See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/311479470

Effect of load increase and power system stabilizer on stability delay margin of a
generator excitation control system

Article  in  Turkish Journal of Electrical Engineering and Computer Sciences · January 2016

DOI: 10.3906/elk-1504-226

CITATIONS READS

5 1,363

2 authors:

Şahin Sönmez Saffet Ayasun

Ömer Halisdemir University Ömer Halisdemir University
37 PUBLICATIONS   451 CITATIONS    84 PUBLICATIONS   1,270 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Delay-Dependent Stability Analysis of Load Frequency Control Systems with Electric Vehicles and Dynamic Demand Response, and Robust Controller Design View project

All content following this page was uploaded by Saffet Ayasun on 03 April 2019.

The user has requested enhancement of the downloaded file.

 Turkish Journal of Electrical Engineering & Computer Sciences Turk J Elec Eng & Comp Sci
(2016) 24: 5183 – 5194
http://journals.tubitak.gov.tr/elektrik/
⃝c TÜBİTAK
Research Article doi:10.3906/elk-1504-226

Effect of load increase and power system stabilizer on stability delay margin of a
generator excitation control system

Şahin SÖNMEZ, Saffet AYASUN∗

Department of Electrical and Electronics Engineering, Niğde University, Niğde, Turkey

Received: 27.04.2015 • Accepted/Published Online: 03.11.2015 • Final Version: 06.12.2016

Abstract: This paper studies the impact of load increase and a power system stabilizer (PSS) on the stability delay
margin of a single-machine-infinite-bus system including an automatic voltage regulator. An analytical method is
proposed to determine the stability delay margin of the excitation control system. The proposed method first eliminates
transcendental terms in the characteristic equation of the excitation system without making any approximation and
transforms the transcendental characteristic equation into a regular polynomial. The key result of the elimination
process is that the real roots of the new polynomial correspond to the imaginary roots of the transcendental characteristic
equation. With the help of the new polynomial, it is also possible to determine the delay dependency of system stability
and the root tendency with respect to the time delay. Delay margins are computed for various loading conditions and
PSS gains. It is observed that the delay margin generally decreases as the PSS gain and load demand increase, resulting
in a less stable system.

Key words: Generator excitation control, power system stabilizer, time delay, stability, delay margin

1. Introduction

This paper investigates the impact of load increase and power system stabilizer (PSS) on the time delay margin of
generator excitation control system. The phasor measurement units (PMUs) and open communication networks
have been extensively used in wide-area measurement/monitoring systems. This causes inevitable time delays,
which include measurement and communication delays [1,2]. Such time delays may reduce the control system
damping performance and could even cause instability if they exceed the upper bound or delay margin for
stability [3–5]. Therefore, large time delays must be taken into account in stability analysis and controller
design, and practical methods that enable us to determine delay margins should be developed.
In electrical power systems, each generator is equipped with a load frequency control (LFC) system and
an automatic voltage regulator (AVR) to keep the system frequency and generator output voltage magnitude at
a desirable range when changes are observed in real/reactive load demands [6–8]. Figure 1 shows the one-line
diagram of the SMIB system that includes both the AVR and PSS, and Figure 2 illustrates its block diagram
with a time delay block in the voltage feedback loop [8].

∗ Correspondence: [email protected]

5183
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

Figure 1. One-line diagram of SMIB with AVR and PSS [8,28].

Figure 2. Block diagram of the SMIB system with AVR and PSS [8,28].

Time delays observed in power systems consist of measurement and communication delays. The use of
PMUs causes measurement delays, including voltage transducer delay and processing delay. The processing
delay is the amount of time required in converting transducer data into phasor information with the help of
the discrete Fourier transform (DFT) [1,9]. In power system control, the total measurement delay is reported
to be on the order of milliseconds [1–3]. The total communication delays, especially in multiarea LFC systems,
can vary from 10 to several hundred milliseconds [10–12]. There are several factors that affect the size of
communication delays. These include communication means such as fiber-optic cables, power line carriers,
telephone lines, etc., and phasor package size, transmission protocol used, and network load.
This work primarily aims to investigate how the load increase and addition of a PSS will affect the delay
margin of the SMIB with AVR and focuses on the time delay observed in the terminal voltage feedback loop,
shown as exponential block e−sτ in Figure 2. It must be stated here that there will be a certain amount of
time delay (τ1 ) in the PSS loop. This delay may or may not be equal to the delay in the voltage control loop.
The inclusion of the delay in the PSS loop will result in another exponential term (e−sτ1 ) in the characteristic
equation. In this case, the characteristic equation will include incommensurate delay terms like e−sτ and e−sτ1
when τ1 ̸= τ and commensurate delay terms like e−sτ and e−s(2τ ) when τ1 = τ . Delay margin computation of
such a characteristic equation, including commensurate and incommensurate delay terms, is a difficult task and
requires a significant amount of computational effort. Therefore, the delay in the PSS loop was neglected in this
study to simplify the delay margin computation and to investigate the effect of the single voltage measurement
delay on the dynamics of voltage control.
There are several methods for computing delay margins for stability of time-delayed dynamical systems.
These methods can be grouped into 2 main types, namely frequency-domain direct and time-domain indirect
methods. The main goal of frequency domain approaches is to compute all critical purely imaginary roots of

