Synchrophasor Measurement Technology in Power Systems: Panorama and State-of-the-Art

Phasor measurement units (PMUs) are rapidly being deployed in electric power networks across the globe. Wide-area measurement system (WAMS), which builds upon PMUs and fast communication links, is consequently emerging as an advanced monitoring and control infrastructure. Rapid adaptation of such devices and technologies has led the researchers to investigate multitude of challenges and pursue opportunities in synchrophasor measurement technology, PMU structural design, PMU placement, miscellaneous applications of PMU from local perspectives, and various WAMS functionalities from the system perspective. Relevant research articles appeared in the IEEE and IET publications from 1983 through 2014 are rigorously surveyed in this paper to represent a panorama of research progress lines. This bibliography will aid academic researchers and practicing engineers in adopting appropriate topics and will stimulate utilities toward development and implementation of software packages.

SECTION I. Introduction

Considerable efforts have been devoted to the subjects of synchrophasor measurement technology (SMT), phasor measurement unit (PMU), wide-area measurement system (WAMS), and their applications to power systems.

The literature explored in this paper preparation includes publications from the following IEEE and IET journals and standards in the period of 1983 through 2014:

1. IEEE Transactions on Power Systems (formerly IEEE Transactions on Power Apparatus and Systems)

2. IEEE Transactions on Power Delivery

3. IEEE Transactions on Smart Grid

4. IEEE Transactions on Energy Conversion

5. IEEE Transactions on Sustainable Energy

6. IEEE Transactions on Instrument and Measurement

7. IEEE Transactions on Circuits and Systems

8. IEEE Transactions on Systems, Man, and Cybernetics

9. IEEE Transactions on Industrial Electronics

10. IEEE Transactions on Industrial Informatics

11. IEEE Transactions on Industry Applications

12. IEEE Transactions on Control Systems Technology

13. IEEE Transactions on Signal Processing

14. Proceedings of the IEEE

15. IEEE Power and Energy Magazine

16. IEEE Communications Magazine

17. IEEE Signal Processing Magazine

18. IEEE Industry Application Magazine

19. IEEE Systems Journal

20. IEEE Journal of Selected Topics in Signal Processing

21. IEEE Journal of Emerging and Selected Topics in Power Electronics

22. IEEE Signal Processing Letter

23. IEEE Computer Applications in Power

24. IEEE Communications Surveys and Tutorials

25. IEEE Potentials

26. IEEE Security and Privacy

27. IEEE Intelligent Systems

28. IEEE Standards

30. IET Science, Measurement, and Technology

31. IET Generation, Transmission, and Distribution

31. Electronics Letters

32. Power Engineering Journal

The research issues elaborated in these publications are classified into nine topics as follows:

1. SMT algorithms and PMU/WAMS structural issues

2. Generic introductions over SMT, PMU, and WAMS

3. PMU placement techniques

4. State estimation (SE) consisting of or based on PMU data

5. Model validation, calibration, and extraction via PMU data

6. Fault/event detection and location using PMU data

7. WAMS-based dynamic/stability monitoring and prediction

8. WAMS-based control strategies

9. WAMS-based protection schemes

Among those listed above, categories 5 to 9 mainly correspond to the applications of PMU data from either local or wide-area perspectives. Fig. 1 shows the distribution of papers with respect to each aforementioned category. It can be observed that the most research has been devoted to the PMU constructional issues, category 1, a prerequisite for the SMT evolutions and developments of WAMS applications. The next prolific research area belongs to dynamic/stability monitoring and prediction. This observation is justified given the PMU prominent capability in capturing real-time and high-resolution information in a continuous data-stream format.

Fig. 1.

Distribution of papers with respect to each category.

Fig. 2 illustrates the distribution of papers versus the publication year. As shown, the chronological trend is highly increasing, which reveals the enthusiasm of research and engineering community in developing and materializing PMU and WAMS applications.

Fig. 2.

Distribution of papers with respect to the publication year.

The rest of this paper digests each aforementioned category in a specific section and cites the relevant references.

SECTION II. Category 1: SMT Algorithms and PMU/WAMS Structural Issues

Along with the liberalization of global positioning system (GPS), the precise system-wide synchronized measurements ($\approx 1~\mu $ s) turned out to be feasible. Combination of this opportunity with the phasor measurement algorithms, which are based on the symmetrical component analysis and have already been developed for digital protection purposes, led to the emergence of SMT. Subsequently, PMU was designed, manufactured, and successfully examined.

The accuracy of SMT has a direct bearing on the effectiveness of PMU applications and relevant functions. Accordingly, significant efforts have been dedicated to the development of efficient and precise measurement algorithms. Note as well that the majority of algorithms were based on preliminary achievements of digital protection relaying and were utilized by laboratory-scale or industrial PMUs.

Algorithms for calculation of phasors and/or local system frequency and rate of change of frequency are mainly based on discrete Fourier transform (DFT) [1]–​[5], Kalman filtering method [6], [7], optimized observer-based filters [8], coordinate transformation [9], modal Kalman filtering [10], zero-crossing method [2], quadrature demodulation [2], demodulation [11], direct-quadrature transformation [12], Newton type algorithm [13], least-square filtering [14]–​[17], short-time Fourier transform [18], Prony’s estimation method [3], [19], $z$ -transform and median filtering [20], neural networks [21], Newton type algorithm combined with least squares [22], constrained maximum likelihood estimation [17], enhanced phase-locked loop system [23], Blackman-Harris windowing technique [16], fast recursive Gauss-Newton adaptive filter [24], [25], and other various types of digital filters [26]–​[39]. Augmented versions of DFT consisting of smart DFT [40], [41], phasor-based DFT [42], and shifting window average method DFT [43], [44] were developed with the purpose of accuracy improvement. Sample value adjustment was another idea in promoting the SMT accuracy grade [45]. Other algorithms such as enhanced DFT [18], [46]–​[48], adaptive dynamic phasor estimator [49], partial sum based [50], multistage least-square [50], modified empirical mode decomposition filtering [51], and least-square method joint with short time Fourier transform [52] were proposed with the objective of diminishing the impact of decaying dc component. Filtering out the subsynchronous frequency components from the signals associated with series-compensated transmission lines is another effort conducted in this context [53]. A full cycle data is a common requirement in most reviewed algorithms, while phasor estimation based on phasorlets, which are signal segments of a fraction of a cycle, was established as well [54], [55]. A recursive wavelet transform was also developed which is capable of estimating the phasor parameters in a quarter cycle of an input signal [56]. As sinusoidal waveform distortions are the major difficulty in the phasor estimation, a wavelet analysis could first discriminate the discontinuities followed by an adaptive window algorithm to estimate the phasor quantities [57].

