You are on page 1of 8

This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON SMART GRID 1

Stochastic Energy Scheduling in Microgrids With


Intermittent Renewable Energy Resources
Wencong Su, Member, IEEE, Jianhui Wang, Senior Member, IEEE, and Jaehyung Roh, Member, IEEE

Abstract—Renewable energy resources such as wind and solar Cost function ($/h) of buying/selling
are an important component of a microgrid. However, the inherent kWh from/to the utility grid.
intermittency and variability of such resources complicates micro- Power output (kW) of the th unit under
grid operations. Meanwhile, more controllable loads (e.g., plug-in scenario at the th hour.
electric vehicles), distributed generators (e.g., micro gas turbines
Power from/to the utility grid (kW) at the
and diesel generators), and distributed energy storage devices (e.g.,
battery banks) are being integrated into the microgrid operation. th hour.
To address the operational challenges associated with these tech- Start-up/shut-down cost ($) of the th
nologies and energy resources, this paper formulates a stochastic unit at the th hour.
problem for microgrid energy scheduling. The proposed problem Overall electrical conversion efficiency
formulation minimizes the expected operational cost of the micro- (%) of battery charger.
grid and power losses while accommodating the intermittent na- Rated energy capacity of the th battery
ture of renewable energy resources. Case studies are performed on
bank (kWh).
a modified IEEE 37-bus test feeder. The simulation results demon-
strate the effectiveness and accuracy of the proposed stochastic mi- Total energy delivered or absorbed by the
crogrid energy scheduling model. th battery bank (kWh) under scenario .
Actual battery life cycle of the th battery
Index Terms—Microgrid, plug-in electric vehicle (PEV), renew-
able energy, smart grid, stochastic programming. bank under scenario .
Actual battery life (kWh) of the th
battery bank under scenario .
NOMENCLATURE Capital cost ($) of the th battery bank.
Battery degradation cost ($/kWh) of the
th battery bank under scenario .
Indices: Production cost function coefficients of
Distributed generator. the th unit.
Degradation cost function coefficients of
Battery bank.
the th battery bank.
Wind scenarios. Grid electricity price ($/kWh) at the th
Solar scenarios. hour.
Power flow equations under scenario at
Hour.
the th hour.
Wind generator. Power losses (kW) under scenario at the
Solar generator. th hour.
Base load (kW) at the th hour.
PEV.
Aggregate PEV charging load (kW) at the
Bus.
th hour.
Variables and Charging load (kW) of the th PEV at the
Functions: th hour.
Probability of scenario . State-of-Charge (%) of the th battery
Cost function ($/h) of generating kWh bank under scenario at the th hour.
from the th unit. Voltage (per unit) of the th bus under
scenario at the th hour.
Constants:
Manuscript received June 08, 2013; revised nulldate; accepted August 18, Minimum power generation (kW)
2013. Paper no. TSG-00438-2013. requirement if the th unit is on.
W. Su is with the Department of Electrical and Computer Engineering, Uni- Maximum power generation (kW)
versity of Michigan-Dearborn, MI 48128 USA (e-mail: wencong@umich.edu).
J. Wang is with Argonne National Laboratory, Argonne, IL 60439 USA
requirement if the th unit is on.
(e-mail: jianhui.wang@anl.gov). Minimum State-of-Charge requirement
J. Roh is with Department of Electrical Engineering, Konkuk University, (%) for the th battery bank.
Seoul, Korea (e-mail: jhroh@konkuk.ac.kr). Maximum State-of-Charge requirement
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. (%) the th battery bank.
Digital Object Identifier 10.1109/TSG.2013.2280645

1949-3053 © 2013 IEEE


This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON SMART GRID

Final State-of-Charge requirement (%). linear programming problem, which is difficult to solve,
Minimum voltage (per unit) requirement the problem is decomposed into one master problem for en-
at the th bus. ergy scheduling and one subproblem for power flow com-
Maximum voltage (per unit) requirement putation. The two problems are solved iteratively as dis-
at the th bus. cussed in detail in Section III.
3) Determine optimal energy scheduling for distributed
generators (DGs) and distributed energy storage devices
I. INTRODUCTION (DESDs) considering battery degradation costs.

