20 IEEE TRANSACTIONS ON SMART GRID, VOL. 1, NO. 1, JUNE 2010

Synchronized Phasor Measurement Applicationsin Power Systems

Jaime De La Ree, Senior Member, IEEE, Virgilio Centeno, Senior Member, IEEE, James S. Thorp, Life Fellow, IEEE,and A. G. Phadke, Life Fellow, IEEE

AbstractSynchronized phasor measurements have become amature technology with several international manufacturers of-fering commercial phasor measurement units (PMUs) which meetthe prevailing industry standard for synchrophasors. With the oc-currence of major blackouts in many power systems around theworld, the value of data provided by PMUs has been recognized,and installation of PMUs on power transmission networks of mostmajor power systems has become an important current activity.This paper provides a brief introduction to the PMU and wide-areameasurement system (WAMS) technology and discusses the uses ofthese measurements for improved monitoring, protection, and con-trol of power networks.

Index TermsElectrical transients, global positioning system,phasor, phasor measurement unit, power system, power systemcontrol, power system monitoring, power system protection, real-time measurements, synchronized measurements, synchrophasor

.

I. INTRODUCTION

S YNCHRONIZED phasor measurement units (PMUs) werefirst introduced in early 1980s, and since then have be-come a mature technology with many applications which arecurrently under development around the world. The occurrenceof major blackouts in many major power systems around theworld has given a new impetus for large-scale implementationof wide-area measurement systems (WAMS) using PMUs andphasor data concentrators (PDCs) in a hierarchical structure.Data provided by the PMUs are very accurate and enable systemanalysts to determine the exact sequence of events which haveled to the blackouts, and help analyze the sequence of eventswhich helps pinpoint the exact causes and malfunctions thatmay have contributed to the catastrophic failure of the powersystem. As experience with WAMS is gained, it is natural thatother uses of phasor measurements will be found. In particular,significant literature already exists which deals with applicationof phasor measurements to system monitoring, protection, andcontrol.

This paper is organized as follows. Section II reviews thebasic concepts of phasor representation of power systemvoltages and currents under normal and abnormal operatingconditions. Section III discusses many of the applications ofthe phasor measurements which have been discussed in theliterature. Section IV provides concluding remarks with someassessment of the direction in which short-term and long-termapplications of this technology will emerge.

Manuscript received January 23, 2010; revised February 09, 2010. Date ofpublication April 15, 2010; date of current version May 21, 2010. Paper no.TSG-00011-2010.

The authors are with the Electrical and Computer Engineering Department,Virginia Tech, Blacksburg, VA 24060 USA.

Digital Object Identifier 10.1109/TSG.2010.2044815

Fig. 1. Phasor representation of a sinusoidal signal. (a) Sinusoidal signal. (b)Phasor representation.

II. PHASOR MEASUREMENT UNITS

A. Classical Definition of a PhasorA pure sinusoidal waveform can be represented by a unique

complex number known as a phasor. Consider a sinusoidalsignal

(1)The phasor representation of this sinusoid is given by

(2)

Note that the signal frequency is not explicitly stated in thephasor representation. The magnitude of the phasor is the rmsvalue of the sinusoid , and its phase angle is , thephase angle of the signal in (1). The sinusoidal signal and itsphasor representation given by (1) and (2) are illustrated inFig. 1.

Note that positive phase angles are measured in a counter-clockwise direction from the real axis. Since the frequency ofthe sinusoid is implicit in the phasor definition, it is clear thatall phasors which are included in a single phasor diagram musthave the same frequency. Phasor representation of the sinusoidimplies that the signal remains stationary at all times, leading toa constant phasor representation. These concepts must be mod-ified when practical phasor measurements are to be carried outwhen the input signals are not constant, and their frequency maybe a variable. This will be discussed in the next section.

B. Phasor Measurement ConceptsAlthough a constant phasor implies a stationary sinusoidal

waveform, in practice it is necessary to deal with phasor mea-surements which consider the input signal over a finite datawindow. In many PMUs the data window in use is one periodof the fundamental frequency of the input signal. If the power

1949-3053/$26.00 2010 IEEE

DE LA REE et al.: SYNCHRONIZED PHASOR MEASUREMENT APPLICATIONS 21

system frequency is not equal to its nominal value (it seldomis), the PMU uses a frequency-tracking step and thus estimatesthe period of the fundamental frequency component before thephasor is estimated. It is clear that the input signal may haveharmonic or nonharmonic components. The task of the PMU isto separate the fundamental frequency component and find itsphasor representation.

The most common technique for determining the phasor rep-resentation of an input signal is to use data samples taken fromthe waveform, and apply the discrete Fourier transform (DFT) tocompute the phasor. Since sampled data are used to represent theinput signal, it is essential that antialiasing filters be applied to thesignal before data samples are taken. The antialiasing filters areanalog devices which limit the bandwidth of the pass band to lessthan half the data sampling frequency (Nyquist criterion).

