IEEE TRANSACTIONS ON SMART GRID, VOL. 1, NO.

1, JUNE 2010

57

A Reliability Perspective of the Smart Grid

Khosrow Moslehi, Member, IEEE, and Ranjit Kumar, Senior Member, IEEE

AbstractIncreasing complexity of power grids, growing

demand, and requirement for greater reliability, security and
efficiency as well as environmental and energy sustainability
concerns continue to highlight the need for a quantum leap in harnessing communication and information technologies. This leap
toward a smarter grid is widely referred to as smart grid. A
framework for cohesive integration of these technologies facilitates
convergence of acutely needed standards, and implementation
of necessary analytical capabilities. This paper critically reviews
the reliability impacts of major smart grid resources such as
renewables, demand response, and storage. We observe that an
ideal mix of these resources leads to a flatter net demand that
eventually accentuates reliability challenges further. A gridwide
IT architectural framework is presented to meet these challenges
while facilitating modern cybersecurity measures. This architecture supports a multitude of geographically and temporally
coordinated hierarchical monitoring and control actions over time
scales from milliseconds and up.
Index TermsArchitecture, autonomous system, coordinated
operation, distributed intelligence, distributed system, execution
cycle, fast local control, global coordination, IT infrastructure,
power grid, power system control, power system operation, power
system
security, reliability, self-healing grid, smart grid, software
.
agent, temporal coordination

I. INTRODUCTION

DVANCES in communication and information technology have always been exploited by the utility industry
for improving efficiency, reliability, security and quality of
service. Increasing complexity in managing the bulk power
grid, growing concerns for environment, energy sustainability
and independence, demand growth, and quest for service
quality continue to accentuate the need for a quantum leap in
application of such technologies. This leap toward a smarter
grid is widely referred to as smart grid.
Smart grid is envisioned to take advantage of all available
modern technologies in transforming the current grid to one that
functions more intelligently to facilitate:
better situational awareness and operator assistance;
autonomous control actions to enhance reliability by increasing resiliency against component failures and natural
disasters, and by minimizing frequency and magnitude of
power outages subject to regulatory policies, operating requirements, equipment limitations, and customer preferences;
efficiency enhancement by maximizing asset utilization;

Manuscript received December 15, 2009; revised February 16, 2010. Date of
current version May 21, 2010. Paper no. TSG-00027-2009.
K. Moslehi is with ABB Network Management, Santa Clara, CA 95050 USA
(e-mail: Khosrow.Moslehi@us.abb.com).
R. Kumar resides in Cupertino, CA 95014-2843 USA.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TSG.2010.2046346

Fig. 1. Examplevariability of wind resource output.

improved resiliency against malicious attacks through

better physical security and state-of-the-art cybersecurity
to maintain data integrity, confidentiality, and authenticity,
and to facilitate nonrepudiation even in the presence of
adversaries in parts of the system;
integration of renewable resources including solar and
wind at levels from consumer premises to centralized
plants to advance global energy sustainability;
integration of all types of energy storage and other resources such as plug-in electric vehicles (PEVs) to counter
the variability of renewable resources (e.g., wind, Fig. 1)
and demand;
two-way communication between the consumer and utility
so that end users can actively participate and tailor their energy consumption based on individual preferences (price,
environmental concerns, etc.);
improved market efficiency via innovative bundled products of energy, ancillary services, risks, etc., made available to consumers and other market participants;
higher quality of servicefree of voltage sags and spikes
as well as other disturbances and interruptionsto power
an increasingly digital economy.
The momentum for realizing the smart grid vision has
increased recently due to policy and regulatory initiatives for
advancing and deploying relevant technologies as exemplified
by [1][4]. These initiatives can be categorized into five trends:
reliability, renewable resources, demand response, electric
storage, and electric transportation. These trends are also
recognized by the Federal Energy Regulatory Commission
(FERC) [1] and recent funding by the U.S. Department of
Energy (DOE) [5].
System reliability has always been a major focus area for the
design and operation of modern grids. The other trends involve
distinct smart grid (SG) resource types with diverse impacts
on reliability. Renewable resources, while supplementing the

