1 Applcaton of Mult-Agents for Fault Detecton and Reconfguraton of Power Dstrbuton Systems K. Nareshkumar, Member, IEEE, M. A. Choudhry, Senor Member, IEEE, J. La, A. Felach, Senor Member, IEEE Abstract--The electrc power system has become a very complcated network at present because of re-structurng and the penetraton of dstrbuted generaton and storage. A sngle fault can lead to massve cascadng effects affectng power supply and power qualty. An overall systematc soluton for these ssues could be obtaned by an artfcal ntellgent mechansm called the mult-agent system. Ths paper presents a mult-agent system model for fault detecton and reconfguraton based on graph theory and mathematcal programmng. The mult-agent models are smulated n Java Agent Development Framework and Matlab and are appled to a power system model desgned n the commercal software, the Dstrbuted Engneerng Workstaton. The crcut that s used to model the power dstrbuton system s a smplfed model of the Crcut of the Future, developed by Southern Calforna Edson. Possble fault cases were tested and a few crtcal test scenaros are presented n ths paper. The results obtaned were promsng and were as expected. Index Terms Mult-agent System, Fault Detecton, Power System Reconfguraton T I. INTRODUCTION HE electrc power system has become a very complcated network at present because of restructurng and the penetraton of dstrbuted energy resources. In addton, due to ncreasng demand for power, ssues such as transmsson congeston have made the power system stressed. A sngle fault can lead to massve cascadng effects, affectng the power supply and power qualty. The massve 2003 Northeast Blackout s a very good example for ths type of falure. An overall soluton for these ssues can be obtaned by a new artfcal ntellgent mechansm called the multagent system. An agent s defned as an autonomous computatonal entty such as a software program that can be vewed as percevng Ths work was supported n part by grants from the US DEPSCoR/ONR grant No. N00014-03-1-0660 and the US DoE grant No. DE-FC26-06NT42793. ts envronment through sensors and actng upon ths envronment through ts effectors [1]. A mult-agent system s a collecton of agents, whch senses the envronmental changes and acts dlgently on the envronment n order to acheve ts objectves. Due to the ncreasng speed and decreasng cost n communcaton and computaton of complex matrces, mult-agent system promse to be a vable soluton for today s ntrnsc network problems. Conventonally, many applcatons n power system were solved by human actons by obtanng nformaton from the system. These same actons when operated by agents, can gve tmely and more relable decsons wth less human nterventon. In addton, agents can learn from ther prevous experences and act accordngly durng dfferent fault stuatons. Mult agent system have been appled to several areas n power systems, such as reconfguraton and restoraton, fault detecton, protecton coordnaton, voltage stablty control, reactve power control, electrcty market prcng etc. These theoretcal based researches are beng appled, at present, to terrestral power systems to test the relablty of MAS. Recently more research work based on complcated applcatons n the real world s beng carred out. In ther paper, Cartes and Srvastava analyze the potental of agent applcaton and ther future n power ndustry [2]. They present a SWOT Analyss (Strengths, Weaknesses, Opportuntes and Threats) framework for agent based applcatons n the area of power systems. Moreover, several authors have addressed the ssue of fault dagnoss and reconfguraton specfcally. The paper by L Lu et al. [3] descrbes the need for fault detecton n naval shpboard power system. The paper dscusses the practcal fault detecton and dagnoss problem along wth prognostcs from control engneerng perspectve. In ther fault dagnostc framework, the task has been defned as to constantly montor the process and from the avalable observatons, to dentfy an ndcaton to decde whether there s a fault or not, and to dentfy the fault locaton. 978-1-4244-4241-6/09/$25.00 2009 IEEE