The accuracy of phasor measurement under dynamic conditions was improved by dynamic phasor estimates based on least-square method [58]–​[61], weighted least square [62], Shanks’ method with autoregressive moving average [63], DFT with Taylor derivatives [64], subspace based approach [65], enhanced Complex Kalman filter [66], consecutive DFT method [67], adaptive cascaded filters [68], Taylor-Fourier transform [69], [70], interpolated dynamic DFT-based estimator [71], and applying Taylor-LQG-Fourier filters [72]. Additionally, a quasi-positive sequence DFT algorithm was proposed as a single-phase measurement system [73]. Inaccurate frequency calculation under electro-mechanical transient circumstances and particularly at load buses has also motivated development of a new method on the basis of simplified model of generators and loads and with the objective of simulation studies [74]. Recently, a PMU simulator package built on the MATLAB-SIMULINK platform has been designed with education purposes [75]. This simulator yields a flexible environment for teaching phasor measurement and frequency estimation algorithms and phenomena. In addition to simulated signals, real-world measured point-on-wave data can be fed into this package.

Although the above mentioned measurement techniques are mainly tailored for the transmission system conditions, those suited for continuous monitoring of electrical quantities in distribution networks in terms of synchronized phasors have also been developed [76]–​[79]. Harmonic distortion and the impact of switching loads on the frequency estimation were of particular concerns in these techniques [80]–​[83]. Extension of synchrophasor algorithms to provide PMUs with the capability of accurately estimating harmonic phasors has been recently carried out in order to proliferation of PMU applications in active distribution networks [84], [85].

The IEEE standard C37.118 for synchrophasors was compiled for many years and revised several times [86]–​[93]. Since PMUs are subjected to inputs under transient conditions [94], the last version of this standard has covered such attributes. In spite of the compatibility of commercial PMUs with the IEEE standard C37.118, laboratory testing exposed noticeable disparities among the phasor estimates of PMUs associated with various manufacturers. This observation, indeed, is mainly due to the phasor computation algorithms and necessitates quality assessment techniques [68], [95]–​[106], compliance analysis [107], calibration of synchronized phasor measure- ments [108], [109], and field testing [110], [111]. The IEEE standards C37.242 for synchronization, calibration, testing, and installation of PMUs and C37.244 for phasor data concentrator (PDC) requirements have been recently published to assist users to specify the performance and functional requirements of typical PMUs and PDCs [112], [113].

Having experienced the effectiveness of synchrophasor applications in strengthening the power system security, new intelligent electronic devices (IEDs) are being enabled to work as PMU devices [114]. Miscellaneous substation devices such as protection relays, digital fault recorders, PMU, etc. are being integrated by various vendors. Amongst, multipurpose platform (MPP) is a newly introduced electronic equipment that combines the abilities of a power quality (PQ) analyzer, an event logger, a PMU, and an inter-area oscillation identifier, all in one device [115]. The closed source of commercial PMU devices is perceived as an obstacle to emerging innovations in the synchrophasor development and application context. As an alternative, open platforms for the development of PMU technology specifically by university researchers have been proposed, amongst is OpenPMU [116].

The stochastic assessment of PMU availability is a crucial requirement due to the growing reliance on the WAMS infrastructure within the operation and control of power systems [117]–​[122]. Besides, spoofing attacks on the GPS receiver of a PMU are a potential threat for the correct phase angle reporting of voltage or current measurements provided by the PMU [123]–​[125]. Such spoofing attacks can induce serious malfunctions in control center applications, e.g., angular stability monitoring, relying on the phase information from measurements provided by PMUs.

Recent investigations on blackouts have revealed the indispensable role of the malfunction of monitoring/control infrastructures. In many occasions, the consequence of a severe event could have been significantly reduced if the system operators had a relatively accurate situational awareness on the stressed power grid. Hence, the power system reliability assessment, which was previously being fulfilled with the assumption of complete observability and controllability, should be revised incorporating the SCADA or WAMS networks deficiencies [126].

Development of new applications to enhance the operation of the electric grid based on high-reported rate PMU data necessitates exploitation of high-bandwidth and networked communication systems. For any particular transmission grid, a comprehensive analysis to simulate, design, and test the adequacy of a communication and computation system is of an extreme importance [127]–​[129]. Advanced comprehensive sampling techniques can be adopted to minimize communication bandwidth and delays while sending synchrophasors [130].

SECTION III. Category 2: Generic Introductions Over SMT, PMU, and WAMS

Following the actualization of precise measurement devices, the first requirement for realizing the WAMS was the time synchronized sampling over an entire power system in order to simultaneously capture voltage and current phasor quantities at specific instants of time [131]. Although most communication systems such as leased lines, microwave, or AM radio broadcast were envisaged as potential means for the wide-area sampling synchronization, their accuracy levels were deemed too coarse for the practical usage [132]. The application of multiplexed fiber optics for the sampling synchronization purpose also suffered from insufficient precision [132]. In contrast, exploiting satellite signals transmitted by GPS or its Russians counterpart GLONAS was considered as a feasible solution [133]. The precise satellite signals available all over the world also let PMUs furnish the measured data with time tags denoting the exact time of the corresponding measurement.