1949-3053/$26.00 2010 IEEE

58

IEEE TRANSACTIONS ON SMART GRID, VOL. 1, NO. 1, JUNE 2010

generation capability and addressing environmental concerns,

aggravate reliability due to their volatility. Demand response
and electric storage are necessary for addressing economics of
the grid and are perceived to support reliability through mitigating peak demand and load variability. Electric transportation
is deemed helpful in meeting environmental targets and also has
the potential to mitigate load variability. Balancing the diversity of the characteristics of these resource types presents challenges in maintaining reliability and requires a quantum leap
in harnessing communication and information technologies. A
common vision for cohesive integration of these technologies
expedites their deployment and facilitates the convergence of
acutely needed standards.
This paper develops such common vision through a systematic approach based on an understanding of grid reliability challenges as well as fundamental impacts of the evolving smart
grid resource mix. We first provide an overview of the reliability challenges and then present a critical review of the salient
reliability impacts of the four resource types identified above.
We observe that an ideal mix of these resources that flattens
net demand would eventually accentuate reliability challenges
even further. Meeting these challenges requires a gridwide IT
infrastructure that provides coordinated monitoring and control
of the grid. An architectural framework for such infrastructure
should support a multitude of geographically and temporally
coordinated hierarchical monitoring and control actions over
time scales ranging from milliseconds to operational planning
horizon. Such capability is necessary to take full advantage of
the modern measurement technologies (e.g., PMUs) and control
devices (e.g., FACTS). This paper presents such architecture to
serve as a concrete representation of a common vision that facilitates the development of various components of the IT infrastructure as well as the emergence of necessary standards.
II. GRID RELIABILITY CHALLENGES
Reliability has always been in the forefront of power grid design and operation. In the United States, the annual cost of outages in 2002 is estimated to be in the order of $79B [6], which
equals about a third of the total electricity retail revenue of
$249B [7]. Much higher estimates have been reported by others.
Meeting reliability objectives in modern grids is becoming
increasingly more challenging due to various factors such as:
aggravated grid congestion, driven by, among others, uncertainty, diversity, and distribution of energy supplies due
to environmental sustainability concernswith some realtime power flow patterns significantly different from those
seen in offline design analyses;
more numerous and larger transfers over longer distances
increasing volatility and reducing reliability marginsaggravated by markets;
the grid being operated at its edge in more locations and
more often because of:
insufficient investment and limited rights of way;
increasing energy consumption and peak demand;
aging infrastructure;
maximizing asset utilization using modern tools for
monitoring, analyzing, and control;

Fig. 2. Exampleimpact of 18 GW of wind power capacity.

consolidation of operating entities giving rise to larger

footprints, and more complex problems with shorter
decision times and smaller error margins;
massive utilization of distributed resources blurring the
distinction between transmission and distribution, and accentuating complexity and volatility of the grid.
III. RELIABILITY IMPACTS OF MAJOR SG RESOURCE TYPES
The reliability impacts of the four resource types cited above
are discussed below.
A. Renewable Resources
Most rapidly expanding renewable resources are expected to
be wind and solar. In the United States, wind is expected to grow
from 31 TWh in 2008 (1.3% of total supply) to 1160 TWh by
2030 (wind energy target of 20% of total supply of 5800 TWh)
[8]. The unpredictability of wind energy resources is indicated
by their low capacity factors (typically 20% to 40% [9]) which
are much lower than conventional generators. This creates challenging problems in the control and reliability of the power grid.
As shown in Fig. 2, the variability of wind energy has little correlation to the variability of the load and hence contributes only a
little towards meeting ERCOTs peak load despite the expected
18 GW of wind capacity.
The variability of wind power is impacted by the design
of the equipment as well as their geographical distribution.
Large scale wind resources are typically far away from loads
and consequently face various transmission limitations including thermal, voltage and stability issues. The wind power
forecasting errors also present scheduling problems. The forecasting errors could be in excess of 25% depending on the
terrain, forecast horizon and forecasting methodology [10].
Wind generators also present problems regarding low voltage
ride through (LVRT). Wind power variability has a relatively
small adverse impact on regulation requirements [11].
The abundant solar energy reaching the surface of the earth is
about 1000 times the current worldwide fossil fuel consumption
each year [12]. Cumulative installed solar capacity is expected
to reach 16 GW by 2020 [13]. The two prevailing technologies
to harness this energy are photovoltaic and thermal. The variability of solar resources is very much impacted by climate and
sunlight availability. The capacity factors for photovoltaic are