T HE increasing penetration of intermittent and variable re-


newable energy resources (e.g., wind and solar) has sig-
nificantly complicated energy system management for micro-
4) Investigate the impact of PEVs on microgrid energy sched-
uling under various charging schemes.
Section II proposes the formulation of the two-stage sto-
grids [1]. Unlike conventional generation sources, wind and chastic microgrid energy scheduling model. Section III intro-
solar power output is highly uncertain and unpredictable. Even duces the test platform and presents the details of the proposed
a small error in wind and solar power forecasting may result decomposition scheme to minimize the expected operational
in great uncertainties for real-time operations of a microgrid cost while reducing power losses in the interconnected mi-
given its limited scale and size. Moreover, in microgrids, the crogrid mode. Section IV provides a detailed case study and
customers play a more important role by managing the control- compares the simulation results with those obtained using a
lable loads compared with conventional power distribution sys- traditional deterministic approach under various operational
tems, in which consumers have little flexibility to fully partici- conditions. Section V concludes the paper.
pate in electricity markets. The term “controllable load” refers to
a type of non-critical load that can adjust its own electric energy II. PROBLEM FORMULATION
usage on the basis of a real-time set point. One special class of A two-stage stochastic microgrid energy scheduling model
controllable loads is a Plug-in Electric Vehicle (PEV) fleet. Un- is proposed in this paper to make an optimal decision on the
like other controllable loads, these vehicles can be connected to day-ahead energy transactions in the first stage while mim-
power grids any place and any time, allowing more spatial and icking the real-time operations with the wind/solar power
temporal diversity but also more uncertainty. A sophisticated variability in the second stage. The idea of establishing such a
microgrid energy scheduling model that takes these unique fac- modeling framework is motivated by the increasing challenge
tors into consideration is urgently needed. of addressing the variability and uncertainty from renewable
Most published works focus on deterministic microgrid op- energy resources on a microgrid. Since the generation resources
erations [2]–[6]. However, the stochastic energy management are usually limited on a microgrid, renewable generation units
methods that have been widely researched in transmission-level such as wind and solar may account for a large portion of the
energy management [7]–[11] demonstrate promising results in total generation portfolio. In the meantime, as a microgrid is
capturing the uncertainty associated with renewable energy re- self-confined and self-balanced, one of the primary goals of
sources and considering worst-case scenarios. The focus of this operating a microgrid is to minimize the operating costs with
paper is to apply stochastic methods to microgrid energy sched- high reliability, by reducing the uncertainty and variability of
uling problems. To the best of the authors’ knowledge, micro- on-site renewable generation and the risk of energy transactions
grid energy scheduling using stochastic methods has not been with the utility grid. By the proposed two-stage framework,
well documented in the literature. Details about the classical we can determine the day-ahead energy scheduling in the first
two-stage or multi-stage stochastic programming can be found stage with the objective to identify the optimal amount of
in [12]. In [13], the authors applied online stochastic optimiza- electricity to be purchased from/sold to the utility grid and
tion to residential appliance energy management with privacy the commitment of distributed generation units over the next
protection. In [14], a stochastic method for the hourly sched- 24 hours. These first-stage decisions do not vary across the
uling of optimal reserves is presented to consider the hourly scenarios in the second stage.
forecast errors of wind energy and load. In [15], a stochastic en-
ergy scheduling model is developed for a local-area smart-grid A. Objective Function
system with a single energy source and multiple energy con- The objective function minimizes the expected operational
sumers. In [16], a stochastic model predictive control method cost in the interconnected microgrid mode over the next 24
is proposed for microgrid operations. However, the paper does hours. In this study, the authors use scenarios to model the uncer-
not consider the network constraints, which may not be real- tain power output of wind and solar resources. In other words,
istic. In [17], a chance-constrained programming framework is the power output of wind and solar generating units in one sce-
proposed to model stochasticity rather than the scenario-based nario is different from another. The production cost of the re-
approach proposed in this paper. However, the power flow con- newable energy resource (i.e., wind and solar energy) is as-
straints are not considered in [17]. sumed to be negligible
The major contributions of this paper include the following:
1) Propose a two-stage stochastic microgrid operation model
for the optimal decisions on the day-ahead energy transac-
tions in the first stage while mimicking the real-time wind
and solar power variability in the second stage.
2) Develop a decomposition scheme to solve the proposed
model. As the proposed model is a mixed integer non-
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

SU et al.: STOCHASTIC ENERGY SCHEDULING IN MICROGRIDS 3

1) Power balance
(1)