5184
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

the characteristic equation for which the system will be marginally stable [13–15]. The frequency domain direct
methods can obtain accurate delay margins for constant delays. However, they cannot be applied to stability
analysis of time-delayed systems that include time-varying delays. This is a drawback of such methods. The
indirect time-domain method utilizes the Lyapunov stability theory and linear matrix inequalities technique [16–
18]. Such methods have been applied to estimate the delay margins of the wide-area damping controller [12,19]
and LFC systems [11]. These methods can deal with both constant and time-varying delays. However, their
delay margin results are more conservative as compared with ones obtained by frequency-domain direct methods
[11,20]. Our previous studies clearly indicated that the direct method reported in [13] correctly estimated the
delay margins of a generator excitation control system with a constant single delay [4,5], delay margins of load
frequency control systems with constant communication delays [20], and a time-delayed DC motor speed control
system [21]. In addition to these applications, this frequency-domain direct method was successfully applied
to investigate the stability of other time-delayed systems such as mechanical systems [22–24], predator–prey
systems [25,26], and a logistic model [27]. Such successful applications, together with correct estimation of delay
margins, have motivated us to apply this method to a SMIB system with AVR and PSS.
The proposed method first eliminates exponential terms in the characteristic equation of the SMIB system
without making any approximation and transforms the transcendental characteristic equation into a regular
polynomial. The key result of the elimination process is that the real roots of the new polynomial correspond
to the critical imaginary roots of the transcendental characteristic equation exactly. With the help of the new
polynomial, it is also possible to determine the delay dependency of system stability and root tendency with
respect to the time delay. An analytical formula is then derived to compute delay margins. The effect of load
increase and PSS gain is investigated, and it is found that delay margin generally decreases as the PSS gain
and load increase, indicating a less stable system.

2. Single-machine infinite bus system model

For delay margin computation, a linearized model of the SMIB system is required. Figure 3 shows the block
diagram of the linearized flux-decay model with a fast exciter [8,28]. The expression for the parameters K1 − K6
can be found in [8]. Figure 4 illustrates the block diagram of the PSS that includes gain, washout filter, and
lead-lag compensator [7,8,28,29]. The basic function of the PSS is to increase the stability limits by modulating

Figure 3. Detailed block diagram of the SMIB system Figure 4. Block diagram of PSS [7,8,28].
with AVR including a time delay [8,28].

5185
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

the generator excitation to support damping for the rotor oscillations of synchronous machines. Using Figures
2–4, a time-delayed state-space equation model of the system could be easily obtained as

ẋ(t) = A0 x(t) + Aτ x(t − τ ) + Bu(t), (1)

[ ]T
where x(t) = ∆δ ∆ω ∆Eq′ ∆Ef d ∆Vw ∆VP SS defines the state variable vector. The meaning of
the state variables can be found in [8]. The system matrices are given as
 
0 ω0 0 0 0 0
 
 −K 1
−M
D
−K2
0 0 0 
 M M 
 
 − TK′4 0 − TK′3 1
′ 0 0 
 d0 d0 Td0 
A0 =  − T1A KA ;
 0 0 0 0 
 TA

 − KPMK1 − KM
PD
− KPMK2 0 − T1w 0 
 
 ( ) 
− KPMKT12T1 − KM
P DT1
T2 − KPMKT22T1 0 1
T2 − T1
T2 Tw − T12

 
0 0 0 0 0 0
 
 0 0 0 0 0 0 
 
 0 0 0  [ ]T
 0 0 0 
Aτ = 

;B = 0
 0 0 0 0 KA
.
 − KTAAK5 0 − KT6 K A
0 0 0  TA
 A

 0 0 0 0 0 0 
 
0 0 0 0 0 0

3. Delay margin computation

For stability analysis and delay margin computation, it is required to determine the characteristic equation of
Eq. (1). This could easily be obtained by using the following:

∆(s, τ ) = det(sI − A0 − Aτ e−sτ ) = P (s) + Q(s)e−sτ = 0, (2)

where τ the total is time delay, and P (s) and Q(s) are polynomials in s with real coefficients given below:

P (s) = p6 s6 + p5 s5 + p4 s4 + p3 s3 + p2 s2 + p1 s + p0
.
Q(s) = q4 s4 + q3 s3 + q2 s2 + q1 s + q0