Having realized the wide-area sampling synchronization, preliminary successes of prototyped implemented systems were reported [134], [135], and a broad spectrum of PMU and WAMS applications was gradually opened up [136]–​[138]. These judicious applications were recognized as effective helps in reducing the frequency and severity of catastrophic failures [139]. Moreover, the potential capabilities of PMU data in both local and system-wide perspectives were identified among various solutions to detect the initiation and propagation of severe disturbances and consequently to avert brownouts or even blackouts [140]. This enthusiasm was intensified by other basic facts consisting of (i) economic pressures on the electricity market and on the grid operators to maximally utilize the high voltage equipments, (ii) rather quick changes in the system operating conditions rendered by market-driven environments, and (iii) highlighted concerns for enhancing the physical security of power grids [141].

To achieve the aforementioned potential benefits, advancements in other areas such as WAMS structural architecture as well as control center hardware facilities and software packages were further focused to match with proven advancements in SMT, PMU manufacturing, and communication facilities. In addition to technical design issues [142]–​[144], these efforts comprised a process of business justification of WAMS applications, a qualitative cost/benefit analysis with a particular emphasis on the beneficiaries, and a roadmap for WAMS implementations and its applications development [145].

To date, various publications have reported the evolution status of WAMS practical applications in different utilities and countries. Among them are North America [146]–​[149], continental Europe [146], [147] Brazil [146], [147], China [146], [147], [150], India [146], [147], Mexico [146], Nordic countries [146], Russia [146], [147], Korea [151], British Colombia Transmission Corporation and British Colombia Hydro [152], and Bonneville Power Administration and Tennessee Valley Authority [153]. These reports revealed the maturity of the elementary local and system-wide SMT applications [154]–​[156].

In contrast, more advanced functions of WAMS realizing smart transmission grids are still in conceptual stages [157], [158]. These functions could be categorized into two different classes:

• Some applications, although are theoretically well developed, need widespread deployment of PMUs making the whole power network observable.

• Others are practically in the infancy period and are currently under investigation and research.

For the sake of illustration, a complete WAMS based state measurement (which is going to serve instead of SE) belongs to the first class and wide-area automatic control of power systems lies within the second one. Obviously, with the current high rate of PMU deployment, the first class applications would actualize in the near future; while the practicality of the second class depends on the further advancements in all power engineering, control strategies, communication facilities, and the information technology discipline. In this respect, there are some propositions consisting of a flexible integrated phasor system to offer transmission owners access to their phasor data [159], an IP-based decentralized and data centric information that can securely support the next generation grid [160], an adaptive routing algorithm over packet switching networks to guarantee the quality of service for the essential measured data [161], a novel cyber-physical model through a cooperative information exchange between the controllers embedded in the system and the network operator in future energy systems [162], and a set of strategies to protect the anonymity of synchrophasor data against passive traffic analysis attacks [163], [164]. The WAMS dependency on high-performance communication systems and the impact of ICT architecture on WAMS reliability were speculated likewise in the existing literature [165]–​[170].

Due to the early high-cost PMU manufacturing process as well as its expensive installation costs in high voltage substations, the wide-area frequency monitoring network (FNET) was developed in 2003 as a low-cost and rapidly deployable alternative [171]. FNET’s building block is the frequency disturbance recorder (FDR), which is readily installable at the 120-V distribution level. This infrastructure is mainly to monitor the frequency of the network, and its potential is expectedly limited compared with the WAMS. Although FNET showed various advantages [172] in North America, it is not broadly deployed in other countries. This observation could be due to limited applications of FNET along with today’s lowered PMU manufacturing and installation expenses.

SECTION IV. Category 3: PMU Placement Techniques

Optimal PMU Placement (OPP) problem was focused in early years after confirmation of WAMS potentials [173]. This task was known as an underlying requirement in effective and consistent developments of WAMS infrastructures. Though the OPP problem was justified in early stages by an emphasis on the considerable expenses of vast PMU installations, the overwhelming set of power system data associated with a large number of PMUs thereafter highlighted the OPP necessity more and more [141], [174].

The OPP is inherently an NP-complete problem with a solution space of $2^{N}$ possible combinations for an $N$ -bus electric power system [175]. Therefore, the OPP problem is considered as a combinatorial optimization problem and valuable research has been accomplished in this area since 1993 [173]. Although the reported works could be categorized based on the optimization methods employed, they are reviewed in the following with the viewpoint of problem definition and investigated technical aspects. A more comprehensive taxonomy of PMU placement methodologies was presented in [176].

Traditionally, the basic constraint of monitoring infrastructures was the holistic network and elements observability. In this sense, majority of OPP algorithms have considered the network base case observability as the optimization constraint. These efforts comprise a dual search algorithm along with simulated annealing [174], genetic algorithms (GA) [175], [177], the immunity GA [178], integer linear programming [179]–​[181], weighted least squares algorithm [182], and recursive Tabu search method [183]. Further researches argued that the OPP problem might have multiple optimal solutions among which the one with maximum measurement redundancy is more desirable [184]–​[187]. Such a solution, in essence, leads to a more precise SE and has higher robustness against element outages. Other efforts underlined the growing reliance of power system operation on the WAMS applications and having a more reliable and robust WAMS infrastructure. These works addressed the OPP problem with a set of constraints assuring the network complete observability at the event of transmission line and/or PMU outages [184], [185], [188]–​[195]. Ensuring the network observability in both normal condition and controlled islanding mode was accommodated in the OPP problem as well [196]. Other sorts of technical constraints were also added to the OPP problem with an intended specific functionality such as fault location observability [197], bad data detection in SE [198], [199] parameter error identification [200], reducing the variances of the SE errors and increasing the local redundancy [201], enhancing topology error processing [202], defending against data injection attacks [164], generating the best estimate of the swing model [203], minimizing the error of SE [204]–​[206], and optimizing a useful metric which accounts for three important requirements in power system state estimation: convergence, observability, and performance [207]. PMUs could also be strategically placed in power networks to offer the detection and identification of parameter error on a single-edge cutset (critical branch) or double-edge cutsets (critical branch pairs) [208].