Frequent charging and discharging may significantly affect


the operational life of DESDs (e.g., battery banks). Therefore,
it is important to consider and formulate the additional battery
degradation cost for microgrid energy scheduling. The battery (10)
degradation cost can be expressed as a function of the actual
battery cycle life [18]. In this paper, all the DESDs are assumed For any scenario at any time step, the left-hand term is the
to be identical with the same charging/discharging efficiency power output from wind, solar, DG, utility grid, and DESD,
. The depth of discharging of the th battery bank respectively. The right-hand term includes the power loss,
is defined as base load, and PEV charging load, respectively.
2) Technical limits of DG

(2) (11)

The battery cycle life can be formulated as a function of For any scenario at any time step, the power output of DG
depending on the type of battery. lies in a certain operating range.
3) Technical limits of DESD
(3)
(12)
For example, the relationship between Lead-Acid battery (13)
cycle life and can be expressed as a linear function [19] (14)

(4) For any scenario at any time step, the power output of
DESD lies in a certain operating range. To avoid any over-
Then the actual battery life (kWh) is expressed as charging or overdischarging, once reaches the upper/
lower bounds, the DESD (e.g., battery bank) switches to a
(5) stand-by mode.
4) Technical limits of the PEV charging load
The th battery degradation cost ($/kWh) under each indi-
vidual scenario is determined by the battery capital cost and the (15)
actual battery life.
(16)
(6)
Similarly, the PEV battery chargers have limitation on the
amount of charging power at any given time step.
Therefore, the battery degradation cost ($) for the th battery Power flow constraints are taken into consideration as well,
bank under each individual scenario is formulated as as shown in (17), (18). In this paper, AC power flow is
calculated by the embedded solver in EPRI’s Open-source
(7) Distribution System Simulator (OpenDSS) V7.6.1 [20],
[21]. Therefore, (17) is a general mathematical formulation
to implicitly represent a variety of power flow constraints.
For distributed generators, the cost function can be formu- 5) Power flow constraints
lated as
(17)
(8)
6) Limits of bus voltage
Because small DG units have negligible start-up/down times,
the start-up cost can be simplified as a constant for each unit. (18)
For the main utility grid, the energy transaction payment can
be formulated as
III. SOLUTION ALGORITHM
(9)
Because the formulated problem is a mixed-integer nonlinear
programming problem and AC power flow is performed in each
scenario, we decompose the original problem into a master
B. Constraints
energy scheduling problem and a power flow subproblem.
The system constraints considered in this paper include the The master problem is to solve the energy scheduling problem
following. (1)–(16), without considering the power flow constraints (17),
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON SMART GRID

Fig. 3. Hourly load profile and grid electricity prices.

2) Matlab passes the optimal control variables to


OpenDSS which is dedicated to the advanced analysis of
Fig. 1. Flowchart of the proposed decomposition scheme.
distribution systems. OpenDSS runs the sequential power
flow of the proposed distribution system over successive
time intervals and check the power flow constraints shown
in (17), (18).
3) We calculate the power losses of the feasible solu-
tions obtained from the Step (2). The most updated power
loss is fed back to the master problem and used to
update the total power loss in (10) as long as the change of
power loss is noticeable (i.e., ). Here,
is a small pre-defined value. If the results from
Step (1) do not lead to the convergence of power flow, we
add a small random vector to the last feasible solution
to recalculate the loss and rerun the master
problem with updated power balance constraints (10) in
the master problem. This kind of local search method can
help the algorithm converge to at least a local optimal so-
lution. The master problem and subproblem are solved it-
eratively. The power flow solutions vary in each scenario.
Each solution corresponds to a probability that quantifies
the likelihood of the scenario. Accordingly, is the
expectation of power losses over all the scenarios.
Fig. 2. Framework of co-simulation platform.
4) After a few iterations, the system converges to a new
steady-state condition in which the best available optimal
(18). Then the dispatch solutions from the master problem energy scheduling solutions are found. The simulation
are tested in the power flow subproblems to see whether the results in Section IV will also show the convergence of the
power flow constraints (17), (18) are satisfied in each scenario. proposed decomposition scheme, which has been applied
The two problems are solved iteratively until the difference of to many other power system applications [23].
the overall expected loss calculated across all the scenarios is
small enough between two adjacent iterations. The output of IV. CASE STUDIES
the model is to decide how much energy can be purchased or In this section, we examine three cases to test the proposed al-
sold in the day-ahead market and the commitment of on-site gorithm and discuss the simulation results. All simulations were
distributed generators at the day-ahead stage. Fig. 1 shows a run on an Intel(R) Core i5 CPU M450@2.40 GHz computer
flowchart of the proposed decomposition solution algorithm. with a 6.00 GB memory.
Fig. 2 illustrates the proposed software-based simulation
testbed architecture. A. Case Configuration
After initialization ( , are the iteration numbers for 1) Base Load and Electricity Price Data: For the sake of
the two iterative loops, respectively), the principal procedure is simplicity, the base load profiles of all the uncontrollable micro-
described as follows. grid loads are assumed to be the same, as shown in Fig. 3. Fig. 3
1) The formulation of the master problem including the ob- also shows the hourly electricity prices of the utility power grid.
jective function and the system constraints (10)–(16) The data is derived from [24].
is implemented in Matlab and solved by the IBM ILOG 2) PEV Charging Load: This case study mainly focuses on
CPLEX Optimizer 12.2 [22]. The optimal control variables level 2 charging which is typically described as the “primary”
found by CPLEX are expressed as a vector of and “standard” method [25] for both private and public charging
, , and . facilities during early adoption of PEVs. The PEV charging load
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

