1 The Dynamc Utlzaton of Substaton Measurements to Mantan Power System Observablty Y. Wu, Student Member, IEEE, M. Kezunovc, Fellow, IEEE and T. Kostc, Member, IEEE Abstract-- In a power system State Estmator (SE), a Network Topology Processor (NTP) determnes the Crcut Breaker (CB) status n real-tme to obtan electrcal network topology. In a conventonal NTP, many substaton measurements are smply dscarded because ther postons n the smplfed bus-branch network model are lost. These measurements cannot be used n the network observablty analyss and when some of the used measurements are lost, an estmate of system states may not be obtaned. Ths paper proposes an nnovatve method to utlze these redundant measurements. The new method uses a numercal matrx to represent the physcal connectvty of substaton devces, and then dynamcally searches for solutons to calculate branch and bus njecton power flow measurement data usng the lnear combnaton of the avalable substaton measurement data. Test cases on IEEE-30 bus system verfy that the proposed method s very effectve n makng the network observable, wthout the need to nstall new measurement devces. Index Terms energy management systems, network topology processor, observablty analyss, state estmaton A I. INTRODUCTION State Estmator (SE) s an essental applcaton n the Energy Management Systems (EMS). The state estmator estmates the system states, whch allows other EMS functons such as securty assessment to be relably deployed n order to analyze contngences, as well as to predct correctve actons [1]. An mportant requrement for state estmaton s the network observablty. Due to the frequent change of network topology, many measurements that provde data to the state estmator may no longer be avalable n the network model. The prevously observable system may turn nto unobservable. An advanced real-tme network modelng tool s needed to handle the loss of observablty problems. In ths paper, a new method to enhance the ablty of exstng network topology processor (NTP) s proposed. The method tres to recover the network observablty by dynamcally provdng more Ths work was funded by ABB Corporate Research Center n Baden- Daettwl, Swtzerland (project CRID 30293). Y. Wu and M. Kezunovc are wth the Department of Electrcal and Computer Engneerng, Texas A&M Unversty, College Staton, TX 77843-3128, USA (e-mals: wuyang@ece.tamu.edu, kezunov@ece.tamu.edu). T. Kostc s wth ABB Swtzerland Ltd - Corporate Research n Baden- Daettwl, Swtzerland (emal: tatjana.kostc@ch.abb.com). measurements to the SE. Secton II talks about what an NTP does n the preprocessng of the raw measurement data; secton III s a bref ntroducton to the concept and methods of network observablty analyss; secton IV explans what the proposed method does to effcently use the substaton measurements n real-tme. Tests results are shown n secton V. Conclusons are presented n secton VI. II. CONVENTIONAL NETWORK TOPOLOGY PROCESSING In ths paper, the term connectvty refers to the statc physcal layout of devces (transmsson lnes, bus-bars, swtches, etc.) n a power system network; the term topology refers to the dynamc structure of a network determned upon the status of swtches and crcut breakers (CBs). Connectvty s usually fxed over longer perods of tme whle topology changes relatvely frequently over tme. The term node refers to an electrcal node n the detaled substaton model; the term bus refers to an electrcal node n the bus-branch model after the processng of an NTP. A bus usually conssts of one or several nodes that are connected by closed CBs or swtches. An NTP takes care of the frst stage of data processng n a state estmaton functon. Its task s to determne the network topology (usually n the form of a bus-branch model) based on the detaled descrpton of network connectvty and the realtme CB status [2-3]. In a conventonal NTP, the nput data consst of the followng. A. The descrpton of the connectvty of the physcal devces n the network The physcal devces nclude generators, loads, CBs, transmsson lnes, transformers, current transformers (CTs), voltage transformers (VTs), etc. These devces can be grouped nto four categores based on ther characterstcs that affect the determnaton of network topology: 1) CBs and other swtchng devces: these devces have two termnals a from-node and a to-node. Ther states are ether open or closed. An open CB corresponds to an open crcut, or an nfnte mpedance branch; a closed CB corresponds to a short crcut, or a zero mpedance branch. 2) Nodal njecton devces (generators, loads, etc): these devces have one termnal the node that they are connected to.
