Systems and Control Opportunities in the

Integration of Renewable Energy into the

Smart Grid ?
E. Bitar ∗ P. P. Khargonekar ∗∗
K. Poolla ∗∗∗
∗
E. Bitar is with the Department of Mechanical Engineering, U.C.
Berkeley ebitar@berkeley.edu (corresponding author)
∗∗
P. P. Khargonekar is with the Department of Electrical and
Computer Engineering, University of Florida ppk@ece.ufl.edu
∗∗∗
K. Poolla is with the Department of Electrical Engineering and
Computer Science, U.C. Berkeley poolla@eecs.berkeley.edu

Abstract: The Smart Grid is among the most important and ambitious endeavors of our time.
Deep integration of renewable energy sources is one component of the Smart Grid vision. A
fundamental difficulty here is that renewable energy sources are highly variable – they are not
dispatchable, are intermittent, and uncertain. The electricity grid must absorb this variability
through a portfolio of solutions. These include aggregation of variable generation, curtailment,
operating reserves, storage technologies, local generation, and distributed demand response.
The various elements in this portfolio must be dynamically coordinated based on available
information within the framework of electricity grid operations. This, in turn, will require critical
technologies and methods drawn from optimization, modeling, and control, which are the core
competencies of Systems and Control. This paper catalogues some of these systems and control
research opportunities that arise in the deep integration of renewable energy sources.

Keywords: Smart Grid, Renewable Energy Integration, Uncertainty, Systems, Control,

Modeling

1. THE OPPORTUNITY instruments, advances in materials, sensors, and power-

electronics, dedicated communication networks with se-
What is the Smart Grid? cure protocols, and targeted infrastructure investments.
The Smart Grid is a vision of the future of the electric Renewable Energy Integration
energy system. There are many visions of this future, The scope of Smart Grid objectives and technologies de-
but perhaps these are best summarized by US Energy lineated above is indeed daunting. In this paper we will
Secretary Steven Chu who writes that “the Smart Grid confine ourselves to one of the many components of the
is the key enabler for: integration of renewable energy Smart Grid vision: deep integration of renewable energy
sources into the grid, management and deployment of into the electric grid. Renewable energy sources such as
energy storage, load management, system transparency, wind and solar are fundamentally different from conven-
and cyber and physical security of the electric energy tional generation such as coal, nuclear, natural gas. The
system”. Concerns over global climate change and carbon energy production from these renewable sources is not dis-
emissions have motivated development of integration of patchable [cannot be controlled on demand], intermittent
renewable energy generation from sources such as wind [exhibits large fluctuations], and uncertain [random or not
and solar to meet our electric energy needs. Another known in advance]. We will use the term variability to
particularly compelling motivation behind the Smart Grid encompass these three characteristics of renewable gener-
initiative is that the use of advanced communication, in- ation [33]. Variability is the most important obstacle in
formation, and control technologies may reduce the need integrating renewable generation into the electric energy
for expensive investments in the physical infrastructure – system at deep penetration levels.
generation, transmission, and distribution. Realizing these
visions of the Smart Grid and its benefits for combating The Variable Generation Problem
global climate change will require a portfolio of solutions
For solar power, there is the natural diurnal cycle of
such as deep integration of renewable resources, grid scale
variability in insolation [incident solar radiation]. Beyond
storage technologies, incorporation of micro-generation,
this, cloud-cover can cause significant ramps in solar
demand response and consumer participation, new market
insolation and electric power output. Solar insolation can
? Supported in part by OOF991-KAUST US LIMITED under award change by more than 80% of the peak insolation in
number 025478, the UC Discovery Grant ele07-10283 under the a matter of seconds [31]. The corresponding ramp in
IMPACT program, and NSF under Grant EECS-0925337, and the the solar power plant output is also very large but it
Florida Energy Systems Consortium.
also depends on panel size and geographical spread. In What is in this paper?
