Event Detection and Localization in Active Distribution Networks using µPMUs

Full text

Event Detection and Localization in Active
Distribution Networks using µPMUs
Rajarshi Dutta, Shreyasi Som, Saikat Chakrabarti,
Ankush Sharma
Department of Electrical Engineering
Indian Institute of Technology
Kanpur, India 208016
Email: rdutta@iitk.ac.in, shreyass@iitk.ac.in
Anurag K Srivastava
Washington State University
Pullman, Washingtion, 99163
Email: anurag.k.srivastava@wsu.edu
Abstract—Event detection and localization is at the heart of
all automated system restoration processes. Localizing an event
can help in alleviating the root cause behind such disturbances.
During an event, the operating states of the power distribution
network may undergo significant changes. These variations are
more prominent in buses which are close to the source of the
event. Thus, events in a distribution system can be localized by
analyzing the changes in the system states. This paper proposes
an event locator based on the results of a event triggered
distribution network state estimator. The state estimation is
performed using measurements from a limited number of micro
phasor measurement units (µPMUs). An l1regularization based
state estimator is designed to estimate the change in system states,
using limited number of µPMU measurements. The results of the
state estimation process are further analyzed to locate any event.
The algorithm is designed for active distribution networks with
radial and meshed topologies. The proposed method is validated
on a 13-node test distribution feeder simulated using OPAL-RT
real-time simulator.
Index Terms—Active distribution network, event locator,
µPMU.
I. INTRODUCTION
The drive for clean and sustainable energy has lead to an
increase in the penetration of distributed energy resources
(DERs) in the conventional power grid. The increased level
of DERs penetration and installation of controllable loads in
distribution networks has changed the conventional passive
distribution system into a dynamic network of active sources
and interacting loads. Thus, an active distribution network has
high probability of being subjected to large variations. The
variations in network operating conditions may occur due to
large changes in DERs generation/ loads, maloperation of a
capacitor bank, faults, etc.. These changes pose challenges to
reliable operation and control [1] of the network. Thus, proper
monitoring and visualization of the dynamic changes in the
grid during large variations in network operating conditions
are crucial to take effective control and preventive actions.
The µPMUs can play a crucial role in enabling proper mon-
itoring of the dynamic changes in the distribution network
by providing high-resolution GPS synchronized voltage and
current phasor measurements with a reporting rate of 120
samples per second [2]. Situational awareness and reliable
operation of power distribution systems can be enhanced by
performing real-time monitoring and analysis of the network
using µPMUs.
During normal operating conditions of the distribution net-
work the bus voltage phasors vary steadily due to gradual
changes in load demand and DER generation. However, a few
of the bus voltages and current phasors in the network may
undergo large variations because of disturbances like faults,
large load/capacitor switching, and large change in DER out-
put power. These large variations in the network are referred
to as events [3]. Some real-time events in distribution systems
like voltage sag/swell, harmonics, transients, interruption, etc.
have respective root causes [4]. Event source localization plays
a critical role in understanding and alleviating the cause of
a possible power interruption/outage. Thus, event localization
can increase system reliability and resiliency by enabling faster
system restoration [5].
Several researchers have contributed to the area of event
localization in distribution networks. However, most of the
existing methods do not consider the presence of DERs [6],
[7], [8] and meshed topology. Many of the existing methods
are based on high frequency energy content of the voltage
and current signals. These methods requires that measuring
devices having very high reporting rate are located at all or
most of buses in the network [9]. Such methods are expensive
and may not be practical for distribution networks. In [6],
[7], authors have proposed an impedance-based fault locator
for permanent fault location. However, these are not suitable
for DER/capacitor switching events, as they assume purely
resistive fault impedance. In [9], multiple real-time events are
detected, classified, and localized in a transmission system
based on TeagerKaiser energy operator. This work utilizes
frequency deviation as an indicator for event detection and
localization. It may not be applicable for distribution networks
as frequency deviations are less prominent in distribution
systems. Authors in [8] have utilized synchronized phase angle
measurements for events localization in distribution feeder
without considering DERs penetration and meshed topology.