It must be noted that all the coefficients of the polynomial above will depend on the system parameters. The
expressions for coefficients p0 − p6 and q0 − q4 are given in the Appendix.
The main purpose of the stability analysis of time-delayed systems is to obtain conditions on the delay
for any given set of system parameters that will ensure the stability of the system and to compute the delay
margin for stability. Depending on the system parameters, a time-delayed system could be delay-independent
or delay-dependent stable. For delay-independent stability, the system remains stable for all finite values of
time delay. In a delay-dependent stability case, asymptotic stability holds for τ < τ ∗ , where τ represents
the delay and τ ∗ is the critical delay, called the delay margin. If the delay exceeds the margin τ > τ ∗ , the
system becomes unstable. It is well known that all the roots of the characteristic Eq. (2) must be located

5186
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

in the left half of the complex plane for a stable system. The characteristic Eq. (2) clearly indicates that
∆(s, τ ) = 0 is an implicit function of s and τ . For simplicity, it is assumed that a delay-free (τ = 0) system
is stable. In other words, all the roots of ∆(s, 0) = 0 are in the left half-plane. This is a realistic assumption,
since a delay-free excitation control system is stable for practical values of the parameters. Suppose that the
characteristic equation ∆(s, τ ) = 0 has a root on the imaginary axis at s = jωc (where subscript c refers to
“crossing” the imaginary axis) for some finite values of the time delay τ . Because of the complex conjugate
symmetry of complex roots, the equation ∆( − s, τ ) = 0 will also have the same root at s = jωc for the same
value of the time delay τ . Consequently, the problem is now reduced to finding values of time delay τ such that
both ∆(s, τ ) = 0 and ∆( − s, τ ) = 0 have a common root at s = jωc . This result could be stated as follows:

P (jωc ) + Q(jωc )e−jωc τ = 0

. (3)
P ( − jωc ) + Q( − jωc )ejωc τ = 0

By eliminating the exponential terms in Eq. (3), the following new polynomial in ωc2 is obtained:

W (ωc2 ) = P (jωc )P ( − jωc ) − Q(jωc )Q( − jωc ) = 0

, (4)
= t12 ωc12 + t10 ωc10 + t8 ωc8 + t6 ωc6 + t4 ωc4 + t2 ωc2 + t0 = 0

where
t12 = p26 ; t10 = p25 − 2p6 p4 ;
t8 = p24 − q42 − 2p5 p3 + 2p2 p6 ;
t6 = p23 − q32 − 2p6 p0 + 2p5 p1 − 2p4 p2 + 2q4 q2 ; .
t4 = p22 − q22 + 2p4 p0 − 2p3 p1 − 2q0 q4 + 2q3 q1 ;
t2 = p21 − q12 − 2p2 p0 + 2q2 q0 ; t0 = p20 − q02 ;

It must be noted that the characteristic equation with an exponential term given in Eq. (2) is now transformed
into a regular polynomial not including any exponential terms given by Eq. (4). The positive real roots of Eq.
(4) correspond to the imaginary roots of Eq. (2) exactly. One could easily compute these real roots by standard
methods. Depending on the roots of Eq. (4), the following situations might be observed [13]:

1. The polynomial of Eq. (4) does not include any positive real roots. This implies that the characteristic
equation of Eq. (2) does not contain any roots on the jω -axis. Therefore, the excitation control system
is delay-independent stable for all finite delays τ ≥ 0 ;

2. The polynomial of Eq. (4) may have at least 1 positive real root. This implies that the characteristic
equation of Eq. (2) contains at least a pair of complex roots on thejω -axis. Thus, the excitation control
system is delay-dependent stable.

For a positive real root ωc of Eq. (4), the delay margin τ ∗ could be computed using Eq. (3) as follows [5,13]:
( )
a9 ωc9 +a7 ωc7 +a5 ωc5 +a3 ωc3 +a1 ωc
τ∗ = 1
ωc T an
−1
a10 ωc10 +a8 ωc8 +a6 ωc6 +a4 ωc4 +a2 ωc2 + 2rπ
ωc ;
(5)
r = 0, 1, 2, ....., ∞

5187
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

where
a10 = p6 q4 ; a9 = p5 q4 − p6 q3 ; a8 = p5 q3 − p4 q4 − p6 q2 ;
a7 = p4 q3 + p6 q1 − p3 q4 − p5 q2 ;
a6 = p6 q0 + p4 q2 + p2 q4 − p5 q1 − p3 q3 ;
a5 = p5 q0 + p3 q2 + p1 q4 − p4 q1 − p2 q3 ; .
a4 = p1 q3 + p3 q1 − p4 q0 − p2 q2 − p0 q4 ;
a3 = p0 q3 − p3 q0 − p1 q2 + p2 q1 ; a2 = p2 q0 + p0 q2 − p1 q1 ;
a1 = p1 q0 − p0 q1 ; a0 = −p0 q0