2 Dfferent approaches are followed for dynamc fault detecton process, such as modelng and estmaton method [4], [5], usng analytcal redundancy [6], [7], factorzaton approaches [8], etc. More recently, Huang et al. [9] have presented a mult agent approach for fault detecton. The fault s detected by usng nonlnear parameter dentfcaton technques. Once the fault s detected, the dagnoss agent makes further decsons on the fault mode, fault locaton and fault severty. Ths study has been performed by usng a partcle swarm optmzaton approach. Many authors have conducted research n the area of reconfguraton and restoraton of power systems, prmarly for shpboard power systems. Solank et al. [10], [11], have presented dfferent approaches for restoraton of power system by dstrbuted reconfguraton. In ths paper, a new approach based on graph theory and mathematcal programmng s presented for fault detecton and reconfguraton of power dstrbuton systems. Secton II wll portray the ssue analyss and how t s addressed. Secton III wll dscuss the mathematcal model and fnally, Secton IV wll present the results of the smulatons. II. PROBLEM STATEMENT Ths paper wll focus on the applcaton of MAS for fault detecton, reconfguraton and restoraton. The study s based on a proto-type crcut, the Crcut of the Future (CoF) developed by Southern Calforna Edson (SCE). The algorthm ensures that the mult-agents, whch are nstalled at all the nodes, sources, loads and swtches, wll communcate and co-ordnate wth ther neghborng agents, n order to provde a relable power supply. Ths s acheved by re-routng the power flow when there s a lne fault, and supplyng the hgh prorty loads, when there s a shortage of power supply and also swtchng on the Dstrbuted Generaton (DG) when the need arses. The crcut s modeled usng the Dstrbuted Engneerng Workstaton (DEW), a power dstrbuton software developed by Electrcal Dstrbuton Desgn Inc. The software used to desgn the MAS s the Java Agent Development (JADE) framework and Matlab. Fault detecton algorthm s developed by JADE and reconfguraton algorthm s mplemented n Matlab agent model to allow condtonal swtchng. Dfferent fault scenaros were studed and the crcut was analyzed for better performance. III. MATHEMATICAL MODEL An agent applcaton has to be frst modeled mathematcally, whch can be then translated to software packages for mplementaton. Thereby, ths secton presents a new algorthm for fault detecton and reconfguraton, based on graph theory. A. Graph Theory Graph Theory s a study of graphs and mathematcal structures that can represent any network applcatons. Graph s a context, refers to a collecton of vertces or nodes and a collecton of edges or arcs that connect pars of vertces. Many practcal applcatons, whch can be represented as networks, can be modeled based on graph theory. Therefore, a power system, whch s a very complcated network, s modeled as a graph n ths work. The power system network s modeled as a graph, G, wth a sngle root, as shown n Fgure 1. A spannng tree, T of G represents a radal power dstrbuton feeder. Each node of the graph represents a power source, a node or a load and each edge of T represents a physcal connecton of dfferent nodes through dstrbuton lnes. The edges n G, but not n T, represent swtches, whch are normally n an open state. Notatons: G - graph, represents the power system network T - spannng tree, represents a radal feeder of the power system N - set of nodes n the network that s modeled as a graph G, represents source, buses, swtches or loads E - set of drected edges, usually called arcs of G, represents the dstrbuton lnes s - power source, represents a substaton or a DG S(E)- set of edges n G, but not n T, represents the set of swtches n the system An agent s placed at every node and every swtch. Fg. 1. Dagram of the notaton of a Graph In the above graph, G, each branch, s n1 n2 n, 3 s n4 n5 n, 6 s n7 n8 n s a tree. The edges 9 whch are n the graph, but not n the trees are the swtches, S1-S6. Orentaton of T: The tree, T, s orented n such a way that every node other than the source node (root) has n-degree 1. That s, every node n n T can be reached by a drected path n T from the source s to N.
