IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO.

2, MARCH/APRIL 2013

899

Comprehensive Real-Time Simulation

of the Smart Grid
Feng Guo, Student Member, IEEE, Luis Herrera, Student Member, IEEE, Robert Murawski, Member, IEEE,
Ernesto Inoa, Student Member, IEEE, Chih-Lun Wang, Philippe Beauchamp,
Eylem Ekici, Senior Member, IEEE, and Jin Wang, Member, IEEE

AbstractThis paper presents a real-time simulation platform

for smart grid applications. The developed platform is capable
of simulating complex smart grid models with large numbers
of high-speed switching devices at real time. Furthermore, an
integrated approach is adopted to combine real-time simulations
of communication systems and electric power systems together,
which provides an effective approach to examine communication
and distributed control related issues in smart grids. With the
flexibility in representing a wide range of communication network
configurations, the developed platform can also be used to evaluate
reconfiguration strategies of communication networks in smart
grids. A case study is demonstrated based on this platform. Simulation results validate the capability of the platform and also show
the importance of the proposed comprehensive approach for the
study of smart grids.
Index TermsCommunication network, real-time simulation,
smart grid.

I. I NTRODUCTION

HE TERM smart grid encompasses up-to-date technologies in areas such as power electronics, communication,
and renewable energy resources in order to establish a more
secure, reliable, economic, and eco-friendly electric power
system [1], [2]. A smart grid usually contains different kinds of
renewable energy resources, such as photovoltaic (PV), wind
turbine, battery storage system, etc., and a large number of
power electronics interface circuits. Thus, experimental studies
of large-scale smart grids are usually not economically feasible.
As an evidence, even for microgrid, where the number of
distributed energy sources and intelligent loads is quite limited,
there are only a handful test platforms around the world [3], [4].
Therefore, simulation becomes a powerful and convenient tool
in this research area.

Manuscript received May 24, 2011; revised December 3, 2011; accepted

December 12, 2011. Date of publication January 16, 2013; date of current version March 15, 2013. Paper 2011-IPCC-299.R1, presented at the
2011 IEEE Energy Conversion Congress and Exposition, Phoenix, AZ, USA,
September 1722, and approved for publication in the IEEE T RANSACTIONS
ON I NDUSTRY A PPLICATIONS by the Industrial Power Converter Committee
of the IEEE Industry Applications Society. This work was supported by the
U.S. Department of Energy under Project DE-OE-0000402.
F. Guo, L. Herrera, R. Murawski, E. Inoa, E. Ekici, and J. Wang are with
The Ohio State University, Columbus, OH 43210 USA (e-mail: guof@ece.osu.
edu; herrera.46@buckeyemail.osu.edu; murawski.8@osu.edu; einoa@ieee.org;
ekici@ece.osu.edu; wang@ece.osu.edu).
C.-L. Wang is with Ford Motor Company, Dearborn, MI 48120, USA
(e-mail: wang.1376@osu.edu).
P. Beauchamp is with Opal-RT Technologies, Montreal, QC H3K 1G6,
Canada (e-mail: philippe.beauchamp@opal-rt.com).
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/TIA.2013.2240642

The integrated power electronics devices in the smart grid introduce two great challenges to the simulation: high-frequency
pulse width modulation signals and large numbers of switching
devices. As a result, traditional offline simulators are time
consuming for long period simulations and not suitable to study
slow phenomena [5]. Moreover, this kind of simulation usually
lacks the hardware interface with real devices; thus, it cannot
be used at all the development stages of the smart grid, such as
prototyping and testing.
As early as 20 years ago, manufactures of large thyristorbased high-voltage direct current and Flexible AC Transmission
Systems devices have used the digital real-time simulator and
hardware-in-the-loop (HIL) methods to solve the aforementioned problems for power system with relatively slow frequency and small numbers of switches [6]. After two decades
of developments, the digital real-time simulator can run a large
and complex model at the physical time; thus, it is suitable
to simulate a long-time frame process. In addition, with the
support of the real-time simulator, power-in-the-loop can be
utilized for testing of high-power devices that are connected to
complex systems.
The requirement for real-time simulation of high-frequency
switching devices was brought forward by the manufacturers of
fast power electronics devices such as Mitsubishi more than five
years ago [7]. The real-time simulator in [7] can run the simulation with a time step of 10 s for one three-phase inverter. In
[8], a field-programmable gate array (FPGA)-based real-time
simulator can reach a time step of 500 ns for a diode-clamped
three-level voltage source converter. However, for a system
with large numbers of switching devices, it becomes difficult to
achieve such small time step, and specific consideration is still
needed. A microgrid model with two renewable energy sources
and two inverters in [9] is only able to run with a switching
frequency of 1 kHz at real time. Therefore, sufficient research
and new methods are still needed to speed up the adoption of
real-time simulation on smart grid.
On the other hand, the communication network is essential
to effectively incorporate many of desired features of the smart
grid such as distributed automated system, distributed energy
resource protection, islanding, and display of network state and
performance [1]. Thus, another challenge in smart grid study
is to integrate different kinds of communication networks and
simulate their influence to electric power systems at real time.
In [10], Nutaro et al. developed an offline software simulation
of a 17 bus power network with communication between loads
and generators, investigating the effects of factors such as