addition, inverter tripping adds to the variability in solar
plant output. In this paper we will explore those aspects of variable gen-
eration integration where systems and control technologies
Wind speed and direction experience significant variations will play a critical role in obtaining effective solutions. Our
at multiple time scales: minutes, hours, daily, seasonally, objective is to catalogue an incomplete list of such prob-
and annually. In addition to the slow and gradual varia- lems and analyze the underlying issues. The opportunities
tions, wind plant output can ramp up or down in minutes abound for our community of researchers, and the time is
[23]. For example, on December 29-30, 2008, wind power ripe to seize them.
output experienced 95% drop in a two hour period in the
BPA balancing authority [35]. Also, severe weather events 2. SYSTEMS AND CONTROL OPPORTUNITIES
can cause wind speed to become dangerously high making
operation of wind turbines unsafe, forcing shutdown. Ag-
We now discuss a number of specific problems which
gregation over large areas tends to reduce the variability in
need to be solved in order to achieve deep penetration
wind power production. These difficulties are compounded
of renewable energy sources where systems and control
by the fact that it is difficult to obtain reliable and accurate
technologies can make critically important contributions.
forecasts of wind power production. Forecasting difficulty
increases with forecast horizon. A. Forecasting
Dealing with Variable Generation Challenges in electric grid operations with deep renewable
penetration become much more tractable if accurate fore-
As we consider the variable generation characteristics of
casts of wind and solar energy production are available in
renewable energy sources, it should be noted that electric
advance. Accurate forecasts can alleviate negative impacts
power system operations are designed to accommodate
on the required spinning reserves for reliable operation
the natural (and very large) variability in load demand as
of the grid. They reduce the total cost of integration of
well as planned and unplanned contingencies. This is done
renewable energy into the grid. Wind power forecasting is
at different time-scales through load-frequency control,
a very active area of research [see [8] for a recent survey
operational reserves, scheduling and unit-commitment, de-
and references to the literature]. Wind power is a non-
mand response, and load shedding. At deep penetration
stationary stochastic process. There are two main ap-
levels, renewable generation add significantly to the extent
proaches to this problem: (a) physical approach where nu-
of the overall variability that must be handled. It has been
merical weather predictions are coupled with a collection
suggested that existing mechanisms to handle demand
of wind turbine physical models and data to predict the
fluctuations are adequate for renewable generation vari-
wind power output of the wind farm, and (b) techniques
ability up to 20% penetration levels [15, 50]. Reconfiguring
that combine empirical time-series models such as ARMA,
these existing mechanisms to handle the increased vari-
statistical forecasting, pattern matching, neural networks,
ability comes with significant cost issues and operational
fuzzy systems, etc. with numerical weather predictions to
challenges [12].
forecast the wind farm power output. Commercial wind
Systems and Control Opportunities power forecasting tools include combinations of these ap-
proaches as well. Spatial diversity can decrease the total
At deep penetration levels, dealing with variable gener- uncertainty in the output of a wind farm. Moreover, the
ation requires a portfolio of technologies which, in turn, forecasting error generally decreases with forecast horizon.
introduce operational challenges for the electricity grid. At short time-scales [minutes to hours], time-series mod-
Improved forecasting of wind and solar energy at multi- els are among the best performers. State-of-the-art tools
ple time-scales reduces uncertainty and permits efficient provide error rates of 4-6% for short term forecasts. For
scheduling of these variable resources into the genera- longer forecast horizons [24-48 hours], the prediction errors
tion mix. Dynamic optimization through optimal resource are large and can exceed 20-30%. A special issue in wind
scheduling and control can reduce the need for various forecasting is the occurrence of wind ramps. In addition
types of operational reserves thereby reducing integra- to point forecasts, newer approaches to dealing with wind
tion cost and increasing the carbon reduction benefits uncertainty require forecasts of probability distribution
of renewable generation. Optimal operation of grid scale functions. Similar comments apply to forecasting methods
storage can reduce the curtailment of wind energy when for solar energy production. Satellite and ground-based sky
the production of wind energy is too high. Smart charging imaging tools can be used to monitor and forecast cloud
of plug-in electric vehicles can make greater use of wind movements.