SU et al.: STOCHASTIC ENERGY SCHEDULING IN MICROGRIDS 5

Fig. 4. Scenarios of wind power output over 24 hours. Fig. 5. Scenarios of solar power output over 24 hours.

can be derived from the PEV data estimation. In order to sim-


ulate PEV data that incorporate some uncertainty, a number of
factors including plug-in time, expected plug-out time, initial
battery state-of-charge , and battery capacity need to be
considered. Ideally, the proposed energy scheduling model re-
quires the real-time estimation and prediction of PEV charging
load. The prediction problem is to forecast the PEV charging
load over the whole prediction horizon utilizing the historical
dynamic pattern. A detailed description of the PEV load data es-
timation can be found in [26], [27]. One of our future research
topics [28] is to investigate a robust real-time estimation and
prediction method of PEV load under real world traffic condi-
tions and traveler choice behaviors.
3) Wind and Solar Scenarios: In this paper, we use the time-
stamped wind power generation output data from several hypo-
thetical sites in Illinois based on the National Renewable Energy
Laboratory’s (NREL’s) Eastern Wind Integration and Transmis-
sion Study [29]. Those wind data were generated by a weather
simulation model and a composite power curve of wind farms.
Fig. 6. Microgrid test system based on a modified IEEE 37-node test feeder.
Multiple day-ahead wind power forecasts were aggregated into
one time series. The forecasting error varies with the geograph-
ical location, weather conditions, and time horizon. In order to uling model can be easily extended using real-world utility’s
capture the uncertainty of the wind power predictions, Monte distribution system data.
Carlo simulations [30] were performed to generate a pool of
wind power scenarios each day. Some scenario reduction tech- B. Simulation and Analysis
niques can be used to further reduce the computational cost [31]. In the case studies, the point forecasts of wind and solar power
Fig. 4 shows a set of 10 probabilistic wind power scenarios over output are also adopted as a deterministic alternative for com-
a period of 24 hours. parison purposes. Actually the deterministic approach can be
The solar data were extracted from the System Advisor considered as a special case of the stochastic one with only one
Model (SAM) developed by NREL [32]. Due to the lack of the scenario which is corresponding to the point forecast. The total
existing scenario-based solar forecasting model, we generate a power loss of the microgrid system over the 24-hour period is
set of solar scenarios by assuming the forecasting errors follow shown in Fig. 7, which illustrates how the entire problem is it-
a normal distribution. Fig. 5 shows a set of 10 probabilistic eratively solved to converge.
solar power scenarios over a period of 24 hours. Initially, the total power losses over the 24-hour period are
Each wind and solar power scenario corresponds to a proba- assumed to be zero for both deterministic and stochastic en-
bility that quantifies the likelihood of this scenario. In this case ergy scheduling models. Then the total power losses change to
study, the wind scenarios are assumed to be independent with 2,374.2 kW and 2,417.3 kW, respectively. The most updated
solar scenarios. Therefore, there are 100 independent scenarios , is fed back to the power balance constraint (10) as long as
in total. The probability of each independent scenario is 0.01. the change of losses between two iterations, , is notice-
A modified IEEE 37-bus distribution test feeder [33] shown able. The master problem is performed again to obtain the new
in Fig. 6 was used to verify the proposed stochastic microgrid energy scheduling solution. Gradually, the total power losses
energy management model. The microgrid DGs, wind gener- converge (i.e., 2,372.3 kW and 2,278 kW respectively for deter-
ators, solar generators, DESDs, and PEVs are connected with ministic and stochastic approaches) as becomes smaller
nodes 701, 722, 730, 720, and 737, respectively. It is worth men- and smaller. After four iterations, the system converges to a
tioning that the proposed stochastic microgrid energy sched- new steady-state condition in which the optimal solutions are
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON SMART GRID

Fig. 7. Total power losses at each iteration over 24 hours.