0093-9994/$31.00 2013 IEEE

900

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 2, MARCH/APRIL 2013

bandwidth and latency. RTDS-based systems for testing protective relays in power networks have been discussed in [11] and
[12], but with a separate study on communication networks.
In [13], a real-time HIL model of a power network was implemented along with a physical network of two nodes: the
real-time simulation and a physical remote control center. The
control center served as a means of monitoring and controlling
the overall system, and testing cyber vulnerabilities of it. In
sum, the importance of emulating communication network
in smart grids becomes apparent. An effective modeling of
communication between control centers along with present
real-time simulation of power networks could show more realistic results and bring great benefits to related studies in
smart grids.
This paper presents a comprehensive platform to overcome
current challenges in smart grid related real-time simulation. In
Section II, theoretical analysis showcases the reason of detailed
modeling communication network in smart grid related study.
Section III introduces the platform, and the solutions to the
challenges. Section IV describes a case study, and Section V
illustrates the simulation results. Finally, in Section VI, summary and conclusion are presented.

Fig. 1.

Communication network.

Fig. 2.

Average latency versus per node arrival rate.

II. C OMMUNICATION N ETWORK M ODELING

AND S IMULATION
Communication networks are expected to be integrated components of smart grid systems. The coexistence of communication and power networks and their joint operation necessitate
accurate modeling of communication events. While at the first
glance, it is tempting to model the communication network
as a black box that introduces delays between its input and
output, the complex interactions among network components
and between data sources and the network make it less tractable.
To illustrate this, consider a case that requires dynamic reconfiguration of the communication network based on changing
operating conditions of the power network. Assume that the
operation of a microgrid is controlled by a Microgrid Control
Center (MCC). To this end, the controller relies on data periodically collected from power sources and loads. Since most data
sources are distributed over large geographic areas, at this point,
it is assumed that the said data is delivered to a centralized access point (AP) via wireless links using Carrier Sense Multiple
Access/Collision Avoidance (CSMA/CA) principles, which is
the most widely used channel access scheme in todays wireless
devices including IEEE 802.11 compliant ones. The information is then relayed to controllers via a wired connection. The
portion of interest of the communication network is shown
in Fig. 1, which consists of 20 reporting nodes, one AP, and
the controller. Consider two operation modes for the network:
A. Regular Operation Mode: Data sources report their readings
at a rate of A packets/s/source, which are collected at AP and
delivered to the controller in TA s; B. Exception Operation
Mode: Events in the power network causes the reporting rate
to rise to B packets/s/source. Messages must be reported to
the controller in TB s.
Consider the transition from the Regular to the Exception
Operation Mode and its effects on the communication net-

work. Assume that this transition results in the changes to the

reporting rate only such that B > A , while the reporting
delay remains constant, i.e., TB = TA . As a general principle,
the latency of message delivery in random access networks
increases exponentially as the load offered to a wireless network operating under CSMA/CA rules increases. The goal
is to control the topology and channel access rules while
retaining the CSMA/CA principles to ensure stability of the
power network. Assume that there is a function F based on
existing non-saturation CSMA/CA performance models such
as [14] that can estimate the average service rate 0 based on
the number of competing nodes N and the per node packet
arrival rate as 0 = F (N, ). Then, the average packet delay
D0 including the queuing time can be estimated using the
M/M/1 queuing model as D0 = 1/(0 ) s. Now, consider
the option of grouping the nodes into K groups such that, at any
given time, only N/K of the nodes contend for channel access
while others wait for their turn to contend. This can be achieved
by dividing the time into slots of length t. Each group can access
the channel every K time slots, where t is long enough to serve
several packets. In Fig. 2, two cases are identified where K = 1
(group marked with dotted lines) and K = 4 (groups marked
with red dashed lines). Using the model represented by the
function F, the service rate of single region can be estimated as
k = F (N/K, ). Then, the average packet delay Dk for a K
group arrangement can be estimated as Dk = t((K 1)/2 +
"1/(t(s ))#K + 1), to account for the delay experienced
during inactive slots and while contending for the channel.
The average packet delay is shown as a function of per node
arrival rate in Fig. 2 for t = 35 ms for K = 1 and K = 4.

