Prevention of Power System Blackouts

Using Windowed-Prony Method

Kasimala Venkatanagaraju Monalisa Biswal
Department of Electrical Engineering Department of Electrical Engineering
National Institute of Technology National Institute of Technology
Raipur - 492010, India Raipur - 492010, India
[email protected] [email protected]

Abstract – Zone-3 unit of conventional distance relay is zone-3 results into relay maloperation. To mitigate this issue
vulnerable to voltage instability and load-encroachment which and to assure the secure operation of the relay, an adaptive
are known as system stress conditions. For such conditions, the technique is proposed by using the rate of change of voltage
impedance trajectory encroaches zone-3 reach. Based on the [4]. Control strategies are also implemented in [5] for
setting parameters of the relay, it is identified as a fault and enhancing the relay performance. A voltage stability index
relay may maloperate. However, the relay may maloperate for [6] and a synchrophasor based voltage instability monitoring
any three-phase fault during such stress events as the change in index [7] are proposed for early detection of voltage
system parameters are identical for both the conditions. Zone-3 instability. However, such methods may fail to detect the
maloperation is the root cause behind 2012 Indian blackout.
fault during voltage instability. In [8], steady state and
So, it is necessary to introduce a supporting algorithm to avoid
relay maloperations and to curtail the occurrence of any
transient components through state diagram are utilized to
blackout. This paper introduces windowed-Prony method to discriminate fault from other power system events.
detect and discriminate fault from system stress conditions. However, it is difficult to extract high frequency transient
This method will be activated when the impedance trajectory components precisely during parallel feeders switching.
enters the setting zone-3 region. It is used to approximate the Rate of change of voltage and current based approach [9]
input signal and to estimate the error existing between actual and active power-based technique [10] are utilized to
and approximated signals. The squared error is considered as distinguish the fault from the voltage degraded conditions.
a fault detection index (FDI). The performance of this method This may not work effectively due to large load switching.
is tested on IEEE 39 bus test system.
During heavily loaded situation, zone-3 reach suffers
Keywords—Zone-3, catastrophic consequences, blackout, from encroachment problem and hence relay maloperates.
voltage instability, load encroachment. To address the aforementioned issue an adaptive protection
scheme based on steady state security analysis is proposed
in [11]. A communication assisted detection scheme [12] is
I. INTRODUCTION suggested based on the synchrophasor data to distinguish
The modern power system is highly complex and three-phase fault from load encroachment. Synchrophasor
integrated. So, power failure in any section may translated to based techniques are reported in [13]-[14] for load
the interconnected areas and this may further lead to large encroachment condition. Even though, all these methods are
economic loss. Repetition of such events may further affect efficient but latency issue due to communication channel
the sustainable growth of the society. Grid disturbances or cannot be avoided.
blackouts happen only due to the occurrence of cascaded Only few algorithms [15]-[22] are able to discriminate
failures. These blackouts are rare but substantially hazardous system stress events from three-phase fault. Synchrophasor
due to their catastrophic consequences. For instance, 2003 based technique [15]-[16], transient monitoring function and
North America blackout affected approximately 50 million impedance angle based technique [17], support vector
people and 61,800 MW load; 2003 Italy blackout affected 58 machines based protection scheme [18], rate of change of
million people and 1421 MW load; 2009 Brazil blackout active power based technique [19], wavelet coefficients
affected 700 million people and 24,450 MW load; further, based approach [20]-[21] and Teager energy-based approach
2012 India blackout affected 700 million people and [22] are proposed with fixed threshold setting to detect the
40,000 MW load [1]. As per the NERC report, 70% of the three-phase fault during system stressed conditions.
disturbances are pertaining to protection failures.
Maloperation of the distance relay plays a vital role in the In this paper a Windowed-Prony analysis based
occurrence of catastrophic consequences [2]-[3]. Distance supporting algorithm is introduced to prevent zone-3
relays are widely used to provide primary as well as back-up maloperation during system stress conditions and three-
protection to the transmission lines. However, they may phase fault. This algorithm utilizes the instantaneous current
cause nuisance operations under system stress conditions signals as the actual input signals and it approximate them.
such as voltage degraded and heavy loading situations. The The fault detection index (FDI) is defined based on squared
reason for that is inherent protective zone settings of error. It will be demonstrated that when three-phase fault
distance relay. Particularly, zone-3 unit of distance relay has occurs during voltage instability or load-encroachment, the
long reach area and large-time delay that makes it FDI will be more than the threshold. Whereas, the FDI will
vulnerable to system stress events. be less than the threshold during pure voltage instability or
load-encroachment conditions. Based on this feature the
An increase in reactive power demand causes a decrease three-phase fault can be reliably detected during system
in system voltage. Further with the continuation of such an stress events. The methodology and simulation results are
event, the impedance trajectory gradually enters into the discussed in subsequent sections.

