Recent Application Examples of FNET/GridEye

Yong Liu1, Yi Cui1, WenpengYu1, Yao Zhang1, Ling Yilu Liu1,2

Wu1, Shutang You1 2. Division of Energy and Transportation Sciences
1. Department of Electrical Engineering and Power and Energy Systems Group
Computer Science Oak Ridge National Laboratory
University of Tennessee Oak Ridge, USA
Knoxville, USA

Abstract—As a pilot wide-area monitoring system (WAMS),

FNET/GridEye monitors the bulk power systems using a II. OVERVIEW
network of GPS-synchronized sensors and a wide variety of
FNET/GridEye applications have been spawned in the past years. A. FNET/GridEye Introduction
In this paper, following a brief description of the FNET/GridEye As mentioned earlier, the FNET/GridEye system currently
historical developments, several recent examples of
consists of more than 300 FDRs and two data centers. Fig. 1
FNET/GridEye applications are introduced.
gives a picture of the FDR unit. Inside each FDR, there is a
Keywords— Applications, Data analytics, Disturbance, GPS signal receiver that regulates the waveform sampling
Oscillation, Time error process for high-precision phasor measurement. Then an
improved discrete Fourier transformation (DFT) algorithm is
I. INTRODUCTION used to calculate the voltage waveform phase angles and
frequencies in real-time [6]. The measurement data are then
Wide area monitoring system (WAMS) has attracted many time-stamped, packaged, and transferred to the data centers. It
researchers’ attention recently due to its huge potential in should be noted that, since its birth, many important updates
power system monitoring, operation, and control [1][2][3]. As have been applied to the FDR unit, including more functions,
a newly-developed power grid measurement tool, it provides
higher accuracy, and better hardware etc. [7][8][9][10].
the unprecedented high-resolution monitoring of power system
operation conditions and is widely expected to benefit the grid
operators [4].
As a WAMS operated by the academia, FNET/GridEye has
been monitoring the bulk power systems in different countries
for more than a decade [5]. Up to now, more than 300
frequency disturbance recorders (FDRs) have been deployed at
multiple locations, including academic organizations, power
utilities, government agencies, and personal residences. Two
data centers have also been configured to receive, process, and
store all the measurements data collected by these FDRs.
In this paper, after an overview of the FNET/GridEye
architecture and existing applications, the latest application
Fig. 1. Photo of the FDR unit
examples will be introduced in more details. This paper is
structured as follows: Section II presents the FNET/GridEye The FDR-collected measurement data are all received by
sensor and server designs as well as existing applications; the data center which contains a series of data conditioning,
Section III describes two newest online applications, including application, and storage functions. Fig.2 shows a typical
the hybrid computational intelligence-based disturbance architecture of the FNET/GridEye data center. The data
localization and forced oscillation detection using adaptive concentrator layer extracts the measurement data from the
thresholds; two offline data analytics applications of received TCP/IP packages and condition them by aligning the
FNET/GridEye are presented in Section IV, and Section V time stamps and abandoning wrong data. The real-time
concludes the paper. application components and the data storage components are
included in the second layer while the third layer includes only
the non-real-time applications.
 hold true and can lead to substantial errors in practice. To
achieve a more accurate location estimation, a method that
utilizes advanced computational intelligence techniques has
been developed and implemented by the FNET/GridEye team,
which will be introduced in this subsection.
A flow chart of this disturbance localization is presented in
Fig.3, which basically includes three major steps. Firstly,
recurrence quantification analysis (RQA) [30] is employed to
identify the set of FDRs that are relatively close to the
disturbance location. Then a minimum volume enclosing
ellipsoid (MVEE) method [31] is utilized to extract informative
features from the FDRs measurements identified by RQA.
Fig. 2. Multi-layer architecture of FNET/GridEye data center
Finally, the extracted features are fed into a multivariate
random forest regression (MRFR) algorithm for estimating the
The data center was constructed in this way so that the
disturbance location and power mismatch.
different data applications can run independently and a large
number of functionalities can be accomplished.
B. Application Overview
In this part, the most important real-time and non-real-time
FNET/GridEye applications are listed. Details of these
applications can be found in the corresponding literature. It
should be emphasized that all these data applications have
been continuously run on the data center for years, therefore
their feasibility and effectiveness had been well demonstrated.