MOSLEHI AND KUMAR: A RELIABILITY PERSPECTIVE OF THE SMART GRID

typically 10% to 20%. For solar thermal plants with storage, this
may reach over 70% [14]. Large scale solar resources could be
far away from loads and consequently face various transmission
limitations. However, solar resources have a positive correlation
with air conditioning loads.
From the reliability perspective, renewable resources such as
geothermal and biofuels behave similar to conventional generation. In contrast, wind and solar generally have adverse impact
on grid reliability due to:
variability and low capacity factors making the net demand
profile steeper (as depicted in Fig. 2);
low correlation with load profiles especially for wind;
relatively larger forecast errors for longer horizons;
transmission congestion due to large installations;
distribution congestion due to dispersed resources;
operational performance issues such as voltage and regulation.
Conventionally, hydro, pumped storage, and gas turbines
have been used as a remedy to address the variability of the
net demand. As renewables grow over the long run, increased
penetration of demand response, storage ,devices and PEVs
will complement the conventional remedies.
B. Load Management/Demand Response
Load management involves reduction of load in response to
emergency and/or high-price conditions. Such conditions are
more prevalent during peak load or congested operation. Reduction initiated by the consumer is usually referred to as demand response. Nonemergency demand response in the range
of 5% to 15% of peak load can provide substantial benefits in reducing the need for additional resources and lowering real-time
prices [15]. Demand response does not substantially change the
total energy consumption since a large fraction of the energy
saved during load curtailment is consumed at a more opportune
timethus a flatter load profile.
Load rejection as an emergency resource to protect the grid is
well understood and is implemented to operate either by system
operator command or through underfrequency and/or undervoltage relays. In a smart grid, this can be enhanced to allow
more intelligence and wider customer participation. Price-based
demand response as a system resource to balance demand and
supply has not been widely adopted yet. Contract-based participation has been typically below 5% (with MISO below 8%) of
peak load [15]. In a smart grid, real-time prices enable wider
voluntary participation by consumers through either automatic
or manual response to price signals, or through a bidding process
based on direct communication between the consumer and the
market/system operator or through aggregators and/or local utilities (Fig. 3). In addition to capability to flatten the load profile, demand response can serve as an ancillary resource to help
reliability.
C. Storage Devices
Most of the existing storage resources are hydro and pumped
storage. However, growth potential for these resources is much
smaller than the need for storage necessary to counter growing
net demand variability presented by new wind and solar resources. Various storage technologies are emerging to fill the

59

Fig. 3. Communications for demand response.

gap. Battery storage appears to be most promising due to improvements in technology as well as economies of scale. Storage
tends to make the net demand profile flatter and, as such, is expected to improve reliability. In addition, most battery storage
devices can respond in subsecond time scales. Hence they can
be valuable as enablers of fast controls in a smart grid. Storage
of various sizes can be distributed throughout the grid ranging
from end-use customer premises to major substations and central power stations. This can alleviate congestion in both transmission and distribution.
D. Electric Transportation
Electric vehicles (PEV, eCAR, etc.) continue to become
more popular as environmental concerns increase. They are a
significant means to reduce reliance on fossil fuels and emission
of greenhouse gases (GHG). They will be a major factor in load
growth with a potential to eventually consume 600 TWh/year.
This estimate assumes 30 kwh for a 100-mile trip [16], and
10 000 miles per year for 200 million vehicles in the United
States. From a purely reliability viewpoint, electric vehicles
have features similar to both demand response and storage resources. However, as a significant factor of load growth, PEVs
can aggravate demand variability and associated reliability
problems depending on the charging schemes and consumer
behavioral patterns. Long recharge times lead to unacceptable
vehicle unavailability and short recharge times have potential
to increase congestion at the distribution level.
IV. ULTIMATE RELIABILITY IMPACT OF SG RESOURCES
As depicted in Fig. 4, under ideal conditions, demand response, storage, and electric vehicles will be closely coordinated
with all other resources such that the net load profile would
be nearly flat. This implies that the grid would be operated
closer to near-peak load conditions most of the time. Initially,
the flattened profile tends to improve reliability by decreasing
the peak. However, over time, as the flattened load grows,
forces of optimal asset utilization will push the system closer
to the edge more often and thus make it more susceptible to
failure; hence, the need for a smart grid solution from a reliability perspective.

60

IEEE TRANSACTIONS ON SMART GRID, VOL. 1, NO. 1, JUNE 2010

Fig. 4. Ultimate reliability impact of SG resources.