Concurrent optimization of placement of PMUs and communication links is another suggestion for the systems in which the communication infrastructure is mainly expanded for the synchrophasors collection [209], [210]. In the technique proposed in [209], permanent and random intermittent PMU outages are taken into account and the steady-state availability of synchrophasor data at each bus meets a prescribed level.

Optimizing the number and location of PMUs and smart metering devices in future active distribution networks is an urgently requested task [211]. Such a robust measurement infrastructure is essential for secure and optimal smart distribution grid control and operation.

All of the aforementioned studies brought about a considerable evolution of the OPP algorithms. However, it would take a rather long time for the deployment of all PMUs required for the network basic/contingency-constrained observability. Hence, staging (phasing) the OPP in time was investigated as a pragmatic requirement [174], [212]–​[215]. The multi-stage OPP was the name assigned to the envisaged problem. In these studies, determination of PMU placement schemes with a given limited depth-of-unobservability associated with subsidiary stages was aimed. These techniques were further extended by giving a higher priority to key buses such as those required for transient and dynamic stability monitoring and buses with high connectivity [216].

The OPP problem in either single- or multi-stage definition has also been tackled in probabilistic manners taking the probability of elements outages into account [120], [217]–​[219]. More installation of PMU devices in a power network would assure higher levels of system security and electricity reliability. Having converted the reliability enhancement to monetary measures, the OPP problem can be analyzed within a cost/benefit framework [220].

The monitoring of power system dynamics with a few PMUs was another perspective of the OPP problem [221]–​[224]. In this sense, dynamically coherent areas of a power system were determined. Thereafter, the sitting process of a given number of PMUs was performed with the objective of maximizing the information content of captured signals while minimizing their cross correlation.

As the notion of phasors does not correspond to dc transmission circuits, the OPP problem in hybrid ac/dc systems should be accordingly revised [225]. Another practical aspect regarding the limitation of PMU measurement channels is letting the optimization algorithm to place more than one PMU in a given substation. This relaxation can be economically desirable due to some common expenses of installing two or more PMU devices in a same substation, such as communication infrastructure procurement.

SECTION V. Category 4: State Estimation (SE) Consisting of or Based on PMU Data

Power systems were traditionally monitored through the supervisory control and data acquisition (SCADA) infrastructure and energy management system (EMS) installed at control centers. Remote terminal units (RTUs) located in substations and power plants nourishes SCADA by their scalar data such as magnitude of voltage and current quantities, active and reactive powers, present settings of tap changing and phase shifting transformers, and breakers status. These raw datasets are refined by SE for reducing their errors, compensating for unavailable values, and eliminating bad data. The power system operators monitor and control the system with the help of SE programs embedded in EMSs. SMT helps conventional SE engines in two ways:

• Its measurements are precisely synchronized and can significantly eliminate the effect of time skew error in the SE solution.

• Its direct measurement of state variables, i.e., phase angles in addition to magnitudes, brings a remarkable enhancement in SE convergence and accuracy.

In the context of PMU data application in SE, the preliminary tasks are focused on the inclusion of PMU data in the conventional SE program [226], [227]. Subsequently, various modified versions of the SE formulation were investigated to comprise the voltage and current phase angles in the estimation process of the system true state [228]–​[242]. Observability/criticality analyses and bad data processing have also been revisited in the presence of PMU measurements [243], [244]. In SE algorithms, each measurement has a specific weight, which is inversely proportional to the uncertainty associated with the measurement. Hence, further research covered the computation of uncertainty associated with two types of PMU measurements, designated as direct measurement of quantities at the PMU-bus and indirect measurements of buses adjacent to the PMU-bus [245]. The other technical concern in adding the PMU data to the conventional SE algorithms arises from dissimilar reporting rates of SCADA and WAMS [246]. To overcome, the buffer length of PMU measurements should be optimally determined which can be also helpful in specifying the corresponding weights in SE [247].

The underlying attribute of the reliability of SE process, which is strongly dependent on the measurement redundancy, was also revised to account for the impact of PMU data [248]. Since in some PMUs it is observed that the voltage and current phasors exhibit phase biases, which degrade the conventional SE performance, phasor-data-based SE algorithms were proposed for phasor angle bias correction [249], [250].

The above techniques necessitated the revision of SE software packages; while the revised ones were still nonlinear and cumbersome to tackle. Other alternatives to include the additional PMU data in SE were developed in which the existing SE software was left in place with no modification [251], [252]. In this method, the solution of SE with traditional measurements is combined with PMU measurements in a post-processing linear estimator. Tracking SE to reconstruct coherent system states can be considerably enhanced by accommodation of PMU data [253].

Other research activities speculated the multi-area SE problem where PMUs, scattered in various control areas of a large-scale power system, improved the estimation accuracy and bad data processing efficiency [254]. Acceleration of a large-scale SE program as another vital challenge was considerably overcome by distributed SE algorithms in which PMUs were considered to coordinate the SE solution of each subsystem [201], [255]–​[257]. The impact of conventional measurement failures on the SE performance was also comprehensively explored where the existence of even few PMUs was recognized very helpful in preserving the SE performance [258].