GUO et al.: COMPREHENSIVE REAL-TIME SIMULATION OF THE SMART GRID

K = 1 provides significantly lower latency to packets when

the reporting rate is below = 2.5 packets/node/s. When the
arrival rate increases further, the delay increases exponentially.
The delay of any node for the K = 4 case is dominated by
the waiting time between active time slots. Grouping the nodes
into smaller contention groups significantly reduces the random
access delay. Hence, K = 4 is more suitable for higher arrival
rates: For > 2.5 packets/node/s, the average delay is significantly lower for K = 4 than for K = 1. Inevitably, the exponential growth of latency occurs as the arrival rate increases.
However, this occurs for > 8 packets/node/s, which is significantly higher than the threshold for K = 1. For TA = TB =
150 ms and A < 2.5 packets/node/s in the regular operation
mode, K = 1 is sufficient. If switching to exception operation
mode results in an arrival rate larger than 2.5 packets/node/s,
then the communication network should be reorganized with
K = 4. It should be noted that K = 1 and K = 4 are not
the only two options. Different reorganizations with different
parameters are also possible. Moreover, it is also possible to
organize the network into a hierarchy deeper than two levels
(i.e., including intermediate levels between reporting nodes and
the AP).
As highlighted in the preceding example, the highly nonlinear relationship between the source rates and the resulting
delays across the network is a significant hurdle to overcome.
In most analytical models that relate arrival rates to network
latencies, the relationship is established among the first-order
statistics of the input and output random variables. Except in
simplest of cases, it is not possible to derive tight bounds on
delays, or other relevant metrics such as loss rates. Moreover,
most analytical models accuracy suffers greatly as the network
size grows from single hop settings to relay networks. On the
other hand, there are a number of simulation platforms that
replicate the behavior of communication networks relatively
accurately. While these simulation tools cannot generate analytical results, their estimation of delay and loss parameters
vis-a-vis actual operation of the communication network is very
accurate. Reserving the use of stochastic model for future studies of network design, this paper concentrates on developing
a simulation platform that mimics both the power network as
well as the communication network operation accurately. With
the ability to simulate a wide range of communication network
configurations, the developed platform can also be used to
evaluate various designs and reconfiguration strategies for an
actual smart grid.
III. R EAL -T IME S IMULATION P LATFORM
The real-time simulation platform presented in this paper
(see Fig. 3) is based on PC technology running a Linux operating system and offers an underlying test bed for the study on
smart grid. It consists of four real-time simulators with a total
of eight CPUs, 48 cores, five FPGA chips, and more than 500
analog and digital inputs/outputs [15]. Dolphin [16] PCI boards
are used to provide an extremely high-speed and low-latency
real-time communication link between simulators.
1) Electric Power System Simulation: With the technology
of switch event interpolation [5], State Space Nodal (SSN)

901

Fig. 3. Real-time simulation platform.

solver [17], and parallel computation, the system provides an

effective way to simulate a system with large numbers of highfrequency switching devices.
The challenge of simulating models with high-speed switching devices is the small step size required to achieve adequate
simulation accuracy. Selecting too large of a step size will
produce an erroneous simulation output and selecting too small
of a step size will not be able to meet the real-time constraint.
The solution to this problem is the switch event interpolation technology, which combines software solving algorithms
with high-end hardware technology. The RT-EVENTS software
library [15] in combination with FPGA is used to handle
discrete events that occur inside one model calculation step,
which allows the model to run with a larger step size without
sacrificing simulation accuracy. The switching devices can also
be modeled directly inside the FPGA in order to obtain higher
switching frequencies for specific applications.
The limitation of the number of switches that can be used
in a real-time model is caused by the O(2N ) computational
complexity. It will increase a great number of inverse matrices to be computed for each additional switch when the
total number of switches becomes large. This problem was
addressed by the SSN solver, which uses a combination of
the nodal approach and the state space solver approach. SSN
solver decouples the whole system state-space matrix into small
groups of matrices, thus reducing the overall computational
complexity. Benchmarks have shown that it is possible to run
simulations with more than 100 switches in a single subsystem.
Furthermore, relying on the parallel computation capability,
this system is able to run a complex power system model
with a maximum of 48 subsystems, which can further increase
the switching frequency and number of switching devices in
the model. In this approach, the smart grid model will be
intentionally divided into small subsystems. Each subsystem
can be either a renewable energy resource unit, such as a PV
model, wind turbine model, or a controller unit. These will
be distributed into cores in different simulators, which will be
executed and synchronized with high precision.
In general, with the help of the presented real-time simulation
system, a switching frequency up to 10 kHz can be achieved,
and a large number of switches can be included in one model.
Moreover, the system allows multiple users to be connected to
the real-time simulator concurrently to perform collaborative
simulation, for which distributed control and HIL operation can
be easily realized.

902

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 2, MARCH/APRIL 2013

Fig. 4. SITL model of the communication network.

2) Communication System Simulation: When combined

with a discrete network simulator, the real-time simulation
system can provide advantages to simulating communication
network and distributed control in the smart grid in a cost
effective manner. Such studies would allow identification of
the influence of key communication factors to the smart grid,
such as protocols, latency and bandwidth requirements, cyber
security, and data management [2].
OPNET [18] is a commercially well-developed network simulator that has been used extensively in research. A systemin-the-loop (SITL) package is available for use with OPNET
that bridges the OPNET network simulation with the hardware
simulation. It has the ability to accept real-time network traffic
and simulate either a wired or wireless network. OPNET allows
significant flexibility in terms of network development; this,
combined with the integrated support of many well-known
communication protocols, makes OPNET an excellent option
for implementation of the proposed communication network.
Fig. 4 shows an example of communication network model
using SITL gateways to link with power network model. Each
key power network device, such as sensors, breakers, control
center, etc., will be assigned its own IP subnet and, therefore,
will require the simulated network to route packets between
them. Within the OPNET environment, IP traffic is accepted
at different SITL gateways by filtering traffic by the source
address. In Fig. 4, a total of six SITL gateways are utilized
within OPNET. Five gateways are connected to the simulated
power network interfaces. The sixth gateway connects to the
MCC which gathers information and distributes control messages to the power network as required. Each device has a
router associated with it which routes the IP layer traffic to its
intended destination based on pre-defined static routing tables.
By placing the SITL gateways, simulated routing, and switching hardware within OPNET, different distances between power