978-1-7281-6664-3/20/$31.00 ©2020 IEEE

Authorized licensed use limited to: National Institute of Technology. Downloaded on December 25,2020 at 01:49:51 UTC from IEEE Xplore. Restrictions apply.
 II. METHODOLOGY  z10 z20  zs0   h1   x(1) 
 1    
A. Protective Zone Settings of the Distance Relay  z1 z12  z1s   h2   x(2)  (6)
=
Settings for all the protective zones of a distance relay          
are to be set as follows;  s −1 s −1     
 z1 z2s −1  zs   hs   x( s) 
• Zone-1 reach K1 is set to cover 80% of the protected
line length and with instantaneous operation.
These equations can be utilized for determining the
• Zone-2 reach K2 is set to cover 100% of the first amplitude vectors. Next, the roots of the polynomial are
protected line plus 20% of the next longest line. The formed based on the known amplitudes. Therefore, the
timer T2 is set to a time delay of 0.35 s. polynomial is represented as
• Zone-3 reach K3 is set to cover 100% of the first
s
protected line plus 100% of the adjacent shortest
line. The timer T3 is set to a time delay of 1 s.
F (z) =  a(l ) zs −l = a(0) zs + a(1) zs −1 + ..... + a(s − 1) z+ a(s) (7)
l =0
Whenever the positive sequence impedance seen by the
distance relay will be less than the set value of K3 then the The parameter a(l) is multiplying on both sides of (5) and k
Prony based supporting algorithm will get activated. is shifted to l.

B. Windowed-Prony Method s

Let x(k) be sampled sinusoidal input signal. It can be

a(l ) x(k − l ) = a (l )  hn znk −l −1 (8)
n =1
approximated by s exponential functions and represented as
Another form of (7) can be represented as
s
y (k ) =  Ane (α n + jωn )(n −1)Ts + jψ n
k = 1, 2, 3, ……. , N (1)
s s  s 
n =1
 a(l ) x(k − l ) =  hn znk − s  a(l) zns −l −1  (9)
l =0 n =1 l = 0 
where, Ts is sampling rate, An is the amplitude, αn is damping
factor, ωn is the angular velocity and ψn is initial phase. The
As per (6), the RHS in (7) is identified as the polynomial
concise form of the approximated signal is expressed as
and estimated at each of its roots zn results zero.
s
y (k ) =  hn znk −1 (2) s
 a(l ) x(k − l ) = 0 (10)
n =1
l =0

where hn = An e jψ n , zn = e(α n + jωn )Ts The equation (6) and (9) can be resolved for determining
coefficients and roots of the polynomial. The damping
Over the N data samples, the y(k) is compared with x(k) factor and frequencies are estimated from roots zn. In
and gives the error ε(k). The estimation problem depends on practice, the number of data samples generally surpasses the
the minimization of squared error. minimum number needed to fit the exponentials of a model.
When the over-determined data situation is exists then the
s equation (9) should be modified as
e( k ) = x ( k ) − y ( k ) = x ( k ) −  hn znk −1 (3)
n =1 s