SCADA scans RTU every 2-5 seconds. Hence, the inclusion of PMU data in the traditional SE, although is able to improve its reliability and accuracy, does not alleviate the SCADA slow and quasi-stationary images over the system status. It is expected that future WAMSs with plenary PMUs would directly measure all state variables and the conventional notion of SE may not exist anymore. These programs with trivial computations will provide a dynamic view of the network under supervision. This assumption was made in few works comprising the calculation of state variable uncertainty in a PMU-based SE algorithm [259], the SE-based line outage detection with only PMU data [260], [261], SE of a system whose parameters are known to be within certain tolerance bounds [262], and comparing some various SE formulations [263].

PMU devices might be synchronized with each other using a precision time protocol (PTP), defined in the standard IEEE 1588, particularly in substations where the series of IEC 61850 standards have been introduced. The impact of this synchronization solution on the measurements quality was examined by applying a SE procedure. It was shown that the accuracy of synchrophasor measurements meets the values usually required for application in power system SE [264].

Distribution class SE for application in smart distribution grids is a new research topic where the SE is customarily formulated in three-phase fashion tailored to the unbalanced operation of distribution systems. The wide deployment of synchrophasor measurements throughout the feeders can bring about remarkable improvements in the distribution system SE performance [265]–​[268]. With the waveform recording capability of nowadays PMU devices (fault recording functionality), the harmonic snapshot of power system turns out to be feasible. These harmonic data can thereafter nourish harmonic SE algorithm which was recently revised based on harmonic phasor data [269].

SECTION VI. Category 5: Model Validation, Calibration, and Extraction via PMU Data

Along with the growing complexities of power systems, rendered by the market-driven operation and integration of intermittent energy resources, the simulation becomes the only means to understand the behavior of such a huge system. The power system operators and planners require realistic component and system models in simulations to take proper decisions in operation processes and planning studies. By the time of installation of primary PMUs, pragmatic observations revealed a noticeable difference between the simulation results and real-world outcomes captured by PMUs [270]. Accordingly, one of the early-realized applications of PMU data was introduced as the improvement of the component and system dynamic and steady-state models for both on-line and off-line analyses.

Following to observations on inaccuracy of the existing models, considerable research activities were dedicated to the problem of model verification and correction based on PMU data. These works used a large library of disturbance recordings to eliminate single event abnormalities and led to revisions of some hydro turbine-governor models [270] and power plant/generator dynamic models and states [271]–​[274]. Finding the erroneous components in a complex and large-scale power system with thousands of elements was renowned as a challenging endeavor and further WAMS-based simulation validation works were conducted [275]. Additionally, PMU measurements were utilized to validate models of fixed speed induction generator-based or converter-based wind farms during frequency transients [276], [277]. These efforts mainly focused on the validation of dynamic models of components because PMUs can record the synchrophasors during the dynamics, something which was not feasible before. Application of PMU measurements in the error identification of steady-state models of other components, such as transmission lines, revealed a significant enhancement too [278], [279].

In addition to the validation and correction of component models, some research efforts were devoted to the identification of the power system equivalent model. The real-time PMU data were employed to build the equivalent model of the power system for transient stability control through flexible AC transmission system (FACTS) devices [280]. A similar effort to develop an HVDC model for online stability monitoring of the integrated ac/dc systems has been carried out as well [281]. Estimating a reduced model to represent the inter-area dynamics of power systems was also investigated by PMU measurements [282]–​[284]. The method elaborated in [282] was then generalized to incorporate the effect of intermediate dynamic voltage control equipments, such as static VAr compensator (SVC) and synchronous condenser [285]. Recent research efforts in this area were devoted to the PMU-based identification of grid impedance matrices, which are necessary for the monitoring and protection of modern power systems [286] and Thevenin equivalent parameters [287]. The developed methods made the dynamic tracking of parameter changes a viable capability. In other works, online identification of parameters associated with the power system equivalent was investigated using the phase-plane trajectories of PMU measured variables [288], or by the application of subspace system identification techniques [289]. Tracking the system dynamic response to generator tripping events is another superior capability of synchrophasors in real-world practices [290].

Power systems dynamic responses cannot be assertively simulated unless accurate load models are available. System-wide load modeling in the transmission and sub-transmission levels is heading toward the application of WAMS synchronized data. The major reason for this tendency is the high resolution of PMU data with phase angle information.

Modelers intend to achieve more accurate models of components and the system, and they usually exploit complex and data-mining-based black box models. On the opposite side, model users which are system operators and planners seek transparent and interpretable models for decision making processes and post-event studies. To this end, a compromised state of transparency versus accuracy of the models should be sought [291].

SECTION VII. Category 6: Fault/Event Detection and Location by PMU Data

Electric power grids, in transmission and distribution levels, are always prone to faults caused by a variety of situations such as equipment failures, accidents, adverse weather, etc. In faulted power systems, either the system is at risk in which the customers’ electricity service is interrupted or the system security is seriously threatened while its entire demands are still being served. Accordingly, precise and fast detection and location of network faults were historically perceived as a crucial task in the system inspection, maintenance, and repair processes. Note that application of phasor quantities in power system protection goes back several years earlier than the PMU emergence; its review is therefore out of the scope of this paper. Here, merely the efforts in which the application of PMU data has been explicitly declared are reviewed.

