A Control and Communications Architecture for a Secure and Reconfigurable Power Distribution System: An Analysis and Case Study

Transcription

1 A and Communcatons Archtecture for a Secure and Reconfgurable Power Dstrbuton System: An Analyss and Case Study Anthony M. Gacomon*, S. Massoud Amn*, and Bruce F. Wollenberg*! *Department of Electrcal and Computer Engneerng, Unversty of Mnnesota, Mnneapols, MN USA (e mal: {gaco0, amn, wollenbe}@umn.edu) Abstract: The transformaton of the end to end power grd to a dgtalzed, ntellgent, self healng system presents many new modelng, sensng, communcatons, and control challenges that must be addressed before extensve deployment can begn. Increasng the securty, robustness, and effcency of electrc power nfrastructure requres utlzng these automaton technologes n order to contnually assess and optmze system performance. In ths paper, an ntellgent dstrbuted secure control archtecture s presented for dstrbuton systems to provde greater adaptve protecton, wth the ablty to proactvely reconfgure, and rapdly respond to dsturbances. Detaled descrptons of functonaltes at each layer of the archtecture as well as the whole system are provded. Applyng ths comprehensve systems' approach, performance results for the IEEE 2 node test feeder are smulated and analyzed. The results show the trade offs between system relablty, operatonal constrants, and costs nvolved. Ths work represents a novel strategy toward developng an analytcal and mult doman methodology to assess the effects of smart grd technologes on dstrbuton system operatons and performance. Keywords: Power Systems, Power Systems, Power Systems Dstrbuton, Smart Power Applcatons, Smart Grds. INTRODUCTION!! Plannng has already begun to replace much of the antquated electrc nfrastructure of the exstng power delvery system wth dgtal systems provdng the grd wth the capablty to reconfgure tself and prevent wdespread outages. Often, ths collecton of dgtal overlad systems s referred to as smart grd. Upgradng the power grd, however, wll present many new securty challenges that must be dealt wth before extensve deployment and mplementaton of smart grd technologes can begn. The dgtalzaton of the electrc grd may enable remote attacks to grow rapdly, potentally spannng countres or even contnents (McDanel & McLaughln, 2009). Moreover, t s rapdly becomng easer to compromse computer systems due to the ncreased avalablty of hacker tools on the Internet and the decrease n techncal knowledge requred to use them to mpose sgnfcant damage (Kropp, 2006). Whle dgtalzaton of the electrc grd wll present many new securty challenges, t wll also provde the grd wth ncreased flexblty to prevent and wthstand potental threats. In ths paper, an ntellgent dstrbuted secure control archtecture s presented for dstrbuton systems to provde greater adaptve protecton, wth the ablty to proactvely reconfgure, and rapdly respond to dsturbances. Secton 2 provdes an overvew of dstrbuton automaton systems (DAS), Secton ncludes detaled descrptons of the functonaltes at each layer of the archtecture as well as the Ths work was supported by the Natonal Scence Foundaton under grant number ! whole system, and Secton 4 formulates the dstrbuton system reconfguraton problem. Fnally, Secton 5 smulates and analyzes performance results for the IEEE 2 node test feeder, and Secton 6 states some conclusons. 2. DISTRIBUTION AUTOMATION SYSTEMS Due to ts sze, complexty, and cost, the transformaton of the exstng electrcal grd to a smart self healng system wll need to occur n several stages over tme. Snce almost 90% of all power outages and dsturbances have ther roots n the dstrbuton network, the transformaton must begn at the dstrbuton level (Farhang, 200) where customers wll see the greatest ncrease n performance. In the Unted States, ntal nvestments n smart grd technologes have hghlghted ths fact. Of the $.4 bllon awarded by the Amercan Recovery and Renvestment Act (ARRA) Smart Grd Investment Grants (SGIGs), announced n October 2009, only $48 mllon went to transmsson related projects (Horowtz et al., 200). Nearly all the rest went to dstrbuton related projects. A frst step n the transformaton wll be n the development and wde scale deployment of DAS. Currently, only a small mnorty of dstrbuton systems worldwde are equpped wth such capabltes. Even n North Amerca, home of one of the world s most advanced power systems, less than a quarter of the dstrbuton system s equpped wth nformaton and communcatons systems, and only about 5% to 20% of the system at the feeder level. As a result, many utltes beleve that ntally nvestng n dstrbuton automaton wll provde them wth ncreasng capabltes over tme.