N
 a(l ) x(k − l ) = e(k ) (11)
l =0
η=  e( k ) 2
(4)
k =1
Therefore, the total squared error can be computed and is
used for estimating the problem.
However, it becomes a complex non-linear problem.
This problem is resolved by the Windowed-Prony method r
[23] which employs linear equation solutions. The FDI =  e(k )
2
(12)
Windowed-Prony method approximate the exact k = s +1
exponential fit to the data samples.
Consider the s-exponent discrete time function x(k). The The flow chart of the proposed method is shown in Fig. 1.
set of linear equations can be expressed in a matrix form as
follows; III. RESULTS AND DISCUSSIONS
s The IEEE 39-bus test system as shown in Fig. 2
x( k ) =  hn znk −1 (5) simulated using EMTDC/PSCAD and data extracted at
different buses are used for result analysis. The operating
n =1
voltage and frequency of the test system is considered as
400 kV and 50 Hz. The voltage and current signals are

Authorized licensed use limited to: National Institute of Technology. Downloaded on December 25,2020 at 01:49:51 UTC from IEEE Xplore. Restrictions apply.
obtained from CT (1000:1 A) and PT (440 kV:110 V). They increase in the load demand, the voltage of the buses
are sampled at the rate of 1 kHz sampling frequency. The reduced to 0.85 p.u. at 4.5 s. The proportionate change is
line 26-27 of length 215 km is being securely protected by also observed in the respective current signals. At the relay
distance relay (R). The relay located near bus 26 is modelled location, change in phase-a current is similar to the other
with quadrilateral characteristics. Parameters of the phases as observed from Fig. 3.
protected line are provided in the Appendix. The reach
setting of zone-1, zone-2 and zone-3 are set to K1=55.05 Ω,
K2=82.58 Ω and K3=165.17 Ω respectively. A three-phase G8
fault (F) is created at 330 km from the relay location as 37 28
indicated in Fig. 2. Based on the various simulated test G10 26 29
R
cases, the threshold is set to 0.1 for reliable and quick 25
30 38
detection of faults. Two-cycle moving window is used for
the analysis. The performance of the proposed supporting 24 G9
27
algorithm is examined for voltage instability, load 2 F
encroachment and three-phase fault cases. 18
1 G6
LOAD
G1 17 16
35
Acquire voltage and current signals 3 LOAD
15
from the relay location 14
39 4 LOAD 21 22
5 12
Compute voltage and current phasors
6 19
using full-cycle DFT
7 23
13
8 11
Compute apparent impedance (Zapp) 10 20 33 36
9 31
34 G4 G7
G2 32
No Is
Zapp < K3 G3 G5

Yes Fig. 2. IEEE 39-bus New England test system.

Estimate the Prony approximated
current signals
Phase-a
Compute squared error (FDI)
ia (kA)

No Is
FDI > 0.1

Yes
Phase-b
Enable zone-3 timer (T3)
ib (kA)

Delay elapsed

Yes
Phase-c
No Is
Zapp < K3
ic (kA)

Yes
TRIP

Fig. 1. Flow chart for supporting algorithm. Time (s)

Fig. 3. Instantaneous three-phase currents available at relay (R) location
under a three-phase fault during voltage instability.

A. Voltage degraded condition

Under such voltage degraded conditions, the apparent
To simulate this condition, the load buses 27, 18, and 16 impedance Zapp seen by the distance relay R moves towards
as highlighted in Fig. 2 are selected. The reactive power the protective zones. Exactly at t = 4 s, the trajectory of Zapp
demand at these busses is increased gradually with a step entered into the reach of zone-3 and the Zapp is less than K3
change of 0.5 s from 1.5 s to 4.5 s. Subsequently, voltages at as noticed from Fig. 4. The apparent impedance measured at
these buses are reduced gradually and the lower limit is different instances is also provided in TABLE I. During
maintained as per NERC standard i.e. 0.85 p.u.. So, due to non-faulty condition also the Zapp trajectory stays within