energy at night when the industrial and residential demand
is low. Demand response through real-time pricing or Renewable generation forecasting is clearly an area where
appropriate contracting mechanisms can reduce the need advanced techniques from linear and nonlinear estimation
for peak generation and improve matching of renewable and prediction can have a major impact.
generation with load. Managing increased variability in the
B. Electricity Markets
electricity grid will need advanced large scale distributed
control and optimization algorithms. Each of these avenues Electricity markets are recognized to be fundamentally
to address the fundamental challenge of renewable resource different than commodity markets. Electricity has phys-
variability requires systems and control technologies. We ical constraints on transmission, generation, scheduling,
submit that system identification, modeling, estimation, and storage, economic constraints on generation costs and
distributed optimization, and optimal control will play system infrastructure, and quality of service constraints
central roles here. demanded by consumers. As a result, traditional com-
modity market structures are not suitable for electricity. air, batteries, flywheels, molten salt, etc. Today, pumped
A certain level of centralized coordination and control is hydro is the most widely used storage modality but it is
necessary to ensure safe, economic, and reliable operation geographically limited. In the future, electric vehicles may
of the electric power system. While there is increasing provide distributed energy storage capacity which could
competition and decentralization in generation and retail be harnessed with Smart Grid technologies. Depending on
distribution, transmission and system operation remain the capacity of the storage device, it can be used on the
centralized. In its most deregulated form, the system is generation and transmission side or on the distribution
managed by an independent system operator [ISO] or a side. From the perspective of renewable integration, the
transmission system operator [TSO]. A common market key consideration is – when should energy storage be
structure [42, 30, 32, 25] consists of two successive ex-ante distributed across the grid versus co-located at generation
markets: a day-ahead (DA) forward market and a real- sites. These are dramatically different visions of large
time (RT) spot market. The DA market permits qualified scale storage deployment. If storage devices are co-located
generators to bid and schedule energy transactions for the with the renewable generation plant, it opens up the
following day. Depending on the region, the DA market possibility of real-time control of storage device to reduce
closes for bids and schedules by 10 A.M. and clears by 1 the variability of the renewable energy output. This can
P.M. on the day prior to the operating day. The schedules be posed as a stochastic optimal control problem to match
cleared in the DA market are financially binding and are a given profile of power output by suitably extracting
subject to deviation penalties. As the schedules submitted from and injecting into the storage device. An important
to the DA market are cleared well in advance of the question then is the minimum storage size needed to
operating day, a RT spot market is employed to ensure achieve a given level of probability of meeting the output
the balance of supply and demand in real-time by allowing power profile. The ultimate objective would be to make the
market participants to adjust their DA schedules based on wind plant dispatchable. This will require high precision
more accurate wind and load forecasts. The RT market control of the power output from the renewable plant. On
is cleared 5 to 15 minutes before the operating interval, the other hand, since storage is expensive [≈ $ 1.5M per
which is on the order of 5 minutes. MW-hr], it is quite possible that the best economic value
can be derived by optimizing its operation at the grid level.
In the context of deep penetration of renewable energy,
On the other hand, in the scenarios involving distributed
many new questions arise regarding these market mecha- renewable energy generation, community scale distributed
nisms and structures [21]. Renewable energy producers are storage may be the optimal choice.
often given favorable treatment via public policy support
through subsidies, price guarantees or other extra-market Optimal sizing and operations of distributed storage modal-
mechanisms. A major unresolved issue in this context is ities with different dynamic characteristics will require
the pricing of carbon emissions by certain types of energy methods from distributed optimization and stochastic de-
producers. Also, costs of integrating variable generation centralized optimal control.