network devices and the delays and bit error rates associated
with transmitting packets can be simulated. In addition, different virtual network types can be chosen between gateways,
such as Ethernet, wireless, and fiber optic, depending on the
transmission distance and data type in a real power network.
OPNET SITL also allows the transmission of individual packets
from the input ports to the output ports of the simulated network
by following the exact steps of buffering, protocol processing,
transmission, and reception.
When combining the power system (continuous time system)
and communication system (discrete time system) together,
the proposed platform runs both systems at real time and
simultaneously solves two distinct models in parallel. This has
an advantage in terms of synchronization of the systems since
both power and communication network models are running
at the same rate as a wall clock. The simulation setup is
shown in Fig. 5. The mathematical model of the power network
is first built in Matlab/Simulink environment, then converted
to real-time model, and finally solved by the Linux based
real-time simulators; whereas the communication network is
modeled using the discrete network simulator OPNET running
on another desktop computer. A second desktop computer is
utilized as the MCC server, and the information sent from the
OPNET network is processed in it.
Meanwhile, the power system model has the capability of
information exchange through an external Ethernet connection,
so power and communication systems can exchange information through physical Ethernet links. An external C code is developed for Linux systems which develops the necessary layers
for real-time information exchange using TCP/IP or UDP/IP
protocols. Therefore, the corresponding data in power system
model is sent to the external computer running OPNET through
physical Ethernet links. Within OPNET, the traffic is filtered
from the interface by the source address and thus creating

GUO et al.: COMPREHENSIVE REAL-TIME SIMULATION OF THE SMART GRID

Fig. 5. Combined simulation platform of power system and communication

system.

different nodes for building an accurate communication network. Information is then processed in real time within OPNET
and sent to the MCC server through another Ethernet link.
After necessary calculation, the MCC sends back the control
command to OPNET, which then distributes the information to
the corresponding destinations within the power system model.
In conclusion, with the cooperation of advanced software
and hardware, and the combination of power network simulator
and communication network simulator, the platform presented
in this paper is able to provide a comprehensive approach for
smart grid-related study.
IV. C ASE S TUDY
A. Model Description
In this section, the smart grid in a small community is
modeled with the developed real-time simulation system. The
model includes a Local Energy Storage (LES) device, a
PV-based generator, a wind turbine, as well as the communication network described in Fig. 4. The system is shown
in Fig. 6.
Each unit in the smart grid is modeled with detailed characteristics. The LES is modeled as recycled electric car batteries.
The impedance and output voltage of LES are not only determined by the State of Charge (SOC), but also the temperature
inside the LES [19][22]. The PV is built with the mathematical
model described in [23]. The output voltage and power of this
unit are related to the irradiation and environment temperature.
The wind turbine is modeled as a doubly fed induction generator (DFIG) with back-to-back inverters. The noncritical load is
modeled as a nonlinear load powered through power electronics
circuits, while the critical load is modeled as a three-phase
resistive load. Additionally, all circuit breakers in the system are
modeled as solid state circuit breakers. In total, there are nearly
100 switches in the model operating at a switching frequency
of 10 kHz. Furthermore, because the switching actions will
influence the whole system characteristics, in order to simulate
the system in a more precise way, the switching frequency
is kept at 10 kHz in both grid-connected mode and islanding

903

mode. This detailed modeling also enables further HIL-based

study.
The MCC within the smart grid is able to serve the following
purposes: changing the settings of protection devices based
on either grid-connected or islanding mode, monitoring values
within the smart grid, and sending signals to reconfigure the
smart grid depending on its current configuration [24]. In the
model, all circuit breakers are equipped with transmitter and
receiver units, which are able to send its status to the MCC as
well as acquire control signals. In order to achieve a low latency,
the circuit breakers of noncritical loads communicate with the
core router at the MCC through simulated fiber optic cabling
using the Synchronous Optical Network standard at OC-1 speed
(51.8 Mbps). The main circuit breaker at the point of common
coupling (PCC) connects to a separate router via WiFi 802.11
wireless communication, since it is usually located far away
from the MCC. While the LES dc bus voltage node connects
to the separate router via Ethernet. As stated in Section III,
the communication network is emulated by OPNET while the
MCC is implemented in an external computer.
The simulation parameters are shown in Table I.
B. Control Strategy
The aim of the controller design is to make sure the system
works properly in both grid-connected mode and islanding
mode. More importantly, the controller should avoid large
voltage and current spikes during the transition from gridconnected mode to islanding mode.
1) Grid-Connected Mode: In grid-connected mode, a multiloop controller is used for all the three grid tied inverters. The
control objective of the inverters is to keep the dc-link voltage
constant. The outer loop of the controller consists of a voltage
loop with dc-link voltage as feedback. The inner loop is a grid
current control loop, with the active current reference generated
by the outer voltage loop and reactive current reference set to 0,
as shown in Fig. 7.
Referring to the circuit diagram in Fig. 6(b), the simplified
current open-loop transfer function in dq-rotating reference
frame is
1
Id
=
.

