See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/261430016

Testing and validation of a fast real-time oscillation detection PMU-based

application for wind-farm monitoring

Conference Paper · July 2013

DOI: 10.1109/BlackSeaCom.2013.6623413

CITATIONS READS

16 199

5 authors, including:

Luigi Vanfretti Iyad Al Khatib

Rensselaer Polytechnic Institute KTH Royal Institute of Technology
236 PUBLICATIONS   2,285 CITATIONS    30 PUBLICATIONS   265 CITATIONS   

SEE PROFILE SEE PROFILE

Muhammad Shoaib Almas

KTH Royal Institute of Technology
38 PUBLICATIONS   407 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

EU funded FP7 iTesla project View project

IDE4L-Ideal grid for all View project

All content following this page was uploaded by Luigi Vanfretti on 08 September 2018.

The user has requested enhancement of the downloaded file.

 Testing and Validation of a Fast Real-Time Oscillation Detection
PMU-Based Application for Wind-Farm Monitoring
Luigi Vanfretti, Maxime Baudette, Iyad Al-Khatib, M. Shoaib Almas and Jan O. Gjerde.

1.03
Abstract—This article provides an overview of a monitoring
application, its testing and validation process. The application 1.02
was developed for the detection of sub-synchronous oscillations

Voltage Magnitude (p.u.)

in power systems, utilizing real-time measurements from phasor 1.01
measurement units (PMUs). It uses two algorithms simultane-
ously to both detect the frequency at which the oscillatory
1
event occurs and the level of energy in the oscillations. The
application has been developed and tested in the framework of
0.99
SmarTS Lab, an environment capable of hardware-in-the-loop
(HIL) simulation. The necessary components of the real-time
0.98
chain of data acquisition are presented in this paper, as well PMU
as testing and validation results, to demonstrate the accuracy SCADA
0.97
of the monitoring tool and the feasibility of fast prototyping 0 5 10 15 20 25 30 35 40
for real-time PMU measurements based applications using the Time (s)

SmarTS Lab environment. Fig. 1. Comparison example between PMU and traditional
Index Terms—Power system oscillations; monitoring application; equipment measurements(SCADA) during a fast dynamic event
BableFish; PMU; Wide Area Monitoring Systems.

I. I NTRODUCTION simulation environment is described in Section III. The vali-

The environmental impact of electricity generation has dation experiment as well as the results obtained are presented
raised concerns in western countries, engaging a shift to in Section IV.
renewable energies among which wind power has become the
fastest growing energy technology since the 1990s [1]. This II. M ONITORING A PPLICATION
expansion will continue, especially in Europe, where several The oscillatory event presented in Fig 1 differs from the
countries have fixed high goals for the coming years [2]. traditional and well studied inter-area or local oscillations.
The increasing amount of wind power, which is one kind It also remained undetected by traditional monitoring tools
among different intermittent generation sources, involves nu- present in the control room of the TSO. An application
merous new challenges for its integration in existing power dedicated to the detection of such phenomena was developed
systems. As such transient, stability issues have already been using LabView at SmarTS Lab. This tool will be introduced
studied [3]. However, it is only recently that some Transmis- in this Section, however, a more detailed presentation of the
sion System Operators (TSOs) have measured, with PMUs [4], algorithms behind the oscillation detection will be the subject
sub-synchronous oscillatory events resulting from interactions of a future publication; the focus here will be put on the tool’s
between wind farms at frequencies around 13-15 Hz. The testing and validation process.
oscillations were even observed at the consumer level in the This monitoring tool has been developed as a real-time
form of flickering [5]. application that exploiting PMU measurements. The detection
Fast dynamics at such frequencies remain unobservable by of oscillations is performed by two separate algorithms, the
most of the standard monitoring tools (SCADA), due to too first one is an adaptation of an algorithm for fast detection of
low sampling frequencies and lack of time synchronisation, low frequency oscillations from the work by Hauer in [6]. It
leaving this kind of phenomenon unstudied. computes the energy content of oscillations within a frequency
Figure 1 presents a comparison between SCADA and PMU range by signal envelope detection. The second algorithm
measurements of this phenomenon. The usage of PMUs en- computes the Power Spectrum Density (PSD) within the same
ables observability for phenomena occurring at frequencies up frequency range with a parametric method assuming an auto
to 15-25 Hz, opening new wide perspectives for monitoring regressive model and a non-parametric method (Welch). The
and control applications based on synchrophasors. These new application also implements several pre-processing steps to
tools will help acquiring a better knowledge on the challenges accommodate for the PMU measurements, which can con-
brought by the increase of intermittent energy sources. tain measurement errors. The pre-processing algorithm was
This paper presents a hardware-in-the-loop (HIL) validation inspired by previous work used for off-line analysis [7], and
process for a PMU-based monitoring application aiming at the adapted for real-time application.
detection of the phenomenon aforementioned. The developed As mentioned previously, the tool is fed by real-time PMU
application is briefly introduced in Section II and the HIL- measurements coming directly from a PDC output stream,
 Fig. 2. Screen shot of the interface of the Monitoring Tool