are often socialized and distributed to other participants
D. Reserves and Local Generation
in the energy markets. However, as the penetration level
increases, such mechanisms are likely to be untenable and Wind farm energy production variability can be com-
the renewable energy producers will have to operate in the pensated by the use of local generation sources such as
market along with traditional energy producers. natural gas or diesel generators. Local generation adds to
What new market mechanisms are necessary to enable capacity costs as well as operational costs of the system.
deep penetration of renewable energy generators? Since Also, as natural gas plants produce carbon emissions, this
renewable generation is inherently variable and uncertain, approach attenuates the clean energy benefit from wind
power production. One can pose a joint stochastic optimal
it might become necessary or advantageous to devise novel
contract structures [ex: interruptible contacts [49]] that control problem involving wind and local generation to
explicitly deal with this variability. Also, more frequent deliver a certain amount of total power in the presence of
intra-day markets where the wind and solar energy pro- stochastic wind variability.
ducers can enter bidding closer to the delivery time will Deep penetration of renewables can have a very large
be better matched to the uncertainty in their production. impact on various different types of reserves needed to
These recourse markets may also help in reducing the operate the grid and meet the relevant reliability criteria.
cost of integrating variable generation by separating load Generally speaking, renewable generation has no signif-
following reserves from regulation reserves and/or reduce icant impact on the contingency reserves as the largest
the need for more expensive regulation reserves. Fast intra- contingencies arise from base-load generating units. By
day markets may be regarded as a form of closed loop contrast, the impact on operating reserves and regulation
control to deal with variabile generation. reserves is substantial. There are many studies to assess
Designing and optimizing new market mechanisms that the additional reserve requirements [15, 50]. Increased need
explicitly account for variable generation is an important for operating and regulation reserves is one of the pressing
research direction that will involve diverse methods from issues in integrating large amounts of renewable energy [9].
control theory and stochastic dynamic programming. They add to the cost of integration and thereby reduce the
beneficial impact of renewables. Since reserve generators
C. Joint Optimization of Wind and Storage often use hydrocarbon fuels, they detract from the pos-
itive impact of renewables on global warming. However,
At deep penetration levels, energy storage is expected to a critical assumption made in many of the current stud-
be a key enabling technology for the Smart Grid [13]. ies on required reserve requirements is that they assume
Energy can be stored using pumped hydro, compressed
no change in the current operating procedure. Hence, it increased occurrences of disconnection from the grid due
may be possible to apply techniques from systems and to voltage fluctuations [22].
control to mitigate the impact of variability in renewables
on reserve requirements. For example, dynamic stochastic Modern commercially available solid state inverters are
optimization of reserves that can meet the required loss capable of supplying (capacitive as well as inductive) re-
of load probability by taking advantage of reduced uncer- active power. At the same time, fast communications of
tainty with shorter forecast horizons. voltage measurements across the distribution network may
allow for coordinated control actions. Thus, it is possible
Progress in both these research areas will use tools from to conceive of control system architectures and algorithms
stochastic optimization. that leverage these actuation capabilities to minimize
voltage fluctuations [34, 39]. In addition, DSTATCOMs
E. Power System Stability (Distributed Static Compensators) are controllable reac-
Experience in island power systems indicates that severe tive power resources that can be employed to deal with
system level problems arise when the penetration of wind voltage quality issues. What is the optimal placement of
power exceeds 15%. In Denmark, where the wind pene- these devices? How should they be controlled to optimize
tration exceeds 20%, this problem has been avoided by metric of voltage quality? These are some of the emerging
the use of interconnections to power systems in Sweden, problems that are beginning to attract attention [41].
