detection of flicker caused by inter harmonics

152 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 58, NO. 1, JANUARY 2009

Detection of Flicker Caused by InterharmonicsTaekhyun Kim, Student Member, IEEE, Edward J. Powers, Fellow, IEEE, W. Mack Grady, Fellow, IEEE, and

Ari Arapostathis, Fellow, IEEE

Abstract—International standard bodies have specified an in-strument capable of detecting and assessing the severity of lightingflicker, which is called a flickermeter. However, a deficiency withregard to flicker caused by high-frequency interharmonics hasbeen reported in the recent literature. In this paper, anotherdeficiency of the International Electrotechnical Commission (IEC)flickermeter with regard to low-frequency interharmonics, whosefrequencies are below the difference between the fundamentalfrequency and the cutoff frequency of a bandpass filter of theIEC standard flickermeter, will be discussed and illustrated basedon analysis and numerical experiments. A new approach basedon down-up sampling in the discrete-time sample domain is alsoproposed to address the flicker-detection problem associated withinterharmonics. We demonstrate the efficacy of the proposedmethod by a comparison with the current flickermeter standardin the presence of interharmonics.

Index Terms—Down-up sampling, flicker, flickermeter,harmonics, interharmonics, peak detection.

I. INTRODUCTION

IN ELECTRIC power distribution networks, various types ofdisturbances may occur, and they can affect electric power

customers in different ways. Among those disturbances, lightflicker might be the most “visible” one for customers. Lightflicker is defined as noticeable illumination changes due to volt-age fluctuations imposed on the fundamental power frequencycomponent. In addition to the fact that the illumination levelchanges may significantly impact work efficiency, they mightalso induce seizures in certain people [1]. Therefore, healthand safety issues with regard to flicker have been raised. Fur-thermore, the voltage fluctuations associated with flicker canalso cause potentially serious problems to critical power systemcomponents such as motors, generators, and transformers.

Therefore, flicker detection and evaluation has garnered theattention of international standardization bodies. The Interna-tional Electrotechnical Commission (IEC) specified an instru-ment capable of measuring the severity of flicker, which iscalled a flickermeter [2], and the IEEE recommended the adop-tion of the IEC flickermeter standard [3]. The IEC standard-compliant flickermeter simulates human eye–brain responses to

Manuscript received July 7, 2007; revised May 13, 2008. First publishedAugust 19, 2008; current version published December 9, 2008. This workwas supported by the U.S. Office of Naval Research under Grant N00014-02-1-0623. This paper was presented in part at the 2007 IEEE Instrumentationand Measurement Technology Conference, Warsaw, Poland, May 1–3, 2007.The Associate Editor coordinating the review process for this paper wasDr. Richard Thorn.

The authors are with the Department of Electrical and Computer Engineer-ing, The University of Texas at Austin, Austin, TX 78712-0240 USA.

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TIM.2008.928413

illumination flux changes of filament (i.e., incandescent) lampsdue to voltage fluctuations.

Flicker is usually produced by large-capacity, time-varyingloads such as adjustable-speed drives (ASDs) and arc furnaces.These loads often generate noninteger harmonics of the fun-damental power frequency, i.e., interharmonics, and it is wellknown that flicker phenomena are closely related to thoseinterharmonics [4]–[6].

However, according to recent studies [4], [7], [8], the currentIEC standard-based flickermeter cannot detect flicker causedby interharmonics whose frequencies are higher than a certainfrequency, i.e., 102 Hz for 60-Hz systems and 85 Hz for50-Hz systems. A real field case documenting the deficiency ofthe current standard is reported in [8]. Utility customers com-plained about light flicker, but standard-compliant flickermeterscould not detect it. To address the flickermeter standards’ defi-ciency, several approaches have previously been proposed [4],[9], [10], but they are based on empirical weighting functionsor deductive expansion from the current flickermeter standards.

In this paper, another deficiency of the current IEC flicker-meter will be presented. Our analysis and numerical experi-ments indicate that, in the presence of interharmonics whosefrequencies are below the difference between the fundamentalfrequency and the cutoff frequency of a bandpass filter of theIEC standard flickermeter, the demodulated beat frequenciesdetected by the IEC flickermeter do not match the actual beatfrequencies associated with the interharmonics. Various typesof frequency converters, including cycloconverters and ASDs,can generate these low-frequency interharmonics [11], [12].

To address the flicker-detection problem, we propose a novelapproach based on down-up sampling of the voltage signal.Our proposed method detects peak fluctuations imposed onthe absolute values of the voltage signal, and our analysisand simulation experiments confirm that the proposed methodis able to detect flicker, regardless of the frequency of theinterharmonics.

The remainder of this paper is organized as follows: Thecurrent IEC flickermeter is reviewed in Section II. Deficienciesin the flickermeter response to interharmonics are discussed inSection III. To overcome these deficiencies in Section IV, wepropose a novel flicker-detection method based on down-upsampling, and the efficacy of the proposed method is verifiedin Section V. Section VI concludes this paper.