Authorized licensed use limited to: National Institute of Technology. Downloaded on December 25,2020 at 01:49:51 UTC from IEEE Xplore. Restrictions apply.
zone-3 reach and for more than the delay time period T3. The quadrilateral characteristics of zone-3 has sufficient
So, for such a condition, the chance of distance relay and independent resistive reach. During this overloading
maloperation is very high. A fault F is created at t = 5 s condition, trajectory of Zapp encroaches the reach of zone-3
during voltage stress. Now for the fault case, the computed at t = 3.7 s and the Zapp is less than K3 as seen in Fig. 7.
Zapp trajectory (dotted line) lies inside the setting Therefore, the relay R misunderstood this non-faulty
characteristics of zone-3 reach, so clear discrimination condition as a fault and may maloperate. A fault F is created
between fault and voltage instability event is difficult. at t = 5.1 s during overloading condition. The Zapp trajectory
Hence, for such a condition the relay will no longer be (dotted line) stays within zone-3 reach but, the relay R fails
secured. to detect the fault and susceptible to maloperation. Table II
provides the apparent impedance measured at various time
intervals.

ia

Current (kA)
iap
Zapp at
X (Ω)

t=4s

(a)

FDI
Time (s)
R (Ω)

Fig. 4. Trajectory of Zapp passing through zone-3 during voltage

instability and three-phase fault. (b)

TABLE I. IMPEDANCE MEASUREMENT AT RELAY LOCATION (R)

Relay output

UNDER VOLTAGE INSTABILITY

S. No. Time (s) Zapp (Ω) K3 (Ω) Status

1. 4.00 162.50 165.17 Zapp < K3
2. 4.38 159.54 165.17 Zapp < K3
(c)
3. 5.00 155.82 165.17 Zapp < K3
Time (s)
4. 5.01 91.68 165.17 Zapp < K3
Fig. 5. Response of the Windowed-prony method under three-phase fault
5. 6.00 85.63 165.17 Zapp < K3
during voltage instability.
6. 6.07 85.78 165.17 Zapp < K3

To overcome relay maloperation under voltage degraded

Phase-a
condition, the Prony based supporting algorithm is
introduced. Results of the supporting algorithm are shown in
ia (kA)

Fig. 5. Fig. 5(a) shows the original (ia) and Prony

approximated current (iap) signals of phase-a at each time
window. This figure reveals that the error between the
original and approximated current signals is large during a
fault. From Fig. 5(b), it is observed that the index FDI is
less than 0.1 during voltage degraded condition. However, it Phase-b
is greater than 0.1 during fault F. Based on this criterion, the
ib (kA)

fault F is detected at t = 5.001 s then the relay R issued a

TRIP signal at t = 6.001 s as shown in Fig. 5(c). From the
discussions, the supporting algorithm is able to avert zone-3
maloperation during voltage instability condition.

B. Load encroachment Phase-c

Overloading of the transmission lines may occur due to
ic (kA)

generation shift or change in system configuration. Thus,

directly influences on impedance seen by the relay. To
simulate this condition, the load buses 27, 18, and 16 as
shown in Fig. 2 are selected. The active power flowing
through the protected line 26-27 is gradually increased by
changing loads from 2 s to 4.9 s. Exactly at t = 4.9 s, Time (s)
overloading condition met the NERC loading criteria. Fig. 6 Fig. 6. Instantaneous three-phase currents available at relay (R) location
shows the incremental change in three-phase currents. under a three-phase fault during load encroachment.

Authorized licensed use limited to: National Institute of Technology. Downloaded on December 25,2020 at 01:49:51 UTC from IEEE Xplore. Restrictions apply.
 According to Fig. 1, the relay R detected the fault F at
t = 5.104 s and then issued TRIP command at t = 6.104 s as
shown in Fig. 8(c). Therefore, zone-3 maloperation under
load encroachment or fault during load encroachment is
avoided by using this supporting algorithm.
X (Ω)