3 Presettng: For each edge e E(T ), let S(e) = {X:X S, such that (T-e) X s a connected graph}. B. Fault Detecton Algorthm The frst step towards the reconfguraton s to dentfy the fault locaton precsely. The fault detecton algorthm constantly montors the power flowng nto each load and follows the algorthm descrbed below to dentfy the fault locaton. Let us assume that the fault locaton detected by the node agents s at n f. Suppose a node agent n dentfes that the power flowng nto ts node s zero, then on the unque drected path from s to n, say s n n n... n, 1 2 3 k where n=n k, n k wll request n k-1, whether t has power. Subsequently, n k-1 wll request n k-2 for ts power. When for some, n does not have power, but n -1 has power, the faulty edge can be dentfed as, e = n, n ). f ( 1 Once the fault locaton s dentfed, the edge e f s solated and agents that are controllng each swtch X S( e f ) wll communcate and coordnate wth each other to decde on the partcular swtch or swtches that have to be swtched on, so that ( T e f ) U X wll contnue to be a fully suppled network. The mxed programmng algorthm for reconfguraton through selecton of proper swtchng s gven n the next secton. C. Reconfguraton & Restoraton Algorthm It s vtal to restore the power supply promptly by reroutng the power flow through a target confguraton, when the power supply s nterrupted by a fault. The problem of obtanng a target system s referred to as power system restoraton [12]. The fault locaton, e f, has been found from the prevous algorthm and a collecton of possble swtches, S ( e f ) have been provded. Thereby, the graph G( e f ) s the graph obtaned by ncorporatng T e and all the edges as a f member n the collecton S ( e f ). Ths algorthm assumes that each edge n the tree can be solated n order to reconfgure the system. Let, P be the total amount of avalable power from the s source n the system. O = {(, j) E : j N}, out flow from node and, I = {( j, ) E : j N}, n flow to node. For an orented edge, e, let P denote the amount of power e flowng through e, and let Y e denote the state ndcator varable: 1 Y e = 0 f e s closed f e s open It s prudent to assume that for each edge e ( E( T ) { e f }), Ye = 1. Therefore, the only fact that has to be decded s whch of the swtches wll be closed and whch should reman open. Objectve Functon: Maxmze W P N L Where, W s the prorty of the load, gven as a weght, and P L the suppled load. N s the number of all load agents. Subject to constrants: P S PL N.e. the total load should not exceed the total capacty of the source Pj P j, max.e. the power flow through any arc or drected edge, e, when t s closed, should not exceed the capacty of the edge. Where P j s the real power flowng through the lne connectng nodes and j and Pj P j, s the maxmum max allowable power through that lne. A node that s not a source node s called an ntermedate node. At a non load ntermedate node, the amount of power flowng n should be the same as the amount flowng out. P e Pe = 0 e I e O At a load node, the amount of power flowng n should be equal to the sum of the amount flowng out and the load at. P P = P e I e e O e L The dstrbuton system s radal,.e., the total number of ncomng edges at any node s at most unty. e 1 k I k For a node, let the load at and, P denote the actve power consumed by L
4 A reconfgured topology s acheved by selectng from the choce of S ( e f ), whch comply wth the objectve functon and the constrants lsted above. IV. SIMULATIONS & RESULTS The new algorthms for fault detecton and reconfguraton have been tested for dfferent scenaros. The fault detecton was smulated n JADE and reconfguraton was tested n Matlab. Both the results were combned wth DEW for power system smulaton and the results were tested for voltage complance. A general model for the applcaton of MAS n the power system engneerng s shown n Fgure 2. The proposed MAS archtecture s llustrated n Fgure 3. It conssts of Load Agents (LAGs), one for each load, and Swtch Agents (SAGs), correspondng to each swtch or dsconnector n the system. Any LAG n the crcut can be assocated wth one or more SAGs. These agents coordnate wth each other n order to detect the faults n the system and re-route the power flow to better serve the customers. total real power load of 24 MW and reactve power load of 12.96 MVar. It has 14 Capactor banks for provdng Var support and two DGs, one provdng real power and the other reactve power. Snce the capactor banks can cater for the entre reactve power loads and losses n the system, only the actve power flow s consdered n the smulatons. For smplcty for smulaton purposes, the orgnal crcut s slghtly modfed by lumpng certan loads wthout affectng all of ts 7 swtchng locatons and the orgnal system