Fig. 9. Trends of SOC under deterministic and stochastic microgrid energy
scheduling.

a relatively large non-zero value of 15.08%. That’s because the


energy scheduling solution from deterministic microgrid energy
scheduling is very sensitive to the point forecasts of wind/solar
power output over the look-ahead horizon. Since electrical gen-
eration and load must remain in balance in real time, the onsite
DESDs have to operate properly in order to compensate for the
wind/solar point forecasting errors. The large non-zero
at the end of a day indicates that the deterministic microgrid en-
ergy scheduling may have large extra operational cost caused by
Fig. 8. Hourly battery power output under deterministic microgrid energy the intermittency and the non-dispatchable nature of wind/solar
scheduling. energy production. In comparison, the end-of-the-day
using stocashtic scheduling is approximately 6.14%, which is
found. The optimal microgrid operational costs are $23,500 and much lower than 15.08% obtained from the deterministic one
$22,276, using deterministic and stochastic approaches, respec- and closer to the ideal final of zero. Therefore, there is
tively. strong evidence that the stochastic microgrid model can achieve
To compare the energy scheduling results from the determin- better energy scheduling for DESDs.
istic (point forecasts) and stochastic (scenarios) approaches, we Also note that is mainly determined by wind power
use the same set of actual wind and solar power output profiles as the battery bank is closer to the wind units than the solar
to simulate the real-time operation of the test system. In other units as shown in Fig. 6. The size of is also affected
words, we first run the deterministic and stochastic methods to by the renewable energy forecasting errors. In this paper, the
obtain the day-ahead energy scheduling solutions, respectively. wind generation capacity is larger than the solar generation ca-
We then fix the amount of energy transactions , replace the pacity. Because the wind forecasting error is typically within
point forecasts with the real wind and solar power output and run 10%–20% while the solar forecasting error is assumed to follow
the deterministic approach again. Fig. 8 shows the hourly bat- a normal distribution with a 5% standard deviation, the wind
tery power output under deterministic energy scheduling. These power has a larger impact on than solar. However, this ob-
values are subject to the battery charger limit (i.e., [ 250 kw, servation may not be generalized to other cases with a different
250 kw]) and the battery requirement (i.e., [10%, 90%]) power generation portfolio and forecasting accuracies although
at any time step. The rated battery capacity is 4,000 kWh. The the simulation method still applies. We have tried to run several
positive/negative power output corresponds to the discharging/ case studies under different operating conditions. The simula-
charging processes, respectively. Because the grid electricity tion results exhibit the similar tendency of the final under
price is cheaper in the early morning, the battery bank starts to different scheduling approaches (deterministic vs. stochastic),
store as much energy as possible. A large number of PEVs start demonstrating the effectiveness of the proposed stochastic en-
to connect to the microgrid at approximately 7:00 a.m when the ergy scheduling. More specifically, under the same operating
grid electricity price increases as well. Accordingly, the battery conditions and forecast data, the stochastic approach always
bank starts to return the stored energy to the grid at peak demand achieves smaller non-zero values of at the end of a day.
time. Due to the limited space, we only presented one representative
It is worth mentioning that, this case study does not con- simulation result here.
sider any additional reward for storing extra energy in battery To further compare the performances of microgrid energy
storage at the end of a day. Ideally, the most cost-effective final scheduling, we perform both stochastic and deterministic ap-
should be close to initial . In other words, the most proaches for a week. Figs. 10 and 11 show the daily operational
cost-effective at the end of a day should be very close cost and power losses over the week under both the stochastic
to zero. Fig. 9 compares the trends of under determin- and deterministic approaches. The stochastic approach outper-
istic and stochastic microgrid energy scheduling. In the deter- forms the deterministic one in terms of the operational cost and
ministic microgrid energy scheduling, the final reaches the total power losses.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

SU et al.: STOCHASTIC ENERGY SCHEDULING IN MICROGRIDS 7

TABLE I
SIMULATION RESULTS UNDER VARIOUS PEV CHARGING SCHEMES

Table I compares the objective function values under the pro-


posed two charging schemes, as well as the deterministic/sto-
chastic microgrid energy scheduling approaches. As most ve-
Fig. 10. Daily operational costs over one week under stochastic and determin-
istic approaches. hicles are expected to park for longer than the actual charging
time, the constrained charging scheme can reduce the opera-
tional cost of the microgrid by taking full advantage of timing
flexibility as shown in Table I.