In [10] power quality events in low voltage AC microgrids
have been localized using signal processing techniques, this
method requires measuring device with a high reporting rate
in the range of 10 kHz for transient features extraction and

event localization.
This paper proposes a novel approach for event localization
using the results of a µPMU-based distribution network state-
estimator. The state estimator is designed to provide an esti-
mate of voltage phasors for all the buses in the system using
measurements from µPMUs connected at generator buses. The
system may not completely observable using µPMU measure-
ments. Therefore, a sparse current injection based distribution
system state estimation (DSSE) is designed in order to have an
estimate of the full state vector. The sparse current injection
based state estimator can provide a good estimate of the
changes in state vector during large disturbances. The change
in state vector can be further analysed to locate the source of
an event. The method is designed considering the presence of
DERs and meshed topology. The proposed method is validated
on a 13-node distribution network simulated using OPAL-RT
real-time simulator.
The paper is organized as follows Section II describes the
problem formulation and the proposed solution, Section III
discuss the test system, Section IV analyzes the algorithm
performance, and Section V concludes the paper.
II. PRO PO SE D MET HO D FO R EVE NT DE TE CT IO N AN D
LOCALIZATION
For a power system network, the vector of voltage mag-
nitudes and angles for all the buses in the system is known
as the state vector of the power network. The operating point
of a power system keeps on varying with changes in load and
generation. These changes in the network operating conditions
are usually described by the changes in the state vector. In
a distribution network, when a large disturbance occurs, the
bus voltage and current phasors for buses close to the event
change more in comparison to voltages at buses, that are far
from the source of the event. Therefore the relative changes
in the element of the state vector can be used to analyze and
locate the source of a disturbance.
In the present work, µPMUs are considered to be present
only at all generator buses. However, the voltage magnitude
and power injections for all the remaining buses in the network
are monitored using remote terminal units (RTUs). The RTU
measurements are communicated to a centralized database
using SCADA. The RTUs typically have a reporting rate of
one sample per minute [11]. As µPMU measurements have
much higher refresh rate than SCADA measurements, several
µPMU measurements are available between two consecutive
SCADA measurements. Therefore, a hybrid DSSE is used
to estimates the system states by making use of forecasted
voltage magnitude and power injection measurements (pseudo
measurements) for the unobserved buses [12]. The voltage and
power forecasting for the buses are done using their respec-
tive past SCADA measurements [13]. However, following a
significant change in the system, the foretasted measurements
using past SCADA data may be far from their actual operating
values. A hybrid state estimator will not be able to reflect these
large changes in states during this period. Thus, the results
of a hybrid DSSE are not reliable during large disturbances/
changes in the system and are not useful for further analysis.
During an event in the absence of SCADA measurements,
this paper proposes to use only the µPMU voltage and current
phasor measurements for estimating the states of the network.
The linear state estimation problem using only µPMU
voltage and current injection measurements can be written as,
z=Hx +(1a)
min
x
(z−Hx)TΣ−1(z−Hx)(1b)
Here, xrepresents the 6N×1vector of system states in
rectangular form. The matrix Hrepresents the measurement
as a linear function of states, zrepresents the m×1vector of
µPMU voltage and current phasor measurements, mrepresents
the number of measurements, and Nis the number of buses in
the system. The vector ∈ N (0,Σ)is the normally distributed
measurement noise vector.
For µPMUs connected at m,m<N, number of buses
in the system, the problem (1) is under-determined, i.e., it has
multiple solutions for a given set of measurements z. A unique
solution can be obtained if enough information is added to the
problem.
