Off the

Beaten Path
Resiliency and
Associated Risk

L
Large-scale electric grids remain
an indispensable critical infrastructure, and the vast
majority of people will continue to receive most of
their electric energy from such grids for decades to
come. The impacts of the loss of a portion of the
electric grid range from minor inconveniences for
most users when the outage is on a small scale and
short lived to potentially catastrophic situations
when the blackout covers a large region for a long
duration. Since the inception of the first electric
grids in the 1880s, much has been done to reduce
the likelihood and extent of blackouts. However,
they cannot be totally eliminated, and there is usu-
ally a tradeoff between reduced blackout risk and
increased cost. The focus of this article is on cost-
effective ways to reduce this risk and, consequently,
improve resiliency.
Keeping the lights on involves designing and
operating the electric grid with the goal of simulta-
neously increasing two related but ultimately quite
different concepts: reliability and resiliency. Mer-
riam-Webster defines reliable as “suitable or fit to
be relied on: dependable” and resilience as “an abil-
image licensed by ingram publishing

ity to recover from or adjust easily to misfortune or

change.” Certainly, to be effective, any large-scale
electric grid must, at least to some degree, be both
reliable and resilient, i.e. available.
Both attributes have been considered either
explicitly or implicitly in grid design almost from
Digital Object Identifier 10.1109/MPE.2017.2780961
Date of publication: 21 February 2018

26 ieee power & energy magazine 1540-7977/18©2018IEEE march/april 2018

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY KURUKSHETRA. Downloaded on March 14,2022 at 08:17:35 UTC from IEEE Xplore. Restrictions apply.
 By Mladen Kezunovic and Thomas J. Overbye

day one. One of the key benefits of interconnecting multiple 1890, and automatic circuit breakers were invented around
generators is improved reliability because the lights can 1900, allowing faulted parts of the grid to be automatically
then stay on even if one generator fails. Papers discussing removed very quickly.
more formal considerations of reliability date to at least the In moving beyond individual devices to consider the sys-
1930s. As for resiliency, any protection device that removes tem as a whole, perhaps a good working description of the
part of the system to save the remainder is contributing to difference between reliability and resiliency comes from the
system resiliency. Thomas Edison patented the fuse in recent U.S. National Academies report on resiliency:
While minimizing the likeli-
hood of large-area, long-duration
outages is important, a resilient
system is one that acknowledges
that such outages can occur, pre-
pares to deal with them, mini-
mizes their impact when they
occur, is able to restore service
quickly, and draws lessons from
the experience to improve perfor-
mance in the future.

Resiliency of the
Physical System:
The Five Operating
States
A major aspect of the idea of resil-
iency can be elaborated upon by
considering the power system op­­
erating states presented in Figure 1
(which reproduces a dia­­gram from
a 1978 IEEE Spectrum article by
Fink and Carlsen). By far, the most
time is spent in the normal (N-1)
state, during which there are no
limit violations for either the pre-
vailing operating point or credible
contingencies (N-1). While the
idea primarily addresses the opera-
tion of the bulk transmission grid,
similar concepts may be applied
to the distribution grid—particu-
larly now, when distributed and
renewable generation are making
the distribution grid resemble the
transmission grid in many respects,
including bidirectional flows and
redundant feeder connections.
Hence, many of the energy
management system (EMS) and
distribution management system
(DMS) tools used in the control
center are focused on normal
operation, and this is the state