V. CONCLUSION
This paper first describes the development of a two-stage
stochastic microgrid energy management model in the inter-
connected mode. A typical microgrid runs in two operational
modes: an interconnected mode linked to the main grid through
the distribution substation transformer vs. an islanded (au-
tonomous) mode when it is isolated from the main grid during a
blackout or brownout. The system operators may have different
Fig. 11. Daily power losses over one week under stochastic and deterministic considerations in interconnected and islanded modes. For ex-
approaches.
ample, in the islanded mode, the microgrid remains operational
and functional as an autonomous entity. Therefore, the overall
system reliability and security should be the top priority. A
number of different objective functions could be formulated to
achieve that. While we focus on the interconnected mode in
this paper, stochastic energy scheduling for islanded microgrid
operations is one of our future research topics.
The proposed stochastic model accommodates the inherent
intermittency and variability of renewable energy resources
(i.e., wind and solar). In addition, the proposed problem formu-
lation minimizes the expected operational cost of the microgrid
while reducing power losses by optimally dispatching the
PEV charging load and scheduling DGs and DESDs. The
simulation results demonstrate the effectiveness and accuracy
Fig. 12. Hourly base load, PEV charging load, and power losses under deter-
ministic microgrid energy scheduling.
of the proposed stochastic microgrid energy systems under
various operating conditions and real-world scenarios. More-
over, the proposed framework can be easily extended to other
Fig. 12 shows the hourly base load, PEV charging load, and microgrid operation applications to accelerate the development
power losses under deterministic microgrid energy scheduling. of full-scale commercial microgrids in the near future. The
In the early morning, the power loss is quite small because of proposed model can also be tailored to take into consideration
the low load level. From 7:00 a.m. to 5:00 p.m., the larger power other uncertainties such as load and customer behavior.
loss occurs as the PEV charging load is heavily introduced into
the microgrid. After 5:00 p.m., the power loss is mainly caused
ACKNOWLEDGMENT
by the base load.
To investigate the impact of PEV charging, we evaluate two The submitted manuscript has been created by UChicago
types of charging schemes. In both charging schemes, all PEV Argonne, LLC, Operator of Argonne National Laboratory
batteries are guaranteed to be fully recharged at plug-out. (“Argonne”). Argonne, a U.S. Department of Energy Of-
• Uncontrolled charging: The charging process starts imme- fice of Science laboratory, is operated under Contract No.
diately when a vehicle arrives at the public charging fa- DE-AC02-06CH11357. The U.S. Government retains for it-
cility (e.g., parking deck). self, and others acting on its behalf, a paid-up nonexclusive,
• Constrained charging: Assuming the total charging time is irrevocable worldwide license to reproduce, prepare derivative
known, the required charging load is equally distributed works, distribute copies to the public, and perform publicly and
over the entire period of parking. display publicly, by or on behalf of the Government.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON SMART GRID

