1

Smart Grid Modeling Approach for Wide Area Control Applications

Aaron St. Leger, Member, IEEE, John James, Senior Member, IEEE, Dean Frederick, Senior Member, IEEE
communication and information networks. For example, results presented in [2] show that an increase in time delay can cause degradation of frequency control using decentralized intelligent loads. Furthermore, too much latency can lead to system instability. As a result, the present state and dynamics of communication networks can be a critical contingency for current and future smart grid applications. The objective of this work is to develop a modeling methodology for analyzing smart grids, control techniques, and identifying important contingencies within cyber and physical elements of the system specifically for wide area control applications. These contingencies could be malicious, for example a cyber or physical attack on a network, or non-malicious. A critical component for modeling smart grids is modeling the interdependencies and interactions between the cyber and physical components. Vulnerability analysis of power systems and information Index TermsSmart Grid, Communication Network Modeling, Power System Modeling, Wide Area Measurement networks is a continuing field of research [3-5]. The focus of many research efforts have been placed on large cascading and Control failures due to impacts of such disruptions. Historically, much research has focused on either the power grid or information I. INTRODUCTION networks [6, 7]. Recently the interdependencies of the two he power grid consists of physical components, which infrastructures been studied [3, 8, 9]. The current state-of-thegenerate and transmit power, and cyber components art techniques rely on qualitative analysis of the systems and which transmit data and control signals. Currently, interdependencies [5] and, as a result, develop approximate operation and control of bulk power generation and results and estimations of the real interdependencies of the two transmission network occurs at centralized control centers and systems. Some work is moving towards a quantitative relies mostly on operator in the loop control and analysis. For approach more suitable for analyzing smart grid applications example, results from state estimation and contingency and contingencies [10]. The approach described in this paper analysis will be reviewed by operators and adjustments system is to develop a unified quantitative methodology of modeling operation made accordingly by the system operator. This both the cyber and physical components of the system, and the control loop relies on human intervention and the time scale is interdependencies between the two. More specifically the on the order of minutes. In addition, some automatic wide area focus is on a unified cyber/physical system model suitable for control, such as automatic generation control (AGC), has been simulation and analysis of the following: implemented and relies on a slow response. More specifically, AGC acts slowly and deliberately over tens of seconds or a Physical contingencies in HV transmission few minutes [1]. Current analytical techniques and models network/bulk power generation and sensors make assumptions that communication lines are in service and Cyber contingencies in smart grid components any latency or bandwidth constraints are negligible and/or related to HV transmission/bulk Generation have no effect on system operation. With the slow response Decentralized local and wide area control and dynamics of current wide area control techniques these Centralized wide area control assumptions are adequate. However, the advancement and implementation of smart grid technology requires more advanced models that factor in the status and performance of Some specific examples of wide area control techniques of interest are voltage control, frequency control, and small signal stability control for inter-area modes. Of particular A. St. Leger and John James are with the Electrical Engineering and interest is how advanced wide area control techniques are Computer Science department at the United States Military Academy, West susceptible to contingencies of physical and/or cyber Point, NY 10996 USA (e-mail: aaron.stleger@usma.edu and components.
Abstract This paper describes an approach for modeling smart grids for wide area control applications. More specifically, it is proposed to model smart grids as a set of interdependent composite networks. A composite network is one whose evolution in time and/or space is described as a composition of more than one category of networks. This work proposes an interconnection of a communication network, information network, and a power system network to model smart grids. More specifically this work proposes a quantitative model focusing on bulk generation and transmission. The resulting model will be used for studying and simulating wide area measurement and control techniques and contingencies of components. The modeling methodology is based on the initial partitioning by the National Institute of Standards and Technology (NIST) of the smart grid domains. A basic example with control of load (demand response) and generator set points over a communication link is presented.

john.james@usma.edu ). Dean Frederick is with Saratoga Control Systems, Inc.

978-1-4673-2729-9/12/$31.00 2012 IEEE

Developing a suitable mode el for smart gr rid simulation is cha allenging as the e smart grid is still emerging and evolving as dev velopment of te echnology and d control techn niques continue es. The e modeling me ethodology pre esented here is s developed in na flex xible fashion to allow for f implemen ntation of ne ew tech hnology and co ontrol schemes s. The smart gr rid as defined by b NIS ST [11] throug gh seven domai ins, shown in Fig. 1, was used as a starting point for this work.