2 DAS are equpped wth nformaton and communcatons systems to provde system dspatchers wth support for dayto day operatons. Accordng to Bassett et al. (988), common functons nclude: " Automatc bus sectonalzng " Feeder deployment swtchng and automatc sectonalzng " Integrated volt/var control " Substaton transformer load balancng " Feeder load balancng " Remote meterng " Load control.. INTELLIGENT DISTRIBUTED SECURE CONTROL The control of DAS can be ether centralzed or decentralzed. In centralzed control, all computng and control functons are based n one centralzed locaton, whle n decentralzed control computng and control functons may be dspersed n many dfferent locatons. Centralzed control, whle easer to mplement than decentralzed control, s unable to respond quckly to adverse events at centralzed control ponts. Decentralzed control s able to respond qucker to adverse events, but the lack of nformaton exchange may lead to unrelable or based decson makng. For deeper and layered protecton, an ntellgent dstrbuted secure control s requred, whch would enable parts of the network to reman operatonal and even automatcally reconfgure n the event of local falures or threats of falure. Wth dstrbuted ntellgence and components actng as ndependent agents, those n each solated area would have the ablty to reorganze themselves and make effcent use of whatever local resources reman to them n ways consonant wth the establshed global goals to mnmze adverse mpacts on the overall network. Local controllers would then be able to gude the solated areas to operate ndependently whle preparng them to rejon the network, wthout creatng unacceptable local condtons ether durng or after the transton. Numerous sources n the lterature proclam how future dstrbuton systems wll employ control and communcatons technologes to acheve such goals. They dscuss numerous operatng capabltes such as how the swtches wll communcate wth each other and, usng preset condtons, or even artfcal ntellgence, wll operate wthout human nterventon (Bouford & Warren, 2007). The lterature, however, provdes few descrptons or models of how such objectves wll be acheved and no analyss of the effects such technology wll have on system operatons. capabltes (Electrc Power Research Insttute, 2002). Several concepts central to the SPID system were utlzed n the desgn. A dagram of the resultng control archtecture s shown n Fg.. Fg.. Intellgent dstrbuted secure dstrbuton system control archtecture (adapted from Electrc Power Research Insttute (2002)). The model utlzes three layers composed of numerous ndependent, ntellgent agents. A thorough descrpton of what ntellgent agents are and how they operate s provded n Amn & Ballard (2000). The agents gather and exchange nformaton wth each other n real tme or near real tme n order to provde coordnated protecton and to optmze system performance. A dagram of example control functons and sgnals beng sent between dfferent agents at each layer of the control archtecture s shown n Fg. 2. In the dagram, each block represents the control functons for the agents at that layer, wth the bottom block representng the reactve layer, the mddle block representng the coordnaton layer, and the top block representng the delberatve layer.. Archtecture To acheve the desred goals stated above for dstrbuton systems, an ntellgent dstrbuted secure control archtecture was developed. The model was based upon the Strategc Power Infrastructure Defense (SPID) system control archtecture produced by the EPRI/DOD Complex Interactve Networks/Systems Intatve (CIN/SI) for systems wth ntellgent wde area sensng, protecton, and reconfguraton Fg. 2. Dstrbuton system ntellgent agent control functons and sgnals..2 Reactve Layer At the lowest control level, the reactve layer s composed of agents located at each smart meter, substaton, and dstrbuted energy resource n the system. The agents gather and