explained further in Section III. It updates the results of the

detection algorithms in real-time, although subject to small
delays from the filters implemented in the algorithms. The
graphs present the results of a buffer of data, which length is
set by the user.
Figure 2 depicts the main Graphical User Interface of the
Monitoring Tool,which is divided in several parts. The top-left
Fig. 3. Screen shot of one Module of the Monitoring Tool
part is the configuration interface, the top-right part contains
an input signal visualization graph and the bottom part is
dedicated to the oscillation detection interface. The monitoring
A. Wide Area Monitoring Systems
interface is comprised by four Modules, see Fig. 3, containing
two graphs and LED indicators. Each Module is configured to Monitoring systems based on PMU measurements are com-
monitor a separate frequency range, targeting several ranges monly referred to as Wide Area Monitoring Systems (WAMS).
of interest (inter-area modes, local modes, faster oscillations The main components of WAMS are:
events, etc.) The graph on the left part of a Module depicts • Phasor Measurement Units (PMUs)
the level of energy computed in the frequency range for the • Phasor Data Concentrators (PDCs)
time window of the buffer. The LED indicators define three • Communication Networks
thresholds of energy, set by the user, they light up according The PMUs are the measuring equipment, which sample the
to the energy computed at the latest time stamp. The graph voltage and current data at a frequency of typically 30 to
on the right shows the Power Spectrum Density (PSD) of the 50 Hz with a time stamp derived from a GPS clock reference.
frequency range, completing the information on any occurring The PDCs are servers responsible for gathering PMU measure-
oscillation. ments, aligning them in time and forwarding them as clustered
output streams (with possible format conversion), they can
III. C HAIN OF R EAL - TIME DATA ACQUISITION also serve as archiving systems. A PMU can usually support
the forwarding of PMU measurements to one or several PDC
Development of new monitoring applications utilizing syn- servers1 . PDCs offer more flexibility, but also have certain
chronized PMU measurements requires a testing platform
which can provide real-time data. This is accomplished with 1 This varies in different commercially available PMUs, with some only
a chain of real-time data acquisition as described below. allowing one connection, and others up to four.
limitations. They can be used to share PMU measurements
with several other PDC servers. The traditional layout of such
monitoring system is depicted in Fig 4. This layout represents
several local PDC servers, which can by owned by different
entities (private owner, electricity generation company, etc.)
The PDCs are then concentrated in a super PDC owned
by the TSO, from which all monitoring tools are fed. All
synchronised phasor data is transported via a communication
network using the IEEE C37.118.2 protocol [8].