2 (a) (b) Fg.1: An example of the detaled substaton model and the used bus-branch model. (a) Detaled substaton model. (b) Used bus-branch model. 3) Transmsson lnes, transformers: these devces are usually represented by non-zero mpedance branches that have two termnals a from-node and a to-node. 4) Measurements: the commonly used measurements nclude CB power flow measurements, nodal power flow njecton measurements and voltage magntude measurements. B. CB Status and analog measurement data The CB status measurement data are provded to the NTP so that t can merge electrcal nodes that are connected by closed CBs nto a sngle bus. After that, the NTP also needs to assgn the nodal njecton devces and branches avalable n the detaled substaton models to the proper locatons n the bus-branch model. The analog measurement data, such as CB power flows, nodal njecton power flows and voltage magntudes also need to be provded to the NTP. These measurement data need to be processed before they can be used by a state estmator. Most state estmators that are avalable n power systems can deal wth three types of measurements n the bus-branch network: bus voltage magntude measurements, bus power flow njecton measurements, and branch power flow measurements. Many of the analog measurement data that are gathered by the physcal devces n the substatons cannot be used drectly n the state estmator, snce the values that they montor do not fall n any of these three categores. These values could be combned and new meanngful measurement values could be calculated. The exstng NTPs use the followng prncples n treatng the raw analog measurement data: 1) A nodal voltage magntude measurement s drectly converted to a bus voltage magntude measurement by mappng the node number to ts correspondng bus number n the bus-branch model. 2) A nodal power flow njecton measurement s converted to ether a branch power flow measurement (f a branch s connected to the node and brngs the njecton) or a porton of a bus njecton power flow measurement (f an njecton devce s connected to the node and brngs the njecton). If a bus s composed of several nodes n the detaled substaton model, a bus njecton measurement s created only f all nodal njecton measurements are avalable. 3) The CB power flow measurements wll be used to calculate the nodal njectons, f possble. The calculated nodal njectons wll be further processed n the way descrbed n 2). An example of how NTP works s shown n Fg. 1. A substaton that has nne CBs and eght nodes s shown n Fg. 1(a). Bus 9, 10 and 11 are from external substatons. Three njecton devces are connected to node 1, 2 and 5. Power flow measurements are nstalled on three transmsson lnes, as well as node 2 and 5. Fg. 1(b) shows the used bus-branch model. It can be seen that two buses exst n ths substaton. Node 1, 3, 4, 5, 7 and 8 are merged nto bus 1, and node 2 and 6 become bus 2. The three branch measurements are preserved n the bus-branch model. The nodal njecton measurement at node 2 s also preserved. The nodal njecton measurement at node 5, however, s elmnated snce bus 1 s njecton equals to the sum of node 1 and 5 s njectons, and node 1 s njecton s unknown. III. NETWORK OBSERVABILITY ANALYSIS A lnear, tme-nvarant (LTI) system s usually descrbed n the followng state space representaton: x& ( t) = Ax( t) + Bu( t), (1) y() t = Cx() t + Du() t where x s the state vector; y s the output vector; u s the nput (or control) vector; A s the state matrx; B s the nput matrx; C s the output matrx; D s the feedthrough (or feedforward) matrx. In power system state estmaton, the state vector x of the system contans the voltage magntude and phase angles of buses. The output vector y (whch s often denoted as z n power system analyss) contans the measurements. Snce state estmaton s a steady state functon, the state vector s constant, and the state matrx and nput matrx are both 0,.e., A=0 and B=0. Also, measurements that are consdered n power system state estmaton have no feedthrough,.e., D=0. Thus the power system state estmaton problem becomes z = Cx. (3) Dfferent from the LTI system, the power system s a nonlnear system. The output matrx C s a functon of x, and (3) can be represented as: z = f ( x). (4) Frst-order Taylor approxmaton of (4) yelds: 0 H Δx = z f ( x ) = Δz, (5) f ( x) 0 0 where H =, evaluated at some x ; Δ x = x x. x Equaton (5) relates all exstng measurements to the state varables, usng the frst-order Taylor approxmaton. An