3 exchange nformaton wth adjacent coordnaton level agents. They respond to ncomng prce sgnals and alarms by performng demand response and load management functons, such as sheddng load or shftng load to lower prce tmes, and connectng or dsconnectng load from the dstrbuton system n response to attacks or natural dsasters. In return, load measurements and alarm sgnals are sent back up to the coordnaton layer.. Coordnaton Layer The coordnaton layer s composed of agents located at each te lne or swtch n the system. The agents exchange nformaton wth each other as well as forward sgnals sent by the reactve and delberatve layer agents to ther approprate destnatons. They make decsons regardng ther connecton status, and take quck acton f faults or attacks are detected. They have the ablty to recognze f they are slanded from the rest of the system and to utlze whatever local resources are avalable to them. In addton, they mplement optmal system confguratons as determned by the delberatve layer agents..4 Delberatve Layer Fnally, the delberatve layer s composed of agents located at the mcrogrd or feeder system level. The agents gather and exchange nformaton wth adjacent coordnaton layer agents and determne the overall system objectves such as ncreased network relablty or mnmzed lne losses. They also determne the optmal system confguraton for each sland n ther system based on the chosen system objectves and send these control sgnals down to the coordnaton layer agents for mplementaton. Furthermore, they perform analyss on ther systems to determne f all operatng constrants are met and aggregate system load n order to submt bds nto realtme electrcty markets at the transmsson system level. 4. DISTRIBUTION SYSTEM RECONFIGURATION Dstrbuton system reconfguraton s one of the most mportant tasks of DAS. The objectve s to determne the status of swtches on the network n order to optmze system performance. The types of swtches nclude both sectonalzng swtches (normally closed) and te swtches (normally open), and large feeder systems can contan several hundred swtches. Normally, dstrbuton system reconfguraton s performed for the followng reasons: ) to reduce lne losses, 2) to allevate network overloads, ) to restore servce to as many customers as possble followng a fault on the system, or 4) to ncrease network relablty, wth the majorty of past work focusng on mnmzng lne losses (Ahuja et al., 2007), (Jazeb et al., 2008), (Karthkeyan et al., 2008), (Shrmohammad & Hong, 989). Dstrbuted ntellgent secure control, however, provdes the grd wth the ablty to dynamcally optmze ts confguraton n the event of local falures or the threat of falures. Thus, new objectves can be developed for dstrbuton system reconfguraton to take nto account current system condtons. 4. Objectve Functon A smple objectve functon s proposed to mnmze the expected mpact of cyber and physcal dsturbances on a system by takng nto account ts relablty, and the avalablty of ts sensng, communcatons, and control systems. The latter s accounted for usng the avalablty of each ntellgent agent, whch s contnually changng due to the effects of cyber and physcal attacks. The avalablty of each agent s calculated by fndng the percentage of tme t s operatng n the up state over the total tme beng measured. Such values can be determned from operatng records for each ndvdual agent. The expected mpact of dsturbances on a system s measured by computng the loss of energy expectaton (LOEE) for a gven confguraton. The LOEE s calculated by fndng the sum of the probabltes that the path from each load to ts source s unavalable weghted by ts load. The probablty of each path beng unavalable s one mnus the product of the relabltes and avalabltes of each lne and ntellgent agent respectvely encountered n the path. The LOEE and the path avalablty calculatons are shown n () and (2) respectvely. LOEE ( F # N, Pload, T $ ( + # ) Ns $ * Pload * T () % & B, ' s,, (2) N ( R * A s jk l % j k& L % l& D, s s Where:! system LOEE " set of all buses # $%&' set of all real power loads # $%&'( real power load at bus ) * set of all path avalablty probabltes * +, path avalablty probablty from bus ) to source bus -. length of tme perod beng measured / +, set of all lnes n the path from bus ) to source bus relablty of the lne from bus to bus 4 5 +, set of all ntellgent agents encountered n the path from bus ) to source bus - 6 $ avalablty of ntellgent agent Problem Formulaton Constrants are added to the problem to ensure that all operatng condtons are met. To mantan standard utlty operatng practces, the system s requred to reman radally connected, and all scheduled loads must be served f possble. Radal system confguratons are characterzed by havng a set of seres components between a substaton and each load pont. Such confguratons account for over 99% of all dstrbuton systems n North Amerca (Wlls, 997). The resultng formulaton for the dstrbuton system reconfguraton problem s shown n () (8). st.. # $ mn F N, P, T () load - * B. N # Pgen ) Pload $ ) Real 4V 2 YkV k 5( 0 #% & B$ / 0 2+ (4) 6 k ( 7 8 9