march/april 2018 ieee power & energy magazine 27

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY KURUKSHETRA. Downloaded on March 14,2022 at 08:17:35 UTC from IEEE Xplore. Restrictions apply.
 with which operators have the most experience. More rarely, An emerging area of concern is what NERC calls high-
the system moves into the alert, emergency, and restor- impact, low-frequency (or HILF) events. These are statis-
ative states. However, such situations are encountered often tically unlikely but still plausible events that, should they
enough that control-room personnel train for them and, for occur, could have catastrophic consequences on the grid and
the most part, have adequate tools to deal with them. Truly thus many everyday lives. Included in this group are large-
enhancing grid resiliency requires tools to deal with the scale cyber or physical attacks, pandemics, electromagnetic
much more difficult in extremis situations. As noted by the pulses (EMPs), and geomagnetic disturbances (GMDs). In
North American Electric Reliability Corporation (NERC) such cases, the length of warning time might be essentially
in its 2012 report on resilience, during such an event, the zero for cyber/physical attacks and EMPs to hours for GMDs
previously interconnected grid may be broken into a number to potentially days or weeks for pandemics.
of electrical islands, and the operation of these islands may The following discussion focuses on time horizons for
need to be performed by entities not normally responsible which predictions are available within time frames that
for grid operations. allow operators or other utility staff to take actions that can
The degree to which blackouts can be minimized or pre- mitigate the impact of catastrophic events by reducing the
vented during such in extremis situations depends on the risk of incurring outages. To illustrate various applications
triggering event itself, as well as the availability of a com- in the planning, operations, and asset and outage manage-
bination of strategies ranging across time frames—from ment time horizons, we provide examples of advanced EMS
real-time operations to asset and outage management to and DMS tools that deal with risk reduction and mitigation.
potentially planning years ahead. A wide variety of differ-
ent events can place a system in this operating state, each Resiliency in the Economics of Energy
having its own horizon of warning time. The most common Exchange: The Market Operating States
would be the more severe manifestations of relatively typical The goal of resiliency for the physical electric grid is to
weather conditions, which could induce large-scale storm ensure continuous energy exchange between producers and
systems—potentially including tornados and ice storms. consumers. While the objective of reliability is to “keep the
Warning times in such situations would be, at most, hours. lights on,” the goal of resiliency in the context of energy
Hurricanes can, of course, cause severe damage, but they exchange is to “keep the markets on” at all times.
usually come with a longer warning period of at least a day The history of market deregulation in the United States
or more. In contrast, high-intensity earthquakes can cause is long, starting as early as 1935 when Congress passed the
widespread damage with essentially no warning time. Public Utilities Holding Company Act. The act included
many new rules regarding the ways in which energy could
be sold. As the oil crisis hit in the 1970s, regulators began to
introduce energy conservation rules (up until 1974). Despite
Normal (N–1 Reliable)
this, the price of oil remained high. As a result, much of the
legislation approved throughout this decade related to uti-
Restorative Alert
lizing other forms of energy to reduce U.S. dependence on
oil or fossil fuels. In 1992, the National Energy Policy Act
allowed for private market competition within the wholesale
In Extremis Emergency generation of electricity. This, in itself, helped pave the way
for true energy deregulation in the United States.
Subsequently, Order 888 in 1996 and Order 2000 in 1999
figure 1. The power system operating states (adapted from
IEEE Spectrum, 1978).
assured “open-access nondiscriminatory transmission ser-
vices” and further deregulation by “creating Regional Trans-
mission Organizations” that replaced state operation and
control over the transmission grid. The Energy Policy Act of
2005 formed the U.S. Federal Energy Regulatory Commis-
sion (FERC) as the primary regulator for energy within every
Normal
Market
state across the country. The Energy Independence and Secu-
Conditions rity Act of 2009 helped further improve electricity delivery
Slow Slow to customers by assuring the development of a “smart grid.”
All these efforts resulted in the creation of wholesale and
Fast retail markets to allow electric energy producers and consum-
Market Market ers to conduct the business of energy exchange for economic
Incomplete Emergency
benefits. The 2009 law imposed the not widely discussed resil-
iency requirement to “keep the markets on” at all times and
figure 2. Market operating states. (Source: EPRI.) minimize the economic losses from grid interruptions. Figure 2,