If are the samples of the inputsignal taken over one period, then the phasor representation isgiven by [1]

(3)

The multiplier in front of the summation sign may need someexplanation. Note that for real input signals, the componentsof the signal at a frequency appear in the DFT at andare complex conjugates of each other. They can be combined,giving a factor of 2 in front of the summation sign in (3). Thepeak value of the fundamental frequency thus obtained is thenconverted to rms value by dividing by . The DFT calcula-tion eliminates the harmonics of the input signal. However, thenonharmonic signals and any other random noise present in theinput signal leads to an error in estimation of the phasor. Theerror of estimation due to these effects has been discussed inthe literature.

C. Synchrophasor Definition and MeasurementsSynchrophasor is a term used to describe a phasor which has

been estimated at an instant known as the time tag of the syn-chrophasor. In order to obtain simultaneous measurement ofphasors across a wide area of the power system, it is necessaryto synchronize these time tags, so that all phasor measurementsbelonging to the same time tag are truly simultaneous. Considerthe marker in Fig. 1 is the time tag of the measurement.The PMU must then provide the phasor given by (2) using thesampled data of the input signal. Note that there are antialiasingfilters present in the input to the PMU, which produce a phasedelay depending upon the filter characteristic. Furthermore, thisdelay will be a function of the signal frequency. The task of thePMU is to compensate for this delay because the sampled dataare taken after the antialiasing delay is introduced by the filter.This is illustrated in Fig. 2.

The synchronization is achieved by using a sampling clockwhich is phase-locked to the one-pulse-per-second signal pro-vided by a GPS receiver. The receiver may be built in the PMU,or may be installed in the substation and the synchronizing pulsedistributed to the PMU and to any other device which requires it.

The time tags are at intervals that are multiples of a period ofthe nominal power system frequency.

Fig. 2. Compensating for signal delay introduced by the antialiasing filter.

It should also be noted that the normal output of the PMU isthe positive sequence voltage and current phasors. In many in-stances the PMUs are also able to provide phasors for individualphase voltages and currents [2].

III. APPLICATIONS OF PHASOR MEASUREMENTS

A. Power System Monitoring1) Early PMU Applications: When the first commercial

PMUs became available, postevent monitoring was their onlyapplication due to the low availability and high cost of commu-nications channels required for real-time monitoring, control,and protection applications. The first recorded wide-areameasurements came from staged tests performed as part ofthe EPRI Parameter Identification Data Acquisition System(PIDAS) project in 1992. The tests consisted of opening andclosing of the Klondike and Bonaire 500 kV lines at PlantScherer in Georgia. PMUs were placed at Plant Scherer andfive other locations in Tennessee, Georgia, and Florida [3].For the four staged states PMU were manually triggered toguarantee wide-area recording of the event. For the sole PMUmanufacturer of that time this was a test of the units capabilityto monitor systemwide events but at the same time. For en-gineers interested in maintaining system models this becamethe first wide-area model validation test. At that time, it took aweek to obtain the plots comparing the measurements with theestimations. Most of the delay came from the mailing of thediskettes with data from each PMU to a common location foranalysis and comparison to the simulation results. This eventproved the capabilities of wide-area monitoring systems andvalidated the system model for the study area.

Following the successful results of the initial PMU tests,small numbers of PMUs were placed in different locationsthroughout systems in the United States and other countriesto validate and correct system models by simulating recordscollected by PMUs. Essentially the PMUs at this stage weredigital system disturbance recorders (DSDRs). Although sev-eral experimental applications were made for protection andcontrol applications, most PMUs functioned exclusively as aDSDR due to limitation in the availability and bandwidth ofcommunication channels. To cope with the limited communica-tion bandwidth, the first PMUs reported only positive sequencephasors and frequency information.

As a DSDR, PMUs have collected interesting informationthat has expanded the monitoring application of the devices.One of these events is the data collected in Texas in July 1993for the Comanche Peak load rejection test [4]. The frequencyoscillations collected by PMUs revealed an electromechanicalwave propagating through the system and easily observed in

22 IEEE TRANSACTIONS ON SMART GRID, VOL. 1, NO. 1, JUNE 2010

the frequency measurements. This and similar measurements inother parts of the country eventually led to the development of afrequency monitoring network, FNET. Although limited in theinformation provided the time synchronized frequency moni-toring network has stimulated high interest among the academicand regulatory communities due to their operation at residentialvoltage level eliminating installation cost and data ownershipissues. PMU and FNET data have shown that any significantloss of generation in a system will cause an electromechanicaloscillation that will propagate through the system at slow speed.FNET data have shown that it is not only possible to detectthese oscillations at residential voltage levels but also to use thefrequency change to predict the amount of generation loss. It isalso possible with some limitation to use the traveling wave todetermine the location of the event [5].