V. IT INFRASTRUCTURE FOR SMART GRID

Realization of the smart grid vision requires meeting the
ever-increasing reliability challenges by harnessing modern
communication and information technologies to enable an IT
infrastructure that provides gridwide coordinated monitoring
and control capabilities. Such IT infrastructure, while facilitating utilization of modern cybersecurity measures, should
be capable of providing fail proof and nearly instantaneous
bidirectional communications among all devices ranging from
individual loads to the gridwide control centers including all
important equipment at the distribution and transmission levels.
This involves processing a vast number of data transactions for
analysis and automation. Managing the communication burden
and resulting data latency is essential for efficient analysis and
fast control responses and calls for distribution of intelligence
throughout the infrastructure, since centralized systems are
too slow for this purpose. A distributed system enables local
data processing and minimizes the need for massive data
exchanges (e.g., bad data detection at substation level, feeder
level forecasts aggregated at substation level). A distributed
system can enable the high performance needed for preventing
or containing rapidly evolving adverse events. We propose a
distributed architectural framework to deliver such performance
using modern enabling technologies:
gridwide distribution of intelligence using multiagent
frameworks for autonomous systems;
better telemetry utilizing PMU technology for faster, timestamped, higher accuracy, subsecond scanning;
faster control, e.g., FACTS-based;
more robust controls through proactive and adaptive adjustment of protection and control settings for wide-area
controls, beyond current ad hoc schemes;
proliferation and coordinated use of advanced sensors and
intelligent embedded devices (IEDs);
integrated and secure communications based on open standards to allow for flexible configurability and fail-proof
communications between all agents;
enhanced computing capabilities, e.g., cloud computing
for fail-proof and secure systems to support operator decisions and autonomous intelligent agents;
Internet technology and protocols to facilitate data
exchange and process control using standards-based distributed service oriented architecture with open interfaces
with plug-and-play hardware and software components;

Fig. 5. Hierarchical architecture for smart grid.

cybersecurity deployed, configured, and maintained consistent with NERC CIP requirements [17] and other
evolving requirements [18], [19] in a built-in rather than
bolted-on implementation.
A. Architecture
A systematic operations driven approach as opposed to an
ad hoc methods driven approach has been adopted for developing the architectural framework proposed above. This approach is based on consideration of all key operating concerns in
categories such as performance enhancement, equipment limits,
operating limits, system protection, and rapid recovery [20],
[21]. The resulting architecture calls for distribution and coordination of the necessary functional tasks in a virtual hierarchy
in three dimensions (Fig. 5):
organizational/control (grid, region, control area,
zone/vicinity, transmission substation, distribution substation, feeder, customer (load, generation, storage), etc.,
representing operational responsibilities;
geographical area (region 1 j, substation 1 n, etc.);
functions (forecasting, alarming, voltage control, etc.).
Autonomous intelligent agents are deployed, as needed,
throughout a gridwide computing infrastructure to provide
services necessary for the functional tasks in the areas of:
data acquisition and model management;
system monitoring (e.g., state estimation, security
analyses, look-ahead/forecasting);
performance enhancement (e.g., efficiency enhancement,
corrective/preventive actions, security constrained dispatch);

MOSLEHI AND KUMAR: A RELIABILITY PERSPECTIVE OF THE SMART GRID

control (e.g., AGC, automatic emergency controls, special

protection schemes).
These functional tasks potentially apply to every level, from
customer resource, feeder, and substation to the entire grid (e.g.,
a substation may perform its own share of state estimation instead of just providing raw data). The agents provide more ubiquitous local controls coordinated by global analysis, real-time
tuning of control parameters, automatic arming and disarming
of control actions, as well as functional coordination in the hierarchy, and in multiple time scales. The virtual architecture allows seamless integration of intelligence at all levels so that the
locations of specific services and data are virtualized and transparent throughout the infrastructure. Such modular, flexible, and
scalable infrastructure meets the global operational needs and
allows for evolutionary implementation on a continental scale.
It can respond to steady-state and transient operating conditions
in real-time more effectively than conventional offline solutions.
The agents operate at different time scales ranging from milliseconds to hours corresponding to the physical phenomena of
the power grid. Their actions are organized by execution cycles. An execution cycle refers to a set of related functional
tasks performed in a temporally coordinated manner. The specific periods and activities of the cycles are configurable according to the operating concerns, physical phenomena, control
response times, computational burden, and engineering practices. In each cycle, at each hierarchical level, an agent is responsible for a specific function and for a specific portion of
the grid, as needed. Each agent is persistent and capable of activating itself depending on the perceived context and deciding
if a task within its purview is to be executed. The agents can execute their tasks synchronously or asynchronously, and access
decentralized databases as needed.
Based on the allowable latency of the tasks, the cycles can be
categorized into slower and faster ones. Communications technology imposes this dichotomy at about 12 s. As such, all subsecond cycles must reside closest to the physical system. Generally, the slower cycles acquire data from larger portions of the
system and perform the more extensive computations required
for systemwide coordination of performance and control strategies. The faster cycles use data from a substation and vicinity to
address local analytical needs to respond to rapid events, subject to the control strategies developed by the slower cycles. The
execution cycles interact with each other through exchange of
event triggers, control parameters, performance indicators, etc.
A representative set of execution cycles for covering time scales
ranging from 10 ms to 1 h is depicted in Fig. 6. The specific data
and algorithms required to perform a given task (e.g., demand
forecast) can vary for different cycles and hierarchical levels.
Depending on the hierarchical position of a cycle, the specific
tasks assigned to it may address any or all of its objectives. For
example, the objectives at the slower cycles may include various contingency analyses and resource dispatching/scheduling
activities. In the 1-s cycle, the objectives may include mitigation
of slow extended oscillations. The 100-ms cycle may be focused
on detecting and containing instability, while the 10-ms cycle is
dedicated to executing intelligent RAS designed in slower cycles and deployed subject to the defined guidelines.