28 ieee power & energy magazine march/april 2018

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY KURUKSHETRA. Downloaded on March 14,2022 at 08:17:35 UTC from IEEE Xplore. Restrictions apply.
 used in presentations by D.J. Sobajic and J. Douglas of the sures the economic impact of the loss of resiliency. Such a
Electric Power Research Institute (EPRI) in early 2003, shows framework is shown in Figure 3.
an example how the wholesale markets may go into different This framework is illustrated in several applications dis-
market states depending the ways in which the physical system cussed later in this article, when asset and outage impacts on
states unfold. resiliency are introduced. By selecting proper data analyt-
ics, the risk can be predicted, and associated risk prediction
Resiliency Quantification: maps can be generated that provide guidelines to operators
Risk-Based Maps for mitigating risk and, hence, improving resiliency. (The
As the notion of resiliency gets further defined, it is neces- resiliency definition and quantification require further stud-
sary to introduce quantitative ways of measuring resiliency. ies beyond the scope of this article.)
One possible approach is to define risk as a measure of resil-
iency. A well-established risk definition commonly used in Resiliency in Control Strategies:
the engineering fields defines risk as Hierarchically Coordinated Protection
When the electricity grid is in the in extremis state, fast control
Risk = Hazard # Vulnerability # Impacts, actions are needed; in most cases, these cause protective relay-
ing systems to “trip” circuit breakers and disconnect faulted
where parts of the system from service. The protective relaying
✔✔ intensity T is the threat intensity function is decentralized to substations for faster action and
✔✔ hazard is the probability of threat with intensity T operates in the millisecond time frame. Control center per-
✔✔ vulnerability is the probability of a consequence C if a sonnel do not get involved in initiating the relaying actions
threat with intensity T occurs because of the subsecond time-frame response required,
✔✔ impacts are the stimulated economic and/or social im- but they are greatly concerned with the outcomes of such
pacts if consequence C has occurred. actions to enable mitigation of impacts and bring the power
This approach not only allows the resiliency to be quantified grid back to its normal operating state. To improve resil-
but also defines a framework to assess and mitigate threats iency going forward, fast control actions will also have to be
for the elements of the grid at risk. Most importantly, the risk redesigned to allow for hierarchically coordinated protec-
can be expressed in monetary values, which further mea- tion (HCP). Such a concept was introduced following the

Weather
Risk Mitigation Risk Threats

Temperature Wind Precipitation Clouds Lightning Humidity

Hazard
Real-Time Market Forward Market

Mitigation
Monitoring
Control and Protection
Flexible Load
Vulnerability

Power Generation
Renewables:
Elements at Risk

Hydro Power Transmission Power Distribution Customers

Solar Substations Substations Residential
Wind Lines Lines Commercial
Fossils: Cables Cables Industrial
Gas
Coal
Risk Assessment

Probabilistic Spatiotemporal
State of Risk
Risk

Probabilistic Regression
Causal Probabilistic Risk Assessment

figure 3. A risk assessment and mitigation framework for resiliency quantification.

march/april 2018 ieee power & energy magazine 29

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY KURUKSHETRA. Downloaded on March 14,2022 at 08:17:35 UTC from IEEE Xplore. Restrictions apply.
 grid modernization study funded through the Power Systems Resiliency in Planning: Using Rare Event
Engineering Research Center by the U.S. Department of Criteria when developing Modeling
Energy (DOE) in 2003, as depicted in Figure 4. and Simulation tools and studies
The most relevant implication of the proposed HCP ap­­­ Resiliency starts with planning and design, answering the ques-
proach is that the resiliency improvement in fast control actions tion of what needs to be done well ahead of any possible event
requires coordination among predictive, adaptive, and cor- to reduce its adverse magnitude and duration. The electric util-
rective protection actions. While adaptive protection has ity industry has a long history of planning, and the high levels
been discussed for many years, particularly in two key ini- of reliability today attest to its success in this area. However,
tial studies funded by the DOE in the mid-1980s (led by the majority of this work has been directed toward improving
A.G. Phadke and G.D. Rockefeller, respectively), it contin- system reliability, mostly focused on designing a system for
ues to be explored not only at the transmission level but also optimal operations during normal conditions and responding to
for distribution systems and microgrids. Corrective relay- events similar to those previously encountered. Here, we pres-
ing actions are also being contemplated, as experience has ent several ideas for increasing system resiliency by incorporat-
shown that relay misoperations can cause major blackouts. ing criteria that deal with rare events.
The most promising, and least explored, option is to try to From the start of the power industry, modeling and simu-
predict faults based on historical data and prevailing weather lation have played important roles. With the introduction of
conditions, which will give operators an option to mitigate digital computers, much of this power system understanding
consequences by undertaking various preemptive actions to has been integrated into software of increasing complexity, the
reduce risk of outages and major blackouts. The following capabilities of which include short-circuit analysis, power flow,
sections explore some innovative ways of getting control cen- contingency analysis, security-constrained optimal power
ter operators involved in planning, tracking, and mitigating flow, and transient stability. Modeling and simulation occur
unforeseen grid conditions that may impact the resiliency of over many time frames, ranging from decades in the future for
the physical system and electricity markets. long-term planning to real-time for operations.