In the 2003 Northeastern U.S. blackout and the 1996 U.S. WestCoast blackouts, the PMU monitoring capabilities were essen-tial for the quick and accurate postmortem analysis of the events.OneoftherecommendationsfromtheUnitedStatesCanadaTaskForce on the 14 August 2003 blackout is to require use of time-synchronized data recorders to all utilities [6]. This and otherrecommendations led to the creation of the Eastern Interconnec-tion Phasor Project (EIPP), now known as North American Syn-chrophasor Project (NASPI). The EIPP performed the first real-timewide-areamonitoring in theUnitedStatesafter solvingsomeminor but interesting problems, such as the determination of acommon phase a for the whole eastern grid. NASPI continuestosolvesomeof theexistingproblemsfor thecreationofa reliablereal-timesynchronizedmonitoringsystemsuchastheavailabilityof a commercial data concentrator capable of handling hundredsof PMUs at the high communication rates required for most pro-posed applications.

With PMU installation cost ranging from 10 k to 70 K (de-pendingontheutility, location,andavailabilityofcommunicationchannels) placing PMUs in the optimum locations is one of thefirst steps of a wide-area monitoring system [7]. PMU placementaimed at system monitoring is usually developed for full systemobservability. Concepts such as depth of observability have beendeveloped to allow for a maximum utilization of data through astaged deployment of a monitoring system for full observability[9]. Other placement algorithms have been developed for spe-cific applications such as Islanding and enhanced state estima-tion. PMU placement studies require the development of specificplacement system models. System models used in commercialanalysis software include virtual buses that either do not exist orare not practical locations to install PMUs, such as tapped linebuses, shunt elements and series capacitor nodes. These virtualbuses may add up to 1/3 more nodes to the model and increase thecomplexity of the problem. Not eliminating these nodes from theplacement model may result in PMUs being placed in unrealisticlocations. This problem is solved in small models by manuallyeliminating virtual buses. In the case of real system models theproblem is solved by developing systematic methods to eliminatevirtual buses from commercial system data models [8].

2) State Estimation: Since the 1965 Northeast blackout,state estimation has become a critical application function atenergy control centers. Existing state estimation algorithms [1]use measurements of line flows and injections, both real and

reactive power, to estimate all bus voltage angles and magni-tudes. The complex bus voltages are the state of the system.The complex power flows in lines and the complex powerinjections at buses can be determined from the bus voltages andan accurate model of the network. Prior to synchronized phasormeasurements, the state could not be measured directly butonly inferred from the unsynchronized power flow measure-ments. This fact and the process of getting large numbers ofmeasurements into the control center (a scan) forced early stateestimators to make compromises that persist today and havean influence on how phasor measurements are integrated intoexisting state estimation algorithms. The most significant ofthese assumptions is that the system did not change during thescanthat the system was static. Another view of the resultingestimate of the state is that it is the state of a hypotheticalsystem which could support the complete set of measurements.Depending on how long the scan takes and the changes in thesystem that took place during the scan, the hypothetical systemmay not exist or may be quite different from the real system.While scans have become quicker, the introduction of phasormeasurements forces reconsideration of the static assumption.

The direct integration of a few synchronized phasor measure-ments into an existing nonlinear estimator is straightforward butresults in an estimator that has most of the limitations of the orig-inal conventional estimator. The new estimator is still recursiveand still has the assumptions required for the static state esti-mator. The issue of how to place the precisely timed PMU datainto the scanned data can also introduce a skew.

If an estimate could be formed with only PMU data, then theissues of data scan and time skew could be eliminated. The PMUdata would be time tagged and the static assumption removed.Since the voltage and current measurements are linear functionsof the system state, the estimation problem is simple a linearweighted least squares problem which requires no iterations.Given PMU data can be reported at rates as high as 60 timesa second, a truly dynamic estimate would be available. An es-timate of a dynamic system at an instant in time correspondingto the measuring instant would be obtained. Issues that must beaddressed include the need for redundancy to eliminate bad dataand determination of how many PMUs are required.

Recognizing that a PMU in a substation would have access toline currents in addition to the bus voltage reduces the number ofPMUs needed. Measuring line currents can extend the voltagemeasurements to buses where no PMU is installed. Using amodel of the transmission line, the line current can be used tocompute the voltage at the other end of the line. With a largenumber of PMUs the redundancy issue is addressed. On theother hand, the smallest number of PMUs needed to indirectlymeasure all the bus voltages and the optimum PMU location toachieve this has been a subject of a number of papers. [9][11].With different approaches, treatment of zero injection buses,and differing assumptions about types of PMU used, the con-sensus is that the minimum number of PMUs is approximatelyone-third the total number of buses. A compromise between thefull nonlinear and linear formulations is described in [12].

An interesting subproblem is the issue of sequentially addingPMUs to a system. With limited annual investments it wouldbe desirable to add a limited number of PMUs until a final goal

DE LA REE et al.: SYNCHRONIZED PHASOR MEASUREMENT APPLICATIONS 23

was achieved. Initially there will not be enough PMUs to havea linear estimator. The attempt is to place the PMUs so thatat each stage the selection satisfies some design criteria. Onecriterion is the degree of unobservability [11], which measureshow far away a bus is from a PMU. The solution to this problemis available using the technique from [12]

(4)

The busbus incidence matrix is a square matrix with the di-mension of the number of buses. There is a one on each diagonaland a one in the th position if bus is connected to bus . Thematrix for the IEEE 14 bus system is given in (4). If we imagineplacing a PMU at bus 2, for example, we would learn the volt-ages at buses 1, 2, 3, 4, and 5, which are the nonzero entries incolumn 2 of A.