4 - * B. N # Qgen ) Qload $ ) Imag 4V 2 YkV k 5( 0 #% & B$ mn / 0 2+ (5) 6 k ( # $ #, $ V : V % & B (6) j max j S : S % j & L (7) System s radally connected (8) Where: # 89:( real power generaton at bus ) ; 89:( reactve power generaton at bus ) ; $%&'( reactve power load at bus ) " < number of buses = +2 ), 4 term of the bus admttance matrx > + complex voltage at bus ) > + voltage magntude at bus ) > +?+: mnmum voltage magntude lmt at bus + complex power flow on lne from bus ) to +?&A maxmum complex power flow lmt on lne from bus ) to bus / set of all lnes Constrant (4) represents the real power equalty constrants, (5) the reactve power equalty constrants, (6) the bus voltage magntude lmts, and (7) the lne complex power flow lmts. To mplement constrants (6) and (7), they are added as penalty factors to the objectve functon. The penalty factor formulatons used for the bus voltage magntude lmts and the lne complex power flow lmts are shown n (9) and (0) respectvely. ; mn V ) V < voltage mag. lmts penalty factor ( max =,0> (9) + mn & B = V ; max Sj ) S < j lne flow lmts penalty factor ( + max =,0> (0) max, j& L =? Sj 4. Dstrbuton system reconfguraton represents a dscrete optmzaton problem. Snce typcal dstrbuton systems can nclude hundreds of swtches, exhaustve enumeraton of all possble combnatons (2 : ) would quckly become computatonally nfeasble. Ths problem s further complcated by the addton of constrants such as those shown n (4) (8). Nevertheless, the performance of the optmzaton, and ultmately the potental cost savngs, ncreases dramatcally as the number of swtches ncreases (McDermott et al., 999). Thus, the analyss technque must be able to handle large systems. Both genetc algorthms (Jazeb et al., 2008) and smulated annealng have been appled to smlar dscrete optmzaton problems, but they encounter dffculty wth the radal structure of dstrbuton systems. Brown (200) states two reasons for ths beng that: ) most of the generated swtch poston combnatons wll not represent feasble solutons, and 2) generatng a new radal tree structure for each combnaton of swtch postons s computatonally ntensve. In addton, branch and bound methods have been attempted by Shrmohammad and Hong (989), but they provde no assurance that convergence wll be reached, and for the cases where convergence s reached the computatonal burden to solve them s extremely hgh. To determne the optmal radal confguraton for the problem formulaton descrbed n Secton 4.2, annealed local search (ALS) was used. ALS takes advantage of the radal structure of dstrbuton systems and overcomes the shortcomngs encountered by genetc algorthms and smulated annealng. The algorthm used was based on the one descrbed by Brown (200) for dstrbuton system relablty optmzaton problems, whch had been successfully appled to topologcally dverse systems wth up to 45 swtches. To adapt the algorthm for the current problem, a few modfcatons were made. The orgnal algorthm made use of the te swtch shft, where a normally open swtch was closed and a nearby upstream swtch was opened, to make ncremental changes to the radal system structure. However, to ensure that all feasble swtch combnatons were searched utlzng ntellgent dstrbuted secure control, search tables comprsng the swtches currently n the closed poston, and those avalable n the opened poston were generated. Each of the feasble swtch combnatons was then searched. The resultng method was found to be effcent, and straghtforward to mplement. 5. NUMERICAL CASE STUDY To nvestgate the performance of the ntellgent dstrbuted secure control archtecture descrbed n Secton wth ALS descrbed n Secton 4., the IEEE 2 node test feeder was smulated usng MATLAB. Results were compared to those obtaned usng the sequental swtch openng (SSO) method, a prevalent mnmum loss reconfguraton algorthm for normal operatng condtons, and decentralzed and centralzed control archtectures as shown n Fg. and Fg. 4 respectvely. The SSO method s descrbed by Shrmohammad & Hong (989). Fg.. Decentralzed dstrbuton system control archtecture. Fg. 4. Centralzed dstrbuton system control archtecture. The decentralzed control archtecture does not utlze delberatve layer agents for centralzed decson makng. Thus, coordnaton and reactve layer agents utlze only local nformaton and reconfguraton s performed usng preprogrammed swtchng prortes. For the smulatons, swtches were prortzed by ) beng n the mnmum loss