Before Fault Occurrence During Fault Occurrence After Fault Occurrence

Predictive Protection Adaptive Protection Corrective Protection

Inputs (Historical): Inputs: Inputs:
Weather Data Local Measurements Relay Trip
Outage Data System Wide Auto Reclosing Action
Equipment Measurements Fault Location
Deterioration Status

Adequacy Learning Real-Time

Legacy Protection Check of Process to Fault Analysis
Inputs: Existing Adapt to to Determine
Fixed Settings Settings Prevailing Accurate Fault
Local Measurements Conditions Location

Tripping for Faults

in the Protection Fast Recalculation of Supervised Operation Supervised Reclosing
Zone Followed by Relay Settings of Legacy Protection Action
Reclosing
...

Before Relay Has Relay Is Getting After Relay Has

Operated Ready to Operate Operated

figure 4. HCP for improved resiliency.

30 ieee power & energy magazine march/april 2018

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY KURUKSHETRA. Downloaded on March 14,2022 at 08:17:35 UTC from IEEE Xplore. Restrictions apply.
 Building on this foundation, enhancing resiliency presents ure, the green arrows show the real power flow, and the contour
the challenge of modeling and simulating systems in very shows the per-unit substation voltages.
unusual and often highly stressed situations. There is also the Enhancing power grid resiliency requires being able to
need for more multidimensional modeling. Severe events are accurately simulate the impact of a wide variety of events on
likely to affect not just the electric grid but also other infra- the power grid—and, potentially, on its coupled infrastruc-
structures as well. At times, there is a need to model some of tures. The events most likely to stress power system resil-
the underlying dynamics of the disturbance itself, such as in iency share two characteristics.
the case of severe storms and GMDs. This leads to the need First, they have a significant impact. From a modeling per-
to develop cosimulation platforms that can model interactions spective, this means that they strain the power grid in new
between the power system and other critical infrastructure, and often unexplored areas. A consequence is that they will
including control systems. also mostly likely stress the power system modeling software.
Key to the research and development needed for the cre- The degree of power system impact often requires detailed
ation of simulation tools for improved resiliency is access to modeling of physical systems associated with the initiating
large-scale, realistic models of electric grids. While some event. For example, correctly modeling the impacts of large
of this information has been available in the past, because earthquakes requires coupled modeling between the power
of the U.S. Patriot Act of 2001, data relating to the U.S. elec- grid and seismic simulations.
tric power grid are now considered critical energy/electricity Second, because the events are low frequency, there may
infrastructure information (CEII), with access much more be little historical information to accurately quantify the
restricted. While most researchers can obtain some infor- risk. From a model perspective, some of the more extreme
mation, e.g., with nondisclosure agreements, these restric- events could be considered extreme manifestations of more
tions can sometimes hinder the free exchange of models common occurrences. Thus, a large-scale physical attack
and results. could be considered a more severe manifestation of more
A solution is to develop entirely fictional (synthetic) models regular disturbances, such as those due to weather. Others,
that match the complexity of the actual grid models but con- however, such as the grid impacts due to an electromagnetic
tain no CEII. This is now starting to occur, due in large part pulse caused by a high-altitude nuclear explosion, would be
to the DOE’s Advanced Research Projects Agency–Energy (or entirely novel.
“ARPA-E”) Grid Data program.
The challenge for this research is
to determine the wide multi-
tude of germane characteristics
of actual grid models and then
mimic these in entirely synthetic
models that can be freely shared.
A quite useful characteristic of
such synthetic models would be
for them to include realistic geo-
graphic coordinates so as to allow
coupling between the power grid
and either other infrastructures or
the actual geography.
One approach is to use an elec-
tric load distribution that matches
the actual population in a geogra­
phic footprint, then employ public
data on the actual generator loca-
tions, and finally use algorithms to
Per Unit Voltages
create an entirely synthetic trans-
1.08
mission grid. As an example, Fig-
(PU)

ure 5 shows a 10,000-bus model 1.03

entirely synthetically sited geo­­
0.98
graphically in the western United
States; the system has a total of
seven different nominal transmis- figure 5. Synthetic models, such as this 10,000-bus model located in the western
sion voltages (765, 500, 345, 230, United States, allow for the study of power grid resiliency without the disclosure of
161, 138, and 115 kV). In the fig- confidential information about the actual grid.

march/april 2018 ieee power & energy magazine 31

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY KURUKSHETRA. Downloaded on March 14,2022 at 08:17:35 UTC from IEEE Xplore. Restrictions apply.
 While the goal of reliability is to “keep the lights on,”
the goal of resiliency in the context of energy exchange is
to “keep the markets on” at all times.