The placement problem for complete observability is thengiven by the integer programming problem in (5) [14]. It is inter-esting to note that the powers of the incidence matrix have an in-terpretation that helps extend the integer programming problemto the partial observability problem

subject to or(5)

Theorem: The ij entry in the th power of the incidence matrixfor any graph or diagraph is exactly the number of differentpaths of length n, beginning at vertex i and ending at vertex j[13] [see (6)].

sign

(6)

Fig. 3. The graph for the incidence matrix for the IEEE 14-bus system alongwith the branches added when the matrix is squared.

TABLE IPMU LOCATION FOR THE IEEE 14-BUS SYSTEM

The proof is by induction. The signum of is the incidencematrix of another graph that has branches added to the graphin Fig. 3. The dashed branches in Fig. 3 represent connectionsbetween nodes that pass through another node, i.e., those nodesare one step removed. If we imagine the complete observabilityproblem is solved [ is the solution to (5)] the one degree ofunobservability problem is the solution of

subject to sign or(7)

And the addition of the constraint forces thesolution to be chosen from the PMU location in [8]. So ingeneral the sequence of integer programming problems in (8)

subject to sign(8)

produce a nested sequence of PMU locations. For the 14-bussystem the locations are given in Table I.

Solutions of (8) for a 1400-bus system with a depth of 5 aregiven in [1]. The first problem is a large amount of computationfor large systems. The subsequent problems are much easier.

B. Power System ProtectionSynchronized phasor measurements offer solutions to a

number of complex protection problems. In general, phasormeasurements are particularly effective in improving protectionfunctions, which have relatively slow response times. For suchprotection functions, the latency of communicating informationfrom remote sites is not a significant issue. A few examplesof protection systems that could benefit from remote phasormeasurements information could include (B-1), control ofbackup protection of distance relays (B-2), protection functionsconcerned with angular voltage stability of networks (B-3),

24 IEEE TRANSACTIONS ON SMART GRID, VOL. 1, NO. 1, JUNE 2010

Fig. 4. Loadability limit imposed by zone-3 setting of a distance relay. Theillustration shows a mho characteristic, which is commonly used in many relays.As the load increases along the bold arrow, it would enter the tripping zone ofthe relay and cause an inappropriate trip.

Fig. 5. Supervisory control for backup protection.

voting schemes where dependability and security could be re-assessed based upon the stress of the system. A number of otherWAMS-based protection improvements can be found in [1].

1) Control of Backup Relay Performance: Backup zones ofdistance relays are prone to tripping due to load encroachmentduring power system disturbances. This has led to a call forreassessing and sometimes abandoning the use of backupzonesin particular zone-3 of distance relays (see Fig. 4),which is used to protect downstream circuits in case theirprotection systems fail to remove a fault on those circuits.[15] Abandoning Zone-3 may be too drastic, and should notbe applied as a blanket policy. The remote backup policy isdesigned to cover certain contingencies [16] for which no otherprotection is available. Under these circumstances, it becomesnecessary to consider ways in which the loadability limitsimposed by the remote backup zones can be circumvented [17].

Wide-area measurements offer a possibility for restrainingthe remote back-up relays in the event that the load swing is

being interpreted by the relay as a fault [18]. Consider the con-ditions illustrated in Fig. 5. Zone-3 of relay is assumed to bepicked up. If a significant negative sequence current is present(indicating an unbalanced fault), the zone-3 pick up is appro-priate, and no further action is necessary. However, if the cur-rents in the line are balanced, either a three phase fault on theneighboring circuits or a possible loadability violation may beinferred. The PMUs at the buses corresponding to the terminalsof lines which are to be backed-up by relay may then deter-mine if any of them see a Zone-1 three-phase fault. If none of thePMUs indicate that a Zone-1 three-phase fault exists, then theZone-3 pick-up of relay must be due to loadability limit vio-lation. If tripping on this condition by relay is to be avoided,it would then be possible to block its operation by supervisorycontrol of its output.

2) Adaptive Out-Of-Step Protection: It is recognized thata group of generators going out of step with the rest of thepower system is often a precursor of a complete system collapse.Whether an electromechanical transient will lead to stable orunstable condition has to be determined reliably before appro-priate control action could be taken to bring the power system toa viable steady state. Out-of-step relays are designed to performthis detection and also to take appropriate tripping and blockingdecisions.

Traditional out-of-step relays use impedance relay zones todetermine whether or not an electromechanical swing will leadto instability. In order to determine the settings of these relays, itis necessary to run a large number of transient stability simula-tions for various loading conditions and credible contingencies.Using the apparent impedance trajectories observed at locationsnear the electrical center of the system during these simulationstudies, two zones of an impedance relay are set, so that the innerzone is not penetrated by any stable swing.