3 estmate for Δx can be obtaned as long as the rank of H s equal to the dmenson of Δx or x. Therefore, the necessary and suffcent condton for a power system to be observable s: rank H = n, (6) where n s the dmenson of the state vector x. It should be noted that the system observablty s ndependent of the branch parameters as well as the operatng state of the system. So, all system branches can be assumed to have an mpedance of j1.0 per unt (p.u.) and all bus voltages can be set equal to 1.0 p.u. for the purpose of observablty analyss. It can be shown that n such a power system network, H can be calculated by: T H = M A, (7) where M s the measurement-branch ncdence matrx, 1 If measurement s ncdent to bus j at the "from end". M j = 1 If measurement s ncdent to bus j at the "to end". 0 If measurement s not ncdent to bus j. A s the branch-bus ncdence matrx, 1 If branch s ncdent to bus j at the "from end". A j = 1 If branch s ncdent to bus j at the " to end". 0 If branch s not ncdent to bus j. The method that uses (6) and (7) to decde whether a network s observable s call the numercal method. Observablty analyss can also be carred out by usng a topologcal method. If a tree can be formed such that each branch of ths tree contans a power flow measurement, then the phase angles at all buses can be determned,.e. the system wll be fully observable. The avalable measurements should be assgned to the branches accordng to the followng rules: 1) If the branch flow s measured, the branch s assgned to ts flow measurement. 2) If an njecton s measured at a termnal node of a branch, the branch can be assgned to that njecton. 3) Once a branch s assgned to a measurement, t can not ba assgned to any other measurement. The essental steps of the algorthm can be summarzed as follows: 1) Frst assgn all the flow measurements to ther respectve branches. 2) Then, try to assgn the njecton measurements n order to reduce the exstng forest by mergng exstng trees. Note that there s no way to predct the correct sequence for processng njectons. Implementaton of the method requres proper back-up and re-assgnment of njectons when necessary. The network observablty analyss determnes f a state estmaton soluton for the entre system can be obtaned usng the avalable set of measurements, therefore t s a very mportant component n the EMS and t s usually carred out before the executon of state estmaton. IV. THE DYNAMIC UTILIZATION OF SUBSTATION MEASUREMENTS In a physcal substaton, CB statuses may be changng relatvely frequently, ether due to faults, or because of operator commands. The network topology changes accordngly. The changes n topology may have the followng potental mpacts on the NTP: 1) The mergng or splttng of buses may cause some substaton measurements to become useless n the changed topology, durng the processng of measurement data as descrbed n secton II. 2) Some measurements may be dsconnected from the rest of the network. For example, when the CBs dsconnect a transmsson lne, the branch power flow measurement on ths lne s also dsconnected and wll not appear n the bus-branch model. 3) The total number of avalable measurements n the busbranch model may be reduced and the locatons of measurements may change, due to the change of network topology. Because the network observablty s hghly related to the number and locatons of measurements n the network, the network may become unobservable after the change of topology, and therefore an estmaton of system states cannot be obtaned. The exstng approach to deal wth the loss of observablty s to suggest new locatons for addtonal measurements [4-6]. However, nstallng new measurements may be costly and can only be done off-lne. A way of utlzng the currently avalable measurements n the substatons to recover the network observablty on-lne s presented n ths secton. The new method s called the dynamc utlzaton of substaton measurements. (DUSM). A. Calculaton of nferred substaton measurements Lke a conventonal NTP, the frst step of DUSM s to read n the statc connectons of devces and CB statuses, and then store