61

Fig. 6. Temporal coordination by execution cycles.

Subsecond control actions required in the 10- or 100-ms

cycles are made possible by the advent of synchronized measurements and FACTS-based fast repetitive control actions. The
measurements are validated using various filtering/regression
approaches (10- or 100-ms state estimation agent) at the
lowest level possible (equipment level, bay level, etc.). The
calculations are fundamentally different from the conventional
state estimation because only about half a power frequency
cycle is available for observation. The estimated data may
include the amplitude, frequency, and phase angle associated
with the instantaneous values of individual phase currents and
voltages. Estimating the rates of changes of these parameters
may also be critical. Nonconventional data errors may arise
from phase imbalances, saturation, and noise from a variety
of other control actions throughout the system (e.g., switching
transients). Communication delays are a significant part of
the total response time. The location of each agent in these
cycles is assigned to minimize the delays (assuming one-way
delays of about 0.2 ms within a substation and 6 ms between
substations). See [21] and [22] for further details of the tasks in
various cycles and hierarchical levels.
B. Adaptive Model Updates
The access to accurate global data synchronized to a microsecond presents a challenge to adaptively identifying the
required models and data for analysis in each hierarchical level
and execution cycle [20]. Based on the severity, rank, and type
of actual or potential problems in the operating conditions
to be analyzed, the models (extent of the network, detail of
generator and load models, etc.) required for calculating corrective/preventive actions are adaptively determined. As part
of these adaptive models, appropriate external equivalents are
calculated.

62

IEEE TRANSACTIONS ON SMART GRID, VOL. 1, NO. 1, JUNE 2010

C. Power Quality
In addition to reliability and efficiency issues, power quality
issues will assume importance at the distribution level. Accordingly, the criticality of functional tasks and control actions in
each execution cycle can be different based on the characteristics of the loads and dispersed resources connected to the relevant feeders. For example, a feeder in a residential area with
a significant level of incandescent lights and heating loads may
not require faster cycles, while another feeder in an industrial
area with a significant level of motor loads may critically depend
on faster cycles to maintain feeder level stability and voltage
quality during motor startups, etc. Similarly, a feeder supplying
a life-support system may need to maintain a higher level of
power quality than others.
D. Cybersecurity in a Gridwide Infrastructure
A distributed cybersecurity system monitors security
throughout the architecture to maintain data integrity, confidentiality, and authentication, and to facilitate nonrepudiation.
Data critical for grid reliability and efficiency is delivered only
to authorized agents, preventing unauthorized modifications,
and guaranteeing that delivered information is authentic while
it traverses through the infrastructure. Security in depth is
provided through such mechanisms as:
segmentation using firewalls, gateways, etc., for quick isolation of security-breached components and/or classes of
applications and services;
role-based management of identity/authentication, access,
and command level filtering;
evolving security life cycle [25] in response to evolving
threats and infrastructure components through sufficiently
frequent secure remote updates;
efficient and scalable policy and key encryption mechanisms, resilient in the presence of active adversaries
[26][28];
systemwide time synchronization for event correlation.
E. Technical Feasibility
The technical feasibility of the proposed architecture relies
on recent advances in the areas of sensors, telecommunications,
computing, Internet technology, power equipment, and power
system analysis. The flexibility and scalability of the design
has been established through a quantitative analysis of a large
example power grid [22][24]. This analysis includes requirements for monitoring, analysis and control. According to this
analysis, the data exchange volumes at various levels of the infrastructure are entirely feasible with contemporary technologies. The latency for an exchange ranges from a few milliseconds at the substation to several seconds at the grid level. However, it is possible to provide a small selected subset of the information at the grid level with a 1-s delay. In spite of the large
range in the latency, using the PMU timestamps, it is possible to
limit the time skew of the data at any level to 1 ms or even less
if so desired.
To establish the financial feasibility of the proposed IT infrastructure, a scalable methodology is developed [23] for assessing the costs and benefits using published statistics. In addition to the cost of necessary IT hardware, the cost models