Problems With Traditional Out-of-Step Relays: Traditionalout-of-step relays are found to be unsatisfactory in highly in-terconnected power networks. This is because the conditionsand topologies assumed when the relay characteristics are de-termined become out-of-date rather quickly, and in reality theelectromechanical swings that do occur are quite different fromthose studied when the relays are set. The result is that tradi-tional out-of-step relays often misoperate: they fail to determinecorrectly whether or not an evolving electromechanical swingis stable or unstable. Wide-area measurements of positive se-quence voltages at networks (and hence swing angles) providea direct path to determining stability using real-time data insteadof using precalculated relay settings.

It should be noted that a related approach was developed fora field trial at the FloridaGeorgia interface [19][22] where theinterface was modeled as a two machine system. The machinesin Fig. 6 are equivalents of the eastern interconnection on theleft and Florida on the right with the four buses being phys-ical buses in the interconnection. The equation of motion of theangle difference between the two rotors of the two machinesis given by (9) where . and are the tworotor inertias, and the remaining terms in (6) are obtained fromthe equivalent system. As the system undergoes changes due toa fault and its clearing the parameters of the differential equa-tion and change and the classical equal-area criterion

DE LA REE et al.: SYNCHRONIZED PHASOR MEASUREMENT APPLICATIONS 25

Fig. 6. Reduced FloridaGeorgia system.

Fig. 7. The equal area criterion.

can be used to determine stability. That is, the area A1 mustbe smaller than the area A2 for stability (see Fig. 7). The issuein adaptive out-of-step is to determine the new parameter values

and from real time measurements. A least squares esti-mate of [22] from samples of is used in [19]. The estimateis obtained from five or six consecutive measurements of

(9)

It is of course possible to determine whether or not a swing isunstable by waiting long enough and observing the actual swing.However, in order to take appropriate control action it is essen-tial that a reliable prediction algorithm be developed which pro-vided the stableunstable classification of an evolving swing in areasonable time. In the FloridaGeorgia experiment a period ofobservation of actual angular swings for a maximum of 250 mswas used to obtain a reliable prediction of the outcome.

For large power networks it is not possible to apply the equalarea criterion as used in the example of FloridaGeorgia inter-face. As a first step after a transient stability event begins, groupsof coherently swinging machines are identified. After two dis-tinct groups of coherent machines are determined, they may beconverted to a system of two equivalent generators connectedby a network. At this point one may apply an equal area crite-rion to this system, or using a time-series prediction algorithmdetermine whether or not the evolving swing is stable or un-stable. Such studies are currently under way, and it has beenfound that in a majority of cases the prediction can be accom-plished in about 250 ms after the start of the event, which is asimilar time frame as observed in the FloridaGeorgia interfaceproblem. Much work remains to be done in this area.

3) Security-Dependability: It should be recognized that arelay has two failure modes. It can trip when it should not trip (afalse trip) or it can fail to trip when it should trip. The two typesof reliability have been designated as security and depend-ability by protection engineers. The existing protection systems

Fig. 8. Adjustment of dependabilitysecurity balance under stressed systemconditions.

with their multiple zones of protection and redundant systemsare biased toward dependability, i.e., a fault is always cleared bysome relay. The result is a system that virtually always clears thefault but as a consequence permits larger numbers of false trips.High dependability is recognized as being a desirable protectionprinciple when the power system is in a normal healthy state,and high-speed fault clearing is highly desirable in order toavoid instabilities in the network. The consequent price paid inoccasional false trip is an acceptable risk under system normalconditions. However, when the system is highly stressed falsetrips exacerbate disturbances and lead to cascading events.

An attractive solution is to adapt the securitydepend-ability balance in response to changing system conditions as de-termined by real-time phasor measurements. Adaptive relayingwith digital relays was introduced on a major scale in 1987 [23],[24]. With three primary digital protection systems it is pos-sible to implement an adaptive securitydependability schemeby using voting logic (see Fig. 8). The conventional arrangementis that if any of the three relays sees a fault, then the breaker istripped. A more secure decision would be made by requiring thattwo of the three relays see a fault before the trip signal is sentto the breaker. The advantage of the adaptive voting scheme isthat the actual relays are not modified but only the tripping logicresponds to system conditions.

C. Power System ControlPrior to the introduction of real-time phasor measurements

power system control was essentially local. Some subsystemssuch as machines were controlled with only local signals. Othercontrol action was taken based on local measurements and amathematical model of the external world. External equivalentsare such models. The introduction of phasor measurements of-fers the possibility of control based on measurement value of re-mote quantities. Latency of the phasor measurements is an issue,but the fact that many of the processes are in the 0.22.0 Hzrange is encouraging. The phasor data will be time tagged sothat control could be based on the actual state of the system ashort time in the past. The frequencies involved are represen-tative of transient stability, electromechanical oscillations, andcertain overload phenomena. The frequency of measurements ofthe order of 1560 Hz is certainly sufficient to handle the con-trol task.

26 IEEE TRANSACTIONS ON SMART GRID, VOL. 1, NO. 1, JUNE 2010

Fig. 9. A typical control based on remote synchrophasor measurements.