Jaehyung Roh’s work is sponsored by [22] IBM ILOG CPLEX Optimization Solver 12.2 [Online]. Available:
KETEP(2001T100100424). http://www-01.ibm.com/software/integration/optimization/cplex-op-
timizer/
[23] A. J. Wood and B. F. Wollenberg, Power Generation Operation and
REFERENCES Control, 2nd ed. New York, NY, USA: Wiley, 1996.
[1] W. Su and J. Wang, “Energy management systems in micrgrid opera- [24] Midwest Independent Transmission System Operator (MISO), Inc.
tions,” Electricity J., vol. 25, no. 8, pp. 45–60, Oct. 2012. [Online]. Available: https://www.misoenergy.org/Pages/Home.aspx
[2] A. Chaouachi, R. M. Kamel, R. Andoulsi, and K. Nagasaka, “Multiob- [25] W. Su, H. Rahimi-Eichi, W. Zeng, and M.-Y. Chow, “A survey on
jective intelligent energy management for a microgrid,” IEEE Trans. the electrification of transportation in a smart grid environment,” IEEE
Ind. Electron., vol. 60, no. 4, pp. 1688–1699, Apr. 2013. Trans. Ind. Inf., vol. 8, no. 1, pp. 1–10, Feb. 2012.
[3] C. Chen, S. Duan, T. Cai, B. Liu, and G. Hu, “Smart energy manage- [26] W. Su and M.-Y. Chow, “Performance evaluation of an EDA-based
ment system for optimal microgrid economic operation,” IET Renew- large-scale plug-in hybrid electric vehicle charging algorithm,” IEEE
able Power Generation, vol. 5, no. 3, pp. 258–267, May 2011. Trans. Smart Grid, vol. 3, no. 1, pp. 308–315, Mar. 2012.
[4] S. X. Chen and H. B. Gooi, “Jump and shift method for multi-ob- [27] W. Su and M.-Y. Chow, “Computational intelligence-based energy
jective optimization,” IEEE Trans. Ind. Electron., vol. 58, no. 10, pp. management for a large-scale PHEV/PEV enabled municipal parking
4538–4548, Oct. 2011. deck,” Appl. Energy, vol. 96, pp. 171–182, August 2012.
[5] W. Su, Z. Yuan, and M.-Y. Chow, “Microgrid planning and operation: [28] W. Su, J. Wang, K. Zhang, and M.-Y. Chow, “Framework for investi-
Solar energy and wind energy,” in Proc. 2010 IEEE Power Energy Soc. gating the impact of PHEV charging on power distribution and trans-
General Meeting, Minneapolis, MN, USA, Jul. 25–29, 2010. portation networks,” in Proc. 38th Ann. Conf. IEEE Ind. Electron. Soc.,
[6] J. Wang, C. Liu, D. Ton, Y. Zhou, J. Kim, and A. Vyas, “Impacts of Montreal, QC, Canada, Oct. 25–28, 2012.
plug-in hybrid electric vehicles on power systems with demand re- [29] NREL, “Eastern Wind Integration and Transmission Study,” [Online].
sponse and wind power,” Energy Policy, vol. 39, no. 7, pp. 4016–4021, Available: www.nrel.gov/wind/systemsintegration/ewits.html
Jul. 2011. [30] P. Pinson, G. Papaefthymiou, B. Klockl, H. A. Nielsen, and H. Madsen,
[7] C. Liu, J. Wang, A. Botterud, Y. Zhou, and A. Vyas, “Assessment of “From probabilistic forecasts to statistical scenarios of short-term wind
impacts of PHEV charging patterns on wind-thermal scheduling by sto- power production,” Wind Energy, vol. 12, no. 1, pp. 51–62, 2009.
chastic unit commitment,” IEEE Trans. Smart Grid, vol. 3, no. 2, pp. [31] “Use of Wind Power Forecasting in Operational Decisions,” Argonne
675–683, Jun. 2012. National Lab. Tech. Rep., Lemont, IL, USA, 2011.
[8] J. Wang, A. Botterud, R. Bessa, H. Keko, L. Carvalho, D. Issicaba, [32] NREL, System Advisor Model (SAM) 2012 [Online]. Available:
J. Sumaili, and V. Miranda, “Wind power forecasting uncertainty and https://sam.nrel.gov/
unit commitment,” Appl. Energy, vol. 88, no. 11, pp. 4014–4023, Nov. [33] “Distribution test feeders,” IEEE Power Energy Society [Online].
2011. Available: http://ewh.ieee.org/soc/pes/dsacom/testfeeders/index.html
[9] R. Jiang, J. Wang, and Y. Guan, “Robust unit commitment with wind
power and pumped storage hydro,” IEEE Trans. Power Syst., vol. 27,
no. 2, pp. 800–810, May 2012. Wencong Su (S’06–M’13) received the B.S. degree (with distinction) from
[10] Q. Wang, Y. Guan, and J. Wang, “A chance-constrained two-stage Clarkson University, Potsdam, NY, USA, in May 2008, the M.S. degree from
stochastic program for unit commitment with uncertain wind power Virginia Tech, Blacksburg, VA, USA, in December 2009, and the Ph.D. degree
output,” IEEE Trans. Power Syst., vol. 27, no. 1, pp. 206–215, Feb. from North Carolina State University, Raleigh, NC , USA, in May 2013, respec-