Norway, Germany, etc. Spain has almost 100% reserve Mitigating undesirable voltage fluctuations will require the
capacity. Analysis of the impact of high penetration of use of techniques from large scale distributed control and
variable generation sources on the stability and perfor- optimization.
mance [voltage, frequency, etc.] of the power system is a
difficult problem. Transient frequency response, regulation G. Matching Variable Generation to Adjustable Demand
response, automatic generation control, load following,
and unit commitment processes, operating at different Demand side actions offer another strategy to absorb
time scales [seconds to hours], together determine the injected variability from renewable generation. There is
overall impact on the voltage and frequency of fluctuations a considerable effort in developing demand response as
in the renewable power production. Also, wind turbines a Smart Grid technology for this (and other) objective.
are variable speed generators as opposed to synchronous See for example [7, 29, 44]. Profitable business models for
generators for traditional power generation. These vari- demand response with commercial customers (ex: refriger-
able speed generators [ex: doubly fed induction generator] ated warehouses) have been realized by companies such as
may adversely impact the transient stability of the power EnerNOC Corporation [14]. Much of the existing research
studys implementation aspects (ex: communication infras-
system [3] as they reduce the inertia of the system by
displacing synchronous generators. Analysis and design of tructure, appliance control, price signalling) of demand
suitable control systems to enhance power system stability response. The operational aspects of matching variable
at high penetration levels is a key problem where systems generation with adjustable demand in transmission con-
and control techniques will prove invaluable. strained power systems remain largely unexplored.

Power system stability is a networked control problem, There are a diversity of consumers with different elec-
the scale of which will require the creative development of tricity needs. Some of these are power consumers: they
require power on demand and offer little opportunity for
computational tools for stability analysis.
demand shaping [ex: continuous manufacturing facilities
F. Voltage Support that cannot defer their demand for electricity]. Others are
energy consumers: they can accept variable power within
Maintaining tight voltage control across the distribution certain constraints with advance notice and coordination.
system is an important part of delivering power quality. Examples here include warehouses which can chill their re-
Tight voltage control is necessary for efficiency in many frigeration space on variable cooling schedules, and electric
end-use applications. Distribution system voltage levels vehicle charging stations which can use variable charging
are traditionally controlled by adjusting load tap changes
schedules.
in substation transformers, and through capacitor banks to
provide reactive power support locally. Integration of large What are generic models for the electricity needs of energy
amounts of renewable power in the distribution system consumers? What is the set of power profiles P that
[ex: rooftop solar] will create new sources of variation in they can accept for the period [0, T ]? Once we have
the voltage level in the distribution system. The random models of acceptable power profiles from certain classes
locations of these generation devices can also lead to of energy consumers, we can formulate various flavors of
increased imbalance operation in the 3-phase distribution demand matching problems. Consider a variable generator
network. These variations and imbalances change along and an energy consumer [EC] who can absorb arbitrary
the length of the feeder circuit. power variability and assume we have no transmission
constraints. In this happy situation, all variable generation
There is active discussion of modifications to the standards can be sold to the EC, through bilateral contracts or
(IEEE 1547-1547.8) which regulate connection of renew- other arrangements. However, if the transmission network
able generators to the distribution system. There are also is constrained, or if the energy consumer can accept only
considerable variations in local rules that also constrain certain power profiles, then some variable generation must
the connection of renewable generators to the distribution be curtailed and/or absorbed by operating reserves. How
system. Among other issues, there is concern that unless much variable generation can be absorbed by adjustable
these regulations are properly conceived, they may lead to
loads? This problem is compounded by the fact that the phase measurement units [synchro-phasors], control of
wind w(t) is random. asynchronous wind turbines, power electronics for optimal
operation of large PV arrays, reactive power compensation
The ideas above suggest introduction of the concept of
issues in wind farms, micro-grid control, distributed energy
Quality-of-Service [QoS] for power. On the generation side,
generation, etc. Even in the confines of our narrow focus,
QoS is intended to capture reliability in supply, while
it is evident that our community of researchers in systems
on the demand side, QoS is meant to capture variability
and control will be major contributors to realizing the
in acceptable power profiles. Pricing mechanisms can be
vision of the Smart Grid.
introduced to reflect QoS in both generation and demand,
and may prove to be the natural mechanism to treat
variability in competitive markets [46].
