 .005584
2 2
-h
0 0 c,+_h_ h VCl (I) 0 +j .076683 +j .076694
2 2R, 2R, VC2(I) 0 2 .961297 .961245 .005584
-h h h h
c,+...!!.- -j .076683 -j .076694
2 2 2 2R, 2R,
3 .945634 .945648 .001454
4 .992219 .992219 .0
£,_ hR, 0 0
-h -h 5 .957334 .957341 .000692
2 2 2
£ _ hR. h iU(I-h) !!.V(I-h)
0 , 2 0 0
2 2
it,l,-h) 0 From the above tables, the largest error ratio in table
L, _ hR, h
0 0 0 i L3 (I-h) +
2 2 0 3 is 0.005584 %. Therefore, as in the complex system
h c,_....'!...- -h VCI(/-h) 0
0 0
2 2R, 2R, Vc,(/-h) 0
without switching element, the eigenvalues of state
h -h -h -h c, _....'!...-
2 2
space method and RCF method are almost the same
2 2R, 2R,
results.
From the above equations, the transition matrix can
3.2 Complex system with switching elements.
be calculated as (6).
The application system with switching elements is
Comparison of eigenvalues
shown in Fig. 2. The parameters of the application
system are:
The eigenvalues of state space method and transition
matrix are compared from table 1 to table 3. In this
example, the time step h is defined as .0001 sec and RI = 20[O],R 2 = 40[O],R) = 30[0],L1 = 0.05[H],
the eigenvalues of state space method are L 2 = O.l[H],C I = 0.2[F],V, = 1l0[V]
transformed into unit circle

Table 1 Comparison of eigenvalues, t=.0001 sec

Mode RCF Method State Space Error

Method Ratio (%)
.987609 987609 .000208
+ j .026216 + j .026216
2 .987609 .987609 .000208
-j .026216 -j .026216
3 .981543 .981544 .000055
4 .997399 .997399 .000005 Fig. 2. Complex system with switching elements
5 .985573 .985573 .000023

705
3.2.1 Case 1 (SW 1 : close, SW 2: open) h
1- hR2 0 -
State space method
2L.
1- hR3
2L.
h
[,,,(I-h)]
= 0
2L2 2L2
' L2 (t - h)
The state space equations of Fig. 2 are: Vcl(t-h)
-h -h
- - 1
2C1 2C.
-RI
0 0 Vs
iLl LI From the above equations, the transition matrix can
-R3 LI be calculated as (6).
1
i L2 = 0 ['LO
i L2 ] + 0
L2 L2 Comparison of eigenvalues
-1 VCI 0
Vel
0 - 0
Cl The eigenvalues of state space method and transition
matrix are compared from table 4 to table 7. In this
RCFmethod example, the time step h is also defmed as .0001 sec
and all the eigenvalues of state space method are
From the circuit diagram, the state transition transformed into unit circle in table 4 and table 5
equations are: while all the eigenvalues of RCF method are
transformed into S-plane in table 6 and table 7.
I hR. 0 Table 4 Comparison of eigenvalues. t=O.0006 sec
+ - 0 -hV.(t)
2L1
0 1 + hR 3 -h [,,,(I)]
i L2 (t) +
1£.
0
inCase 1

2L 2 2L 2
h VCI(t) 0 Mode RCFMethod State Space Error
0 - 1 Method Ratio (%)
2C1
1 .99989 .99989 .0
2 .83534 .83535 .00134
1- hR. 3 .78660 .78662 .00318
[~ v,(t-hl]
0 0
2L.
,,,(I-hl ]
1- hR3 h In table 4, the eigenvalues are calculated at t=O.0006
0 IL2(t-h) + 0
2L 2 2L 2 sec in Case 1 (SWl is closed and SW2 is opened)
-h VCI(t-h) 0
0 - which is from 0 to 0.0006 sec with a time step 0.0001
2e, sec. It is assumed that the errors of transition matrix
by sequential substitution of state variables in each
3.2.2 Case 2 (SW 1 : open, SW 2 : close) time steps will be the largest one at 0.0006 sec. In this
case the largest error ratio between state space
State space method method and RCF method is 0.00318 % which means
that the eigenvalues from the two method are almost
The state space equations of Fig. 2 are: the same.