One example is the study of the impact of GMDs on the consider tools to enhance the capabilities of the operators
high-voltage electric grid. GMDs, which result from corona and engineers running the system.
mass ejections from the sun, cause low-frequency (much One of the undesirable consequences of large-scale inter-
lower than 0.1 Hz) variations in the Earth’s magnetic field. connects is that disturbances in one portion of the system can
The changing magnetic field then induces electric fields on rapidly affect the entire system. The normal operating state
the Earth’s surface, with the magnitude and direction of can rapidly become an emergency or in extremis state, dur-
these electric fields determined by the conductivity of the ing which quick, informed intervention by a human operator
Earth’s crust going down several kilometers. These electric is essential. Hence, operators need to maintain situational
fields then cause low-frequency, geomagnetically induced awareness. As shown by the loss of situational awareness
currents to the flow in the high-voltage transmission sys- that was one cause of the 14 August 2003 North America
tem, potentially causing half-cycle saturation in the high- blackout, there is a need to develop better techniques to help
voltage transformers. A GMD with a maximum electric human operators manage the unique operating challenges
field of about 2 V/km caused a blackout in Quebec, Canada, posed by in extremis conditions. A key resiliency need in
in 1989. Much larger GMDs occurred in North America in the operations area is better data analytics and visualizations
1859 and 1921, with magnitudes estimated by some as being to help operators manage the potentially quite unusual con-
up to five times those of the 1989 event. Such a GMD occur- ditions they might encounter during in extremis situations.
ring today could cause a severe event, with potentially long- The degree to which operator action can prevent or mini-
term power outages. mize a blackout depends on the severity of the event and its
The potential for GMDs to impact the electric grid has time frame. Some large-scale blackouts cannot be prevented
been known since at least the early 1940s, and the incor- by operator action. For example, during an earthquake, an
poration of GMD analysis within the power flow was first unanticipated event can cause severe damage within seconds.
proposed in 1981. However, power grid GMD assessment is Here, visualization would be most helpful in the restorative
still an active area of research and development, with much state because there is nothing the operator can do to prevent
progress in the last few years through interdisciplinary work and physical damaged due to an earthquake. Conversely, slow-
active industry and government involvement. As a result, moving weather systems, such as hurricanes or ice storms,
GMD analysis has now been integrated into commercial give operators plenty of time to act, but blackouts still cannot
power system planning tools, including power flow and tran- be fully prevented. For example, 2102’s Superstorm Sandy
sient stability analysis software. However, determining the in the eastern United States caused 8.5 million customer
magnitudes of the severe events to model can be challeng- power outages with damage estimated at $US65 billion.
ing because historical records are often incomplete or non- Many, if not most, potential large-scale blackouts have time
existent. Determining the scenarios to consider for human- frames that could allow for effective operator intervention. A pri-
caused severe events, such as a combined cyber and physical mary reason for this is the underlying power system dynamics,
attack, is even more challenging. including the time constants associated with thermal heating on
transmission lines and transformers, the operation of load-tap-
Resiliency in Operations changing transformers, and generator overexcitation limiters.
and Operations Planning: During the unusual situations associated with severe events,
Visualizing Massive Amounts of Data wide-area power system visualization will be crucial for provid-
In addition to planning considerations, much can be done ing operators and engineers with the big picture of a grid that
in the area of real-time operations of the electric grid to may be operating in a state they have not previously encountered.
enhance resiliency. With the advent of the smart grid, the There may be multiple electric islands, transmission line flows
electric grid is getting more intelligent, offering more sens- may differ substantially from normal, and the voltage profile
ing and embedded controls. This is certainly beneficial, but could be quite unusual.
a consequence is increased grid complexity. While this auto- Over the years, much has been done in power system ana-
matic control is helpful, any consideration of power system lytics and visualization to improve situational awareness, and
operations needs to recognize that human operators are still this remains an active area of research. While a full discus-
very much “in the loop” and will continue to be so for many sion is beyond the scope of this article, one approach that
years. Therefore, enhanced operational resiliency needs to has been helpful is the use of dynamically formatted one-line