Fig. 7. Analogy for transformational technology.

account for developing and deploying necessary new software

innovations. It is assumed that once the necessary new techniques are prototyped, demonstrated, and implemented in the
context of a selected function, they can easily be adopted for
other functions. Cost of communication links and human interfaces such as visualization, and intelligent alarming were not
included. Benefit models quantify selected significant benefits:
energy cost savings and value of reduction of service interruptions. The methodology is fully scalable and can be adopted
for a preliminary cost/benefit assessment for a system of any
size. It has been applied to a generic example system and the results indicate that the benefits significantly outweigh the costs.
Once the R&D entry barrier is overcome, the costs of subsequent implementations are of the same order of magnitude as
for conventional control centers. The recent stimulus funding in
the smart grid arena [5] is a harbinger of the future investments
at the levels suggested in [23].
VI. SYNERGIES WITH CURRENT PRACTICES
The proposed architecture provides a generalized framework
for the design and development of various components of the IT
infrastructure and emergence of necessary standards and protocols needed for the smart gridespecially with regard to reliability issues. The essentiality of an architectural approach in the
transformation of the grid to a smart grid is analogous to the
role of the iPhone paradigm in the transformation of the phone
system from its intelligent form of the 20th century (represented
by the touch-tone phone shown in Fig. 7) to its current state. It
was not because of a few specific applications that iPhone revolutionized the phone but for its architecture that led to an
explosion of functionality.
The proposed architecture is intended to enable a similar
transformation of todays grid to a smarter grid. It provides
a framework for a systematic development of innovative applications and the integration of new and existing applications
to meet various reliability concerns, and as such facilitate
integration of various smart grid resources.
A. Industry Trends
The proposed architecture is in synergy with current industry
practices. Many of the smart grid technologies are already in
place in various ad hoc implementations. Examples of such implementations include wide-area monitoring and control, special protection schemes, state estimation, and forecasting. These
are briefly reviewed below.

MOSLEHI AND KUMAR: A RELIABILITY PERSPECTIVE OF THE SMART GRID

Wide-area monitoring and control has been gaining worldwide interest. This involves gathering data from and controlling a large region of the grid through the use of time synchronized phasor measurement units (PMUs). Currently efforts are
underway for the design and development of a robust and secure data highway for the synchronized phasor data in the North
America [29]. Some key applications dependent on such data include [30]:
phase angle monitoring;
slow extended oscillation monitoring;
voltage stability/transfer capability enhancement;
adaptive line thermal monitoring /dynamic rating;
PMU augmented state estimation;
geomagnetic disturbance recognition.
Special protection/remedial action schemes (SPS/RAS) are
proliferating. They can be seen as precursors of intelligent
agents. The current customized schemes are too expensive
to build and maintain. Additionally, arming and disarming
of these schemes is not adaptive. The proposed architecture
will improve their effectiveness by frequent parameter updates from a higher level and greater use of local intelligence;
hence intelligent SPS/RAS or iSPS/iRAS. This together with
plug-and-play components allows real time coordination of
numerous schemes/control actions at lower costs. Such coordination is already pursued in an ad hoc manner [31], [32].
State estimation provides reliable knowledge of the current
state of the power system for use by the operator and other analytical functions as needed. In current practice, since all analytical functions are centralized, a typical state estimator is also
centralized. To provide intelligence throughout the grid, timely
state estimation must be available at local levels for all required
execution cycles/time scales (including subseconds). As such, a
distributed state estimator with functional agents at every level
of the three dimensional hierarchy can enable local analysis. For
example, a transmission substation level agent retrieves necessary data from the local substation and other substations within
the electrical vicinity. It resolves topology errors, identifies
and rejects erroneous measurements, and when necessary, obtains substitute data from other functional agents (e.g., bus load
estimation or forecast) at the substation level or other levels.
Other higher level agents have to coordinate their estimation
with lower level solutions. Similarly, state estimation agents at
various distribution levels feed processed information to higher
distribution level agents and ultimately to the appropriate cycles of the transmission agents. This enables a well-coordinated
scalable methodology.
Demand and resource forecasting is usually done at a macroscopic level such as control area and load zone. However, as
need for more discrete and intelligent local control increases
with distributed resources, accurate forecasts at the local level
will be required. The proposed architecture provides for plugand-play forecasting agents throughout the grid to access required information to produce such load and generation models.
For example, at the substation level, the agents may forecast data
for a bus subject to operating constraints suggested by higher
level agents who in turn can have lower level data folded into
their own forecasts.