For any positive roots of Eq. (4), we should also investigate if at s = jωc , the root of Eq. (2) crosses the
imaginary axis with increasing τ . The necessary condition for the existence of roots crossing the imaginary axis
is that the critical characteristic roots cross the imaginary axis with a nonzero velocity defined as follows:
[ ]
ds
Re ̸= 0, (6)
dτ s=jωck

where Re(•) represents the real part of a complex variable. The sign of root sensitivity is defined as root
tendency (RT) [5,13,15]:
{ [ ] }
ds [ ]
RT |s=jωc = sgn Re = sgn W ′ (ωc2 ) , (7)
dτ s=jωc

where the prime represents differentiation with respect to ωc2 . The derivation of Eq. (7) could be found in [5].
The RT expression gives a practical tool to evaluate the direction of transition of the roots at s = jωc as τ
increases from τ1 = τ ∗ − ∆τ to τ2 = τ ∗ + ∆τ, 0 < ∆τ << 1. The root s = jωc crosses the imaginary axis
to either the unstable right half-plane when RT = +1 or to the stable left half-plane when RT = −1 . For the
excitation control system, by applying Eq. (7), we obtain the following equation that allows us to compute the
RT for each crossing frequency:

W ′ (ωc2 ) = 6t12 ωc10 + 5t10 ωc8 + 4t8 ωc6 + 3t6 ωc4 + 2t4 ωc2 + t2 . (8)

Another frequency domain method known as Rekasius substitution [14,15] could also be used to compute the
delay margins of the SMIB system. This method also eliminates the exponential term e−sτ in the characteristic
equation of Eq. (2) by using an exact substitution given by

1 − Ts
e−sτ = , (9)
1 + Ts

where T ∈ ℜ is called the pseudo-delay. Using this substitution, the transcendental characteristic Eq. (2) is
converted into a polynomial without transcendentality similar to the one given in Eq. (4) such that its purely
imaginary roots, determined by the Routh–Hurwitz stability criterion, coincide with the purely imaginary roots
of the characteristic Eq. (2) exactly. The corresponding delay margin is then determined by the following
formula [15]:
2 [ ]
τ∗ = T an−1 (ωc T ) ± rπ , r = 0, 1, 2, . . . (10)
ωc

5188
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

It must be noted that both the proposed method [13] and the Rekasius substitution method [15] eliminate the
transcendental term in the characteristic Eq. (2) using different substitutions that are both exact and obtain
a new polynomial without transcendentality. In the proposed method, the real roots of this new polynomial,
if they exist, coincide with the imaginary roots s = ±jω of the characteristic equation exactly. As compared
to the proposed method, the Rekasius substitution method requires the introduction of a pseudo-delay (T )
and an additional step, which is the Routh–Hurwitz stability criterion, to determine the pseudo-delay T
and the imaginary roots of the characteristic equation. Therefore, the Rekasius substitution method is not
a computationally attractive method when there is a single time delay term in the characteristic equation.
Moreover, our previous studies [20,30] on the stability of the time-delayed LFC systems clearly indicated that
the Rekasius substitution method gives the same delay margin results as the proposed method since they are
both exact methods. However, the proposed method cannot be used when independent unequal time delays
are observed in the PSS loop and voltage control loop. Such a delay situation is known as an incommensurate
delays case, and recent studies on the Rekasius substitution method show that this method could be extended
to delay margin computation of systems with multiple incommensurate delays [31,32].

4. Results
Case studies are performed for the SMIB system shown in Figure 1. Delay margins are computed for various
PSS gains in the range of KP = 0 − 30 and load values ranging from PL = 0.1pu to PL = 1.0pu. The damping
is chosen as D = 0 . The power factor of the load remains unchanged at pf = cos ϕ = 0.9 lagging during the
load variation. Other system parameters are as follows: xd = 1.60 pu, xq = 1.55 pu, x′d = 0.32 pu, M = 6.0,
′
Tdo = 6.0 s, ω0 = 377rad/s , KA = 50, TA = 0.05 s, T1 = 0.5 s, T2 = 0.1 s, Tw = 2.0 s, re = 0 , xe = 0.4 pu.
The variation of delay margin with respect to PSS gain for 3 different values of load (PL = 0.1, 0.3, and
0.5pu) is shown in Figure 5. It is obvious that the impact of PSS gain on the delay margin has 2 tendencies.
For small gains, the delay margin increases as the gain increases. This implies that the addition of PSS into the
excitation system improves system stability, since the delay margin increases for smaller gains. However, for
larger gains, the delay margin decreases as the gain increases, causing a less stable system. Delay margin results
are also presented in the Table for these load values. It must be mentioned here that the new characteristic Eq.
(4) has 3 real positive roots at each load level. For this reason, delay margins for all 3 real roots are computed
using Eq. (5) and presented in the Table. The minimum of 3 delay margins, shown in bold in the Table, is the
system delay margin. It is clear from the Table that the delay margin decreases for all values of PSS gain as
the load increases, indicating a less stable system.