Large disturbance in the system such as load/generation
change, capacitor switching, and faults have a very low prob-
ability of co-occurrence [14]. Moreover, any event will have
significant effect on injections at buses which are near to the
location of the event. Thus, during an event at a particular bus,
the change in bus current injection vector can be approximated
as a sparse vector with non-zero elements for buses which
experience large change in their injections, e.g, for a large load
change at a bus, the injection at that bus along with injection at
few other generator buses may experience significant change.
The bus voltage phasors may change for other nearby buses,
with the maximum change in the voltages of the buses which
are closest to the source of the event.
As for an event, the vector of change in current injections is
approximated by a sparse vector. The change in state vector,
∆x= [∆vRe ∆vIm ]T, can be expressed as a sparse vector
using the linear transformation given bellow,
∆iRe
∆iIm =YRe −YIm
YIm YRe ∆vRe
∆vIm (2a)
or,
∆i=φ∆x(2b)
Here, YRe,YI m represents the real and imaginary parts of
the bus admittance matrix of the distribution network and ∆i
is the vector of change in bus current injections. With a priori
knowledge that the change in current injection vector ∆iis
sparse, an l1-norm regularized state estimator is designed to
detect a unique ∆x, using the µPMU measurements only[15].
The estimation problem can be formulated as,
min
∆x
||(p(Σ)−1)(∆z−H∆x)||2+λ||φ∆x||1(3)
Here λis the regularization parameter, and ∆zis the vector
of change in µPMU measurements. The l1regularization pro-
motes sparsity [15] and leads to the finding of a change in state

vector ∆xwhich has a sparse ∆i, and minimum mismatch
between the calculated and actual µPMU measurements.
A. Selecting the regularization parameter
The selection of regularization parameter λplays a crucial
role in obtaining the best solution which can minimize both
the objectives. An efficient method of choosing λis based on
the level of measurement noise variances σi[15], where λis
chosen as,
λ= max
i=1,...,m{σi} × λmax (4)
where,
λmax =||2(Hφ−1)T∆zk∞(5)
The optimization problem (3) is a convex problem which is
solved by an interior-point based solver [16]. On obtaining the
change in state vector ∆x, the bus voltage magnitudes, ∆v
and current injections ∆iare calculated and the bus which has
the largest change in voltage/current magnitude is considered
as bus nearest to the source of an event.
The first step towards event localization is event detection.
Therefore, in the present work a simple auto-regressive state
prediction model (6) is used to detect the occurrence of an
event.
ˆxt=At−1xt−1+βt−1(6)
Here, At−1is the 6N×6Nstate transition matrix and βt−1
is the 6N×1trend setting vector. The state transition matrix
and the trend setting vectors are calculated from past esti-
mated states during steady state condition using exponential
smoothing regression [17].
The state prediction model provides the predicted states, ˆxt,
using the past estimated states, xt−1. The predicated states are
used to calculate the predicted voltage and current injection
phasors, ˆzt, for buses connected with µPMUs. The phasor
measurements can be predicted using (7), as shown below.
ˆzt=Hˆxt(7)
The mismatch between predicated phasor measurements, ˆzt
and phaors measured by µPMUs, is known as the measurement
innovation vector, ∆ˆzt[18], as given below,
∆ˆzt=zµP M Ut−ˆzt(8)
During normal operating conditions the absolute value of
the elements of the measurement innovation vector lie well
below a threshold, η. However, for any event, the generators
would responds to the sudden change in the system. Thus,
during an event, some of the elements of the innovation vector,
∆ˆzt, are significantly affected and exceeds the predefined
threshold, η. Therefore, an event can be detected by monitoring
the µPMU measurement innovation vector [18]. In the present
case the threshold level, η, is selected as three times of the
standard deviation of the changes in the elements of the
innovation vector during normal, quasi-steady state, operating
conditions. The algorithm for event source localization is
described in Fig. 1.