5 confguraton and 2) from node. In contrast, the centralzed control archtecture does not utlze coordnaton layer agents for dstrbuted decson makng. Thus, reconfguraton capabltes are not avalable n real tme, and can only be mplemented wth advanced plannng, such as for mantenance or planned outages. 5. Test Case A one lne dagram of the IEEE 2 node test feeder s shown n Fg. 5, and key system characterstcs are lsted n Table. Data for the IEEE 2 node test feeder and other test feeder cases are avalable from Kerstng (200) Fg. 5. IEEE 2 node test feeder one lne dagram (Kerstng, 200). Table. IEEE 2 Node Test Feeder Key System Characterstcs Value Substatons 4 Swtches 2 Lnes 8 Load (4E) Base Voltage (4>) 4.6 Base Complex Power (D>6) 0 It was assumed that all elements were balanced n both mpedances and loadngs, whch has tradtonally been chosen as the best compromse between avalable resources and requred results for such an analyss (Wlls, 997). Each lne was set to have a relablty of 97%, each te lne/swtch was set to have a relablty of 00%, and the ntal avalablty of each ntellgent agent was set to 00%. The mnmum bus voltage magntude for each bus was set to 0.94 BC, and the maxmum complex power flow for each lne was set to 2,4964>6 based on the standard practce by electrc utltes of desgnng man feeder lnes wth an emergency ratng of 6006 (Short, 2004). 5.2 Customer Load Model Each customer was modeled to have the load demand curve shown n Fg. 6, whch s dvded nto three levels. The lowest level represents load that a customer absolutely requres n order to mantan basc lvng functons or crtcal busness operatons. It s assumed that ths type of load comprses one tenth of a customer s total electrcty demand, and t s served regardless of electrcty prce. The next level represents nondscretonary load. It ncludes load that s necessary for a customer to mantan hs or her basc qualty of lfe or normal busness operatons, but t can Fg. 6. Customer load demand curve. be done wthout for short perods or n the event of an emergency. It s assumed that ths type of load comprses four tenths of a customer s electrcty demand, and t s served as long as the electrcty prce s below some upper prce lmt, whch s the same for all customers. The last level represents dscretonary or supplemental types of load that can be scheduled n advance or are unnecessary to mantan one s basc qualty of lfe or normal busness operatons. Ths type of load s assumed to comprse one half of a customer s electrcty demand and t s served only f the electrcty prce s below one s wllngness to pay (WTP). The WTP for each customer was randomly generated from a unform probablty dstrbuton n the range [0,00] $/DEF and remaned constant throughout the smulatons. 5. Smart Meter Agents Each smart meter agent s desgned wth demand response capabltes to shft dscretonary and supplemental load from perods when the electrcty prce s above ts owner s WTP or s unavalable to perods when the electrcty prce s below ts owner s WTP and servce s avalable. Furthermore, several protectve measures were bult nto each smart meter agent to combat key threats to the smart grd as descrbed by Wnkler (2009). To prevent abnormal loads from overburdenng the system, each smart meter caps ts owner s load demand durng each hour to three tmes ts average peak load based on past data. If an owner s ntal load demand for one hour s below ths lmt, then addtonal load may be shfted to that hour untl the lmt s reached, as long as the electrcty prce s below hs or her WTP. To prevent brownouts from occurrng, each smart meter s programmed to serve only necessary or Must Have load as shown n Fg. 6 when the prce of electrcty rses above some predefned upper lmt set by the local electrc utlty or publc utltes commsson. The above actons help prevent adversares from compromsng the system and ultmately undermnng consumer confdence. 5.4 Smulatons Smulatons were executed for a length of,68 hours, and the electrcty prce and load demand curve data were