the network topology nformaton n an organzed way for easer processng. The devces (CBs, branches, loads, generators, etc.) are grouped nto dfferent substatons. Each devce s assgned a vrtual measurement that supposes to measure the power flow of ths devce, and a measurement vector can be as created usng the followng equaton: z [ ( ) ( ) ( )] T = z devce1 z devce 2 L z devce n, (8) where s the substaton number, devce 1,2, n are the devces of substaton, and z(devce j) s the power flow measurement of devce j. The followng assumptons are made regardng the drectons of the measurements: 1) A CB power flow measurement s drecton s always the same as the CB s. 2) A branch power flow measurement s drecton s always gong nto the node that the branch s connected to. 3) A power njecton measurement s drecton s always gong nto the node that t s connected to,.e., for a generator, the measurement value s postve; for a load, the measurement value s negatve. If the drectons of measurements are dfferent from the
4 Substaton 1 TABLE I TOPOLOGICAL INFORMATION STORAGE FOR FIG. 1 (A) SUBSTATION 1 Devces CB1 CB2 CB3 CB4 CB6 load1 load2 b1 b2 Nodes 0101-1 0 0-1 0 0 0 0 0 0102 1-1 0 0 0 1 0 0 0 0103 0 1-1 0 0 0 0 1 0 0104 0 0 1 0 1 0 0 0 0 0105 0 0 0 1 0 0 0 0 1 0106 0 0 0 0-1 0 1 0 0 Fg. 2. Sample detaled substaton model of a 3-bus system. above assumptons, they can be easly modfed to conform to the assumpton by changng the sgns of the measurement values. DUSM uses a three-dmensonal ncdence matrx M to store the topologcal nformaton, as llustrated n Fg. 2 and Table I. The element of the ncdence matrx M can be expressed as 1 If measurement x's drecton goes nto node y, M( y, x) = 1 If measurement x's drecton goes out of node y, 0 If measurement x s not ncdent to node y, where s the substaton number. Accordng to Krchhoff s current law, we have M z = 0, (9) where s the substaton number. In a practcal system, some of the elements n z are measured, and others are not. The measured elements can be replaced by ther measurement values, whle other elements reman as unknown. What we are nterested n s to nfer as many measurement values as we can by usng (9). It can be seen that an nferred measurement can be calculated when the measurements of all other devces that are connected to the same node are avalable. Ths can be llustrated by the followng example. In Fg. 2, suppose the power flow of CB8 s measured and ts value s z 8. Applyng (9) to node 0202, we get z8 zcb9 1 0 0 1 0 z =, (10) [ ] 0 CB10 or z b 1 = z 8, whch means the power flow of b1 equals the power flow of CB8. Now that has been calculated, both z8 and z b 1 can be used to calculate other nferred measurements, untl no more measurements can be nferred. z b1 z z b1 b3 Nodes Substaton 3 Nodes Substaton 2 CB11 (B) SUBSTATION 2 Devces CB8 CB9 CB10 b1 b3 0201 0-1 0 0 0 0202-1 0 0 1 0 0203 1 0 1 0 0 0204 0 1-1 0 1 CB12 (C) SUBSTATION 3 Devces CB13 CB14 0301-1 0 0-1 0 0 0 0 0 0 0302 1-1 0 0 0 0 0 1 0 0 0303 0 1-1 0 0 0 1 0 0 0 0304 0 0 1 0 0 1 0 0 0 0 0305 0 0 0 1-1 0 0 0 1 0 0306 0 0 0 0 1-1 0 0 0 1 CB15 CB16 B. Calculaton of bus-branch measurements Once all possble nferred measurements are obtaned, the next step s to calculate the values of the bus voltage magntude measurements, bus njecton power flow measurements and branch power flow measurements. The calculaton of bus voltage magntude measurements s straght-forward. It can be done by a drect mappng of the substaton node to the correspondng bus, as show below: V = V n, (11) where n s the node number n the substaton model, s the bus number of n n the bus-branch model. The calculaton of branch power flow measurement uses the followng rules: 1) If there s only one branch (sngle lne) between two buses, map the branch measurement n the detaled substaton model to the correspondng branch measurement n the bus-branch model by changng the node numbers to the bus numbers. 2) If more than one branch (multple lnes) exsts between two buses, sum up all branch measurements n the substaton model to get the branch measurement n the bus-branch model. The bus njecton power flow measurement can be calculated by addng all nodal njecton measurements n a generator Load3 b2 b3