Vdl
Lf s + R f

(1)

The response of the inner current loop should be much faster

than the outer voltage loop, hence a proportionalintegral (PI)
controller is used to achieve a good system performance. The
design of the voltage control loop is based on the assumption
that the inner current loop is ideal. The open-loop transfer
function is
Vdc
1
.
=

Id
Cs

(2)

Again, a PI controller is applied to get the required system

response characteristics.
The control of the rotor side inverter of the DFIG follows
the strategy in [25]. A decoupled control between the electric
torque and rotor excitation current is achieved by a rotating

904

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 2, MARCH/APRIL 2013

Fig. 6. Illustration of the case study: a small community microgrid. (a) System diagram; (b) detailed circuit topology.
TABLE I
S IMULATION PARAMETERS

Fig. 7. Control strategy of the LES-connected inverter during grid-connected

mode.

dq axis frame with the d-axis oriented along the stator flux
vector position.
Considering the converters, the LES-connected converter and
noncritical-load-connected converter are working in current
control mode, while the PV-connected converter is working
with Maximum Power Point Tracking (MPPT) strategy.
2) Islanding Mode: In islanding mode, the inverters and
converters connected to PV and DFIG are still controlled current sources as in the grid-connected mode. However, the LESconnected inverter, because of the loss of the grid, is controlled
to be a voltage source to function as the swing bus of the system.
The same multi-loop control structure is utilized for the LES
inverter, but with the ac bus voltage as feedback in the outer

voltage loop, as illustrated in Fig. 8. The inner loop is the same

current control loop as in grid-connected mode.
The current open-loop transfer function in islanding mode is
slight different from grid-connected mode, which becomes
Id
Cf s
.
=
Vdl
Lf C f s 2 + R f C f s + 1

(3)

A PI controller is implemented for the current control. At the

same time, the open-loop transfer function for the outer voltage
loop becomes
Vd
Kcp s + Kci
(4)
=
Id Lf Cf s3 + (Rf Cf + Kcp Cf )s2 + (Kci Cf + 1)s

GUO et al.: COMPREHENSIVE REAL-TIME SIMULATION OF THE SMART GRID

Fig. 8.

Control strategy of the LES-connected inverter during islanding mode.

where Kcp and Kci are the proportional gain and integral gain
of the current PI controller. A PI controller is applied as well to
reach the voltage control requirement.
As mentioned above, in islanding mode, the LES-connected
inverter is used to control the ac bus voltage. Therefore, the control purpose of the LES-connected dc/dc converter is changed to
control the dc-link voltage. Considering that the output current
of the LES should also be tightly controlled, a multi-loop control structure with outer voltage loop and inner current loop is
designed for this converter. The outer voltage control loop uses
the dc-link voltage as feedback, which will generate a reference
current value for the inner control loop.
As shown above, the only noticeable difference between
control strategies in grid-connected and islanding mode is that
the voltage reference of LES-connected inverter is changed
from the dc bus voltage to ac bus voltage. Therefore, there
is no sudden change on the local controllers, and a seamless
transition from grid-connected mode to islanding mode can be
achieved.
The whole system operation indeed requires a high level
of coordination, which is realized with the MCC. When the
maximum power from the PV and wind turbine could be
fully absorbed by the load and the battery, the PV and wind
turbine will be commended to operate in MPPT mode. If the
LES battery reaches its upper limit of SOC and the power
supply from renewable energy sources is higher than the load
demand, which usually is rare, the PV and wind turbine will
be commended to reduce power output. During the transient of
dynamic islanding, load shedding and reconnection will also be
performed if the MCC predicts that the power generation is less
than the demand.
V. S IMULATION R ESULTS
A. Results Without Considering the Latency of the
Communication Network
Real-time simulation was performed for the case where the
system transfers from grid-connected mode to islanding mode.
First, only the power network and associated control strategies
were simulated and evaluated. In the simulation, a three-phase

905

short circuit occurs on the grid side. A protection scheme to

detect loss of main, which is developed based on the control
discussed in [26], detects the failure and opens the main circuit
breaker at the PCC. It is assumed that during islanding mode,
the LES, PV, and wind turbine can provide enough power for
both critical and noncritical loads.
Since the simulation is running at real time, the simulation
results could be observed via oscilloscope. Oscillograms of
simulation results of the dc bus voltage within LES inverter
system, ac bus voltage, and LES output ac current are shown
in Fig. 9(a), and ac currents from grid side, PV, wind turbine,
critical load, and noncritical load are shown in Fig. 9(b).
The results illustrate that when the system changes from
grid-connected mode to islanding mode, the dc bus voltage,
ac bus voltage, and load currents remain constant. Moreover,
the voltage and current spike during this transition are within
two times of their normal values, which is not significant.
Therefore, the proposed control strategy is effective, and the
real-time simulation platform successfully handles the complex
smart grid model with a large number of switching devices that
are operating at high frequency.
B. Comprehensive Results With Communication Latency
Real-time simulation was then performed to test the influence
of the communication on the power network. In this case, both
power network and communication network were simulated
and evaluated. It is assumed that during islanding mode, the
LES, PV, and wind turbine cannot provide enough power for
all loads; consequently, load shedding is necessary to maintain
the system stable. In islanding mode, the LES can provide a
maximum power of 8.3 kW, the PV 5.3 kW, and the wind
turbine 10 kW. On the load side, the power consumption of
each noncritical load and critical load is 5 kW and 10 kW,
respectively. The dc bus threshold voltage is 730 V.
Considering the communication effect, during a fault on the
grid side, the main circuit breaker at the PCC will open and send
a signal to the MCC which initiates islanding mode and disconnects the three noncritical loads simultaneously. Furthermore,
the LES dc bus voltage is monitored by the MCC and gives
a status of the power capability of the smart grid. The MCC
will start reconnecting noncritical loads one at a time every
0.1 s. If the monitored LES dc bus voltage is constant, the smart
grid is able to sustain itself. However, if the voltage begins to
decrease, the MCC will disconnect the last load added to the
system, and the load shedding process stops. The effects of
latency, bandwidth, and packet losses will introduce behaviors
in the overall network which can be studied and compared.
Fig. 10 shows the real-time simulation results without the
communication effect, Fig. 10(a), and with the communication
effect, Fig. 10(b). The local dc bus voltage value and the
dc bus value detected by the MCC were monitored with the
same digital oscilloscope; they are both shown in the figures
to identify the total communication delay. In the first case,
the communication between nodes is ideal and instantaneous
data transfer. All the noncritical loads were cut off at the
beginning of the dynamic islanding at t1. Then, one by one,
each noncritical load is reconnected after every 0.1 s. At the