PMUs deployed at transmission line ends, in addition to positive, negative, or zero sequence phasors, compute three-phase values of bus voltage and line current as well. Furthermore, the sampling clocks of various PMUs, located several kilometers away, are precisely synchronized. These features of PMU outputs actualized an advanced fault detection and location algorithm, which has an adaptive inherent and an ample robustness against system noise and measurement errors [292], [293]. Additionally, the deviation in line parameters caused by simplified calculation methods, ambient temperature, or aging phenomena, was eliminated by an effective estimation technique based on the PMU accurate data available at both ends of transmission lines [292], [293]. These efforts, which were devised for the situation of having permanent faults, were next revised for arcing transient faults [294], [295]. Furthermore, the proposed algorithms in [292] and [293] were further extended and tailored for three-terminal and N-terminal (N$\ge 3$ ) transmission lines [296]–​[298]. Fault detection and location in multi-terminal transmission lines were focused by other research activities too [299]–​[304]. Wide-area synchrophasor data collected from many PMUs are materializing the system-wide fault diagnosis and location algorithms [305].

The crucial task of fault detection and location was explored by a multistate network inference in which phase angles across the buses were modeled through a Markov random field [306]. It was also argued that when a power system is under dynamic conditions, such as power oscillations, the fault voltage and current cannot be expressed as standard sinusoidal functions [307]. Hence, a fault location estimator was devised in which the voltage and current waveforms were considered as time-variant signals whose magnitude and frequency are changing against time.

Fault detection and location in parallel transmission lines, as a common practical configuration, were known as a more complicated problem due to some phenomena such as mutual coupling effect, shunt capacitance, etc. Algorithms based on PMU data demonstrated acceptable performances for parallel transmission lines in either transposed or untransposed arrangements [308], [309]. Series compensation of transmission lines has always been desired for some advantages such as the increasing power transfer capability, controlling the network power flow, damping dynamic oscillations, etc. Since the voltage drop of a series device is unknown during fault periods, the protection of series-compensated transmission lines has been a challenging subject to date. PMU measurements at both ends of transmission lines offered a great potential for detection and location of series-compensated lines without requiring the voltage drop values based on the compensation device model [310]. Although transmission line parameters are required in some techniques, PMU data can be directly used for online estimation of series-compensated line parameters and Thevenin’s equivalent of the system to ensure considering the actual operating conditions [311].

Phasor estimation algorithms embedded in PMU devices offer calculation of signal harmonic contents up to higher orders. This information, along with the symmetrical component data, was known very helpful in fault detection and location of transmission lines as well as in discrimination between permanent and transient faults [312]. This discrimination can be used in blocking reclosing of transmission lines when permanent faults occur.

The majority of the aforementioned works used voltage and current measurements at one or both ends of transmission lines. In another research, it was criticized that PMU current inputs are usually fed through the protection cores of current transformers; hence, the accuracy of PMU current measurements might be inadequate for a precise fault location. This work, assuming a given bus admittance matrix at the time of fault, has developed a new fault detection and location approach which uses only PMU voltage measurements [313]. This approach is applicable in both transposed and untransposed transmission lines and was then extended for multi-terminal lines [314]. The existing algorithms, although achieve a high accuracy in locating faults, necessitate a broad deployment of PMUs which is not the case in many designed WAMS networks or at least in the transition period toward a plenary WAMS infrastructure. Therefore, a fault-location algorithm based on fewer voltage measurements for large-scale transmission networks has been proposed which can be promising due to its low cost and practicality [315]. Phasor measurements could also be utilized in order to unveil power-line outages which is extremely vital not only to prevent cascaded events but even for routine monitoring and control tasks in smart power system control centers [316]–​[320].

Due to relatively large capacitance values in underground cables, the detection and location of their faults were deemed to be of technical difficulties compared to overhead transmission lines. PMU measurements, including voltage and current phasors at a single end of a cable, were employed as well in an efficient fault location algorithm [321]. Likewise, two-terminal multi-section compound transmission lines, which combine overhead lines with underground power cables, suffer from effective protection solutions. These compound transmission lines are usually adopted to achieve a compromise between construction cost and environmental concerns. PMU data proposes powerful algorithms to locate a fault; no matter the fault is on the overhead line or underground power cable [322].

A dominant portion of electricity interruptions is originated by the distribution system failures. Accurate locating of faulted segments can significantly expedite the dispatch of repair crews and restoration process. However, this task is getting more difficult with modernization of distribution grids by growing penetration of distributed energy resources. Synchrophasor measurements, with superior resolution and accuracy attributes, offer a practical and effective solution to this challenge [323].

While a dynamic event can be initiated by a fault in the transmission network, it may be caused by tripping a generator or switching a transmission line in or out. In addition to the fault, the PMU data has demonstrated a salient potential for event detection and location. A method was devised for detection and location of generator trips based on generator rotor angle measurements obtained using PMU data [324]. Unintentional islanding is among power system disturbances in which a part of the network becomes isolated from the rest. PMU phase angle measurements and frequency estimations are of great values for timely detection of islanding events [325], [326]. Additionally, PMU data revealed prosperous outcomes in locating disturbance sources for low-frequency oscillations in a real-world application [327]. Along with the quick proliferation of synchrophasor data and recordings, advanced and intelligent techniques are being proposed with the purpose of single/multiple events/faults early detection and clustering [328]–​[330].

SECTION VIII. Category 7: Wams-Based Dynamic/Stability Monitoring and Prediction

Today’s SCADA systems are neither fast enough to track dynamic events nor capable to monitor key stability indicators such as phase angles. Besides, computation of phase angles with SE algorithms is a time consuming process and is executed every several minutes. Since the advent of WAMS infrastructure, its salient feature in furnishing control centers with real-time dynamic situational awareness has remarkably advanced the power system frequency/voltage stability monitoring and prediction. Real-world evidences and laboratory-based platforms have, thus far, confirmed such expected advancements [331]–​[340]; nonetheless, undergoing research and development activities around the world would lead to forthcoming supplementary advancements and achievements.