REFERENCES
Matching variable generation to distributed demand re-
sponse resources will require techniques drawn from model [1] J. L. Angarita, J. Usaolab, J. Martinez-Crespob,
predictive control, and stochastic dynamic programming. “Combined hydro-wind generation bids in a pool-
based electricity market,” Electric Power Systems
H. Optimal Coordination Research, vol. 79, pp. 10381046, 2009.
Today, wind power is commonly treated as a negative load [2] G. N. Bathurst, J.Weatherill, and G. Strbac, “Trading
and the utility (or system operator) must accept all wind wind generation in short term energy markets,” IEEE
power produced at a fixed price (feed-in tariff). This forces Trans. Power Syst., vol. 17, pp. 782-789, 2002.
the utility to use reserves to compensate for the injected [3] A.R. Bergen and V. Vittal, “Power Systems Analy-
variability in the wind power w(t). sis,” Prentice-Hall, Englewood Cliffs, NJ, 2000.
[4] http://www.transmission.bpa.gov/business/ opera-
Reserve generation imposes both a capacity (reservation) tions/Wind/default.aspx
and energy (use) cost. There are many different types [5] E. Bitar et al., “Bringing Wind Energy to Market”,
of reserves, each with its own start, stop and ramping Submitted to IEEE Transactions on Power Systems,
characteristics. Reserve capacity is traditionally procured 2011.
well in advance (∼ 1 hr) of the delivery time. Greater [6] E. Bitar et al., “Optimal Contracts for Wind Power
forecast uncertainty therefore increases reserve capacity Producers in Electricity Markets,” Proc. of the Conf.
costs. This can be mitigated through: (a) improved fore- on Decision and Control, pp: 1919-1926, Atlanta,
casting, (b) reserve capacity purchases over short horizons, 2010.
and (c) matching adjustable demand to variable gener- [7] S. Borenstein, M. Jaske and A. Rosenfeld, “Dynamic
ation [see subsection G above]. These strategies require Pricing, Advanced Metering and Demand Response
coordination: optimal reserve power purchases and load in Electricity Markets,” Center for the Study of En-
adjustment strategies are dynamically coupled through ergy Markets Working Paper, University of California
forecast information. at Berkeley, October, 2002.
[8] C. Monteiro, R. Bessa, V. Miranda, A. Botterud, J.
How can we dynamically aggregate variable generation
Wang, and G. Conzelmann, “Wind Power Forecast-
with various types of reserves so that the portfolio col-
ing: State-of-the-Art 2009,” Argonne National Labo-
lectively behaves as reliably as dispatchable thermal gener-
ratory, ANL/DIS-10-1, November 2009.
ation. In this dynamic portfolio problem we must select the
[9] California Independent System Operator, “Integra-
most economical mix based on current generation forecasts
tion of Renewable Resources, August 2010.”
and current cost information of available firming resources.
[10] A. Cavallo, “High capacity factor wind energy sys-
Implementation of dynamic portfolios will leverage the
tems,” J Sol Energy Eng, vol. 117, no. 5, 1995.
Smart Grid infrastructure – sensors and communications
[11] A. Cavallo, “Controllable and affordable utility-
to provide the necessary real-time information.
scale electricity from intermittent wind resources
The problem of optimally computing resource acquisi- and compressed air energy storage (CAES),” Energy,
tion and dispatch decisions is a challenging constrained 2007;32(2):1207.
stochastic optimal control problem. [12] J. DeCesaro and K. Porter, “Wind energy and power
system operations: A review of wind integration stud-
3. CONCLUSIONS ies to date,” The Electricity Journal, vol. 22, pp. 34-
43, 2009.