Cyber com mponents: o S Smart grid con ntrol logic (e.g g. wide area c control logic/de ecision making g). Transfer of information n between o T c components (in nformation netw work)

The p physical and cy yber componen nts are modele ed separately and the en linked together to model l the interactio ons between them. M More specifica ally, the separ rate models ar re combined into uni ified set algebr raically constr rained ordinary y differential equation ns:
x (t ) = f ( x (t ) , u )

g ( x (t ) , u ) = 0

(1)

where:
x
u

f g
Fig g. 1 Actors in the Seven Domains of f the Smart Grid

set of the sta ate variables of f the system set of system m parameters an nd inputs set of ordina ary differential equations set of algebra aic constraints

In the next section s we wil ll provide an overview o of our o prop posed methodo ology for mod deling the smar rt grid, includin ng specific details on mode eling the power p system m, com mmunication, and a information n network com mponents. This is foll lowed by a se ection showin ng some initia al modeling an nd sim mulation of som me components followed by a conclusion. II. SMA ART GRID MOD DELING METHO ODOLOGY Only a subset t of Smart Grid d components, as defined by the sev ven domains in Fig. 1, is pertinent to HV bulk pow wer tra ansmission netw work. As a result, only comp ponents applica able to the previously y outlined analysis are mode eled in this wo ork. cally, this model inc cludes controls, More specific ommunication network dy ynamics, and power syst tem co ynamics. Thi is aligns with w the Bu ulk Generati ion, dy Tr ransmission, an nd Operations actors in the NIST N framewo ork. Inf fluence at the HV transmiss sion level from m customer lo oads an nd the distribut tion network can c be modele ed aggregately y at HV V substations. Additional stu udies are requir red to sufficien ntly an nalyze the im mpact of smart grid com mponents in the dis stribution syst tem at the tra ansmission lev vel. Physical and a cy yber compone ents within and a between the generati ion, tra ansmission and d operation actors are modele ed. This work has foc cused on the fo ollowing: Phys sical componen nts: o Generator rs, loads, transm mission networ rk o Communi ication devices (e.g. modems) ) o Communi ication links (e.g. fiber op ptic cable) e.g. phasor mea asurement unit ts) o Sensors (e o Controller rs (e.g. vol ltage regulato ors, governors s).

nterdependenc cies between n cyber and d physical The in compon nents are accou unted for by the set of system m parameters and inpu uts, u. More s specifically, in nputs and/or pa arameters of physical l components s are modeled such tha at they are depende ent upon the sta components and d vice versa. ates of cyber c For exa ample, a dece entralized wid de area voltag ge controller input, cu urrent state (vo oltage(s)) of th he power system m, is not the power system state but the stat te of the cy yber/physical nents which ar re measuring, p processing, an nd delivering compon the volt tage(s) inform mation to the c controller. As a result, the dynamic cs associated with measu urement, processing, and commun nication comp ponents which connect the d decentralized controll er to the syste em are explicitly modeled. In addition, as contingencie es of any cyber r or physical discrete events, such a nents, are mo odeled discretely as chan nges to the compon differen ntial equations and/or algebr raic constraint ts using the previous itions. s system state as initial condi A ge eneral framewo ork of the overall smart gr rid model is in Fig. 2. The e model incorp porates the po ower system shown i model, consisting of generato ors, transmission lines, rmers and load ds, the commu unication netw work model, transform consistin ng of comm munication lin nks between and within compon nents, and local/wide are ea control. The Local Commu unication Netw work and Contr rol (LCNC) mo odels control actions distributed thr roughout the gr rid that are tak ken at a local ol actions c could depend d on local level. These contro ement, wide ar rea measureme ents or both. F For example, measure a smart substation is m modeled as a L LCNC model. This model orithms gove erning smart substation would include algo behavio or, local measu urement/control techniques, a and interface with ext ternal compon nents via a com mmunication li ink/network. Remote System Opera ation and Cont trol (RSOC) is s represented in a sim milar fashion and allows fo or modeling o of wide area control and operation n. This model l structure pa asses control nds to the syste em via the com mmunication ne etwork. comman

communication links (e.g. modems, fiber optic networks, etc). Initial work has incorporated a frequency shift key (FSK) modem to transmit control signals between components. The second approach is a generic communication link model incorporating bandwidth and latency which are the two most inherent properties for smart grid communication as discussed in [13]. The initial model incorporates a variable time-delay to the data sent over a communication link. This allows variation of the time-delay to model constant or variable latency in data transmission. Work is ongoing to develop time-delay based models to represent specific communication hardware and protocols. However, the initial time-delay model can be used to study the effects of latency on wide area control techniques and other smart grid functions.