Fig. 1. Event localization algorithm
III. TES T SYS TE M DESCRIPTION
The test distribution feeder has been developed in OPAL-
RT hypersim real time simulator [19]. The OPAL-RT OP5600
digital simulator is capable of running multiple real-time sim-
ulations which enables Software and Hardware-In-the-Loop
testing. The simulator runs on Intel Xeon E5 processors, up
to 32-cores, with a clock frequency of 3.2GHz. The real-time
simulations are performed with a time-step of 50 micro sec-
onds. All practical considerations of a distribution system, like
unbalanced loading, untransposed lines, are taken into con-
sideration while simulating the test network. The test system
comprises of three solar PV generators, one compressed heat
and power (CHP) plant, and one small hydro plant. The solar
PV panel are operating in constant PQ mode i.e., fixed active
and reactive power delivery mode. Any voltage sag/swell in
the network has direct impact on the DER injections. The quasi
steady-state voltage magnitude fluctuations in the network is
introduced by adding gradual small changes to loads and solar
insolation on PV panels. The network has two radial-feeder
with possible mesh interconnection among themselves. During
mesh interconnection, any event occurring on one feeder
would impact the other. Therefore, it is necessary to assess
how this interconnection affects the event localization process.
Fig. 2 shows the test system with all DERs, loads and µ-
PMUs placement (indicated with circled nodes). The µ-PMUs
are placed at generator buses, i.e., at buses: {1,4,6,8,9,13}.
IV. PERFORMANCE ANA LYSI S
The performance of the proposed algorithm has been val-
idated by creating several scenarios on the test distribution
system. The algorithm has been tested on both radial and mesh

Fig. 2. Single-line diagram of the distribution system
distribution system test cases. The three test scenarios from
mesh distribution network load switching, capacitor switching
and faults in the distribution system are discussed in this
section. The event localization is carried out on the basis of
change in the magnitude of bus voltages/ current injections.
The highest change experienced by any bus indicates the event
occurrence location.
A. Load Switching Localization
Large load of 200 KW and 300 KW has been switched on
Bus-5 and Bus-10 at 1sec and 6sec, respectively. Fig. 3 shows
the µPMUs voltage measurements from bus-8and bus-13. The
012345678
Time (sec)
0.7
0.8
0.9
1
1.1
1.2
(P.U)
Bus-8 Phase A Volatge Magnitude
012346678
Time (sec)
-15
-10
-5
degrees
Bus-8 Volatge Phase Angle of Phase A
012345678
Time (sec)
0.6
0.7
0.8
0.9
(P.U)
Bus-13 Phase A Volatge Magnitude
012345678
Time (sec)
-16
-14
-12
-10
-8
-6
degrees
Bus-13 Volatge Phase Angle of Phase A
Fig. 3. µPMU measurements from Bus-8and Bus-13
proposed algorithm is utilized for switching localization, using
the the change in voltage magnitudes for all buses. The bus
with maximum change in voltage magnitude/current injection
indicates the most affected bus following an event. However,
some other nearby buses may also experience significant
change in their voltage magnitudes as shown in detailed view
in Fig.4. Thus, large load switching events at Bus-5and
Bus-10 could be accurately detected and localized in the
mesh distribution system by identifying the bus which shows
maximum change in its voltage magnitude.
Fig. 4. Localization of load switching using estimated change in voltage
magnitudes
B. Capacitor Switching Localization
Capacitor switching is normally carried out in small steps
for feeder voltage regulation. However, large capacitor switch-
ing can happen because of a malfunctioning of capacitor banks
or miss-operation of automatic voltage regulators. Locating
large capacitor switching would help in root cause analysis
behind any sudden and significant voltage sag/swell in the
system. In order to create a large capacitor switching, the
capacitor bank at Bus-7is switched on at 8sec and switched
off at 8.5sec. Fig. 5 shows p.u. change in estimated voltage
magnitude for all buses. Bus-7responds with highest change
in estimated voltage magnitude among all other buses during
the switching instant.