906

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 2, MARCH/APRIL 2013

Fig. 9. Simulation results for intentional islanding. (a) dc bus voltage within LES subsystem, ac bus voltage, and LES output ac current; (b) ac currents from
grid side, PV, wind turbine, critical load, and noncritical load.

Fig. 10. Simulation results for load shedding. (a) Ideal communication (no delay); (b) communication with delay.

same time, a sensing unit measures the LES dc bus voltage

and sends it to the MCC. When the last noncritical load was
reconnected at t4, the dc bus voltage of LES starts to drop
due to power imbalance. When the dc bus voltage drops to the
threshold voltage at t5, CB3 is reopened by the MCC, and the
dc bus voltage of LES returns to its normal value.
In the second case, the communication network is not ideal.
It is assumed the MCC sends out data every 10 ms. Moreover, because of the distances between different nodes and the
MCC, different latencies during the data transmission paths
are produced. Compared to the first case, the influence of the
communication network can be found as follows: 1) Since the
MCC sends out data every 10 ms, the noncritical loads cannot
be disconnected from the grid at the same time the main circuit
breaker opens, which results in a larger disturbance during the
transition; 2) The distance of each circuit breaker to the MCC
is different, thus the latency in different routes is also different.

Though the MCC sends out the reclose command every 0.1 s,
the noncritical loads are not reconnected to the smart grid after
each 0.1 s; 3) There is delay between the real dc bus voltage
value and the dc bus voltage detected by the MCC, which
causes the reopen time of the CB3 to be delayed. As a result,
the dc bus voltage drops below the threshold voltage and causes
a bigger ac bus voltage drop.
In conclusion, the communication network brings great benefits to the operation of the smart grid in allowing for implementation of new and different types of protection schemes
and control. However, it is seen from Fig. 10(a) and (b) that
without simulating the communication network, the actual
system response cannot be accurately predicted. The real-time
simulation platform presented in this paper integrates both the
power system and the communication system, thus provides an
effective method to study the communication system in smart
grids.

GUO et al.: COMPREHENSIVE REAL-TIME SIMULATION OF THE SMART GRID

VI. C ONCLUSION
This paper presents the latest state-of-the-art technology
on real-time simulation of smart grids. On one hand, compared to the real-time simulation of traditional power grid and
simple power electronics circuit, the large numbers of highspeed switching devices in smart grids introduce significant
challenges to the simulation. The real-time simulation platform presented in this paper utilizes switch event interpolation,
SSN solver, and parallel computation technologies; combined
with the advanced hardware platform, this system can achieve
real-time simulation with hundreds of switches at switching
frequencies up to 10 kHz. On the other hand, the characteristics
of smart grids are greatly depending on the communication
network. When combining the advanced power systems simulator with a state-of-the-art network simulator, this system can
simulate the communication network in smart grids and provide
great convenience for related study.
A microgrid model of a small community is built with the
aforementioned technologies. The real-time model contains
most of critical units in smart grids, including LES, PV, wind
turbine, power conditioning circuits, solid state switches, and
a comprehensive communication network. First, simulations
were carried out without communication latency. It is verified
that the system is capable of simulating detailed simulation
models with a large group of switches at 10 kHz. Also, the
proposed control strategies for seamless transition from gridconnected mode to islanding mode were verified. The simulation only utilizes no more than 10% of the calculation capability
of the proposed platform.
Then, with the SITL technology provided by OPNET, a
communication network is simulated together with the power
network at real time. An external MCC is utilized as the
system level controller with the help of the communication
network, which can be easily adjusted based on different smart
grid communication network structures. The distributed local
control of inverters and converters is embedded in the model
of each energy source and load. The simulation results show
the effectiveness of the presented comprehensive simulation
platform as well as the importance of an accurate and detailed
simulation approach to evaluate the interactions between the
communication and power networks in smart grids. As explained and demonstrated in the paper, the platform can be
utilized to study difference scales of smart grid applications.
New control strategies, such as dynamic reconfiguration of both
communication and power network within a smart grid, could
also be verified with this platform.
R EFERENCES
[1] R. Brown, Impact of smart grid on distribution system design, in Proc.
IEEE Power Energy Soc. Gen. Meeting, 2008, pp. 14.
[2] P. Parikh, M. Kanabar, and T. Sidhu, Opportunities and challenges of
wireless communication technologies for smart grid applications, in
Proc. CCECS Power Energy Soc. Gen. Meeting, 2010, pp. 17.
[3] R. H. Lasseter, J. H. Eto, B. Schenkman, J. Stevens, H. Vollkommer,
D. Klapp, E. Linton, H. Hurtado, and J. Roy, CERTS Microgrid laboratory test bed, IEEE Trans. Power Del., vol. 26, no. 1, pp. 325332,
Jan. 2011.
[4] M. Mao, M. Ding, J. Su, L. Chang, M. Sun, and G. Zhang, Testbed for
microgrid with multi-energy generators, in Proc. IEEE Can. Conf. Elect.
Comput. Eng., 2008, pp. 637640.