0.24
0.22 PL = 0.1 p.u.
PL = 0.3 p.u.
0.2 PL = 0.5 p.u.
0.18
0.16
τ * (s)

0.14
0.12
0.1
0.08
0.06
0.04
0 5 10 15 20 25 30
KP

Figure 5. Variation of delay margin with respect to the PSS gain for different loads.

5189
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

Table. Delay margin with respect to load increase and PSS gain.

PL (pu)
τ (s)
0.1 0.3 0.5
0.2023 0.1789 0.1788
KP = 0
0.6263 0.5477 0.4579
0.3138 0.3518 0.3678
0.2230 0.1801 0.1632
KP = 5
0.4848 0.4303 0.3774
0.3525 0.4039 0.4262
0.2098 0.1549 0.1289
KP = 10
0.4580 0.3945 0.3539
0.3652 0.4247 0.4508
0.1962 0.1315 0.1010
KP = 15
0.4402 0.3735 0.3407
0.3764 0.4440 0.4738
0.1832 0.1114 0.0786
KP = 20
0.4260 0.3592 0.3320
0.3867 0.4623 0.4958
0.1710 0.0943 0.0600
KP = 25
0.4141 0.3488 0.3258
0.3964 0.4799 0.5171
0.1596 0.0796 0.0439
KP = 30
0.4038 0.3408 0.3214
0.4056 0.4970 0.5378

Figure 6 shows the variation of delay margin with respect to the load for 3 different values of PSS gain,
KP = 15, 20, and 25. Note that an increase in load results in a decrease in the delay margin for all PSS gains.
Delay margin curves shown in Figure 6 will enable us to determine how much the load could be increased for a
given delay such that the system preserves stability. For example, suppose that the maximum amount of delay
observed in the system is 0.1 s. From the delay curve shown in Figure 6 for KP = 15 , the corresponding amount
of load is then found to be PL = 0.51pu. This is the maximum loading point for stability. If the load is further
increased, the system will be unstable since the delay observed in the system becomes larger than the delay
margin. Evidently, the system will be stable for all selected load values if the delay is ignored in the system. As
a result, it is obvious that the delay reduces the maximum loadability of the system and the stability margin.

0.25

KP = 15
0.2
KP = 20
KP = 25
0.15
τ* (s)

0.1

0.05

0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
PL (p.u.)
Figure 6. Variation of delay margin with respect to the load for different PSS gains.

5190
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

The region under the delay curves could be considered as the stability region containing the pairs of time delay
and load (τ, PL ), for which the excitation system is stable.
Finally, using MATLAB/Simulink [33], time-domain simulations are carried out to verify the theoretical
delay margin results. Figure 7 shows the simulation model in Simulink. For simulation, load and PSS gain are
selected as PL = 0.5pu , KP = 20. The constant parameters K1 − K6 computed are given as: K1 = 1.0058,
K2 = 0.8441, K3 = 0.36, K4 = 1.0805, K5 = 0.0468 , K6 = 0.4991 . For this case, the new characteristic
equation in Eq. (4) has 3 real positive roots. These are ωc1 = 13.1187 , ωc2 = 11.0473 , and ωc3 = 2.5140 rad/s.
The corresponding delay margins and root tendencies (RT s) are computed as τ1 ∗ = 0.0786 s , τ2 ∗ = 0.3320 s ,
τ3 ∗ = 0.4958 s, and RT1 = +1, RT2 = −1, RT3 = +1 , respectively.

Figure 7. MATLAB/Simulink model of the SMIB system with AVR and PSS.