TSO CONTROL ROOM

SUPER
PMU PDC PMU

PDC PMU
PDC
PMU
PDC
PMU PMU

PMU
Fig. 5. Photo of the set-up of the experiment at SmarTS Lab
PMU
PMU

Fig. 4. Layout of the communication and computer architecture of developed (BableFish) that handles the communication with
wide-area monitoring systems a PDC server and delivers the PMU measurements to the
LabView environment after decoding the standard protocol.
This kind of set-up is replicated at SmarTS Lab, with BableFish is comprised by several modules with different
some simplifications due to the impracticality of building an functions. The core of the application consists of Visual C++
entire power system in the lab. The set-up built in the lab libraries that handle the communication with the PDC server
makes use of HIL simulation, replacing the power system by to establish the connection and perform real-time acquisition
models executed in real-time on a simulator equipped with of new measurements. Interactions with the core application
analog reconfigurable inputs/outputs. The rest of the WAMS are possible using the ActiveX technology that interfaces the
architecture described above is replicated, building a testing Visual C++ libraries with a LabView VI, depicted in Fig 6.
platform which is depicted in Fig. 5, with the following It allows for the configuration of the connection to the PDC,
elements: selection of the measurements to be delivered and saving of
1) The real-time simulation target from Opal-RT. the configuration. The measurements are then forwarded to
2) An oscilloscope connected to an analog output of the a LabView sub-VI, which can be integrated in an external
target, showing the three-phase signal produced by the application (such as the monitoring tool presented in this
model (only for lab. set-up purposes). article) for delivering the decoded measurements to processing
3) A National Instruments cRIO PMU connected to the algorithms.
analog output aforementioned. The monitoring application has also been adapted to Stat-
4) The connection cables of the PMU for acquiring voltage nett’s PMU Recorder Light (PRL), which provide similar
measurements, connected to the back of the simulator. functions for real-time data mediation within LabView. More
5) The network connection of the PMU to stream the data information about the PRL are available in [9].
to the PDC server of the lab.
6) The PDC server of the lab receiving the PMU data and IV. T ESTING AND VALIDATION
broadcasting an output stream to itself for this photo. The development of the application was carried out with a
In the PMU configuration a scale-up factor is added for the test output stream containing the PMU measurements from a
measurements so that the simulated values are well transcribed PMU connected to the distribution network. This set-up was
in the phasor calculation, despite the low ratings of the analog primarily used to develop the integration of the chain of real-
output of the simulator. time data acquisition in the application and did not provide any
behaviour of interest apart from a local harmonic at 16.7 Hz
B. Real-time data mediators created by the feeding system of a railway nearby the lab. This
The monitoring application development has been carried frequency component was used to develop and configure the
out in the LabView language, which doesn’t provide libraries spectrum analysis algorithm. Figure 2 depicts the monitoring
for the IEEE C37.118.2 standard to connect to either PMUs tool running in this set-up and the local harmonic is well
or PDC servers. An additional software has therefore been detected on the bottom right Module.
 Fig. 6. Screen Shot of the main interface of BableFish

TABLE I. Time-line of the experiment

with 10 % turbulence. The loads are also configured to have
Start Oscillation Additional Major Fault End a sinusoidal profile at different frequencies, according to the
Injection Perturba-
tions following parameters:
Random Set pertur- Generating Three phase End of the • Load in Area 1:
load bation at minor Fault and oscillation – Active power 1 MW random load variation and
variation 10.83 Hz faults line opening injection
2 MW of sinusoidal load at 0.4 Hz.
– Reactive power 1 MW random load variation and
For validating the monitoring tool, HIL simulation was used 2 MW of sinusoidal load at 0.8 Hz.
with the development of a power system model capable of • Load in Area 2:
recreating the event described in Section I and additional – Active power 1 MW random load variation and
perturbations [10]. The power system model is equipped with 2 MW of sinusoidal load at 1 Hz.
two variable loads introducing random variation and sinusoidal – Reactive power 1 MW random load variation and
variation that will excite lower dynamics oscillations in the 150 MW constant load.
power system. The model performs the simulation of wind The processing algorithms of the Monitoring Tool detect
farms interaction by injecting oscillations at 10.83 Hz, which the slow dynamic activity resulting from both load variations
was chosen as the equivalent of the 13 Hz oscillations in and wind turbulences, see Fig. 3, where the spectral estimator
a 50 Hz system (the original case occurred in the USA, highlights both frequency components at 0.4 Hz and 0.8 Hz.
where the nominal frequency is 60 Hz). Finally, the model Forced oscillations are then injected at first with 0.05 p.u.
also includes the generation of faults in the system, with amplitude at the point of common coupling of the first wind
perturbation of the mechanical input of one of the generators farm. They can be observed in the frequency graph of the
and a three phase fault followed by line tripping on an heavily tool, see Fig. 7(a). It can be noticed that the frequency range
loaded line. containing 10.83 Hz is active with the flag Danger!!!, see
The validation of the algorithms is performed with a sce- Fig. 7(b), while the other frequency ranges remain inactive or
nario including all of the events aforementioned according to with a low activity, see Fig 7(c). This shows the fast reaction
the time-line summarized in Table I, enabling the testing of of the tool, its selectivity and its ability to estimate in real-
the tool’s performance in different situations. The validation time the level of energy in these oscillations. The frequency
experiment has been performed several times to ensure a estimation algorithm does not update as quickly as the energy
satisfying configuration of the numerous parameters of the detection algorithm, the frequency of the injected oscillations
processing algorithms. is thus not detected as quickly. However, it can be noticed that
The simulation of the power system model is started with the parametric method starts to show distinctively a peak at
both wind farms receiving an average wind speed 12 m/s the right frequency, see Fig. 7(b).
 (a) Frequency of the system (a) Frequency of the system