2012. tively.
[11] Q. Zheng, J. Wang, P. Pardalos, and Y. Guan, “A new decomposition He is currently an Assistant Professor in the Department of Electrical and
approach to stochastic security-constrained unit commitment in elec- Computer Engineering at University of Michigan-Dearborn, MI, USA. He
trical power generation,” Ann. Oper. Res., to be published. worked a Research Aide at Argonne National Laboratory from January to
[12] J. R. Birge and F. Louveaux, Introduction to Stochastic Program- August 2012. He also worked as a R&D engineer intern at ABB U.S. Corporate
ming. : Springer, 2011. Research Center in Raleigh, NC, from May to August 2009. His research
[13] Z. Chen and L. Wu, “Residential appliance DR energy management interests include Smart Grid, grid integration of plug-in electric vehicles
with electric privacy protection by online stochastic optimization,” and renewable energy, microgrids, distribution system analysis, intelligent
IEEE Trans. Smart Grid, 2013, to be published. energy management, computational intelligence, power system optimization,
[14] C. Sahin, M. Shahidehpour, and I. Erkmen, “Allocation of hourly re- modeling and simulation.
serve versus demand response for security-constrained scheduling of
stochastic wind energy,” IEEE Trans. Sustainable Energy, vol. 4, no.
1, pp. 219–228, Jan. 2013.
[15] C. Gong, X. Wang, W. Xu, and A. Tajer, “Distributed real-time en-
ergy scheduling in smart grid: Stochastic model and fast optimization,” Jianhui Wang (M’07–SM’12) received the Ph.D. degree in electrical engi-
IEEE Trans. Smart Grid, vol. 4, no. 3, pp. 1476–1489, Sep. 2013. neering from Illinois Institute of Technology, Chicago, IL, USA, in 2007.
[16] A. Hooshmand, M. H. Poursaeidi, J. Mohammadpour, H. A. Malki, Presently, he is a Computational Engineer with the Decision and Information
and K. Grigoriads, “Stochastic model predictive control method for Sciences Division at Argonne National Laboratory, Argonne, IL, USA. He is
microgrid management,” in Proc. IEEE PES Innovative Smart Grid also an affiliate professor at Auburn University.
Technol, Washington, DC, USA, Jan. 16–20, 2012, pp. 1–7. Dr. Wang is the chair of the IEEE Power & Energy Society (PES) power
[17] Z. Wu, W. Gu, R. Wang, X. Yuan, and W. Liu, “Economic optimal system operation methods subcommittee. He is an Editor of the IEEE
schedule of CHP microgrid system using chance constrained program- TRANSACTIONS ON POWER SYSTEMS, the IEEE TRANSACTIONS ON SMART
ming and particle swarm optimization,” in Proc. IEEE Power Energy GRID, an Associate Editor of Journal of Energy Engineering, an Editor of the
Society General Meeting, July 24–29, 2011, pp. 1–11. IEEE PES Letters, and an Associate Editor of Applied Energy.
[18] C. Zhou, K. Qian, M. Allan, and W. Zhou, “Modeling of the cost of EV
battery wear due to V2G application in power systems,” IEEE Trans.
Energy Conversion, vol. 26, no. 4, pp. 1041–1050, 2011.
[19] “Performance Characterization—GM EV1 Panasonic Lead Acid Bat- Jaehyung Roh (M’05) received the B.S. degree in nuclear engineering from
tery,” Electric Transportation Division, Southern California Edison, Seoul National University Seoul, Korea, in 1993 and the M.S. degree in elec-
2000. trical engineering from Hongik University, Korea, in 2002. He received Ph.D.
[20] Electric Power Research Institute (EPRI), “Open-Source Distri- degree in electrical engineering from Illinois Institute of Technology, Chicago,
bution System Simulator (OpenDSS),” 2012 [Online]. Available: IL, USA.
http://sourceforge.net/projects/electricdss/ During 1992–2001, he was with Korea Electric Power Corporation, and for
[21] J. Taylor, A. Maitra, M. Alexander, D. Brooks, and M. Duvall, “Evalu- 2001–2010, he was with Korea Power Exchange. Since 2010, he has been with
ations of plug-in electric vehicle distribution system impacts,” in Proc. Electrical Engineering Department, Konkuk University, Seoul, Korea, as an As-
2010 IEEE Power Energy Society General Meeting, Minneapolis, , sistant Professor. His research interests include power systems restructuring,
USA, Jul. 25–29, 2010. smart grid and resource planning.

You might also like