Primary research in this is focused on the application of PMU data in determining whether an evolving event will eventually be stable or unstable. Most of the transient stability assessment techniques, while simple in off-line applications, are too complicated for real-time uses. Decision trees, as a type of classifier that can be constructed off-line from a training set of examples, were used to classify a transient swing as either stable or unstable on the basis of real-time PMU data [341]–​[343]. Moreover, similar efforts were conducted based on the largest windowed Lyapunov exponent index [344], [345], clustering estimation integration along with a reduced-order modeling [344], [346], decoupled PQ integration [347], artificial neural networks (ANN) [348], fuzzy neural networks [349], ball-on-concave-surface mechanics system [288], support vector machine classifier [350], [351], unscented Kalman filter [351], and derivative of total angle separation index [277].

Historically, power system operations have been challenged by stability margins, and dynamic stability (security) assessment has been a vital function for operation centers. Application of PMU data in this context was widely investigated and various practical experiences across the world demonstrated its effectiveness and worth. Among pioneer works was the use of a time-frequency-based approach to classify the contingencies with respect to the loss of voltage or frequency stability [353]. Other research on the analysis, discrimination, interpretation, and visualization of wide-area PMU measurements suited for dynamic phenomena was conducted through various methods including the numerical algorithm for subspace state space system identification [354], recursive adaptive stochastic subspace identification [355], Hilbert technique [356], energy function method [357], empirical mode decomposition approach [358], [359], hybrid DFT, finite-impulse response, prototype algorithm, and least-square error methods [360], decision trees [361]–​[365], regression analysis [366], autoregressive moving average model [367], correlation and autocorrelation functions analysis [368], [369], correlation and partitioning around medoid clustering [370], regularized robust recursive least squares method [371], wavelet shrinkage analysis [359], adaptive sampling Prony analysis [372], a model-based approach suited for unbalanced power systems [373], and a characteristic ellipsoid technique [374]. As a crucial function, monitoring of inter-area oscillations through the real-time PMU data and estimation of oscillation and damping characteristics were broadly investigated as well [375]–​[383]. Identification of dominant inter-area oscillation paths is another rigorous task which can be simply executed from ambient measurement of PMUs [327], [384].

Determination of security regions requires computational efforts which grow dramatically with the number of critical parameters to be evaluated. This difficulty could be tackled by advanced boundary search methods based on PMU information [385]. The generator coherency determination in an inter-area oscillation, which is based on the estimation of generator rotor angle and speed variables [386], [387], was suited for online applications based on WAMS infrastructure [388]. Additionally, voltage across an area of a network is a notion to capture synchrophasor measurements at the border buses and is used to monitor area stress and whether line trips occur or not [389].

Voltage instability, caused by either an unexpected raise in the load level or a loss of an important transmission line or generator, has been concerned for many decades as the origin of several collapses in many countries. The majority of techniques developed for voltage security monitoring (VSM) were usually computationally cumbersome for on-line applications. However, real-time PMU-data realized the on-line VSM techniques, which were mostly based on the artificial intelligence methods such as fuzzy neural networks [390]. It was also discovered that the voltage stability margins derived on the basis of a constant-power load may be pessimistic for the systems with mixed load types including the voltage-sensitive loads. For such situations, the wide-area PMU data showed that voltage phasors contain enough information to detect the voltage stability margin directly from their measurements [381], [391]–​[396]. Further research activities were also dedicated to this area, including the application of Tellegen’s theorem to determine Thevenin’s parameters [397], decision trees [342], [398], [399], algebraic modeling of the system trajectory [400], [401], ANNs [402], and a network decoupling transformation [403].

One of essential tools for network security monitoring is the $N-1$ contingency analysis. The associated results help system operators specify whether or not the power system will remain intact at the event of single-order contingencies. In case of any probable violation, the operator needs to apply a set of preventive actions such as generation re-dispatch. Resulted by computational burden of repeated power flow studies, contingency analysis is usually fulfilled through linear-sensitivity distribution factors (DFs), such as injection shift factors, power transfer distribution factors, and line outage distribution factors. Conventionally, DFs are computed by model-based approaches and might not absolutely correspond to the system real situations. Instead, wide-area PMU data actualize near real-time computation of adaptive DFs [404].

SECTION IX. Category 8: WAMS-Based Control Strategies

Succeeding monitoring, trending, and prediction of transient, dynamic, and voltage stability metrics, proper control and remedial actions are to be taken in time for preserving the power system security and averting major supply interruptions. PMU measurements are instantaneous and have a high resolution; hence, they can be used to take local/centralized corrective/preventive control actions for transient, dynamic, and voltage stability, as well as thermal constraints [405]–​[407]. It is worth noting that the standard message format of PMU and PDC data, described in the IEEE standard C37.118, has a command type which can convey the control signals toward the PMU locations. This category has the least mature suite of applications and presently very few WAMS-based control applications are in use. Any malfunction of automated controls could produce unintended consequences for grid operations. Prosperous deployments of these controls require rigorous control system engineering with approved robustness under a variety of operating conditions, controller tuning and testing, controller self-diagnostics, etc. However, with the enduring research and development activities, the deployment of WAMS-based controls on a wider scale is anticipated in the near future. Comprehensive simulation platforms, such as secure operation of sustainable power systems which offers real-time simulations in closed-loop, are essentially requested for hastening the validation process of wide-area applications [408]. One practical challenge, which has been ignored in majority of research activities, belongs to the signal latencies in the wide-area control. This attribute can degrade or even deteriorate the performance of control systems and has to be accounted for in studies and implementations.

For a long time, it was well perceived that quick measurements in the early stage of a disturbance would estimate its behind transient energy, and a quick control action, e.g., generation dropping or switching, could counteract the disturbance and stabilize the system. Primary research revealed that the transient energy could be quantified through PMU measurements for activating control actions [409], [410].