These results indicate that the system is stable when the delay is less than τ1 ∗ = 0.0786 s . Since
RT1 = +1, a pair of the stable complex conjugate roots of Eq. (2) moves in the left half of the complex
plane and crosses the imaginary axis at s = ±jωc1 = ±j13.1187rad/s for τ1 ∗ = 0.0786 s , passing to the right
half-plane. Therefore, the system becomes unstable for τ > τ1 ∗ = 0.0786 s. A simulation result illustrating this
case is shown in Figure 8 for 3 different delay values (τ = 0.07 s, τ = 0.0786s, and τ = 0.085s) . When the
time delay is further increased, the unstable complex conjugate roots of Eq. (2) move in the right half-plane
and cross the imaginary axis at s = ±jωc2 = ±j11.0473rad/s for τ2 ∗ = 0.3320 s , passing to the left half-plane
since the corresponding root tendency is negative, RT2 = −1. Please note that the system is unstable for
τ1 ∗ = 0.0786 s < τ < τ2 ∗ = 0.3320 s. It then becomes stable again for τ > τ2 ∗ = 0.3320 s. Figure 9 depicts
simulation results for terminal voltage around τ2 ∗ = 0.3320 s, clearly illustrating this stability exchange from
unstable to stable with an increase in time delay. Please note that the system is first unstable for τ = 0.32s ,
then marginally stable for τ2 ∗ = 0.3320 s, and then it becomes stable for τ = 0.35s . Finally, if we continue to
increase the time delay, a pair of the stable complex conjugate roots of Eq. (2) moves in the left half of the
complex plane and crosses the imaginary axis at s = ±jωc3 = ±j2.5140rad/s for τ3 ∗ = 0.4958 s, passing to the
right half-plane since the corresponding root tendency is positive, RT3 = +1 . Therefore, the system becomes
unstable for τ > τ3 ∗ = 0.4958 s and remains unstable if the delay is further increased. A simulation result
illustrating this case is shown in Figure 10 for 3 different delay values, τ = 0.48 s, τ = 0.4958s, and τ = 0.51s .
It is expected that the measurement delay observed in speed deviation in the PSS loop, which has been
neglected in theoretical delay margin computation, will adversely affect stability and will reduce the delay
margin. For this reason, simulation studies are carried out for the case when equal amounts of delay are

5191
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

introduced into both voltage and speed deviation loops. Figures 11a and 11b show the terminal voltage for
τ = 0.0786 s , τ = 0.02255 s , and τ = 0.024 s , respectively. It is clear from Figure 11a that when the same
amount of delay (τ = 0.0786 s) is added to both loops, the system becomes unstable. As seen from Figure 8,
the excitation control system is marginally stable at τ = 0.0786 s when the delay in the PSS loop is not taken
into account. With an additional delay in the PSS loop, the system becomes marginally stable at a lower delay
value, τ = 0.02255 s , as shown in Figure 11b. As a result, the delay margin is reduced from τ = 0.0786 s to
τ = 0.02255 s.

1.05 1.05
τ = 0.07 s τ = 0.32 s
1.04 τ = 0.0786 s 1.04
τ = 0.3320 s
1.03 τ = 0.085 s 1.03 τ = 0.35 s
1.02 1.02
VT (p.u.)

VT (p.u.)
1.01 1.01
1 1
0.99 0.99
0.98 0.98
0.97 0.97
0.96 0.96
0 5 10 15 0 5 10 15
Time (s) Time (s)
Figure 8. Generator terminal voltage for different time Figure 9. Generator terminal voltage for different time
delays for KP = 20 delay values around τ1 ∗ = 0.0786 s delays for KP = 20 and delay values around τ2 ∗ =
for PL = 0.5pu . 0.3320 s for PL = 0.5pu .

1.05
1.04 τ = 0.48 s
τ = 0.4958 s
1.03
τ = 0.51 s
1.02
VT (p.u.)

1.01
1
0.99
0.98
0.97
0.96
0.95
0 5 10 15 20 25 30 35 40 45 50
Time (s)
Figure 10. Generator terminal voltage for different time delays for KP = 20 and delay values around τ3 ∗ = 0.4958 s
when D = 0 and PL = 0.5pu .

5. Conclusions
This paper has investigated the impact of loading and PSS gain on the stability of the generator excitation
control system including a PSS with respect to the time delay using an analytical method. A useful formula
is developed to compute delay margins for stability. Delay margins are computed for a wide range of PSS
gain and load values. It is observed that the delay margin decreases with an increase in PSS gain and system
load demand. Time-domain simulations clearly indicate that the proposed method correctly estimates delay
margins. Finally, it is observed that the maximum loadability of the system decreases when time delays are
taken into account in stability analysis. Together with the maximum loading point, delay margins should be
used as a stability index to measure the degree of stability and to determine stability margins.

5192
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

2 1.02
τ = 0.0786 s τ = 0.02255 s
1.015 τ = 0.024 s
1.5
1.01

1 1.005
VT (p.u.)

VT (pu)
1

0.5
0.995

0.99
0
0.985
a b
–0.5 0.98
0 0.5 1 1.5 2 2.5 0 2 4 6 8 10
Time (s) Time (s)
Figure 11. Generator terminal voltage for KP = 20 , D = 0, and PL = 0.5pu when a delay is added into PSS loop.