IV. CONCLUSIONS
Zapp at In this paper, a supporting algorithm is introduced for
t = 3.7 s the conventional distance relay to avoid zone-3
maloperation during system stress conditions and three-
phase fault. From the simulated results it is confirmed that
the exact detection and discrimination between the fault and
Time (s) stress conditions is possible. The main advantage of this
R (Ω)
method is it improves the dependability and security
Fig. 7. Trajectory of Zapp passing through zone-3 during load features of the conventional distance relay. The
encroachment. maloperation of zone-3 is avoided. Consequently, the
catastrophic consequences are also curtailed.
TABLE II. IMPEDANCE MEASUREMENT AT RELAY LOCATION (R)
UNDER LOAC ENCROACHMENT
APPENDIX
S. No. Time (s) Zapp (Ω) K3 (Ω) Status Transmission line parameters:
1. 3.70 141.60 165.17 Zapp < K3 R = 0.03293 Ω/km
2. 4.00 136.89 165.17 Zapp < K3 XL = 0.3184 Ω/km
3. 5.00 125.08 165.17 Zapp < K3 XC = 3.56×10-6 Ω/km
4. 5.10 124.94 165.17 Zapp < K3
5. 6.00 105.48 165.17 Zapp < K3 Positive sequence impedance of a transmission line
6. 6.20 106.00 165.17 Zapp < K3 = (0.03293+j0.3184) Ω/km

Fig. 8 shows response of the Windowed-Prony method. Zones reach settings:

For instance, analysis of phase-a is considered. During load K1 = 0.8 × 215 × (0.03293+j 0.3184)=55.05∠84.090 Ω
encroachment the Prony approximated current signal (iap) is
exactly replica of the actual current signal (ia). Whereas in K2 = 1.2 × 215 × (0.03293+j 0.3184)=82.58∠84.090 Ω
fault condition both the signals differ from each other and
introduces error as seen in Fig. 8(a). Fig. 8(b) shows the K3 = 1.2 × (215+215) × (0.03293+j 0.3184)=165.17∠84.090 Ω
amplitude of FDI which is defined based on squared error.
The value of index FDI is less than 0.1 during load
ACKNOWLEDGMENT
encroachment and is greater than 0.1 during fault.
The authors would like to acknowledge the Department
of Electrical Engineering, NIT Raipur, for providing the
ia facilities to conduct this research work.
Current (kA)

iap
REFERENCES
[1] M. R. Aghamohammadi, S. Hashemi, A. Hasanzadeh, "A new
approach for mitigating blackout risk by blocking minimum critical
distance relays," Int. J. Electr. Power Energy Syst., vol. 75, pp. 162-
(a) 72, 2016.
[2] North American Electric Reliability Council. Disturbances Analysis
Working Group Database. [Online]. Available: http://www.nerc.com/
~dawg/database.html
FDI

[3] K. Venkatanagaraju and M. Biswal, “Mitigation of power system

blackout with microgrid system,”. In: Microgrid: Operation, Control,
Monitoring and Protection. Lecture Notes in Electrical Engineering,
vol 625. Springer, Singapore, 2020.
(b) [4] M. Jonsson and J. E. Daalder, “An adaptive scheme to prevent
undesirable distance protection operation during voltage instability,”
IEEE Trans. Power Del, vol. 18, no. 4, pp. 1174–1180, Oct. 2003.
Relay output

[5] H. Song, B. Lee, and V. Ajarappu, “Control strategies against voltage

collapse considering undesired relay operations,” in Proc. Inst. Eng.
Technol., Gener. Transm. Distrib., vol. 3, pp. 164–172, 2009.
[6] L. D. Arya, S. C. Choube, and M. Shrivastava, “Technique for
voltage stability assessment using newly developed line voltage
(c) stability index,” Energy Convers. Manage., vol. 49, no. 2, pp. 267–
Time (s) 275, 2008.
[7] R. Sodhi, S. C. Shrivastav, and S. N. Singh, “A simple scheme for
wide area detection of impending voltage insatbility,” IEEE Trans.
Fig. 8. Response of the Windowed-Prony method under three-phase
Smart Grid, vol. 3, no.2, pp. 818 – 827, Jun. 2012.
fault during load encroachment.