[13] P. Denholm, E. Ela, B. Kirby, and M. Milligan, “The
The deep integration of renewable energy resources into
Role of Energy Storage with Renewable Electric-
the electricity grid is a compelling and challenging en-
deavor. It offers a rich set of research problems, many of ity Generation,” Technical Report, NREL/TP-6A2-
which will critically require systems and control methods 47187, January 2010.
in their solution. We have collated a partial list of spe- [14] http://www.enernoc.com/
cific problems driven by the variability and uncertainty of [15] EnerNex Corp., Eastern Wind Integration and Trans-
renewable resources. mission Study, National Renewable Energy Labora-
tory, Report NREL/SR-550-47078, January 2010.
Our focus in this paper has been on the issue of integration [16] W.A. Gardner, A. Napolitano, and L. Paura, “Cy-
of variable generation. We have not discussed many other clostationarity: Half a century of research,” Signal
research opportunities for the systems and control research Processing (Elsevier) (2006) 639–697.
community in the broad topic of Smart Grid. These [17] Global Wind 2009 Report, Global Wind Energy
include cyber-physical security, optimal deployment of Council, Brussels, Belgium, 2010.
[18] T. Gomez, C. Marnay, A. Siddiqui, L. Liew, and M. Journal of Electrical Power & Energy Systems, vol.
Khavkin, “Ancillary Services Markets in California,” 33, no. 3, pp:485 - 492, 2011.
[19] J. B. Greenblatt, S. Succar, D. C. Denkenberger, R. [35] J. M. Pease, “Critical Short-term ForecastingNeeds
H. Williams, and R. H. Socolow, “Baseload wind en- for Large and Unscheduled Wind Energy on the
ergy: modeling the competition between gas turbines BPA System,” 3rd Workshop on the Best Practice
and compressed air energy storage for supplemental in the Use of Short-term Forecasting of Wind Power,
generation,” Energy Policy, vol. 35, pp. 1474-1492, October 2009.
Mar 2007. [36] P. Pinson, C. Chevallier, and G. N. Kariniotakis,
[20] G. Gautam, V. Vittal, and T. Harbour, “Impact of “Trading wind generation from short-term probabilis-
Increased Penetration of DFIG-Based Wind Turbine tic forecasts of wind power,” IEEE Trans. Power
Generators on Transient and Small Signal Stability Syst., vol. 22, pp. 1148-1156, 2007.
of Power Systems,” IEEE Trans. on Power Systems, [37] P. Pinson, H. A. Nielsen,H. Madsen, and K. Karin-
Vol. 24, No. 3, August 2009, pp. 1426-1434. iotakis, “Skill Forecasting from Ensemble Predictions
[21] W. W. Hogan, “Electricity Wholesale Market Design of Wind Power,” Applied Energy, vol. 86, pp. 1326-
in a Low Carbon Future,” in Harnessing Renewable 1334, 2009.
Energy, Jorge Padilla and Richard Schmalensee, Ed- [38] R. Rajagopal et. al., “Risk Limiting Dispatch,” in
itors. preparation.
[22] http://grouper.ieee.org/groups/scc21/dr shared/ [39] K. M. Rogers, R. Klump, H. Khurana, A. A. Aquino-
[23] C. Kamath, “Understanding wind ramp events Lugo, and T. J. Overbye, “An Authenticated Control
through analysis of historical data,” IEEE PES Framework for Distributed Voltage Support on the
Transmission and Distribution Conference, New Or- Smart Grid,” IEEE Transactions on Smart Grid, Vol.
leans, April 2010. 1, pp. 40 - 47, June 2010.