5 sngle merged electrcal bus. If any measurement value s unknown after the addton, the njecton power flow of ths bus cannot be calculated. C. The applcablty n practce The DUSM algorthm uses the followng mplct assumptons: 1) It assumes that the Krchhoff s current law s applcable, whch requres that there s no unknown ground fault or ground leakage current exstng n the substaton. Ths s usually true, snce frstly, the possblty of ground faults n a substaton s low; secondly, n case of a bus ground fault, the bus protecton wll trp all CBs that are connected to the bus and clear the leakage current. 2) The open CBs are excluded from the topologcal matrx and the power flows through them are assumed to be zero. Ths requres that the CB statuses are correctly reported. If the CB status measurements are not accurate enough, the open CBs should stll be ncluded n the topologcal matrx and the power flow through them should be regarded as unknown. The DUSM algorthm can be mplemented as a supplementary functon to the substaton automaton system (SAS), or to the EMS n the control center. The advantages of mplementng DUSM n substatons are: 1) More measurements are avalable n substatons than n the control center. In recent dgtal substatons, besdes the measurements that are gathered by RTUs, many ntellgent electronc devces (IEDs) also record and montor the status of the substaton on-lne. Many measurement data can be obtaned from the recordng of these devces [7]. 2) Because of the ndependent storage of substaton topologcal nformaton, DUSM can be mplemented n any number of substatons n the system. Ths adds to the flexblty of mplementaton. A few substatons may be pcked up and the effectveness of the new algorthm may be tested wthout the need for upgradng the exstng EMS software n the control center or SAS software n other substatons. On the other hand, the advantage of mplementaton n the control center s that the full potental of dynamcally creatng new measurements for the state estmaton purpose can be obtaned. The EMS n control center has the access to the topologcal nformaton from all substatons. The nstallaton of the new algorthm wll enable the new measurements to be calculated from all the measurements that are transmtted to the control center. V. TEST CASES Tests have been run on the IEEE-30 bus system. The data of the IEEE-30 bus system, ncludng the bus-branch dagram, can be obtaned from [8]. It was assumed that 31 measurements were already avalable to the state estmator. The types and locatons of these measurements are lsted n Table II. Measurement Bus voltage magntude Branch power flow Bus njecton power flow TABLE II MEASUREMENT PLACEMENT Locaton 4, 16 29-27, 30-29, 30-27, 25-26, 12-16, 16-17, 1-3, 9-11, 14-12, 12-13, 14-15, 6-8, 28-8, 22-21, 18-19, 17-31 25, 27, 4, 9, 10, 22, 24, 15, 12, 28, 20, 18, 2 Fg. 3. Detaled substaton model of FIELDALE substaton n the IEEE-30 bus system. Measurement V0501 Value 1.01 TABLE III MEASUREMENT DATA VALUES PCB2 45.4 -j13.0 PCB3 45.4 -j22.8 PCB7 34.0 +j6.9 PCB9-53.0 -j10.7 Pb2 14.8 -j10.6 Furthermore, the detaled breaker-and-a-half confguraton was arbtrarly pcked to represent the FIELDALE substaton (Bus 5) n the IEEE-30 bus system, as shown n Fg. 3. Sx measurements were placed n the substaton, ncludng one voltage magntude measurement and fve power flow measurements. The measurement values are lsted n Table III. Usng the conventonal NTP, only two measurements were generated to serve the state estmator. In the bus-branch model, V 0501 became the voltage magntude measurement of Bus 5, and P b2 became the branch power flow measurement of branch 5-7. The conventonal NTP was also able to calculate the njecton power flow of the synchronous condenser. However, the njecton of Bus 5 could not be calculated snce nether of the three loads power flow could be obtaned. A topologcal method as mentoned n secton III was used to evaluate the observablty of the 30-bus network. Addng these two measurements to the system, t was found that the whole network was not totally observable. TABLE IV CONVENTIONAL NTP VS. DUSM Measurement Conventonal NTP DUSM V bus5 1.01, equals to V 0501. 1.01, equals to V 0501. P 5-2 (Unknown) 79.4-j6.1, equals to -P CB2 -P CB7. P 5-7 14.8-j10.6, equals to P b2. 14.8-j10.6, equals to P b2. P nj5 (Unknown) -94.2+j16.7, equals to P CB2 +P CB7-2P CB9 -P b2.