 Real-time power system status visualization[11]

 Disturbance recognition and location[12][13]
 Interconnection-wide oscillation detection and
analysis[14]
 Modal analysis using ambient measurement data
[15][16]
Fig. 3. Flowchart of RQA-MVEE for disturbance location and power
 Islanding and off-grid detection[17][18] mismatch estimation.

 Disturbance replay and analysis[19] Since the implementation on the FNET/GridEye platform,
more than 1300 disturbances have been localized successfully.
 Measurement-aided model validation[20]
Two examples are given below in Fig. 4. In Fig. 4(a), since the
 Electromechanical speed map development[21][22][23] disturbance was surrounded by several FDRs, both the TDOA-
based and the RQA-MVEE method were able to identify the
 Historic data statistical analysis[24]-[29] disturbance location accurately. Comparatively, only the RQA-
MVEE method could locate the disturbance precisely in Fig.
III. RECENT REAL-TIME APPLICATIONS 4(b) while the TDOA-based method failed to give a
In this section, two recent real-time applications of satisfactory result since the FDRs available at the disturbance
FNET/GridEye will be presented, including the hybrid time were only on one side of the disturbance. These two case
computational intelligence-based disturbance localization and studies demonstrate that the RQA-MVEE method is able to
forced oscillation detection using adaptive thresholds. deliver a more accurate and robust disturbance location
estimation result especially when there are not sufficient
measurement locations for the TDOA-based method to work.
A. Electromechanical Disturbance Localization based on
Hybrid Computational Intelligence Techniques
Severe electromechanical disturbances in the power grid
such as generator outages and load shedding affect the power
system dynamic behaviors and threaten system stabilities.
Synchronized measurements taken by FDRs allow for the real-
time detection and localization of such electromechanical
disturbances. However, the previous TDOA-based method
[12][13] assumed the propagation speeds of the disturbance are
identical in each direction. This assumption may not always
 2016/1/15 05:04:23 (UTC) O(n) to O(1) where n is the number of measurements in each
Power mismatch estimated/confirmed: 890 MW / 877 MW
time window.

Fig. 5. Envelop lines of the relative angle

The key of this detection algorithm is the envelope band

(a) case study 1
2016/2/1 15:02:29 (UTC)
Power mismatch estimated/confirmed: 950 MW / 970 MW
threshold selection. In most previous studies, a uniform
detection threshold was applied for the entire grid. However,
the power grid does behave differently at different locations.
As shown in Fig. 6, the angle change ranges in some regions
are much larger than those of the others. So it is unreasonable
to employ the uniform threshold for all the measurement
locations. Instead, we dynamically calculates the threshold for
each FDR. As shown in Fig.7, a threshold is selected based on
the probability distribution of its phase-angle changing range
so that the probability of a measured phase angle larger than
this threshold is less than 0.005.

(b) case study2

Fig. 4. Examples of the RQA-MVEE disturbance localization method

B. Forced Oscillation Detection

Forced oscillations are usually incurred by periodic
perturbations in the system such as generator exciter Fig. 6. Distribution of phase-angle changing ranges at different locations
malfunctions or cyclic load variations [32][33][34]. Forced
oscillation may make power systems susceptible to cascading
outages, mechanical vibrations, equipment failures, and other
disturbances. So it is important to detect them as soon as
possible so that countermeasures can be taken. Many detection
algorithms have been reported in the literature
[35][36][37][38]. However, most of them are difficult to be
implemented online at this stage due to the high computation
burden involved. Instead, an envelope band based method has
been implemented on the FNET/GridEye platform, which
proves to be accurate and fast enough for forced oscillation
detection.
The first step of this algorithm is the envelop band
calculation. Fig. 5 shows the envelopes of three voltage angle Fig. 7. Method to derive the threshold of phase-angle changing range
channels. To obtain such envelopes for hundreds of
measurement channels simultaneously, a regressive calculation This forced oscillation detection algorithm has been
algorithm has been designed, which preserves the result of the implemented online and its accuracy and robustness have been
previous time window and update the envelope line verified by confirmed forced oscillation events. A flowchart of
accordingly after appending the current time window’s new this algorithm is presented in Fig. 8.
measurements. By doing this, the time complexity of the
envelop band calculation algorithm has been reduced from
 the light-load grid conditions can potentially have more
detectable oscillations.