63

In parallel with the above implementation trends, major standards initiatives are also underway sponsored by NIST [23],
IEEE (IEEE 2030) [24], etc. We believe these efforts would converge sooner if a common vision for the smart grid architecture
is shared by all stakeholders.
VII. CONCLUSIONS
Smart grid is envisioned as a quantum leap in harnessing communication and information technologies to enhance grid reliability and to enable integration of various smart grid resources
such as renewable resources, demand response, electric storage,
and electric transportation. Based on a critical review of the reliability impacts of these resources, it is concluded that an ideal
mix of the smart grid resources leads to a flatter net demand that
eventually accentuates reliability issues further. Thus, the centrality of meeting reliability challenges in the realization of the
smart grid is underscored.
Meeting these challenges requires a systematic approach to
develop a common vision for cohesive gridwide integration of
the necessary IT technologies. An architectural framework is
proposed to serve as a concrete representation of such common
vision to facilitate the design, development, and integration
of various components as well as the emergence of necessary
standards and protocols. This architecture supports a multitude of fail-proof geographically and temporally coordinated
hierarchical monitoring and control actions over time scales
ranging from milliseconds to operational planning horizon.
The architecture delivers high performance through a virtual
hierarchical operation of a multitude of software agents and
services in organizational, geographical and functional dimensions. This infrastructure can be thought of as a super EMS
consisting of a network of networks that allows for evolutionary
implementation of the infrastructure.
An architectural approach is essential for transforming the
power grid to a smarter grid as the iPhone architectural paradigm was for transforming the phone. It was not because of a
few specific applications that iPhone revolutionized the phone
but for its architecture that led to an explosion of functionality.
REFERENCES
[1] Smart grid policy U.S. Federal Energy Regulatory Commission
(FERC), Docket PL09-4-000, 2009.
[2] Title XIIISmart Grid, Sec. 1301, Statement of Policy on Modernization of Electricity Grid, Energy Independence and Security Act of
2007 (EISA).
[3] Smart Grid Systems, California Senate Bill 1438 [Online].
Available:
http://info.sen.ca.gov/cgi-bin/postquery?bill_number=
sb_1438&sess=PREV&house=B&site=sen
[4] American Recovery and Reinvestment Act of 2009, Pub. L. No. 111-5.
[5] Recovery ActSmart Grid Investment Grant Program, Electricity
Delivery and Energy Reliability, Research, Development and
Analysis, U.S. Department of Energy, Funding Opportunity No.
DE-FOA-0000058, CFDA#81.122.
[6] K. H. LaCommare and J. H. Eto, Cost of power interruptions to electricity consumers in the United States, LBNL-58164, 2006.
[7] Summary Statistics for the United States U.S. Energy Information Administration, 19962007 [Online]. Available: http://www.eia.doe.gov/
cneaf/electricity/epa/epates.html
[8] 20% wind energy by 2030, U.S. Department of Energy, 2008
[Online]. Available: http://www1.eere.energy.gov/windandhydro/
pdfs/41869.pdf