Applying wide-area PMU measurements to solve special dynamic stability problems is very intricate. In the last decade, an extensive research body was dedicated to this context. Developing a decentralized/hierarchical control approach with significant advantages of reliability and flexibility for inter-area oscillatory modes was among the pioneers [411]. Improving power system dynamic performance was also testified by means of WAMS-based controllers of automatic voltage regulators (AVRs) [107], [412], power system stabilizers (PSSs) [413], high-voltage DC (HVDC) transmission lines [414]–​[417] FACTS devices [418]–​[424], energy storage devices [413], [417], [423], [424], variable-speed wind farms [425], or doubly-fed induction generator (DFIG) wind farms [426]. Some research efforts were dedicated for designing control systems compensating for the communication limited bandwidth and latencies in feedback and command signals [415], [422], [427]–​[435]. Wide-area power system damping controllers resilient to communication failures have also been researched and effectively accomplished [436], [437].

In terms of voltage security control, the PMU data was utilized and revealed a great enhancement in steering the power system away from the critical points of the voltage collapse [438]. This vital task was done by activation of available reactive-power reserves [391], [439], freezing of tap changers [391], adjusting shunt power electronic devices such as SVCs and active/hybrid filters [440], or integrating dynamic voltage restorers [441]. With the raising interest to deploy PMU devices in the future distribution networks, the volt/var control of feeders would be accordingly heightened [442].

Another achievement in the field of WAMS-based control is related to the virtual synchronous islanded operation of an isolated grid while it is not electrically connected to the main network [443]–​[446]. Since the island is held in synchronism, its reconnection to the main system can be performed at any time with minimum transient effects and with no potential out-of-synchronism damages.

Following to any power system black or brownout, the restoration strategies should be taken in use whose main objective is to maximize the restored demand in the shortest period of time. Three different stages of restoration efforts are conceived as black-start of generating units, network reenergizing, and load restoration. SMT can be of a great assistance in either of above stages by accelerating the processes, optimizing the paths, and limiting unsuccessful tries. A sectionalizing method for the build-up strategy in power system restoration is among effective solutions in which each island is preserved fully observable by means of PMU measurements [447]. The amount of restorable load is essential in the proper design of the restoration process and sequences. This task can be promisingly conducted by the access to WAMS data [448]. It should be also noted that due to the limited number of switching devices, the restorable load has a discrete nature and this feature should be necessarily accounted for in the load restoration processes.

SECTION X. Category 9: WAMS-Based Protection Schemes

Recent operational challenges imposed on transmission networks, raised by the restructuring, the intermittent renewable energy resources, series compensation, and highly loaded transmission systems, dictate an evolution in system and component protection wisdoms. The advent of SMT along with computer-based relaying proposed another dimension, namely protection, to the field of WAMS-based applications. Special protection schemes (SPSs), remedial action schemes, and modern adaptive protection algorithms were proposed as direct consequences of integration of communication and protection systems [449], [450]. PMU measurements led to the enhancement of dependability and security of the traditional protection methods, development of adaptive auto reclosing schemes [451], as well as protection schemes based on data-mining [452]. It was also argued that backup protection systems with only local measurements can malfunction when the system is highly stressed, and wide-area PMU data can significantly promote their performances [453]–​[456]. Additionally, the real-time design and operation of system intentional islanding in the face of a potential cascade was realized with the help of PMU data [457]–​[459]. To counteract the system instability, splitting the system into islands is an effective remedial action scheme (RAS) enabled by wide-area synchrophasor data [348]. It has to be emphasized that the most critical barrier against implementation of many theoretically-proven protection algorithms is the signaling latency whose trend is continuously lessening with the advancements in communication and processing technologies and deployment of dedicated communication media.

Early efforts for applying PMU data in the field of power system protection regarded the development of adaptive out-of-step relaying to make the relay blocking and tripping settings more attuned to the prevailing system conditions [460], [461]. Further research extended the PMU-based protection paradigm for the distance protection [462]–​[464], the current differential protection of multi-terminal and series-compensated lines [465]–​[467], the precise discrimination of in-zone or out-of-zone faults [468], and the overhead line protection accounting for the arc resistance [469]. Also, PMU data enhanced the fault type detection by means of the parameter fitting method or ANNs [14].

Load shedding at particular load buses is recognized as the last but effective resort in sustaining the system continuity after a frequency, angular, or voltage instability threat [391], [470]. Two major load shedding schemes are based on under-frequency and under-voltage criteria. Traditionally, these schemes are designed and operate independently from each other. Due to intrinsic coupling of frequency and voltage anomalies and their probable coincidence, it however makes more sense to have an integrated scheme for voltage and frequency load shedding. PMU data from different areas of a power system are helpful in the adaptive integrated frequency and voltage load shedding scheme [471], [472].

Although deployment of PMU devices in distribution systems, because of very close phase angles in a given distribution network, does make less sense, their capability in synchronized measurements and calculation of signal harmonic contents resulted in some research efforts. The development of PMU-based adaptive protection schemes for distribution systems with high penetration of distributed generation has demonstrated satisfactory outcomes [473].

SECTION XI. Conclusions

This paper presents a comprehensive survey on the subject of SMT and its relevant issues and applications. Since the presence of omissions is bound to be there, the authors would like to apologize for any errors or omissions and hope that additional references will be advanced as discussion to this paper. This bibliography illuminates the current engineering wisdom pertaining to advanced sets of PMU and WAMS applications. Presently, a clear consensus is heading toward the WAMS-based dynamic/stability monitoring, trending and prediction, automated control strategies, and wide-area special protection schemes. Such advancements will provide the tools for monitoring, protection, and control of emerging modern power systems in the smart grid paradigm.