T T

T T

Fig. 2. Smart Grid Model Structure

This general framework of LCNC and RSOC linked to physical model of the power system/communication network is generalized to allow for modeling of a wide range of smart grid devices and controls. The next sections discuss the power system, communication network, and control components in more detail. A. Power System Network Components The power system network, which consists of an interconnection of transmission lines and transformers, is modeled by an interconnection of impedances modeling each component. Network equations in terms of the nodal admittance can be written for an n bus system from this as follows [12]:
I1 Y11 I Y 2 = 21 K K I n Yn1 Y12 K Y1n V1 Y22 K Y2 n V2 K K K K K K Ynn Vn

C. Control Components Modeling of control components is broken down into LCNC and RSOC models. RSOC models are used for wide are controllers such as Static Var Compensation (SVC) control proposed in [14]. A model for a SVC in our approach is shown in Fig. 3. Communication links transmit measurements from a phasor measurement unit (PMU) unit embedded in the power system model and deliver it to an algorithm which processes the data, updates the discrete state of the SVC and send a control signal over a communication link. Different control algorithms, communication links, SVC models, etc. can be modeled with this approach.

(2)
Fig. 3. RSOC Model of Wide Area Control SVC

or

I = Ybus V

(3)

where Ybus is the bus admittance matrix, I is a column vector of current injection at the network nodes and V is a column vector of nodal voltages. Generators and loads are modeled as power injections into the system nodes. Generators are modeled as synchronous machines with governors, exciters and power system stabilizers. The mechanical model of the generator is based on the swing equation and governors, exciters and power system stabilizers are based on standard models used in power system analysis. Details on these models can be seen in [12]. Loads are modeled as constant power. Enhancement of this work is ongoing to incorporate ZIP and dynamic load models based on induction machines.

LCNC models are used for localized control which can be based on local and/or wide area measurements. For example, power system stabilizer controls are implemented via local measurement and feedback at the generators. Smart grid components requiring wide area measurement or transfer of information between components embedded in the network are modeled as intelligent agents. An intelligent agent is an autonomous, goal-oriented entity that can interact with its environment [13]. This is modeled here as an algorithm dictating the behavior of the agent with local and wide area measurement as inputs while local control actions and communication with other agents as outputs. Latency of local control actions and measurements for LCNC is assumed to be zero. Latency and transmission of information to and from LCNC is represented by the communication network model. The following section discusses initial efforts toward constructing and simulating the proposed smart grid model.

B. Communication Network Components Communication network modeling consists of two approaches. The first is to model physical devices and

III. SMART GRID MODEL SIMULA ATION MATLAB/Sim mulink [15] ha as been utilized d for construct ting an nd simulating the t proposed smart s grid mo odel in this wo ork. Th his software en nvironment is flexible enoug gh to add cust tom mo odels, adjust pre-existing models, dev velop a cust tom gra aphical user in nterface, and utilize u various solvers availa able in MATLAB. In I addition, co-simulation c of discrete and a co ontinuous syste ems can be im mplemented. This T is critical for sim mulating contingencies amon ng discrete an nd dynamic sta ates pre esent in the pr roposed smart grid g model. Al ll of the propo osed co omponents hav ve not yet been n implemented d; however, so ome ini itial progress has h been made and is presente ed here. Power syste em simulation is han ndled via the Sim mPowerSystem ms toolbox. Controllers C (g generator volt tage reg gulation, powe er system stabilization, etc.) are implemen nted via a Simulink. Co ommunication components are a simulated by b a co ombination of communication toolbox x and cust tom pro ogrammed fun nctions. More advanced a smar rt grid controll lers an nd agents are being develop ped through custom c functions int terfacing wit th power system s and communicat tion co omponents. Presently, the IEEE 14 bus system has been b implemen nted ith remote con ntrol of generat tor set points, power p output and a wi vo oltage magnitu ude, via a com mmunication link and a FSK F mo odem. In addit tion, load cont trol and status of power syst tem co omponents are controllable via FSK mod dem. This ini itial ork shows a pr roof of concept of integrating g communicati ion, wo co ontrol and pow wer system co omponents wh hich comprise the pro oposed smart grid model. An example, , including th hese co omponents, of a single machi ine infinite bus s system is sho own d 5. This exam mple utilizes a RSOC model ling in Figures 4 and ethod to use a remote algori ithm to set gen nerator set poi ints me thr rough a FSK modem m and a communicatio on link. The FSK F mo odem model and a transmissio on of data is discrete d in natu ures wh hile the genera ator, exciter, an nd load are con ntinuous. Anot ther ex xample that has s been implem mented is a rem motely controlla able bre eaker that can vary load lev vels or compon nent status via a an inp put from a com mmunication ne etwork.