Wide-area measurements allowed the system to react tothreatening situations without employing continuous feedbackcontrol. Monitoring of angles to detect possible instabilitiesand taking discrete switching controls in an attempt to militateagainst these events is a form of control made possible withsynchrophasors [25], [26]. The training of the switching logicfor such controllers has used simulation to train support vectormachines [27] and decision trees obtained from data mining[26]. Incorporation of synchrophasor measurements into newsystem integrity protection schemes (SIPS) has also beendescribed. [28]

Early applications with continuous feedback were aimed atproblems where a few phasor measurements could be applied toproblems in which the control objective were global in nature:for example an HVDC controller may be called upon to dampelectromechanical oscillations between two widely separatedareas of a power system. Control of HVDC systems, excita-tion control, power system stabilizers, and FACTS devices con-trol [29][37] were approached in a similar fashion. Because ofthe nonlinearity of these problems, the power system dynamicswere linearized and linear feedback based on various robust con-trol techniques employed. Uncertain parameters were used tomodel unknown time delays [37], unmodeled dynamics wereused to deal with the unmeasured portion of the system [38],and LMI techniques used to design a robust . orcontroller [29], [39].

These techniques can be used to coordinate a number of localcontrollers of different types. For example, a collection of powersystem stabilizers, dc lines, and FACTS devices could be coor-dinated to provide damping for a collection of interarea modeswith a variation of the technique used for PSSs alone [29]. Arepresentative scheme for such controllers is shown in Fig. 9.The local controllers retain their local input signals but have ad-ditional inputs from the wide-area controller. Selection of thePMU locations to be used and verification of the robustness ofthe overall scheme are important issues.

IV. CONCLUSION

With the growing interest in PMUs and WAMS throughoutthe world, it is clear that these systems will be implemented inmost major transmission networks. To a large extent the successof this endeavor depends upon adherence to the industry stan-dard governing the PMUs [2]. This standard is being revised toinclude requirements for compliance when power system is innonstationary state. Compliant PMUs will assure that PMUs ofdifferent manufacture can be used interchangeably.

The earliest applications of PMU technology in most powersystems are likely to be: a) validation of system models, andb) accurate postmortem analysis. As experience with real-timePMU data is gained by system operators, the next logical stepwould be to include PMU data in system state estimation proce-dures in energy management systems. With sufficient number ofPMUs installed, it would be possible to perform linear state es-timation, and it would be possible to track dynamic phenomenain real time. Ultimately, this ability will lead to the use of PMUsfor improving protection and control functions. The goal of suchimprovements is to make the power system less immune to cata-strophic failures and to reduce the severity of such failures whenthey do occur.

REFERENCES[1] A. G. Phadke and J. S. Thorp, Synchronized Phasor Measurements and

Their Applications. New York: Springer, 2008.[2] IEEE Standard for Synchrophasors for Power Systems, IEEE Standard

C37.118-2005 (Revision of IEEE Std 1344-1995).[3] R. O. Burnett, M. M. Butts, T. W. Cease, V. Centeno, G. Michel, R.

J. Murphy, and A. G. Phadke, Synchronized phasor measurement ofa power system event, IEEE Trans. Power Syst., vol. 9, no. 3, pp.16431650, Aug. 1994.

[4] D. Faulk and R. J. Murphy, Comanche peak unit no 2 100% loadrejection testUnderfrequency and voltage phasor measured acrossTU electric system, presented at the 47th Annual Texas A&M Conf.Protective Relay Engineers, College Station, TX, Mar. 1994.

[5] Z. Zhong, C. Xu, B. J. Billian, L. Zhang, S.-J. S. Tsai, R. W. Conners,V. Centeno, A. G. Phadke, and Y. Liu, Power system frequency moni-toring network (FNET) implementation, IEEE Trans. Power Syst., vol.20, no. 4, pp. 19141921, Nov. 2005.

[6] Final Report on the August 14, 2003 Blackout in the United Statesand Canada: Causes and Recommendations Apr. 5, 2004, U.S.CanadaPower System Outage Task Force.

[7] Responses Summary to Questionnaire on PMU Installation and Main-tenance May 2006, Eastern Interconnection Phasor Project (EIPP) Per-formance Requirements Task Team (PRTT).

[8] M. Zhou, V. A. Centeno, A. G. Phadke, H. Yi, D. Novosel, and H.A. R. Volskis, A preprocessing method for effective PMU placementstudies, in Proc. 3rd Int. Conf. Electric Utility Deregulation and Re-structuring and Power Technologies (DRPT 2008), Apr. 69, 2008, pp.28622867.

[9] R. F. Nuqui and A. G. Phadke, Phasor measurement placement tech-niques for complete and incomplete observability, IEEE Trans, PowerDel., vol. 20, no. 4, pp. 23812388, Oct. 2005.

[10] B. Gou and A. Abur, An improved measurement placement algorithmfor network observability, IEEE Trans. Power Syst., vol. 16, no. 4, pp.819824, Nov. 2001.

[11] A. Abur, Optimal Placement of Phasor Measurements Units for StateEstimation PSERC Publ. 06-58, Oct. 2005.