32 ieee power & energy magazine march/april 2018

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY KURUKSHETRA. Downloaded on March 14,2022 at 08:17:35 UTC from IEEE Xplore. Restrictions apply.
 elements, such as dynamic pie charts, to indicate overloaded “X.” A contour is also used to show the voltage variation across
or open transmission lines or transformers. This will cause the region.
the information to “pop out” by taking advantage of pre-atten-
tive processing to ease the search for system limit violations, Resiliency in Asset Management:
such as overloaded lines. The technique is demonstrated in Tracking State of Equipment
Figure 6, which represents a portion of a 2,000-bus synthetic Deterioration and Risk of Asset Failure
model covering much of Texas. Dynamic sizing and coloring While typically not a function of the control center, asset man-
are used so pie charts on lines loaded above 100% of their agement has profound impacts on resiliency. Monitoring of
limits increase in size and are colored magenta, whereas the status of assets—in particular, because the assets in the
open lines are indicated by a large black circle with a green U.S. grid are, on average, 30–40 years old—needs to become

Voltage A
103%
1.08 Amps
(p.u.)

1.03

0.98 A A
111% 110%
Amps Amps
A
106%
Amps

A
151%
Amps
A
126%
Amps

figure 6. A wide-area visualization combining a number of overloaded lines, open lines, and low voltages.

Historical
Lightning Data GCRF Prediction
BIL_new
Model
Historical
Weather Data
SUB2 T1_N T1_1
BIL_old MS1
T1_N
SUB1 T3_1 ... T3_N SUB4
...

Lightning
Detection : Measurement
T1_N SUB : Substation
Network
MS2
... T : Tower
SUB3 T2_N T2_1
MS : Meteorological Station
Weather Stations

figure 7. Graph-based GCRF data analytics for computing BIL_new based on BIL_old and weather conditions.

march/april 2018 ieee power & energy magazine 33

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY KURUKSHETRA. Downloaded on March 14,2022 at 08:17:35 UTC from IEEE Xplore. Restrictions apply.
 an integral part of everyday grid operation. Operators must be
aware of grid component health conditions to develop mitigat-
ing control strategies, should the risk of component failures
start increasing. This leads to a new risk-based framework for
online monitoring of assets. The following example illustrates
the new online approach to tracking the state of transmission
line insulators and developing an operator decision-making
framework for acting to mitigate the risk of insulator failures.
This approach integrates asset management actions to opti-
(a)
mize design and repair strategies as well.
The proposed approach assumes that the deterioration
state of a large number of insulators on transmission lines
is tracked continuously in both time and location. The sug-
gested spatiotemporal approach helps differentiate the insu-
lators that are deteriorating faster and posing a risk of failure
because their operating and environmental conditions cre-
ate a high hazard. To differentiate declining performance
(b)
characteristics, the basic insulation level (BIL) is tracked for
each insulator and correlated to factors causing the insula-
tor to be vulnerable to failures. An example of the computa-
tional framework extracted from an ongoing study at Texas
A&M University is shown in Figure 7. The goal is to cal-
culate BIL_new using BIL_old by taking into account the
historical lightning and weather data that a given insulator
(c) has experienced over time. A particular data analytics frame-
work—Gaussian conditional random fields (GCRF) invented
Risk 0–20% 20–40% 40–60% by a study partner, Z. Obradovic at Temple University—is
60–80% 80–100% used to calculate BIL_new. This is a graph-based calculation
paradigm that processes data in each node of the graph asso-
ciated with a measurement point where data are collected.
figure 8. The risk as of (a) 1 January 2009, (b) 31 December The calculation correlates data at each point with the impacts
2014, and (c) January 2015 (prediction). that data at other points may have on the measurements. The
nodes where data are measured are
shown in yellow. The designation
table 1. Data used in predictive risk data analytics for outage Tx_N relates to the n transmission
management on distribution feeders. lines with m towers each.
For each of the nodes, a set
of variables X is defined, and the
Weather Data: Weather Events:
GCRF data analytics uses the net-
• Temperature • Thunderstorm
Hazard

• Precipitation • Hail work branches to establish the graph

• Wind Data • Flash Flood correlations between the measure-
• Humidity • Drought ment nodes where each individual
insulator is located. As a result,
Historical Outage Data: BIL_new is computed for each
Vulnerability

• Data and Time insulator at any given moment in

• Location Risk
• Duration
Outage Event
Factor
time, which allows operators to see
• People Affected the risk of a specific insulator fail-
• Cause Type ing at a particular time. This type of
information also instructs the asset
Worth of Loss

management planning group to ini-

Consequence Measures:
• Customer Information Customer Outage tiate work orders that mitigate the
• Outage Cost Model Cost situation by replacing all insulators
• Outage Location at a high risk of failing.
Such time-evolving risk maps
may be created to track the status