Fig. 5. S Simulation Results s for Step Change e in Generator Vo oltage via FSK Modem

A rem motely control llable circuit breaker is show wn in Fig. 6. This con nsists of a phy ysical model o of the circuit b breakers, one for each h phase, a c control input, and an inter rface to the commun nication link. T This controllab ble breaker is i implemented in a load d control appli ication in Fig. 7. An input fro om a control algorithm hm is provided to the FSK m modem which t transmits the control signal over a communica ation link to the circuit breaker. . This exampl le is being ut tilized to cont trol demand response e remotely. In n addition, fu uture work w will utilize a similar p physical mode el to control SV VCs as shown i in Fig. 3.

Fig. 6. M Model of Remotely Controllable Circu uit Breaker

Fig. 7. FS SK Modem Contro olling Load

IV. CONCLUS SION

Fig g. 4. Single Mach hine Infinite Bus with w FSK Modem Control of Gener rator Vo oltage

This paper presents an approach h for modeling g smart grid dynamic cs as a set of f interdependen nt power, com mmunication, and con ntrol componen nts. The mode el is quantitativ ve in nature and inc cludes power, , communicat tion, and con ntrol system dynamic cs with a focus s on bulk gene eration and tran nsmission of power. The resulting approach is b being utilized for studying and sim mulating wide a area measurem ment and contro ol techniques

and their associated cyber and physical contingencies. The MATLAB/Simulink simulation environment is flexible to allow a wide range of modeling and simulation of smart grid components and control schemes. Future work will include analyzing the scalability of the proposed approach and refining models. Simulating very large systems in detail will likely be impractical, however, the developed framework allows for implementation of refined or simplified models more appropriate for large system simulation. V. ACKNOWLEDGMENT This work was supported by the Defense Threat Reduction Agency MIPR# 10-2693M, to the United States Military Academy. VI. REFERENCES
[1] [2] N. Jaleeli, et al., "Understanding Automatic Generation Control," IEEE Transactions on Power Systems, vol. 7, pp. 1106-1122, Aug 1992. D. Trudnowski, et al., "Power-system frequency and stability control using decentralized intelligent loads," Proceedings of the 2006 IEEE Power Engineering Society T&D Conference adn Expo, pp. 1453-1459, May 2006. S. Chiaradonna, et al., "On a modeling framework for the analysis of interdepenedencies in electric power systems," Proceedings of the 37th Annual IEEE/IFIP International Conference on Dependable Systems and Networks, 2007. J. C. Laprie, et al., "Modeling cascading and escalating outages in interdependent critical infrastructures," Proceedings of the 2006 IEEE International Conference on Dependable Systems and Networks, pp. 226-227, 2006. J. C. Laprie, et al., "Modelling interdependencies between the electricity and information infrastructures," Proceedings of the International conference on Computer Safety, Reliability and Security (SAFECOMP), pp. 54-67, 2007. P. Crucitti, et al., "Model for cascading failures in complex networks," Physical Review E, vol. 69, Apr 2004. P. Task Force on Understanding, Mitigation and Restoration of Cascading Failures in Electric Power Systems, "Vulnerability Assessment for Cascading Failures in Electric Power Systems," Proceedings of the 2009 IEEE Power Systems Conference and Exposition, pp. 1-9, 2009. A. Z. Faza, et al., "Reliability Modeling for the Advanced Electric Power Grid: A Proposal for Doctoral Research," Proceedings of the 33rd Annual IEEE International Computer Software and Applications Conference, pp. 672-675, 2009. J. Lin, et al., "A General Framework for Quantitative Modeling of Dependability in Cyber-Physical Systems: A Proposal for Doctoral Research," Proceedings of the 33rd Annual IEEE International Computer Software and Applications Conference, pp. 668-671, 2009. J. Nutaro, "Designing power system simulators for the smart grid: combining controls, communications, and electro-mechanical dynamics," Proceedings of the 2011 IEEE Power Engineering Society General Meeting, pp. 1-5, July 2011 2011. "NIST Interagency Report 7628: Guidelines for Smart Grid Cyber Security: Smart Grid Cyber Security Strategy, Architecture, and HighLevel Requirements," 2010. P. Kundur, Power System Stability and Control. New York: McGraw Hill, 1994. C. P. Nguyen and A. J. Flueck, "Modeling of communication latency in smart grid," Proceedings of the 2011 IEEE Power and Energy Society General Meeting, 24-29 July 2011. J. Quintero and V. Venkatasubramanian, "SVC Compensation on a Real-Time Wide-Area Control for Mitigating Small-Signal Instability in Large Electric Power Systems," Proceedings of the 2006 International Conference on Power System Technology, pp. 1-8, Oct 2006. "MATLAB," http://www.mathworks.com, Ed., ed: The Mathworks, Inc.

[3]

[4]

[5]

[6] [7]

[8]

[9]

[10]

[11] [12] [13] [14]

[15]