7.8 8 8.2 8.4 8.6 8.8
Time (sec)
0
0.05
0.1
0.15
0.2
0.25
Change in Volatge Magnitude (P.U)
Capicator Switching
Bus -1
Bus -2
Bus -3
Bus -4
Bus -5
Bus-6
Bus -7
Bus -8
Bus -9
Bus -10
Bus -11
Bus -12
Bus -13
Fig. 5. Localization of capacitor switching using estimated change in voltage
magnitudes
C. Fault Localization
Faults in the distribution system are needed to be detected
and isolated as soon as possible in order to increase system
reliability (by reducing the outage time). The efficacy of the
proposed method is tested for locating faults in the distribution
feeder. A Line to ground fault with 0.1Ω fault resistance has
been created on Bus-2and Bus-10 for the duration of 1sec to
1.1sec and 11 sec to 11.1sec, respectively. Fig. 6 shows the
change in estimated voltage magnitude for all buses. However,
during any fault scenario there is a similar impact on all

0 5 10 15
Time (sec)
0
0.2
0.4
0.6
0.8
1
Change in voltage magnitude (p.u.)
Fault Scenario
Bus-1
Bus-2
Bus-3
Bus-4
Bus-5
Bus-6
Bus-7
Bus-8
Bus-9
Bus-10
Bus-11
Bus-12
Bus-13
1.008 1.01 1.0121.014
0
0.05
0.1
0.15
0.2
Fig. 6. Localization of faults using estimated change in voltage magnitudes
nearby bus voltages. Therefore, during fault in the system the
change in voltage magnitude is unable to give accurate area
localization as shown in Fig.6.
Fig. 7. Localization of faults using estimated change in current injections
Fault localization can be carried out by observing the
estimated change in current magnitude injections, as fault cur-
rents are very high in comparison to normal load/ generation
injections. During fault, the affected area would experience
large change in current injection. Fig. 7 shows all bus current
injection change, with Bus-2showing maximum change at
1.02 sec, which is the first peak after the occurrence of the
fault. Similarly, Bus-10 shows maximum change at 11.02 sec.
Thus, the proposed scheme is efficient in locating the faulted
bus.
V. CONCLUSION
This paper presented an event localization technique for
active distribution networks using an event-triggered l1regu-
larized linear state estimator. The sate estimator uses measure-
ments from a sparse set of µPMUs. Various switching events
like large load switching, faults, and capacitor switching,
are successfully located using the proposed method. A test
distribution system with DER penetration is developed in
OPAL-RT real-time simulator. The test results verify that the
proposed method works in the case of both meshed and radial
networks.
ACKNOWLEDGMENT
This work is supported by the Indo-US Science and
Technology forum in partnership with Department of Sci-
ence and Technology, Government of India, under grant no.
IUSSTF/JCERDC-Smart Grids and Energy Storage/2017.
REFERENCES
[1] D. Allen, A. Apostolov, and D. Kreiss, “Automated analysis of power
system events,” IEEE Power and Energy Magazine, vol. 3, no. 5, pp.
48–55, 2005.
[2] H. Mohsenian-Rad, E. Stewart, and E. Cortez, “Distribution synchropha-
sors: Pairing big data with analytics to create actionable information,”
IEEE Power and Energy Magazine, vol. 16, no. 3, pp. 26–34, 2018.
[3] M. Jamei, E. Stewart, S. Peisert, A. Scaglione, C. McParland, C. Roberts,
and A. McEachern, “Micro synchrophasor-based intrusion detection in
automated distribution systems: Toward critical infrastructure security,”
IEEE Internet Computing, vol. 20, no. 5, pp. 18–27, 2016.
[4] V. G. Werner, D. F. Hall, R. L. Robinson, and C. A. Warren, “Collecting
and categorizing information related to electric power distribution inter-
ruption events: data consistency and categorization for benchmarking
surveys,” IEEE transactions on power delivery, vol. 21, no. 1, pp. 480–
483, 2005.