Fig. 9. Inter-area oscillation detections in the EI.

Dominate mode frequency and damping ratio were also

analyzed year by year but due to the limited page space, only
the analysis results in 2015 is presented in Fig. 10. According
to Fig.10, 0.2 Hz was the most common mode in the EI system
and the damping ratio was larger than 10% in most of the time.
Furthermore, negative damping ratios may result in an increase
in the magnitude of the oscillations. However, it only accounts
for a quite small proportion among all damping ratios.

30

15

0
0

Fig. 8. Forced oscillation detection algorithm (a)

IV. RECENT DATA ANALYTICS APPLICATIONS

In this section, two latest data analytic applications of
FNET/GridEye will be introduced.

A. Inter-area oscillation statistical analysis

Many oscillation detection and analysis studies have been
seen in the literature. However, due to the unavailability of real
oscillation events information to most of researchers, there has
been almost no oscillation analysis from a statistical
perspective. In this part, the FNET/GridEye-detected
oscillation events in one of the world’s largest synchronous (b)
grid – the Eastern Interconnection (EI), will be analyzed Fig. 10. Dominant mode frequency and damping ratio distribution in the EI
statistically. The distribution of oscillation numbers, dominant (2015).
mode frequency, and damping ratio are presented [39].
The EI oscillations’ hourly distribution is presented in Fig. B. Impact of System Frequency on Time Error Service
9. It is clear that all the six years included in this analysis are
Many electric clocks rely on the power grid frequency for
similar in terms of the hourly distribution. Basically, oscillation
synchronization and correction. Therefore, the grid frequency
detection increases around 8 AM (coordinated universal time)
fluctuations will cause time errors of those clocks. In this part,
in the morning and peaks at 10 AM. This trend indicates that
the influence of grid frequency on the clock time error will be
analyzed from a statistical perspective. The FNET/GridEye
frequency measurements are utilized in this study [19].
Fig. 11 presents the time error deviations in the EI system
in 30 days. “Run fast” in Fig. 11 means this clock is faster than
the ideal clock while “Run slow” means it is slower than the
ideal clock. From Fig. 11(a), it is obvious that the time error of
the electric clock may accumulate at different speeds and can
be either positive or negative. However, once the error
accumulated to some extent, it got corrected so that the
maximum error stay within an acceptable range. This indicates
that for certain applications, the electric clock that relies on the
EI grid frequency for synchronization still have a reasonably
good timing accuracy. Fig. 11(b) shows the error distribution in
the same 30 days’ period of time, which demonstrates that the
clock tends to have a higher possibility to run fast than the
ideal clock.
(a)
(b)
Fig. 11. Electric clock time error in 30 days.
Next, the maximum error (maximum running fast error),
the minimum error (maximum running slow error), and
average error were calculated for each month from Jan., 2014
to Apr., 2017 and the results are shown in Fig. 12. The blue
and orange bars show the maximum and minimum time error
respectively while the grey line represents the average error.
One obvious observation is that the maxim error keeps growing
in this time period, which means the electric clock tends to run
faster before it gets correction. In comparison, the minimum
error is decreasing in the same time period, which means the
time when the clock runs slow get shorter and shorter.
Furthermore, the average time error tends to become positive
from being negative although it was always lower than ±3.50
seconds.
 