6 The same measurement data were then processed by the DUSM algorthm and four measurements were generated: the voltage magntude measurement of Bus 5, the branch power flow measurement of branch 5-2 and 5-7, and the bus njecton power flow measurement of bus 5. A comparson of the dfferent results s shown n Table IV. Wth the help of the two extra measurements that were created by DUSM, the whole network became observable, and the state estmaton was then able to be executed. VI. CONCLUSIONS Ths paper explans the mportance of an advanced network topology processor n preservng as many substaton measurements as possble to mantan the network observablty. A new method the dynamc utlzaton of substaton measurements (DUSM) was presented. Instead of seekng the nstallaton of new measurements n the system, ths method tres to calculate meanngful state estmaton measurement values by applyng the current law that regulates dfferent measurement values mplctly. Its processng s at the substaton level and therefore can be mplemented n dfferent substatons. Test cases on the IEEE-30 system show DUSM s advantage n measurement processng over a conventonal network topology processor. VII. REFERENCES [1] A. Abur and A. G. Exposto, Power System State Estmaton, Marcel Dekker, USA, March 2004, p. 5. [2] A. Montcell, "Electrc power system state estmaton," Proceedngs of the IEEE, Vol. 88, No. 2, pp. 262-282, Feburary 2000. [3] S. Pandt, S. A. Soman and S. A. Khaparde, "Object-orented network topology processor," IEEE Computer Applcatons n Power, Vol. 14, No. 2, pp. 42-46, Aprl 2001. [4] F. H. Magnago and A. Abur, Unfed approach to robust meter placement aganst bad data and branch outages, IEEE Trans. on Power Systems, Vol.15, No. 3, pp. 945-949, August 2000. [5] Q. Dng and A. Abur, "An mproved measurement placement method aganst loss of multple measurements and branches," n Proceedng of IEEE Power Engneerng Socety Wnter Meetng, Vol. 1, pp. 27-31, January 2002. [6] M. K. Celk and W.-H.E. Lu, "An ncremental measurement placement algorthm for state estmaton," IEEE Trans. on Power Systems, Vol. 10, No. 3, pp. 1698-1703, August 1995. [7] M. Kezunovc, "Future trends n protectve relayng, substaton automaton, testng and related standardzaton," n Proceedng of IEEE Transmsson and Dstrbuton Conference and Exhbton, Vol. 1, pp. 598-602, October 2002. [8] R. Chrste. Power Systems Test Case Archve, 30 Bus Power Flow Test Case [Onlne]. Avalable: http://www.ee.washngton.edu/ research/pstca/pf30/pg_tca30bus.htm. VIII. BIOGRAPHIES Yang Wu (S 05) receved hs B.S. and M.S. degrees from X an Jaotong Unversty, X an, Chna, both n electrcal engneerng, n 1999 and 2002 respectvely. He has been a Ph.D. student n Texas A&M Unversty snce Aug. 2002. Hs research nterests nclude protectve relayng, substaton automaton and state estmaton. Mladen Kezunovc (S 77, M 80, SM 85, F 99) receved the Dpl. Ing. Degree n electrcal engneerng from the Unversty of Sarajevo, Bosna- Herzegovna, n 1974, and the M.S. and Ph.D. degrees n electrcal engneerng from the Unversty of Kansas, Lawrence, n 1977 and 1980, respectvely. Currently, he s the Eugene E.Webb Professor and Drector of Electrc Power and Power Electroncs Insttute at Texas A&M Unversty. Hs man research nterests are dgtal smulators and smulaton methods for relay testng as well as applcaton of ntellgent methods to power system montorng, control, and protecton. Dr. Kezunovc s also a Fellow of IEEE and a member of CIGRE-Pars. Tatjana (Tanja) Kostc (M 95) receved her BSEE ( 89) and MSEE ( 94) from the Unversty of Belgrade, Yugoslava, and the Dr. of Sc. Techn. degree ( 97) from the Swss Federal Insttute of Technology (EPFL), Lausanne, Swtzerland. After her post-doc year wth Mtsubsh Electrc, Amagasak, Japan, she joned ABB Corporate Research n Swtzerland, where she s currently workng as a prncpal scentst n Utlty Solutons group. Her research nterests nclude IT applcatons for power system operaton and for utlty asset management, standardzed utlty doman models, object-orented analyss and desgn, and artfcal ntellgence. She s a member of the IEEE PES and Computer socetes, a workng member of the Cgré WG C2.01, and an IEC expert n TC57 WG14.