Table 5 Comparison of eigenvalues. t = 0.001 sec

-R 2 1 in Case 2
0 -
iLl LI LI
-R3 1
i L2 0 ['LO
'L2 ] Mode RCFMethod State Space Error
L2 L2 Method Ratio (%)
-I -1 VCI
Vel
- - 0 1 .99988 .99988 .00089
Cl Cl 2 .88697 .88697 .0
3 .72606 .72618 .01706
RCF method
In table 5, the eigenvalues are calculated at t=O.OOI
From the circuit diagram, the state transition sec in Case 2 (SWl is opened and SW2 is closed)
equations are: which is from 0.0006 to 0.001 sec with a time step
0.0001 sec. It is also assumed that the errors of
transition matrix by sequential substitution of state
1+ hR 2 -h variables in each time steps will be the largest one at
0 -
2L.
0 1 +--
hR3
2L1
-h
[,,,(I)] 0.001 sec. In this case the largest error ratio between
state space method and RCF method is 0.01706 %,
i L2 (t) which means that the eigenvalues from the two
2L 2 2L 2
h h Vc.(t) method are almost the same.
- 1
2C1 2C.

706
 Table 6 Eigenvalues of Case 1 and Case 2 by state IEEE power engineering society, "FACfS overview",
space method (S-plane) 95TPI08, 1995.
Yousin Tang, A. P. Sakis Meliopoulos, "Power
Mode 't=.0006 T=.OOI system small signal stability analysis with
1 -.16675 -.29187 FACfS elements", IEEE Trans. of power
2 -299.83323 -299.83312 delivery, Vol. 12, No. 3, pp. 1352-1361, July
3 -399.99999 -799.87498 1997.

In table 6, all the eigenvalues of state space method

in case 1 and case 2 are shown in S-plane. It is clear
that there is no relation between the eigenvalues of
case 1 and those of case 2 caused by the switching
action in state space method.

Table 7 Eigenvalues ofRCF method including

switching effect (S-plane)

Mode "t=.0007 t=.OOO8 t=.OO09 "t=.001

1 -.1498011 -.1792794 -.1876583 -.1954392
2 -262.3736 -299.8555 -299.8555 -299.8555
3 -457.2448 -500.1340 -533.4900 -560.1738

The variation of eigenvalues by RCF method in each

time steps are shown in table 7. All the eigenvalues
are transformed into S-plane. To compare the
eigenvalues of table 6 and 7, the loci of eigenvalues
in table 7 by the RCR method in case 2 start from
those of 0.0006 sec and are getting closer to those of
0.001 sec in table 6. It is clear that the variation of
eigenvalues in each time steps in table 7 is caused by
switching operation. These results are impossible to .
analyse by the state space method. To compare the
eigenvalues of state space method and RCF method
at 0.001 sec, mode 2 is almost the same. But mode 1
and 3 of state space method and RCF method are
different and the errors are not small.

CONCLUSIONS

The eigenvalues from the conventional state space

method and RCF method are compared in small
signal stability analysis. Those are almost the same
in continuous system analysis. But, in non-
continuous systems including switching elements, the
eigenvalues from the state space method and RCF
method are different. The RCF method can calculate
exactly the variation of oscillation modes after
switching action. Therefore, RCF method is a useful
one to analyse a non-continuous systems including
switching elements in small signal stability analysis.

REFERENCES

Bohmann L. J., Lasseter R. H., "Stability and

harmonics in thyristor controlled reactors",
IEEE Trans. on Power delivery, Vol. 5, No. 2,
pp. 1175-1181, Apri11990.
Eugene V. Solodovnik, George J. Cokkinides, A. P.
Sakis Meliopoulos, "On stability of implicit
numerical methods in nonlinear dynamical
systems simulation", Proceedings of the 30th
southeastern symposium on system theory, pp.
27-31, West Virginia, March 1998.

707