Fig. 12. Monthly time error deviation for the EI system from 2014 to 2017.
V. CONCLUSIONS
Several recent FNET/GridEye applications have been
introduced in this paper. The hybrid computational intelligence
based method allowed a more accurate and robust disturbance
location estimation. The forced oscillation detection algorithm
used adaptive thresholds in the detection and achieved an
online implementation on hundreds of measurement channels.
Additionally, applying the FNET/GridEye in two data analytics
applications were also reported. In a word, FNET/GridEye has
developed a large number of real-time and non-real-time
applications and will continue to benefit its various users.
REFERENCES
[1] P. Bonanomi, “Phase angle measurements with synchronized clocks
principle and applications,” IEEE Trans. Power App. Syst., vol. PAS-
100, no. 12, pp. 5036–5043, Dec. 1981.
[2] A.G. Phadke et.al. "A new measurement technique for tracking voltage
phasors, local system frequency, and rate of change of frequency," IEEE
Trans. Power App. Syst., vol. PAS-102, no.5, pp.1025-1038, May 1983.
[3] A.G. Phadke, "Synchronized phasor measurements in power systems,"
IEEE Comput. Appl. Power, vol.6, no.2, pp.10-15, April 1993.
[4] Y. Liu, S. You, and Y. Liu, ‘Smart transmission & wide area monitoring
system. Communication, Control and Security for the Smart Grid. 2017.
[5] Q. Bin, et al. “Internet based frequency monitoring network (FNET),” in
Proc. 2001 IEEE Power Eng. Soc. Winter Meet., pp. 1166–1171.
[6] T. Xia and Y. Liu, "Single-phase phase angle measurements in electric
power systems," IEEE Trans. Power Syst., vol.25, no.2, pp.844-852,
May 2010.
[7] Y. Liu, et al. "Wide-Area Measurement System Development at the
Distribution Level: an FNET/GridEye Example," IEEE Trans. Power
Del., vol. 31, no. 2, pp.721-731, 2015.
[8] Y. Liu, et al. “Recent Developments of FNET/GridEye – A Situational
Awareness Tool for the Smart Grid,” CSEE Journal of Power and
Energy Systems, vol. 2, no. 3, pp. 19-27, 2016.
[9] Y. Liu et al., "A Distribution Level Wide Area Monitoring System for
the Electric Power Grid–FNET/GridEye," IEEE Access, vol. 5, no. , pp.
2329-2338, 2017.
[10] W. Yao, et al. “GPS signal loss in the wide area monitoring system:
Prevalence, impact, and solution”. Electric Power Systems
Research, 147, pp.254-262. 2017.
[11] Y. Zhang, et al. "Wide-Area frequency monitoring network (FNET)
architecture and applications," IEEE Trans. Smart Grid, vol.1, no.2,
pp.159-167, Sept. 2010.
[12] R. M. Gardner, et al. “Non-parametric power system event location
using wide-area measurements,” in 2006 IEEE PES T&D Conference &
Exposition, pp. 1668–1675.
[13] R. M. Gardner et al. “Power system event location
analysis using wide-area measurements,” in Proc. 2006 Power Eng.
Soc. Gen. Meet., pp. 1-7.
[14] Z. Yuan, et al. "Inter-area oscillation analysis using wide area voltage
angle measurements from FNET," in Proc. 2008 IEEE Power Eng. Soc.
Gen.Meet., pp.1-7, 2008.
[15] S. You, J. Guo, G. Kou, Y. Liu, and Y. Liu. “Oscillation mode
identification based on wide-area ambient measurements using
multivariate empirical mode decomposition”. Electric Power Systems
Research, 134, pp.158-166. 2016.
[16] D. Trudnowski et al. "Performance of three mode-meter block-
processing algorithms for automated dynamic stability assessment,"
IEEE Trans. Power Syst., vol.23, no.2, pp.680-690, May 2008.
[17] Z. Lin et al. "Application of wide area measurement systems to islanding
detection of bulk power systems," IEEE Trans. Power Syst., vol.28,
no.2, pp.2006-2015, May 2013.
[18] J. Guo et al. "Design and implementation of a real-time off-grid
operation detection tool from a wide-area measurements perspective,"
IEEE Trans. Smart Grid, vol. 6, no. 4, pp.2080-2087, July 2015.
[19] L. Chen, “Wide-area measurement application and power system
dynamics,” Ph.D. dissertation, Dept. of EECS, Univ. of Tenn.,
Knoxville, TN, 2011.