topology. The modfed CoF wll have 11 loads, 7 swtches, 18 nodes, a Substaton and a Dstrbuted Generatng Source, whch provdes real power. The dstrbuton system s consdered to be a radal system n all system confguratons. For smulaton purposes of servng the hgh prorty loads, when there s a power shortage, three loads have been chosen to be hgh prorty loads and two loads have been chosen as low prorty loads. The rest of the loads are gven medum prortes. A. JADE Fault Detecton Smulaton JADE Fault detecton smulaton can be ntated by creatng JADE agents from readng a text fle, whch s generated by DEW after runnng the load flow applcaton. Test Case 1: Sngle Lne Fault The system s smulated for a sngle lne fault at the very begnnng of one of the three man feeders. Fg. 2. Generc model of the MAS applcaton n Power Dstrbuton System. When the JADE MAS s appled, the relevant load agents communcate wth each other to fnd out the fault locaton. The message passng s clearly shown n the JADE Snffer agent GUI n Fgure 4. Fg. 4. JADE Message Passng to dentfy the fault locaton for a sngle lne fault. Fg. 3. Archtecture of the Proposed MAS. The smulatons were carred out for a proto-type power dstrbuton system, a smplfed model of the Crcut of the Future (CoF), developed by Southern Calforna Edson (SCE). The CoF modeled n DEW, has a sngle substaton, wth three man feeders, whch are also connected for flexble re-routng of power flow, through 7 swtches/dsconnectors whch are normally open. The crcut has 14 loads demandng All the LAGs wll be created n the partcular feeder. But only those LAGs whch dentfy, that they have no power flowng n, wll send a request message to ts ncomng neghborng agents whether the neghborng agents have ther loads suppled or not and wat for ther reply. Snce, ths partcular feeder has only three loads, all three LAGs, do not have power flowng nto ther loads due to the fault at the begnnng of the feeder. Hence the LAGs whch receve the
5 request message reply to the sender wth ther flow values. Snce all the LAGs have zero flows, the fault locaton can be dentfed accordngly. Test Case 2: Multple Faults For testng a multple fault scenaro, let us assume that there are 3 faults n all three feeders at dfferent locatons. Ths stuaton has been run n JADE fault detecton program and the message passng between the LAGs have been observed to be as shown n the followng fgure. constrants, explaned n the prevous sectons. Fgure 6 shows the reconfgured crcut. It can be seen that, from all the swtchng possbltes, swtch 1 and 5 are operated n order to supply the loads n the faulty feeder. However, snce, ths fault s n one of the three man feeders, the total load of 24 MW, cannot be suppled due to capacty constrants of the other two man feeders. Wthn the avalable capacty, all the hgh and medum prorty loads are suppled and only the least prortzed loads are not suppled fully due to capacty constrants. Out of all the possble lne faults that can occur n the CoF, ths partcular fault, s the least desrable fault, where the total load cannot be suppled wthout overloadng the dstrbuton lnes. In all other lne fault cases, combnaton of swtches can be found for system reconfguraton, wthout curtalng the power supply to any loads. Fg. 5. Snffer Agent GUI for Multple Fault Scenaros. B. Matlab Reconfguraton Smulaton Reconfguraton of the CoF has been mplemented n Matlab usng graph theory. As n the Fault Detecton algorthm, each fault locaton wll be dentfed by the relevant LAG, and wll have correspondng combnaton of SAGs for reconfguraton. Each proposal s reconfgured usng the Matlab Energy Management System [13], to fnd the best soluton for CoF. All the possble fault cases have been tested n the model. Reconfguraton s carred out based on maxmum flow algorthm. The maxmum flow prncple s a branch of graph theory and combnatoral optmzaton. It seeks a feasble soluton that sends the maxmum possble flow from a source to a snk node. In the reconfguraton of energy management system, maxmum flow algorthm represents the power system as graph orentatons [13]. A GUI s developed for easy user nteracton for carryng out smulatons. Four dfferent scenaros can be developed. A lne fault can be created by specfyng the from node and the to node agent names. The source capacty can be decreased to test the shortage of supply from the substaton, by enterng the new capacty. It s also possble to change the capacty of the DG and the prortes of the loads for testng purposes. Multple types of fault scenaros can also be smulated. Test Case 4: Shortage of Source Capacty In the event that the source has a shortage of capacty, the new topology of the system s found n such a way to ensure that wth the avalable source capacty, frst the hgh prorty loads are