As future work, the proposed method will be extended to multimachine power systems whose dynamics
are described by ordinary differential or differential-algebraic equations. To do that, load buses should be
eliminated and a linear time-delayed state-space equation model around an operating point should be obtained.
Moreover, an additional delay, which is different from the delay in the voltage control loop, will be added to
the PSS loop, and delay margins will be computed using the Rekasius substitution method.

References

[1] Naduvathuparambil B, Valenti MC, Feliachi A. Communication delays in wide area measurement systems. In:
Proceedings of the 34th Southeastern Symposium on System Theory; 18–19 March 2002; Huntsville, Alabama.
Piscataway, NJ, USA: IEEE. pp. 118-122.
[2] Xia X, Xin Y, Xiao J, Wu J, Han Y. WAMS applications in Chinese power systems. IEEE Power Energy M 2006;
4: 54-63.
[3] Wu H, Tsakalis K, Heydt GT. Evaluation of time delay effects to wide-area power system stabilizer design. IEEE
T Power Syst 2004; 19: 1935-1941.
[4] Ayasun S, Gelen A. Stability analysis of a generator excitation control system with time delays. Electr Eng 2010;
91: 347-355.
[5] Ayasun S. Computation of time delay margin for power system small-signal stability. Euro Trans Electr Power 2009;
19: 949-968.
[6] Saadat H. Power System Analysis. New York, NY, USA: McGraw-Hill, 1999.
[7] Kundur P. Power System Stability and Control. New York, NY, USA: McGraw-Hill, 1994.
[8] Sauer PW, Pai MA. Power System Dynamics and Stability. 1st Indian Reprint. Singapore: Stipes Publishing, 2002.
[9] Phadke AG. Synchronized phasor measurements in power systems. IEEE Comput Appl Pow 1993; 6: 10-15.
[10] Liu M, Yang L, Gan D, Wang D, Gao F, Chen Y. The stability of AGC systems with commensurate delays. Euro
Trans Electr Power 2007; 17: 615-627.
[11] Jiang L, Yao W, Wu QH, Wen JY, Cheng SJ. Delay-dependent stability for load frequency control with constant
and time-varying delays. IEEE T Power Syst 2012; 27: 932-941.
[12] Yao W, Jiang L, Wu QH, Wen JY, Cheng SJ. Wide-area damping controller of FACTS devices for inter-area
oscillations considering communication time delays. IEEE T Power Syst 2014; 29: 318-329.

5193
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

[13] Walton KE, Marshall JE. Direct method for TDS stability analysis. IEE Proc-D 1987; 134: 101-107.
[14] Rekasius ZV. A stability test for systems with delays. In: Proceedings of the Joint Automatic Control Conference;
13–15 August 1980; San Francisco, CA, USA. New York, NY, USA: American Society of Mechanical Engineers.
[15] Olgaç N, Sipahi R. An exact method for the stability analysis of time-delayed linear time-invariant (LTI) systems.
IEEE T Autom Control 2002; 47: 793-797.
[16] He Y, Wang QG, Xie LH, Lin C. Further improvement of free-weighting matrices technique for systems with
time-varying delay. IEEE T Autom Control 2007; 52: 293-299.
[17] Wu M, He Y, She JH, Liu GP. Delay-dependent criterion for robust stability of time-varying delay systems.
Automatica 2004; 40: 1435-1439.
[18] Xu SY, Lam J. On equivalence and efficiency of certain stability criteria for time-delay systems. IEEE T Autom
Control 2007; 52: 95-101.
[19] Yao W, Jiang L, Wu QH, Wen JY, Cheng SJ. Delay-dependent stability analysis of the power system with a
wide-area damping controller embedded. IEEE T Power Syst 2011; 26: 233-240.
[20] Sönmez Ş, Ayasun S, Nwankpa CO. An exact method for computing delay margin for stability of load frequency
control systems with constant communication delays. IEEE Trans Power Syst 2016; 31: 370-377.
[21] Ayasun S. Stability analysis of time-delayed DC motor speed control systems. Turk J Elec Eng Comp Sci 2013; 21:
381-393.
[22] Jalili N, Olgac N. Multiple delayed resonator vibration absorbers for multi-degree-of-freedom mechanical structures.
J Sound Vib 1999; 223: 567-585.
[23] Filipovic D, Olgac N. Delayed resonator with speed feedback-design and performance analysis. Mechatronics 2002;
12: 393-413.
[24] Ji JC. Stability and bifurcation in an electromechanical system with time delays. Mech Res Commun 2003; 30:
217-225.
[25] Yan XP, Li WT. Hopf bifurcation and global periodic solutions in a delayed predator-prey system. Appl Math
Comput 2006; 177: 427-445.
[26] Yan XP, Chu YD. Stability and bifurcation analysis for a delayed Lotka–Volterra predator–prey system. J Comput
Appl Math 2006; 196: 198-210.
[27] Song Y, Peng Y. Stability and bifurcation analysis on a logistic model with discrete and distributed delays. Appl
Math Comput 2006; 181: 1745-1757.
[28] Shahgholian G, Faiz J. The effect of power system stabilizer on small-signal stability in single-machine-infinite-bus.
International Journal of Electrical and Power Engineering 2010; 4: 45-53.
[29] Falehi AL, Rostami M, Doroudi A, and Ashrafian A. Optimzayion and coordination of SVC-based supplementary
controllers and PSSs to improve power system stability using a genetic algorithm. Turk J Elec Eng Comp Sci 2012;
20: 639-654.
[30] Sönmez Ş, Ayasun S, Eminoğlu U. Computation of time delay margins for stability of a single-area load frequency
control system with communication delays. WSEAS Trans Power Syst 2014; 9: 67-76.
[31] Fazelina H, Sipahi R, Olgaç N. Stability robustness analysis of multiple time-delayed systems using building block
concept. IEEE T Autom Control 2007; 52: 799-810.
[32] Liu Z, Zhu C, Miao Y, Jiang Q. Stability analysis of multiple time delayed power systems using building block
concept. In: Proceedings of 2008 IEEE Power Engineering Society General Meeting; 20–24 July 2008; Pittsburgh,
PA, USA. Piscataway, NJ, USA: IEEE. pp. 1-8.
[33] MathWorks Inc. Simulink. Model-Based and System-Based Design: Using Simulink. Natick, MA, USA: MathWorks,
2000.