(b) Module [10 - 12 Hz] (b) Module [3 - 12 Hz]

Fig. 8. Partial screen shot of the Monitoring Tool during the increase of
oscillation injection at 10.83 Hz

(c) Module [0.09 - 1 Hz]

Fig. 7. Partial screen shot of the Monitoring Tool during the oscillation
injection at 10.83 Hz

After the beginning of the injection with 0.05 p.u. amplitude

the injection is increased up to 0.07 p.u. amplitude. The
resulting oscillations have a larger amplitude, as shown in
Fig. 8(a), and the energy detection algorithm identifies an
increase in the energy level in the oscillations. The frequency
estimation algorithm also detects very precisely the frequency
at which the oscillations are occurring as shown in Fig. 8(b). Fig. 9. Partial screen shot of the Monitoring Tool during the oscillation
The injection of oscillations is maintained and the power injection at 10.83 Hz and minor faults
system model stabilizes with these high frequency power oscil- almost all frequency ranges, see Fig. 9.
lations still active. Additional (minor) perturbations performed The major fault performed is a three phase fault on one of
on the system are implemented: the lines between the two sectors of the power system model.
• A step increase in the mechanical power input of one of After the fault is applied, the line is opened and re-closed after
the synchronous generators (0.05 p.u.). clearing the fault, eleven seconds after the application of the
• A three-phase fault at the bus of one of the sub-system fault. The effect on the power system model is strong, but
of the first wind-farm (Three wind turbines affected). stability is maintained after re-closing the line.
The perturbations are performed sequentially, see Fig. 9. The Monitoring Tool is affected by discontinuities in the
The first perturbation is performed at an instant approximately voltage measurements, triggering temporarily a Danger!!! flag
at the middle of the figure and the second at the end of the in all frequency ranges, see Fig. 10 for two Modules. This
figure. behaviour should be eradicated and the tool should be less
The first perturbation induces a small ripple in the frequency sensitive to such behaviours. The only reliable information that
of the system, leading to the detection of activity in almost can be retrieved from the tool during such events is estimated
every frequency range. While the most affected are the lower modes of oscillation at lower frequencies that are the most
frequency ranges, which was expected. The detections mostly affected by such perturbation. This is noticed by the trigger
result in discontinuities in the voltage measurements. The flag being displayed for a longer time in this frequency range.
phenomenon is amplified in the case of the second pertur- The frequency estimation algorithm is also vastly affected
bation, which induces a greater discontinuity in the voltage by the discontinuities in the measurements, making it very
measurements and therefore triggers the Danger!!! flag in difficult to make interpretations during such events.
 ACKNOWLEDGEMENT
The support of Austin White at Oklahoma Gas & Electric in
providing wind farm measurements is sincerely acknowledged.
L. Vanfretti was supported by Statnett SF, the Norwegian
Transmission System Operator and the STandUP for Energy
collaboration initiative.
M. Baudette, I. Al-Khatib and M. Shoaib Almas were sup-
ported by the EIT KIC InnoEnergy through Action 2.6 of the
Smart Power project.
L. Vanfretti and M. Shoaib Almas were supported by Nordic
Energy Research through the STRONg2 rid project.
The economical support of the institutions and funding
bodies listed above is sincerely acknowledged.
Fig. 10. Partial screen shot of the Monitoring tool during a fault
R EFERENCES
[1] J. F. Manwell, J. G. McGowan, and A. L. Rogers, Wind Energy
Explained: Theory, Design and Application. John Wiley & Sons, 2010.