6 obtaned from the Mdwest Independent Transmsson System Operator (MISO) (Mdwest ISO, 200). The electrcty prces used were the real tme market clearng prces (MCPs) for each hour durng the perod from July 6, 2009 August, 2009, and ranged from.79 $/DEF to $/DEF. An electrcty prce of 75 $/DEF was set as the upper prce lmt as shown n Fg. 6. The load demand curve for each customer was generated usng the MISO actual load curve from July 6, 2009 August, 2009 scaled to the value of each customer s peak load. The smart meters were enabled to shft dscretonary or supplemental load as descrbed n Secton 5., and also to serve all dscretonary load n the frst avalable perod regardless of prce, as s the case n conventonal dstrbuton system operatons. Furthermore, random cyber attacks were enabled to occur for each reactve layer agent durng each hour wth a probablty of 20%, and for each coordnaton layer agent durng each hour wth a probablty of 0%. The probablty of a reactve layer agent cyber attack was set to be greater than that of a coordnaton layer agent cyber attack because of ther larger numbers and ther ncreased vulnerablty due to ther tendency to be located n less secure areas. Because of the crtcal functons provded by the delberatve layer ntellgent agent, whch must be secured to ensure 00% uptme, t was assumed that t was protected to wthstand all cyber attacks. A successful cyber attack on a coordnaton layer agent or a reactve layer substaton agent was assumed to mmoblze the agent for the hour durng whch the attack occurred, whle a successful cyber attack on a reactve layer smart meter agent was assumed to trgger the agent nto emergency operaton mode where only crtcal load s served as shown n Fg. 6 for the hour durng whch the attack occurred. In addton, lne falures were enabled to occur, and a lne falure was assumed to remove the lne from operaton for the hour durng whch the falure occurred. In order to account for the varance n each smulaton due to random cyber attacks and lne falures, ten trals were performed, and the mean results were used for analyss. In addton, all smulatons utlzed ntal confguratons wth mnmum lne losses. 5.5 Results The smulaton results are shown below. Fg. 7 shows the change n the LOEE for the dfferent algorthms and control archtectures smulated, Fg. 8 shows the change n lne losses, Fg. 9 shows the cumulatve sum of the voltage volatons, and Fg. 0 shows the cumulatve sum of the lne flow volatons. In addton, Table 2 shows the average dscretonary, nondscretonary, and total energy costs wth system reconfguraton, wth system reconfguraton wthout optmzaton, wthout system reconfguraton, and wth and wthout demand response capabltes enabled. The ntellgent dstrbuted secure control archtecture utlzes system reconfguraton capabltes, the decentralzed control archtecture utlzes system reconfguraton capabltes wthout optmzaton, and the centralzed control archtecture does not utlze any system reconfguraton capabltes. Voltage Volatons (p.u. V) Lne Flow Volatons (kva) Loss of Energy Expectaton (MWh) Lne Losses (MWh) Dscusson Fg. 8. Lne losses comparson. Loss of Energy Expectaton Lne Losses Voltage Volatons Fg. 9. Voltage volatons comparson. Lne Flow Volatons Fg. 0 Lne flow volatons comparson. Fg. 7. Loss of energy expectaton comparson. Table 2. Average Energy Costs Seq. Swtch Openng Method Decentralzed Centralzed Seq. Swtch Openng Method Decentralzed Centralzed Seq. Swtch Openng Method Decentralzed Centralzed Seq. Swtch Openng Method Decentralzed Centralzed ($/GHI) w/dr w/o DR Dscretonary w/reconfguraton w/reconfguraton (w/o opt.) w/o Reconfguraton Nondscretonary w/reconfguraton.4.46 w/reconfguraton (w/o opt.) w/o Reconfguraton.9.47 Total w/reconfguraton w/reconfguraton (w/o opt.) w/o Reconfguraton Fg. 7 shows that the ntellgent dstrbuted secure control archtecture usng both the SSO and ALS methods and the decentralzed control archtecture greatly mproved the relablty and avalablty of the test system compared to the centralzed control archtecture due to the advanced reconfguraton capabltes enabled. Because of these