[20] G. Kou, S. Hadley, and Y. Liu, "Dynamic Model Validation with
Governor Deadband on the Eastern Interconnection," Oak Ridge
National Laboratory (ORNL), Oak Ridge, TN, USA, Feb. 2014.
[21] P. N. Markham, “Data Mining and Machine Learning Applications of
Wide-Area Measurement Data in Electric Power Systems,” Ph.D.
dissertation, Deptment of Electrical Engineering and Computer Science,
University of Tennessee, Knoxville, TN, 2012.
[22] Y. Liu et al. “Observation and Applications of Electromechanical Wave
Propagation Based on Wide-Area Synchronous Measurement”. IEEE
PES T&D Conference & Exposition, 50(1), pp.73-78. 2018.
[23] S. You et al. “U.S. Eastern Interconnection (EI) Electromechanical
Wave Propagation and the Impact of High PV Penetration on Its Speed”,
2018 IEEE PES T&D Conference & Exposition, pp.1-5. 2018.
[24] Y. Ye, “Wide-area Situational Awareness Application Developments,”
Ph.D. dissertation, Deptment of Electrical Engineering and Computer
Science, University of Tennessee, Knoxville, TN, 2011.
[25] Y. Lei et al. “The impact of synchronized human activities on power
system frequency”, in 2014 IEEE Power Engineering Society General
Meeting, pp.1-5.
[26] Y. Liu, et al. “Application of power system frequency for digital audio
authentication,” IEEE Trans. Power Del., vol. 27, no. 4, pp. 1820-1828,
2012.
[27] S. You et al. "Disturbance Location Determination Based on
Electromechanical Wave Propagation in FNET/GridEye--a Distribution-
Level Wide-Area Measurement System." IET Generation, Transmission
& Distribution 2017.
[28] S. You et al. “Non-Invasive Identification of Inertia Distribution Change
in High Renewable Systems Using Distribution Level PMU”. IEEE
Transactions on Power Systems. 2017 Jun 9.
[29] W. Yao et al. “Source Location Identification of Distribution-level
Electric Network Frequency Signals at Multiple Geographic
Scales”. IEEE Access. 2017.
[30] N. Marwan, M. Carmenromano, M. Thiel and J. Kurths, "Recurrence
plots for the analysis of complex systems," Phys. Rep., vol.438, Issue 5-
6, pp. 237-329, 2007.
[31] J. Ma et al. "The characteristic ellipsoid methodology and its application
in power systems," IEEE Trans. Power Syst., vol.27, Issue 4, pp. 2206-
2214, 2012.
[32] C. D. Vournas, N. Krassas, and B. C. Papadias, “Analysis of forced
oscillations in a multimachine power system,” in Proc. Int. Conf.
Control, Mar. 1991, pp. 443–448.
[33] J. E. Van Ness, “Response of large power systems to cyclic load
variations,” IEEE Trans. Power App. Syst., vol. PAS-85, no. 7, pp. 723–
727, Jul.1966.
[34] J. A. Pinneilo and J. E. Van Ness, “Dynamic response of a large power
system to a cyclic load produced by a nuclear accelerator,” IEEE Trans.
Power App. Syst., vol. PAS-90, no. 4, pp. 1856–1862, Jul.1971
[35] J. Follum and J. W. Pierre, “Detection of periodic forced oscillations in
power systems,” IEEE Trans. Power Syst., vol. 31, no. 3, pp. 2423–
2433, May2016.
[36] J. Ma, P. Zhang, H. Fu, B. Bo, and Z. Y. Dong, “Application of phasor
measurement unit on locating disturbance source for low-frequency
oscillation,” IEEE Trans. Smart Grid, vol. 1, no. 3, pp. 340–346, Dec.
2010.
[37] L. Chen, Y. Min, and W. Hu, “An energy-based method for location of
power system oscillation source,” IEEE Trans. Power Syst., vol. 28, no.
2, pp. 828–836, May 2013.
[38] Y. Li, C. Shen, and F. Liu, “An energy-based methodology for locating
the source of forced oscillations in power systems,” in Proc. IEEE Int.
Conf. Power Syst. Tech., Oct.-Nov. 2012, pp. 1–6.
[39] Y. Cui, L. Wu, W. Yu, Y. Liu, W. Yao, D. Zhou and Y. Liu, "Inter-area
oscillation statistical analysis of the U.S. Eastern interconnection," IET
Journal of Engineering, DOI: 10.1049/joe.2017.0243 .
[40] Y. Zhang et al. “Impacts of Power Grid Frequency Deviation on Time
Error of Synchronous Electric Clock and Worldwide Power System
Practices on Time Error Correction,” Energies, vol. 10, no. 9, pp. 1-15,
2017.