suppled. In case, there s stll remanng source power avalable, the lower prorty loads are suppled n the prorty order. In ths test case, t s assumed, that the source capacty s reduced to 3 MW, from 24 MW. From the reconfgured network, shown n Fgure 7, t can be observed that 3 MW s not even adequate to supply the hgh prorty loads. Hence, to supply the remanng hgh prorty loads, the DG s swtched on. However, snce the source and DG have a total source of 4 MW, but the hgh prorty loads add up to 6.56 MW, even the total prortzed loads cannot be suppled fully wth both the substaton and the DG sources. C. Algorthm Executon Duraton Both the fault detecton and reconfguraton algorthms are tested for the duraton of operaton. Tme duratons obtaned from the testng of several executons are plotted. Fgure 8 shows the executon tme taken for 100 samples of fault detecton processes. It can be seen that the average tme taken for executon of fault detecton s 22.4 ms. Fgure 9 shows the executon tme taken for reconfguraton for 50 sample executons. The average tme taken for the reconfguraton process s 14.5 seconds. Test Case 3: Lne Fault Assume that a fault s dentfed between agent numbers 2 and 3. The algorthm s run to reconfgure the system n order to acheve the objectve functon and satsfy all the
6 Snk@15 Demand=2420 Pro= 2 Snk@17 Demand=1260Capacty=2520.0 Pro= 1 Capacty=4840.0 Flow =2420.0 Router@16 Capacty=8925.0 Capacty=8925.0 Flow =2420.0 Capacty=5120.0 Router@7 Flow =2560.0 Router@14 SW3 Capacty=5185.0 Flow =2420.0 Snk@13 Demand=2070 Pro= 2 Capacty=8925.0 Flow =2560.0 Capacty=4140.0 Flow =1610.0 Capacty=4260.0 Router@5 Flow =2130.0 SW2 Router@12 Flow =1610.0 Snk@6 Demand=2130 Pro= 2 Capacty=8925.0 Flow =4690.0 Capacty=6620.0 Flow =3310.0 SW1 SW4 Router@18 Snk@8 Demand=2560 Pro= 3 Router@11 Router@3 Snk@4 Demand=3310 Pro= 2 Flow =9610.0 Router@9 Flow =9610.0 Flow =10540.0 Router@10 Router@2 Capacty=24200.0 Flow =21080.0 Source@1 Supply=24200 Capacty=1860.0 Flow =930.0 Capacty=11135.0 Flow =2420.0 SW7 Snk@30 Demand=930 Pro= 2 Snk@28 Demand=2670 Pro= 3 Router@27 SW6 Capacty=5340.0 Flow =2670.0 Capacty=11135.0 Flow =2670.0 Router@19 Flow =10540.0 SW5 Capacty=2660.0 Flow =1330.0 Snk@20 Demand=1330 Pro= 3 Router@29 Capacty=2975.0 Capacty=11135.0 Flow =7870.0 Capacty=16915.0 Flow =7870.0 Router@23 Router@25 Capacty=5000.0 Flow =1450.0 Snk@26 Demand=2670 Pro= 1 Capacty=1000.0 Source@24 Supply=1000 Router@21 Capacty=5000.0 Flow =2670.0 Snk@22 Demand=2670 Pro= 2 Agents called/ Moves made: Agents called/ Moves made: 10050 / 465 Fg. 6. Reconfgured system after the lne fault. Fg. 7. Reconfgured System after a shortage n source capacty.
7 It has to be noted that the average executon tmes calculated are for the proto type crcut smulated n ths work. But n realty, the executon tme wll be hgher than the software smulaton duratons gven n ths secton. Ths s due to the tme taken for nterfacng the software wth the hardware and also due to the actual tme delay n communcaton lnks that are used for message transportaton. 50 45 40 Executon tme for Fault Detecton data 1 y mean Fg. 10: Voltage profle of the CoF under normal operatons. tme (ms) 35 30 25 20 15 0 20 40 60 80 100 Fg. 8. Plot of Executon Tme for Fault Detecton Algorthm. 20 15 Executon Tme for Reconfguraton Fg. 11: Voltage profle of the CoF under the faulty stuaton n Test Case 3. From the above graphs, t s shown that by reconfgurng the system, the system voltages are mantaned wthn ts lmts of ±10% tolerance. tme (sec) 10 The reason for the voltage peaks and dps observed n the graphs above are due to the capactance of the dstrbuton cables n the system. 5 data 1 y mean 0 0 10 20 30 40 50 Fg. 9. Plot of Executon Tme for Reconfguraton Algorthm. D. Voltage Profle It s mportant to test the voltage profle of the reconfgured system to check whether the customer voltages are mantaned wthn the tolerance lmt. Hence the orgnal system and the Test Case 3 were smulated n DEW to analyze the voltage profle n Fgure 10 and Fgure 11 respectvely. V. CONCLUSION Mult-agent system models for fault detecton and reconfguraton applcatons for a proto-type crcut are presented n ths paper. The model was developed based on graph theory. The crcut concerned s the Crcut of the Future, a power dstrbuton system that has three man feeders, several loads and swtches. All possble fault scenaros were tested n both the models. The results obtaned are promsng and t shows a very good start n the drecton of MAS applcaton n power dstrbuton system. The smulatons reveal that the crtcal lne fault that can occur s n the very begnnng of the frst feeder, whch causes 12.24% of the total unsuppled demand. Ths harms the system relablty and can be avoded f the dstrbuton lne capacty can be ncreased.