V. C ONCLUSION [2] D. Milborrow, “Europe 2020 wind energy targets, eu wind power now
Developing new algorithms or new applications for mon- and the 2020 vision,” Windpower Monthly Special Report, March 2011.
[3] J. Wiik, J. O. Gjerde, and T. Gjengedal, “Impacts from large scale inte-
itoring the power grid is a complex task that faces several gration of wind energy farms into weak power systems,” in Proceedings
challenges such as the access to real-time measurements from of the International Conference on Power System Technology, 2000, pp.
a real power system. On the one hand, actual data from a 49 –54.
[4] A. White and S. Chisholm, “Relays become problem solvers,” Trans-
TSO would not follow an established testing scenario such mission & Distribution world, November 2011.
as the ones presented in Table I without significant plan- [5] A. White (OGE), S. Chisholm (OGE), H. Khalilinia (WSU), Z. Tash-
ning efforts and implementation costs, leaving some cases man (WSU), and M. Venkatasubramanian (WSU), “Analysis of Subsyn-
chronous Oscillations at OG&E,” NASPI-NREL Synchrophasor Tech-
untested. Moreover it is important to emphasize that some nology and Renewables Integration Workshop - Denver, CO., June 7
scenarios may endanger the stability of the power system, they 2012.
could therefore not be envisaged for the purpose of testing [6] J. Hauer and F. Vakili, “An oscillation detector used in the bpa power
system disturbance monitor,” IEEE Transactions on Power Systems,
algorithms. On the other hand, the traditional approach of vol. 5, no. 1, pp. 74 –79, Feb. 1990.
working with off-line simulations and recorded data offers a [7] L. Vanfretti, S. Bengtsson, and J. O. Gjerde, “Preprocessing Synchro-
some flexibility in the calibration of the oscillation detection nized Phasor Measurement Data for Spectral Analysis of Electrome-
chanical Oscillations in the Nordic Grid,” European Transactions on
algorithms. However, it neglects all off the real-time aspects Electrical Power, in press.
such as computing time, rolling data window length and the [8] “IEEE Standard for Synchrophasor Data Transfer for Power Systems,”
quality of the data (no missing data, no bad measurements, IEEE Std C37.118.2-2011 (Revision of IEEE Std C37.118-2005), pp. 1
–53, 28 2011.
etc.). HIL simulation offers in this case a good compromise [9] L. Vanfretti, V. H. Aarstrand, M. S. Almas, V. S. Perić, and J. O.
between the experiment feasibility in a lab and closeness to Gjerde, “A software development toolkit for real-time synchrophasor
an actual WAMS architecture implementation. applications,” in 2013 IEEE PES Granoble PowerTech, 2013.
[10] M. Baudette, “Fast real-time detection of sub-synchronous oscillations
In Section IV the validation experiment of the tool covers in power systems using synchrophasors,” Master’s thesis, KTH, Electric
a wide variety of cases that such a tool could encounter in Power Systems, 2013.
a real power system. The first and primary conclusion is that
the tool was successful in the detection of both underlying
activity and strong oscillatory events in the higher frequency
range, performing according to its design. The underlying
activity was detected in the low frequency range (below 1 Hz),
resulting from the cyclic load variation and higher frequency
range with the detection of the sub-harmonic generated by the
nearby railway system. The injected oscillations at 10.83 Hz,
modelling wind farms interactions were successfully detected.
The detection speed was fast, with negligible delay. However,
the validation process unveiled the sensitivity of the algorithms
to strong discontinuities in the measurements, such as the ones
created by large faults in the power system. The improvements
necessary for handling such situations will be the object of
future work.

View publication stats