[24] A. Kusiak and W. Y. Li, “Short-term Prediction of [40] S. Sethi, H. Yan, J. Yan, and H. Zhang, “An Analysis
Wind power with a Clustering Approach,” Renewable of Staged Purchases in Deregulated Time-Sequential
Energy, vol. 35, pp. 2362-2369, 2010. Electricity Markets,” Journal of Industrial and Man-
[25] M. Korpaas, A. T. Holen, and R. Hildrum, “Oper- agement Optimization 2005, 1, 443463.
ation and sizing of energy storage for wind power [41] F. Shahnia, A. Ghosh, G. Ledwich, and F. Zare,
plants in a market system,” International Journal of “Voltage unbalance reduction in low voltage distri-
Electrical Power and Energy Systems, vol. 25, no. 8, bution networks with rooftop PVs,” Proceedings of
pp. 599-606, Oct. 2003. Australian Universities Power Engineering Confer-
[26] H. Lund and G. Salgi, “The role of compressed air ence : Power Quality for the 21st Century, pp: 5-8,
energy storage (CAES) in future sustainable energy Christchurch, New Zealand, December 2010.
systems,” Energy Conversion and Management 50 [42] F. P. Sioshansi (Ed.), “Competitive Electricity Mar-
(2009) 1172-1179. kets - Design, Implementation, Performance,” Else-
[27] J. Matevosyan and L. Sajder, “Minimization of im- vier, 2008.
balance cost trading wind power on the short-term [43] “Market Operator of the Electricity Market of the
power market,” IEEE Trans. Power Syst., vol. 21, pp. Iberian Peninsula,” OMEL 2008. [Online]. Available:
1396-1404, 2006. http://www.omel.es.
[28] D. L. Hawkins, J. Blatchford, and Y. V. Makarov, [44] K. Spees, L. Lave, “Impacts of Responsive Load
“Wind integration issues and solutions in California,” in PJM: Load Shifting and Real Time Pricing,”
in Proc. IEEE PES General Meeting, Tampa, FL, Carnegie Mellon Electricity Industry Center Working
Jun. 24-28, 2007. Paper CEIC-07-02; Carnegie Mellon University, 2007.
[29] Z. Ma, D.S. Callaway and I.A. Hiskens, “Decentral- [45] S. Succar and R. H. Williams, “Compressed air energy
ized Charging Control for Large Populations of Plug- storage: Theory, resources, and applications for wind
in Electric Vehicles,” Proceedings of the IEEE Con- power,” Energy Systems Analysis Group, Princeton
ference on Decision and Control, Atlanta Georgia, University, 2008.
2010. [46] C.W. Tan, P.P. Varaiya, “Interruptible electric power
[30] Y. Makarov, C. Loutan, J. Ma, and P. de Mello, service contracts,” Journal of Economic Dynamics
“Operational impacts of wind generation on Cali- and Control, Vol. 17, No. 3, pp. 495:517, 1993.
fornia power systems,” IEEE Transactions on Power [47] G.C. Thomann, M.J. Barfield, “The time variation of
Systems 2009; 24: 1039-1050. wind speeds and windfarm output in kansas,” IEEE
[31] J. Marcos, L. Marroyo, E. Lorenzo, D. Alvira and E. Transactions on Energy Conversion 1988.
Izco, “Power output fluctuations in large scale PV [48] P.P. Varaiya, F. Wu, J.W. Bialek, “Smart Operation
plants: one year observations with one second resolu- of Smart Grid: Risk-limiting Dispatch,” submitted to
tion and a derived analytic model,” Prog. Photovolt: Proceedings of the IEEE, 2010.
Res. Appl., DOI: 10.1002/pip.1016, (2010) [49] C.W. Tan, and P.P. Varaiya, “Interruptible electric
[32] J.M. Morales, A.J. Conejo, J. Perez-Ruiz, “Short- power service contracts,” Journal of Economic Dy-
term trading for a wind power producer,” IEEE Trans namics and Control 17 (May, 1993): 495-517.
Power Syst; in press. [50] GE Energy, Western Wind and Solar Integration
[33] NERC Special Report, “Accommodation of High Lev- Study, National Renewable Energy Laboratory, Re-
els of Variable Generation,” April 2009. port NREL/SR-550-47434, May 2010.
[34] M. Oshiro et al., “Optimal voltage control in distri-
bution systems using PV generators,” International