8 VI. REFERENCES [1] G. Wess, Multagent Systems: A Modern Approach to Dstrbuted Artfcal Intellgence, The MIT Press, 2000 [2] D. A. Cartes and S. K. Srvastava, Agent Applcatons and ther future n Power Industry, IEEE Power Engneerng Socety General Meetng, pp 1 6, June 2007. [3] L. Lu, K. P. Logan, D. A. Cartes and S. K. Srvastava, Fault Detecton, Dagnostcs and Prognostcs: Software Agent Solutons, IEEE Transactons on Vehcular Technology, Vol. 56, No 4, pp 1613 1622, July 2007 [4] R. Isermann, Process fault detecton based on modelng and estmaton methods A survey, Automatca, Vol. 20, No. 4, pp 387 404, July 1984 [5] R. Isermann, Model based fault detecton and dagnoss Status and Applcatons, Annual Rev. Control, Vol. 29, pp 71 85, 2005 [6] E. Chow and A. Wllsky, Analytcal Redundancy and the desgn of robust falure detecton systems, IEEE Transactons on Automatc Control, Vol. 29, No. 7, pp 603 614, July 1984 [7] R. J. Patton and J. Chen, Advances n Fault Dagnoss usng Analytcal Redundancy, IEE Proceedngs of Integrated Operatons Management Control, pp 6/1 6/12, 1993 [8] X. Dng and P. M. Frank, Fault Detecton va Factorzaton Approach, Systems Control Lett., Vol. 14, No. 5, pp 431 436, June 1990 [9] K. Huang, S. K. Srvastava, D. A. Cartes and L. Lu, Agent Solutons for Navy Shpboard Power Systems, IEEE Internatonal Conference on Systems Engneerng, pp 1 6, Aprl 2007 [10] J. M. Solank, N. N. Schulz and W. Gao, Reconfguraton for Restoraton of Power Systems usng Mult-Agent System, Proceedngs of the 37th Annual North Amercan Power Symposum, pp 390 395, October 2005. [11] J. M. Solank, S. Khushalan and N. N. Schulz, A Mult-Agent Soluton to Dstrbuton Systems Restoraton, IEEE Transactons on Power Systems, Vol. 22, No. 3, pp 1026 1034, August 2007 [12] C. Rehtanz, Autonomous Systems and Intellgent Agents n Power System Control and Operaton, Sprnger, 2003 [13] P. Tulpule, Multagent Approach for Power System Reconfguraton, M. S. Thess, West Vrgna Unversty, 2007 VII. BIOGRAPHIES Koushaly Nareshkumar (M 2005) receved her undergraduate degree from Unversty of Moratuwa, Sr Lanka n 2002 and Masters Degree from West Vrgna Unversty n August 2008, n electrcal engneerng. Her employment experence nclude beng a Lghtng Desgn Engneer provdng energy effcent lghtng solutons and as a Transmsson Desgn Engneer at the Sr Lankan Power Utlty, The Ceylon Electrcty Board. Muhammad A. Choudhry receved B.Sc. (EE) from Unversty of Engneerng and Technology, Lahore, Pakstan n 1973. He receved M.S. (EE) from the Unversty of Kansas n 1977 and the Ph.D. degree from Purdue Unversty n 1981. From August 1973 to December 1975, he was Assstant Engneer wth Water and Power Development Authorty n Pakstan. He joned West Vrgna n 1981 and s professor n the Department of Computer Scence and Electrcal Engneerng. Hs areas of nterest are HVDC Systems, System Stablty, Optmal Control, and Power Electroncs. Hong-Jan La receved hs Ph.D. n Mathematcs from Wayne State Unversty n 1988. He s a full Professor of the Department of Mathematcs at West Vrgna Unversty. He has been workng on dscrete mathematcs, algorthm and optmzatons for 15 years. Al Felach (M 83, SM 86) receved the Dplôme d Ingéneur en Electrotechnque from Ecole Natonale Polytechnque of Algers n 1976, and MS (1979) and Ph.D. (1983) n EE from Ga Tech. He s the holder of the Electrc Power Systems Char Poston, and the Drector of the Advanced Power & Electrcty Research Center at West Vrgna Unversty. Hs area of nterest s modelng and control of large scale power systems.