[12] M. Zhou, V. A. Centeno, J. S. Thorp, and A. G. Phadke, An alternativefor including phasor measurements in state estimators, IEEE Trans.Power Syst., vol. 21, no. 4, pp. 19301937, Nov. 2006.

[13] N. Korevaar, Incidence Is No Coincidence, Univ. Utah Math Circle,Oct. 2002.

[14] M. Zhou, Phasor measurement unit calibration and applications instate estimation, Ph.D. dissertation, Virginia Tech, Blacksburg, Mar.2008.

DE LA REE et al.: SYNCHRONIZED PHASOR MEASUREMENT APPLICATIONS 27

[15] S. H. Horowitz and A. G. Phadke, Boosting immunity to blackouts,IEEE Power Energy Mag., vol. 1, no. 5, pp. 4753, Sep.Oct. 2003.

[16] S. H. Horowitz and A. G. Phadke, Third zone revisited, IEEE Trans.Power Del., vol. 21, no. 1, pp. 2329, Jan. 2006.

[17] A. G. Phadke, D. Novosel, and S. H. Horowitz, Wide area measure-ment applications in functionally integrated power systems, presentedat the CIGRE B-5 Colloq., Madrid, Spain, 2007.

[18] [Online]. Available: http://www.nerc.com/~dawg/dawg-distur-bancereports.html

[19] V. Centeno, A. G. Phadke, A. Edris, J. Benton, M. Gaudi, and G.Michel, An adaptive out-of-step relay [for power system protection],IEEE Trans. Power Del., vol. 12, no. 1, pp. 6171, Jan. 1997.

[20] V. Centeno, A. G. Phadke, and A. Edris, Adaptive out-of-step relaywith phasor measurement, in Proc. 6th Int. Conf. Developments inPower Syst. Protection, Mar. 2527, 1997, pp. 210213, Conf. Publ.No. 434.

[21] V. Centeno, A. G. Phadke, A. Edris, J. Benton, and G. Michel, Anadaptive out-of-step relay, IEEE Power Eng. Rev., vol. 17, no. 1, pp.3940, Jan. 1997.

[22] V. Centeno, J. De La Ree, A. G. Phadke, G. Michel, R. J. Murphy, andR. O. Burnett, Jr., Adaptive out-of-step relaying using phasor mea-surement techniques, IEEE Comput. Appl. Power, vol. 6, no. 4, pp.1217, Oct. 1993.

[23] S. H. Horowitz, A. G. Phadke, and J. S. Thorp, Adaptive transmissionsystem relaying, IEEE Trans. Power Del., vol. 3, no. 4, pp. 14361445,Oct. 1988.

[24] G. D. Rockefeller, C. L. Wagner, J. R. Linders, K. L. Hicks, and D.T. Rizi, Adaptive transmission relaying concepts for improved perfor-mance, IEEE Trans. Power Del., vol. 3, no. 4, pp. 14461458, Oct.1988.

[25] D. Hu and M. Venkatasubramanian, New wide area algorithms fordetection and mitigation of angle instability using synchrophasors, inProc. Power Eng. Soc. General Meeting, Jun. 2007, pp. 18.

[26] S. Rovnyak, C. W. Taylor, and J. S. Thorp, Real-time transient sta-bility predictionPossibilities for on-line automatic database genera-tion and classifier training, presented at the 2nd IFAC Symp. Controlof Power Plants and Power Syst., Cancun, Mexico, Dec. 7, 1995.

[27] M. Venkatasubramanian, X. Yue, G. Liu, M. Sherwood, and Q. Zhang,Wide-area monitoring and control algorithms for large power systemsusing synchrophasors, presented at the Power System Conf. Exposi-tion, Pullman, WA, Mar. 1518, 2009.

[28] V. Madani, E. Taylor, D. Erwin, A. Meklin, and M. Adamik, Highspeed control scheme to prevent instability of a large multi-unit powerplant, presented at the Clemson Power System Conf., Clemson, SC,Mar. 1316, 2007.

[29] Y. Zhang and A. Bose, Design of wide area damping controllers forinterarea oscillations, IEEE Trans. Power Syst., vol. 23, no. 3, pp.11361143, Aug. 2008.

[30] N. Rostamkolai, Adaptive Optimal Control of AC/DC Systems,Ph.D. dissertation, Virginia Tech, Blacksburg, 1986.

[31] E. C. Manansala, Adaptive Power System Control, Ph.D. Disserta-tion, Virginia Tech, , Aug. 1989.

[32] N. Rostamkolai, A. G. Phadke, J. S. Thorp, and W. F. Long, Mea-surement based optimal control of high voltage AC/DC systems, IEEETrans. Power Syst., vol. 3, no. 3, pp. 11391145, Aug. 1988.

[33] E. C. Manansala and A. G. Phadke, An optimal centralized controllerwith nonlinear voltage control, Electric Mach. Power Syst., vol. 19,pp. 139156, 1991.