907

[5] W. Li, G. Joos, and J. Belanger, Real-time simulation of a wind turbine

generator coupled with a battery supercapacitor energy storage system,
IEEE Trans. Ind. Electron., vol. 57, no. 4, pp. 11371145, Apr. 2010.
[6] D. Brandt, R. Wachal, R. Valiquette, and R. Wierckx, Closed loop testing
of a joint Var controller using a digital real-time simulator, IEEE Trans.
Power Syst., vol. 6, no. 3, pp. 11401146, Aug. 1991.
[7] S. Abourda, J. Belanger, and C. Dufour, Real-time HIL simulation of
a complete PMSM drive at 10 s time step, in Proc. IEEE Eur. Conf.
Power Electron. Appl., 2005, pp. P-3P-9.
[8] M. Matar and R. Iravani, FPGA implementation of the power electronic
converter model for real-time simulation of electromagnetic transients,
IEEE Trans. Power Del., vol. 25, no. 2, pp. 852860, Apr. 2010.
[9] Y. Li, D. Vilathgamuwa, and P. Loh, Design, analysis, and real-time
testing of a controller for multibus microgrid system, IEEE Trans. Power
Electron., vol. 19, no. 5, pp. 11951204, Sep. 2004.
[10] J. Nutaro, P. Kuruganti, M. Shankar, L. Miller, and S. Mullen, Integrated
modeling of the electric grid, communications, and control, Int. J. Energy
Sector Manag., vol. 2, no. 3, pp. 420438, 2008.
[11] M. Kezunovic, Teaching the smart grid fundamentals using modeling,
simulation, and hands-on laboratory experiments, in Proc. IEEE Power
Energy Soc. Gen. Meeting, 2010, pp. 16.
[12] M. Brenna, E. De Berardinis, L. D. Carpini, P. Paulon, P. Petroni,
G. Sapienza, G. Scrosati, and D. Zaninelli, Real time simulation of
smart grids for interface protection test and analysis, in Proc. Int. Conf.
Harmon. Quality Power, 2010, pp. 16.
[13] R. Reddi and A. Srivastava, Real time test bed development for power
system operation, control and cyber security, in Proc. North Amer. Power
Symp., 2010, pp. 16.
[14] E. Felemban and E. Ekici, Single Hop IEEE 802.11 DCF analysis revisited: Accurate modeling of channel access delay and throughput for
saturated and unsaturated traffic cases, IEEE Trans. Wireless Commun.,
vol. 10, no. 10, pp. 32563266, Oct. 2011.
[15] Opal-RT Technologies. [Online]. Available: http://www.opalrt.com
[16] Dolphin. [Online]. Available: http://www.dolphinics.com
[17] C. Dufour, J. Mahseredjian, and J. Belanger, A combined state-space
nodal method for the simulation of power system transients, IEEE Trans.
Power Del., vol. 26, no. 2, pp. 928935, Apr. 2011.
[18] OPNET. [Online]. Available: http://www.opnet.com
[19] Y. Hu, B. J. Yurkovich, S. Yurkovich, and Y. Guezennec, A technique
for dynamic battery model identification in automotive applications using
linear parameter varying structures, IFAC Control Eng. Pract., vol. 17,
no. 10, pp. 11901201, Oct. 2009.
[20] Y. Hu, B. J. Yurkovich, S. Yurkovich, and Y. Guezennec, Electrothermal
battery modeling and identification for automotive applications, J. Power
Sources, vol. 196, no. 1, pp. 449457, Jan. 2011.
[21] Y. Hu and S. Yurkovich, Battery state of charge estimation in automotive
applications using LPV techniques, in Proc. ACC, Baltimore, MD, USA,
Jun. 2010, pp. 50435049.
[22] Y. Hu, B. J. Yurkovich, S. Yurkovich, Y. Guezennec, and
R. Bornatico, Model-based calibration for battery characterization
in HEV applications, in Proc. ACC, Seattle, WA, USA, Jun. 2008,
pp. 318325.
[23] M. G. Villalva, J. R. Gazoli, and E. R. Filho, Comprehensive approach
to modeling and simulation of photovoltaic arrays, IEEE Trans. Power
Electron., vol. 24, no. 5, pp. 11981208, May 2009.
[24] H. J. Laaksonen, Protection principles for future Microgrids, IEEE
Trans. Power Electron., vol. 25, no. 12, pp. 29102918, Dec. 2010.
[25] R. Pena, J. C. Clare, and G. M. Asher, Doubly fed induction generator
using back-to-back PWM converters and its application to variable-speed
wind-energy generation, in Proc. Inst. Elect. Eng.Elect. Power Appl.,
1996, pp. 231241.
[26] S. Conti and S. Raiti, Integrated protection scheme to coordinate MV
distribution network DG interface protections and Micro-grid operation,
in Proc. IEEE Int. Conf. Clean Elect. Power, 2009, pp. 640646.