34 ieee power & energy magazine march/april 2018

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY KURUKSHETRA. Downloaded on March 14,2022 at 08:17:35 UTC from IEEE Xplore. Restrictions apply.
 level—need to change to accom-
modate resiliency requirements
and subsequently offer tools for
9 19
monitoring and maintaining sys-
2 14 7 tem resiliency. EMS and DMS
17
4 designs conceived in the late 1960s
and late 1990s, respectively, are
24 13
20 no longer adequate to deal with
23 unfolding grid dynamics, grid
21 Vegetation Risk
Vege expansions, and emerging electric-
3 Risk (%) ity markets. The set of techniques
1 11 and technologies discussed in this
0–20
0
10 16 21–40
2 article point toward possible oppor-
22
5 tunities in future developments
41–60
4
61–80
6
aimed at improving grid resiliency.
8
81–100
18 For Further Reading
T
Trimming Zones
15 National Academies of Sciences,
6
8 12 Engineering, and Medicine, En­­
hancing the Resilience of the
Nation’s Electricity System. Wash-
figure 9. Predictive risk maps for vegetation-caused outages. ington, D.C.: National Academies
Press, 2017.
of assets (in this case, transmission line insulators), as shown National Academies of Sciences, Engineering, and Med-
in Figure 8. We can observe from the rectangular marker on icine. (2016). Analytic Research Foundations for the Next-
the network that the individual risk changes with time and Generation Electric Grid. Washington, D.C.: National Academies
affects various insulators differently. Press. [Online]. Available: https://doi.org/10.17226/24836
Once the ability to track a large number of assets and North American Electric Reliability Corporation (NERC).
predict their risk of failure becomes a standard EMS/DMS (2012). Severe impact resilience: considerations and recom-
function, the resiliency affected by random asset failures mendations. [Online]. Available: http://www.nerc.com/docs/
could be improved by a decision-making tool that identifies oc/sirtf/SIRTF_Final_May_9_2012-Board_Accepted.pdf
a repair-and-replace strategy. This could become an integral M. Kezunovic, Z. Obradovic, T. Dokic, B. Zhang, J. Sto-
part of the control strategy that keeps the grid in normal janovic, P. Dehghanian, and P.-C. Chen, “Data science and
operating conditions. big data: An environment of computational intelligence,” in
Predicating Spatiotemporal Impacts of Weather on Power
Resiliency in Outage Management: Systems Using Big Data Science, W. Pedrycz and S.-M. Chen,
Predicting Faults and Optimizing Eds. New York: Springer Verlag, 2017, pp. 265–299.
Pro-Active Mitigation Measures L. H. Fink and K. Carlsen, “Operating under stress and
The outage-management risk-based framework can be illus- strain,” IEEE Spectr., vol. 15, pp. 48–53, Mar. 1978.
trated using the vegetation management task in distribution M. Kezunovic, J. D. McCalley, and T. J. Overbye,
systems. Another ongoing study at Texas A&M University has “Smart grids and beyond: Achieving the full potential of
focused on using the data shown in Table 1 to predict outages electricity systems,” Proc. IEEE, vol. 100, pp. 1329–1341,
due to weather impacts. As a result, risk maps with a high May 2012.
risk of outages due to vegetation can be predicted, as shown M. Panteli, D. Trakas, P. Mancarella, and N. D. Hatziar-
in Figure 9 for a town in Florida. The various colors on the gyriou, “Power systems resilience assessment: Hardening
feeder sections indicate the level of risk, and the numbering and smart operational enhancement strategies,” Proc. IEEE,
of various feeder sections indicates an optimal order of tree vol. 105, no. 7, pp. 1202–1213, July 2017.
trimming aimed at reducing the risk of outage. Being predic-
tive in nature, such data analytics techniques would allow Biographies
DMS operators and maintenance crews to coordinate their Mladen Kezunovic is with Texas A&M University,
mitigation actions and keep the distribution grid in normal ­College Station.
operating conditions. Thomas J. Overbye is with Texas A&M University,
In summary, future control centers—whether at the trans- ­College Station.
 p&e
mission, distribution, microgrid, renewables, or any other

march/april 2018 ieee power & energy magazine 35

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY KURUKSHETRA. Downloaded on March 14,2022 at 08:17:35 UTC from IEEE Xplore. Restrictions apply.