[5] A. Shahsavari, A. Sadeghi-Mobarakeh, E. M. Stewart, E. Cortez, L. Al-
varez, F. Megala, and H. Mohsenian-Rad, “Distribution grid reliability
versus regulation market efficiency: An analysis based on micro-pmu
data,” IEEE Transactions on Smart Grid, vol. 8, no. 6, pp. 2916–2925,
2017.
[6] R. Krishnathevar and E. E. Ngu, “Generalized impedance-based fault
location for distribution systems,” IEEE transactions on power delivery,
vol. 27, no. 1, pp. 449–451, 2011.
[7] J. Ren, S. Venkata, and E. Sortomme, “An accurate synchrophasor
based fault location method for emerging distribution systems,” IEEE
Transactions on Power Delivery, vol. 29, no. 1, pp. 297–298, 2013.
[8] M. Farajollahi, A. Shahsavari, E. M. Stewart, and H. Mohsenian-Rad,
“Locating the source of events in power distribution systems using
micro-pmu data,” IEEE Transactions on Power Systems, vol. 33, no. 6,
pp. 6343–6354, 2018.
[9] R. Yadav, A. K. Pradhan, and I. Kamwa, “Real-time multiple event
detection and classification in power system using signal energy trans-
formations,” IEEE Transactions on Industrial Informatics, vol. 15, no. 3,
pp. 1521–1531, 2018.
[10] S. Som, S. Chakrabarti, and S. R. Sahoo, “Wavelet transform based pq
event localization scheme for benchmark lvac microgrid,” in 2018 20th
National Power Systems Conference (NPSC). IEEE, 2018, pp. 1–6.
[11] I. Dˇ
zafi´
c, R. A. Jabr, and T. Hrnji´
c, “Hybrid state estimation in complex
variables,” IEEE Transactions on Power Systems, vol. 33, no. 5, pp.
5288–5296, 2018.
[12] M. Huang, Z. Wei, G. Sun, and H. Zang, “Hybrid state estimation for
distribution systems with ami and scada measurements,” IEEE Access,
vol. 7, pp. 120 350–120 359, 2019.
[13] R. Z. S. Santos and J. R. C. Orillaza, “Distribution system state estimator
using scada and µpmu measurements,” in 2018 IEEE Innovative Smart
Grid Technologies-Asia (ISGT Asia). IEEE, 2018, pp. 558–562.
[14] J. Zhao, A. Gomez-Exposito, M. Netto, L. Mili, A. Abur, V. Terzija,
I. Kamwa, B. C. Pal, A. K. Singh, J. Qi et al., “Power system
dynamic state estimation: motivations, definitions, methodologies and
future work,” IEEE Transactions on Power Systems, 2019.
[15] S.-J. Kim, K. Koh, M. Lustig, S. Boyd, and D. Gorinevsky, “An interior-
point method for large-scale l1-regularized least squares,” IEEE journal
of selected topics in signal processing, vol. 1, no. 4, pp. 606–617, 2007.
[16] M. Grant and S. Boyd, “Cvx: Matlab software for disciplined convex
programming, version 2.1,” 2014.
[17] M. Hassanzadeh, C. Y. Evrenoso˘
glu, and L. Mili, “A short-term nodal
voltage phasor forecasting method using temporal and spatial correla-
tion,” IEEE Transactions on Power Systems, vol. 31, no. 5, pp. 3881–
3890, 2015.
[18] S. J. Geetha, S. Chakrabarti, and K. Rajawat, “Hierarchical parallel
dynamic estimator of states for interconnected power system,” IET
Generation, Transmission & Distribution, vol. 12, no. 10, pp. 2299–
2306, 2018.
[19] HYPERSIM, https://www.opal-rt.com/simulator-platform-op5600/,
[Online; accessed 19-July-2019].