Feng Guo (S09) was born in Henan, China, in 1986.

He received the B.S. degree in electrical engineering
from Wuhan University, Wuhan, China, in 2009.
Currently, he is working toward the Ph.D. degree at
The Ohio State University, Columbus, OH, USA.
His current research interests include large-scale
photovoltaic power plants, hardware-in-the-loop,
and real-time simulation of smart grids, power electronics circuits in hybrid electric vehicles, and energy
harvesting around high-voltage transmission lines.

908

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 2, MARCH/APRIL 2013

Luis Herrera (S10) received the B.S. degree

in electrical engineering from the University of
Tennessee at Martin, TN, USA, in 2010. Currently,
he is working toward the Ph.D. degree at The Ohio
State University, Columbus, OH, USA.
His current research interests include modeling
and networked control of the smart grid, hardwarein-the-loop, and field-programmable gate array
modeling of fast switching power electronics, and
high-frequency battery impedance testing.

Robert Murawski (S10M12) received the B.S.

and M.S. degrees in electrical engineering from The
Cleveland State University, Cleveland, OH, USA, in
2003 and 2004, respectively. He received the Ph.D.
degree in electrical engineering from The Ohio State
University, Columbus, OH, USA, in 2011.
Currently, he is working for QinetiQ North
America at the NASA Glenn Research Center under the Space, Communication, and Navigation
Program. His research interests include wireless networking, cognitive radio networks, and space-link
communication emulation.

Ernesto Inoa (S10) was born in the Dominican

Republic. He received the B.S. degree in electronics
engineering (magna cum laude) from the Pontificia
Universidad Catlica Madre y Maestra, Dominican
Republic, in 2000, and the Masters degree in electrical and computer engineering from the University
of Central Florida, Orlando, FL, USA, in 2005. He
is currently working toward the Ph.D. degree at The
Ohio State University, Columbus, OH, USA.
He worked as an Automation Engineer for
Cerveceria Nacional Dominicana, a Dominican
brewing company, where he helped automate several bottling lines with the use
of programmable logic controllers. His major research interests are related to
the application of novel control techniques to power electronics, motor drives,
and energy conversion systems.

Chih-Lun Wang received the B.S. degree from

National Chiao-Tung University, Hsinchu, Taiwan,
in 2007 and the M.S. degree from The Ohio State
University, Columbus, OH, USA, in 2010, both in
electrical engineering.
His research interests include power system simulation and power electronic circuit design. In
2011, he joined Ford Motor Company, Dearborn,
MI, USA, where he is currently a Power Electronic Engineer, working on power components
characterization.

Philippe Beauchamp received the B.S. degree in

software engineering with a specialization in control
from the cole Polytechnique de Montral, Montral, QC, Canada in 2006. He is currently working
toward the M.B.A. degree at HEC Montreal.
He joined Opal-RT Technologies, Montreal, in
2006 as a Simulation Specialist in the integration
team and held positions in R&D and technical sales.
Currently, he is managing business development for
the academic market in North America. His current
fields of interest include real-time simulation and distributed computing applied to renewable energy, hardware-in-the-loop testing,
and field-programmable gate array co-simulation.

Eylem Ekici (S99M02SM11) received the B.S.

and M.S. degrees in computer engineering from
Bogazici University, Istanbul, Turkey, in 1997 and
1998, respectively. He received the Ph.D. degree in
electrical and computer engineering from Georgia
Institute of Technology, Atlanta, GA, USA, in 2002.
Currently, he is an Associate Professor in the
Department of Electrical and Computer Engineering of The Ohio State University, Columbus, OH,
USA. His current research interests include cognitive
radio networks, vehicular communication systems,
nanoscale networks, and wireless sensor networks, with a focus on routing
and medium access control protocols, resource management, and analysis
of network architectures and protocols. He is an Associate Editor of the
IEEE/ACM T RANSACTIONS ON N ETWORKING, Computer Networks Journal
(Elsevier), and ACM Mobile Computing and Communications Review.

Jin Wang (S02M05) received the B.S. degree

from Xian Jiaotong University, Xian, China, in
1998, the M.S. degree from Wuhan University,
Wuhan, China, in 2001, and the Ph.D. degree from
Michigan State University, East Lansing, MI, USA,
in 2005, all in electrical engineering.
From September 2005 to August 2007, he worked
at the Ford Motor Company as a Core Power Electronics Engineer and contributed to the traction drive
design of the Ford Fusion Hybrid. Since September
2007, he has been an Assistant Professor in the Department of Electrical and Computer Engineering, The Ohio State University,
Columbus, OH, USA. His teaching position is co-sponsored by American
Electric Power, Duke/Synergy, and FirstEnergy. His research interests include
high-voltage and high-power converters/inverters, integration of renewable
energy sources, and electrification of transportation.
Dr. Wang received the IEEE Power Electronics Society Richard M. Bass
Young Engineer Award and the National Science Foundations CAREER
Award, both in 2011. He has over 50 peer-reviewed journal and conference
publications. Dr. Wang has been an Associate Editor for the IEEE T RANSAC TIONS ON I NDUSTRY A PPLICATIONS since March 2008.