64

IEEE TRANSACTIONS ON SMART GRID, VOL. 1, NO. 1, JUNE 2010

[9] Annual report on US wind power installation, cost and performance

trends: 2007 Energy Efficiency and Renewable Energy, U.S Department of Energy, 2008 [Online]. Available: http://www.nrel.gov/docs/
fy08osti/43025.pdf
[10] I. Marti, Evaluation of advanced wind power forecasting models, presented at the European Wind Energy Conf., Athens, Greece, 2006.
[11] The effects of integrating wind power on transmission system planning, reliability and operations, report on phase 2: System performance
evaluation New York State Energy Research and Development Authority, Albany, NY, 2005 [Online]. Available: http://www.nyserda.
org/publications/wind_integration_report.pdf
[12] Waiting for the sunrise (solar energy forecast), The Economist, p. 95,
May 19, 1990.
[13] Solar energy industry forecast: Perspectives on U.S. solar market trajectory Solar Energy Technologies Program, U.S. Department of Energy, 2008 [Online]. Available: http://giffords.house.gov/DOE%20Perspective%20on%20Solar%20Market%20Evolution.pdf
[14] Executive summary: Assessment of parabolic trough and power
tower solar technology cost and performance forecasts National Renewable Energy Laboratory, 2003 [Online]. Available:
http://www.nrel.gov/csp/pdfs/35060.pdf
[15] Harnessing the power of demandHow ISOs and RTOs are integrating demand response into wholesale electricity markets, Markets
Committee of the ISO/RTO Council, 2007 [Online]. Available:
http://www.isorto.org/atf/cf/%7B5B4E85C6-7EAC-40A0-8DC3003829518EBD%7D/IRC_Renewables_Report_101607_final.pdf
[16] R. Gawel, Teslas tests confirm roadsters 245-mile range,
in Electron. Des., Nov. 5, 2007 [Online]. Available: http://
electronicdesign.com/article/power/tesla-s-tests-confirm-roadster-s245-mile-range173.aspx
[17] Reliability Standards, Critical Infrastructure Protection (CIP) North
American Electric Reliability Corporation (NERC) NERC [Online].
Available: http://www.nerc.com/page.php?cid=2|20
[18] Data bus technical specifications for NASPInet; phasor gateway technical specifications for NASPInet U.S. Department of Energy [Online].
Available: http://www.netl.doe.gov/business/solicitations/ NASPI/
[19] Smart grid cyber security strategy and requirements National Institute
of Standards and Technology, U.S. Department of Commerce [Online].
Available: http://csrc.nist.gov/publications/drafts/ nistir-7628/draft-nistir-7628.pdf
[20] K. Moslehi et al., Control approach for self-healing power systems:
A conceptual overview, in Electricity Transmission in Deregulated
Markets: Challenges, Opportunities, and Necessary R&D. Pittsburgh, PA: Carnegie Mellon Univ., 2004.
[21] Transmission fast simulation and modeling (T-FSM)Functional requirements document EPRI, Palo Alto, CA, 2005. 1011666.
[22] Transmission fast simulation and modeling (T-FSM), architectural requirements EPRI, Palo Alto, CA, 2005. 1011667.
Transmission fast simulation and modeling
[23] Intelligrid
(T-FSM)Business case analysis EPRI, Palo Alto, CA, 2005.
1012152.
[24] K. Moslehi et al., Framework for a self-healing power grid, presented
at the IEEE PES General Meeting, San Francisco, CA, Jun. 2005.

[25] R. Creel, Assuring software systems security: Life cycle considerations

for government acquisitions Carnegie Mellon University, Software
Engineering Institute, Pittsburgh, PA, 2007 [Online]. Available:
https://buildsecurityin.us-cert.gov/daisy/bsi/articles/best-practices/
acquisition/892-BSI.html
[26] H. Khurana, Scalable security and accounting services for contentbased publish/subscribe systems, in Proc. E-Commerce Track of the
ACM Symp. Applied Computing (SAC), 2005.
[27] D. Dasgupta, Immuno-inspired autonomic system for cyber defense,
information security, Tech. Rep. 12 (4), 2007.
[28] V. Gorodetski et al., The multi-agent systems for computer network
security assurance: Frameworks and case studies, in Proc. IEEE Int.
Conf. Artif. Intell. Syst. (ICAIS02).
[29] North American SynchroPhasor Initiative [Online]. Available: http://
www.naspi.org/
[30] Working Group 601 of Study Committee C4, Wide area monitoring
and control for transmission capability enhancement CIGRE Technical
Brochure, Final Report, 2007.
[31] P. Arons, Piloting a centralized remedial action scheme (C-RAS)
with emerging telecomm/protection technologies, presented at the
OSISOFT User Conf., Monterey, CA, 2007.
[32] C. W. Taylor et al., WACSWide-area stability and voltage control
system: R&D and on-line demonstration, Proc. IEEE, vol. 93, no. 5,
pp. 892906, May 2005.
[33] Smart Grid Interoperability Standards Project, [Online]. Available:
http://www.nist.gov/smartgrid/
[34] IEEE P2030 Draft Guide for Smart Grid Interoperability of Energy
Technology and Information Technology Operation With the Electric
Power System (EPS) and End-Use Applications and Loads, [Online].
Available:
http://grouper.ieee.org/groups/scc21/2030/2030_index.
html

Khosrow Moslehi (S76M82) received the Ph.D. degree from the University
of California, Berkeley.
He is the Director of Product Development at ABB Network Management in
Santa Clara, CA. He has over 25 years of experience in R&D in power system
analysis and optimization, system integration and architecture, electricity markets, and smart grid.

Ranjit Kumar (S73M78SM84) received the Ph.D. degree from the University of Missouri, Rolla (now known as Missouri University of Science and
Technology).
He has over 30 years of experience in research and development of algorithms
and software for the design, operation, and real-time control of power systems,
markets, and smart grid. He has made several contributions related to power
system stability, fuel resource scheduling, and dynamic security analysis. He is
a Consultant to ABB Network Management, Santa Clara, CA.