Authorized licensed use limited to: National Institute of Technology. Downloaded on December 25,2020 at 01:49:51 UTC from IEEE Xplore. Restrictions apply.
[8] C. H. Kim, J. Y. Heo, and R. K. Aggarwal, “An enhanced zone 3
algorithm of a distance relay using transient components and state
diagram,” IEEE Trans. Power Del., vol. 20, no. 1, pp. 39–46, Jan.
2005.
[9] M. Sharifzadeh, H. Lesani and M. Sanaye-Pasand, “A new algorithm
to stabilize distance relay operation during voltage-degraded
conditions,” IEEE Trans. Power Del., vol. 29, no. 4, pp. 1639–1647,
Aug. 2014.
[10] M. Biswal and K. V. Nagaraju, “Active power based method to
prevent zone-3 maloperation during voltage instability,” in Proc.
IEEE international conference on Control, Instrumentation,
Communication and Computational Technologies (ICCICCT),
December 16-17, 2016.
[11] M. Jin and T. S. Sidhu, “Adaptive load encroachment prevention
scheme for distance protection,” Elect. Power Syst. Res., vol. 78, pp.
1693–1700, 2008.
[12] A. El-Hadidy and C. Rehtanz, “Blocking of distance relays zone 3
under load encroachment conditions - A new approach using phasor
measurements technique,” in Proc. 14th Int. Middle East Power
Systems Conf., Cairo, Egypt, Dec. 19–21, 2010.
[13] D. Pal, B. Mallikarjuna, R. J. Reddy, M. J. Bharata Reddy, and D.K.
Mohanta, “Synchrophasor assisted adaptive relaying methodology to
prevent zone-3 mal-operation during load encroachment,” IEEE
Sensors Journal, vol. 17, no. 23, pp. 7713–7722, Dec. 2017.
[14] T. G. Bolandi, M. R. Haghifam, and M. Khederzadeh, "Real-time
monitoring of zone 3 vulnerable distance relays to prevent
maloperation under load encroachment condition,” IET Gener.
Transm. Distrib., vol. 11, no. 8, pp. 1878-1888, 6 1 2017.
[15] P. Kundu and A. K. Pradhan, “Synchrophasor-assisted zone 3
operation,” IEEE Trans. on Power Del., vol. 29, no. 2, pp. 660–667,
April. 2014.
[16] K. Venkatanagaraju and M. Biswal. “Synchrophasor-Based Wide
Area Protection Scheme for System Stressed Conditions to Avoid
Power System Blackout: Wide Area Protection Scheme for System
Stressed Conditions.” Handbook of Research on Smart Power System
Operation and Control. IGI Global, pp. 1-34. Nov. 2019.
doi:10.4018/978-1-5225-8030-0.ch001
[17] P. Nayak, A. K. Pradhan, and P. Bajpai, “Secured zone-3 protection
during stressed condition,” IEEE Trans. Power Del., vol. 30, no. 1,
pp. 89-96, Feb. 2015.
[18] K. Seethalekshmi, S. N. Singh, and S. C. Srivastava, “A classification
approach using support vector machines to prevent distance relay
maloperation under power swing and voltage instability,” IEEE
Trans. Power Del., vol. 27, no. 3, pp. 1124– 1133, Jul. 2012.
[19] Shalini, S. R. Samantaray, and A. Sharma, “Supervising zone-3
operation of the distance relay using synchronised phasor
measurements,” IET Gener. Transm. Distrib., vol. 13, no. 8, pp. 1238-
1246, 2019.
[20] K. Venkatanagaraju and M. Biswal, “Wavelet transform based zone-3
protection algorithm,” in Proc. International Conference on Electrical,
Electronics, Computers, Communication, Mechanical and Computing
(EECCMC), January 28-29, 2018.
[21] T. Jose, M. Biswal, K. Venkatanagaraju, and O. P. Malik, “Integrated
approach based third zone protection during stressed system
conditions,” Electr. Power Syst. Res, vol. 161, pp. 199-211, Aug.
2018.
[22] K. Venkatanagaraju, T. Jose, M. Biswal, and O. P. Malik, “Third zone
protection to discriminate symmetrical fault and stressed system
conditions,” Int. Trans. Electr. Energ. Syst., vol. 29, no. 2, Dec. 2019.
doi:10.1002/2050‐7038.12121
[23] T. Lobos, J. Rezmer, and H. J. Koglin, ‘‘Analysis of power system
transients using wavelet and Prony method,’’ in Proc. IEEE Power
Tech Conf., Porto, Portugal, Sep. 2001.

Authorized licensed use limited to: National Institute of Technology. Downloaded on December 25,2020 at 01:49:51 UTC from IEEE Xplore. Restrictions apply.