5194
 SÖNMEZ and AYASUN/Turk J Elec Eng & Comp Sci

Appendix
The coefficients of P (s) and Q(s) polynomials in Eq. (2) are given as follows:

′
p6 = TA Td0 Tw M ω0 T2

′ ′ ′ ′
p5 = Td0 Tw M ω0 T2 + K3 TA Tw M ω0 T2 + TA Td0 M ω0 T2 + Td0 Tw DM ω0 T2 + TA Td0 Tw M ω0
′ ′ ′
p4 = K3 Tw M ω0 T2 + Td0 M ω0 T2 + Td0 Tw Dω0 T2 + Td0 Tw M ω0 + K3 TA M ω0 T2 + K3 TA Tw Dω0 T2 +
′ ′ ′ ′
K3 TA Tw M ω0 + TA Td0 Dω0 T2 + TA Td0 M ω0 + TA Td0 Tw Dω0 + K1 T2 TA Tw Td0 ω02

′ ′ ′
p3 = K3 M ω0 T2 + K3 Tw T2 Dω0 + K3 Tw M ω0 + T2 Td0 Dω0 + Td0 M ω0 + Td0 Tw Dω0 + K3 DTA T2 ω0 +
′ ′ ′ ′
K3 TA M ω0 + K3 TA Tw Dω0 + TA Td0 Dω0 + K1 T2 TA Td0 ω02 + K1 Tw TA Td0 ω02 + K1 T2 Tw Td0 ω02 +
K1 K3 T2 Tw TA ω02 + K2 KP KA T1 Tw ω0 − K4 K2 T2 Tw TA ω02

′ ′
p2 = K3 T2 Dω0 + K3 M ω0 + K3 Tw Dω0 + Td0 Dω0 + K3 DTA ω0 + K1 TA Td0 ω02 +
′ ′
K1 T2 Td0 ω02 + K1 K3 T2 TA ω02 + K1 Tw Td0 ω02 + K1 K3 Tw TA ω02 + K1 K3 T2 Tw ω02 +
K2 KP KA Tw ω0 − K4 K2 T2 TA ω02 − K4 K2 Tw TA ω02 − K4 K2 T2 Tw ω02

′
p1 = K3 Dω0 + K1 ω02 Td0 + K1 ω02 K3 TA + K1 (ω02 )T2 K3 + K1 ω02 Tw K3 − K4 K2 ω02 TA −
K4 K2 ω02 T2 − K4 K2 ω02 Tw

p0 = K1 ω02 K3 − K4 K2 ω02

q4 = K6 KA M T2 Tw ω0

q3 = K6 KA DT2 Tw ω0 + K6 KA M T2 ω0 + K6 KA Tw M ω0

q2 = K6 KA M ω0 + K6 KA DT2 ω0 + K6 KA DTw ω0 + K1 K6 KA T2 Tw ω02 − K2 K5 KA T2 Tw ω02

q1 = K6 KA Dω0 + K1 K6 KA T2 ω02 + K1 K6 KA TW ω02 − K2 K5 KA T2 ω02 − K2 K5 KA Tw ω02

q0 = K1 K6 KA ω02 − K2 K5 KA ω02 .

View publication stats