[34] L. Mili, T. Baldwin, and A. G. Phadke, Phasor measurements forvoltage and transient stability monitoring and control, presented at theWorkshop Appl. Adv. Math. Power Syst., San Francisco, CA, Sep. 46,1991.

[35] A. F. Snyder, N. Hadjsaid, D. Georges, L. Mili, A. G. Phadke, O.Faucon, and S. Vitet, Inter-area oscillation damping with powersystem stabilizers and synchronized phasor measurements, in Proc.POWERCON98, vol. 2, pp. 790794.

[36] M. A. Smith, Improved Dynamic Stability Using FACTS DevicesWith Phasor Measurement Feedback, M.S. Thesis, Virginia Tech,Blacksburg, 1994.

[37] G. L. Yu, B. H. Zhang, H. Xie, and C. G. Wang, Wide-area mea-surement-based nonlinear robust control of power system consideringsignals delay and incompleteness, in Proc. Power Eng. Soc. GeneralMeeting, Jun. 2007, pp. 18.

[38] X. Li, Q. Yan, and Y. Hu, Decentralized coordinated robust adaptivecontrol for AC/DC interconnected power system based on WAMS, inProc. APCCAS 2008, Nov. 30Dec. 3 2008, pp. 18961899.

[39] C. Dou, S. Jin, Z. Bo, Q. Jia, and H. Xue, Decentralized stabilizerdesign for model-based uncertain large interconnected power systemsbased on WAMS, in Proc. IPES 2007, Dec. 36, 2007, pp. 799806.

Jaime De La Ree (S80M80SM94) receivedthe B.S.E.E. degree with honors from The InstitutoTecnologico y de Estudios Superiores de Monterreyin 1980 and the M.S.E.E. and Ph.D. degree fromUniversity of Pittsburgh, Pittsburgh, PA, in 1981 and1984 respectively.

In 1984, he joined the ECE Department at VirginiaTech, Blacksburg, where he is currently the AssociateDepartment Head. His primary research area is ad-vanced protection, and control of power systems.

Prof. De La Ree is the recipient of the Sporn andthe Wine Teaching Awards. These are the most prestigious awards granted atVirginia Tech.

Virgilio Centeno (S88M88SM06) received theM.S. and Ph.D. degrees in electrical engineeringfrom Virginia Polytechnic Institute and State Univer-sity (Virginia Tech), Blacksburg, in 1988 and 1995,respectively.

He worked as a Project Engineer at Macrodyne,Inc., Clifton Park, NY, in the development of phasormeasurement units from 1991 to 1997. He joined thefaculty of Virginia Tech as a Visiting Professor in thefall of 1997 and became an Assistant Professor in thefall of 2001 and an Associate Professor in 2006. His

area of interest is wide-area measurement and its applications.

James S. Thorp (S58M63SM80F89LF03)received the B.E.E., M.S. and Ph.D. degrees fromCornell University, Ithaca, NY, in 1959, 1961, and1962, respectively.

He was a Faculty Intern, American Electric PowerService Corporation in 19761977 and an OverseasFellow, Churchill College, Cambridge University in1988. He was the Director of the Cornell School ofElectrical and Computer Engineering from 1994 to2001. He was the Charles N. Mellowes Professor inEngineering at Cornell University from 19942004.

He was the Hugh P. and Ethel C. Kelley Professor of Electrical and ComputerEngineering and Department Head of the Bradley Department of Electrical andComputer Engineering at Virginia Tech from 2004 to 2009.

Prof. Thorp was an Alfred P. Sloan Foundation National Scholar and waselected a Member of the National Academy of Engineering in 1996. He receivedthe 2001 Power Engineering Society Career Service award, the 2006 IEEE Out-standing Power Engineering Educator Award, and shared the 2008 BenjaminFranklin Medal with A. G. Phadke.

A. G. Phadke (M64SM97F80LF04) receivedthe B.Sc., B.Tech. (Hons.), M.S. and Ph.D. degreesfrom Agra University, IIT, Khargpur, IIT, Chicago,and the University of Wisconsin, Madison, in 1955,1959, 1961, and 1964 respectively.

He is a University Distinguished Professor (Emer-itus) at Virginia Tech, Blacksburg. He is a coauthorof three books: Computer Relaying for PowerSystems, Power System Relaying, and SynchronizedPhasor Measurements and Their Applications, andis the editor of and contributor to the book Handbook

of Electrical Engineering Computations. His primary research area is themicrocomputer based monitoring, protection, and control of power systems.

Prof. Phadke was awarded the IEEE Third Millennium Medal in 2000, namedthe Outstanding Power Engineering Educator by the IEEE in 1991, and receivedthe Power Engineering Educator Award of the EEI in 1986. He received theIEEE Herman Halperin Transmission and Distribution award in 2000. He wasthe Chairman of the Technical Committee of USNC CIGRE, and Editor-In-Chief of IEEE TRANSACTIONS ON POWER DELIVERY. He was elected to the U.S.National Academy of Engineering in 1993. He was awarded Honorary Doc-torate by INP Grenoble, France, in 2006. He shared the 2008 Benjamin FranklinMedal with J. S. Thorp.