7 capabltes, however, Fg. 8 and Fg. 9 show that the amount of lne losses and the severty of voltage volatons greatly ncreased. Lne flow volatons dd not have any measured effect on system operatons except for the cases where the centralzed control archtecture was used as shown n Fg. 0. Comparng the ALS method to the SSO method, t can be seen from Fg. 7 Fg. 9 that whle the ALS method further mproved the relablty and avalablty of the system, and decreased the severty of voltage volatons, t resulted n an ncrease n lne losses. Table 2 shows that the use of demand response sgnfcantly decreased the average cost of dscretonary energy served, and Fg. 7 and Fg. 8 show that t decreased or had no apprecable effect on the LOEE and lne losses respectvely for all algorthms and control archtectures smulated. However, t resulted n an ncrease n the severty of voltage volatons for the ntellgent dstrbuted secure control archtecture usng both the SSO and ALS methods, and the decentralzed control archtecture as shown n Fg. 9. It also ncreased the severty of lne flow volatons for the centralzed control archtecture as shown n Fg. 0. Thus, for the test system, the prce for mnmzng cyber and physcal dsturbances usng the ntellgent dstrbuted secure control archtecture s ncreased lne losses and voltage volatons. To determne the most benefcal control archtectures and algorthms to mplement, the benefts from mproved performance must be balanced aganst the operatonal costs. 6. CONCLUSION In order to enhance the relablty, robustness, effcency, and securty of the power grd to meets the needs of today s dgtal socety and those of the future, the end to end electrc nfrastructure must effectvely utlze sensng, nformaton, and control systems technologes to contnually optmze system performance. The contrbutons of ths paper can be summarzed as follows: ) An ntellgent dstrbuted secure control archtecture s presented for dstrbuton systems to provde greater adaptve protecton, wth the ablty to proactvely reconfgure, and rapdly respond to dsturbances. 2) Ths work represents a novel approach toward developng an analytcal and mult doman methodology to assess the effects of smart grd technologes on dstrbuton system operatons and performance. ) The model ntegrates aspects of cyber physcal securty, dynamc prce and demand response, sensng, communcatons, and dynamc optmzaton and reconfguraton. 4) Smulaton results show the trade offs between system relablty, operatonal constrants, and costs nvolved. Future work wll be focused on further mprovng the capabltes of the model, and smulatng and comparng the effects of addtonal control archtectures and technologes on dstrbuton system operatons and performance. REFERENCES Ahuja, A., Das, S., and Pahwa, A. (2007). An AIS ACO hybrd approach for mult objectve dstrbuton system reconfguraton. IEEE Trans. on Power Systems, 22(), pp.0. Amn, M. and Ballard, D. (2000). Defnng new markets for ntellgent agents. IT Pro, July/August, pp Bassett, D.S. et al. (988). Dstrbuton automaton and the utlty system. In Dstrbuton Automaton, pp. 6. IEEE Press, Pscataway, NJ. Brown, R.E. (200). Dstrbuton relablty assessment and reconfguraton optmzaton. In IEEE/PES T&D Conference and Exposton, 2, pp Bouford, J.D. and Warren, C.A. (2007). Many states of dstrbuton. IEEE Power & Energy Magazne, 5(4), pp Electrc Power Research Insttute (2002). Complex Interactve Networks/Systems Intatve: Fnal Summary Report: Overvew and Summary Report for Jont EPRI and U.S. Department of Defense Unversty Research Intatve. EPRI, Palo Alto, CA. Farhang, H. (200). The Path of the Smart Grd. IEEE Power & Energy Magazne, 8(), pp Horowtz, S.H., Phadke, A.G., and Renz, B.A. (200). The future of power transmsson. IEEE Power & Energy Magazne, 8(2), pp Jazeb, S., Hossenan, S.H., Pooyan, M., and Vahd, B. (2008). Performance comparson of GA and DEA n solvng dstrbuton system reconfguraton problem. In th Internatonal Conference on Optmzaton of Electrcal and Electronc Equpment, pp Karthkeyan, S.P. et al. (2008). Assessment of dstrbuton system feeder and ts reconfguraton usng fuzzy adaptve evolutonary computng. In Annual IEEE Inda Conference, pp Kerstng, W.H. (200). Radal dstrbuton test feeders. In IEEE Power Engneerng Socety Wnter Meetng, 2, pp Kropp, T. (2006). System threats and vulnerabltes. IEEE Power & Energy Magazne, 4(2), pp McDanel, P. and McLaughln, S. (2009). Securty and prvacy challenges n the smart grd. IEEE Securty and Prvacy, 7(), pp McDermott, T.E., Drezga, I. and Broadwater, R.P. (999). A heurstc nonlnear constructve method for dstrbuton system reconfguraton. IEEE Trans. on Power Systems, 4(2), pp Mdwest ISO (200). Mdwest ISO Documents. [Onlne] Avalable at: Shrmohammad, D. and Hong, H.W. (989). Reconfguraton of electrc dstrbuton networks for resstve lne losses reducton. IEEE Trans. on Power Delvery, 4(2), pp Short, T.A. (2004). Electrc Power Dstrbuton Handbook. CRC Press, New York. Wlls, H.L. (997). Power Dstrbuton Plannng Reference Book. Marcel Dekker, Inc., New York. Wnkler, I. (2009). Opnon: The hackablty of the smart grd. Computerworld, December.