II. IEC FLICKERMETER REVIEW

The IEC flickermeter estimates the severity of flicker using avoltage input signal. The flicker estimation process is based onthe incandescent lamp model and the lamp–eye–brain response

0018-9456/$25.00 © 2008 IEEE

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE UBERLANDIA. Downloaded on May 11,2010 at 00:31:57 UTC from IEEE Xplore. Restrictions apply.

KIM et al.: DETECTION OF FLICKER CAUSED BY INTERHARMONICS 153

Fig. 1. Simplified diagram of the IEC standard-based flicker estimation blocks [2].

model. Fig. 1 illustrates a simplified block diagram of the IECstandard-based flickermeter [2].

Block 1 first normalizes voltage input using a transformerand a gain controller. The normalized voltage input signal issquared by a multiplier in Block 2 to demodulate the flicker-related voltage fluctuation term.

Block 3 contains two filters. The first filter is a bandpassfilter for extracting the modulating terms from the squaredvoltage signal. The bandpass filter consists of a first-order high-pass filter whose cutoff frequency is 0.05 Hz and a sixth-order low-pass filter whose cutoff frequency is set to 42 for60-Hz systems and to 35 Hz for 50-Hz systems. This cutofffrequency is determined based on the assumption that anyfrequency higher than the cutoff frequency does not cause asignificant illumination change due to the thermal inertia of thefilaments in incandescent lamps. The second filter in Block 3is a weighting filter representing the human visual responseto fluctuations. It is known that the human visual sensitivityto illumination changes reaches a maximum level when thefrequency is between 8 and 10 Hz. Therefore, the filter responsein the second weighting filter of Block 3 has a maximumspectral response at 8.8 Hz.

The signal filtered by Block 3 continues to be processedby Block 4, which emulates human lamp–eye–brain responsesto illumination flux fluctuations. The nonlinear and memoryeffects of the human responses are simulated by squaring andsmoothing filters. The output of Block 4 indicates the instan-taneous flicker level. One unit of the output corresponds toflicker perceptibility threshold and is called the perceptibilityunit (PU).

The last block of the IEC flickermeter is the statistical evalu-ator of the instantaneous flicker level. Based on the cumulativedistribution function of the instantaneous flicker level (PU), theflicker sensation is evaluated over a longer period of time. Oneoutput of Block 5 is the short-term flicker severity index Pst.Pst is a measure of severity based on an observation over 10 min[2]. One unit of Pst corresponds to the threshold of illuminationchanges causing “irritation” to people, which is the so-called“irritation threshold.”

III. FLICKER CAUSED BY INTERHARMONICS

One potential limitation of the IEC flickermeter is its lackof consideration for flicker caused by interharmonics. It has re-cently been shown that the flicker characteristics and generation

process are considerably different in the presence of interhar-monics [4]. As mentioned earlier, the current IEC flickermeteris not capable of detecting flicker caused by high-frequencyinterharmonics [4], [7], [8]. According to [7], the deficiency isdue to the squaring and filtering processes (in Blocks 2 and 3 inFig. 1, respectively) during the flicker-estimation procedure ofthe IEC flickermeter.

In addition to these deficiencies, in this paper, we will presentanother deficiency of the IEC flickermeter with regard to flickercaused by low-frequency interharmonics, whose frequenciesare below the difference between the fundamental frequencyand the cutoff frequency of a bandpass filter of the IECstandard flickermeter. With the low-frequency interharmonics,the modulation (or beat) frequencies represented by the IECflickermeter are not consistent with the actual beat frequency.We will demonstrate the discrepancy with simple analysis andexamples in this section.

The flicker-evaluation process in the presence of interhar-monics can be examined using a signal model with a funda-mental frequency term and an interharmonic term, i.e.,

v(t) = sin(2πf1t) + m sin(2πfIHt + θIH), (1)

where f1, m, fIH, and θIH are the fundamental frequency, in-terharmonic relative magnitude, interharmonic frequency, andphase of the interharmonic, respectively. We assume that theinput signal is already normalized. This interharmonic signalmodel is also used in [6] and [7]. Note that the signal model in(1) is different from the direct amplitude modulation models(sinusoidal and rectangular voltage fluctuations) used in theIEC standard [2].

According to analytical and experimental studies in [4], theactual beat frequency fB due to the interharmonic is known asfollows:

fB = |fIH − hf1| = |fIH − (2m + 1)f1| , m = 0, 1, 2, . . .(2)

where h indicates the closest odd-order harmonic numberto fIH.

In [7], it is shown that if fIH is higher than 102 Hz for 60-Hzsystems or 85 Hz for 50-Hz systems, flicker associated with theinterharmonic is not detected by the IEC standard flickermeter.In this paper, we will focus on the case when fIH is lower than102 or 85 Hz.


Ezio

Realce


Fig. 2. Deficient flicker-detection example of the current IEC standard. (a) Waveform with an 8-Hz interharmonic. (b) DFT magnitude of the Block 3 outputof the IEC flickermeter. (c) Half-cycle RMS value fluctuations. (d) DFT magnitude of the RMS fluctuations. (e) Absolute value peak value fluctuations. (f) DFTmagnitude of the peak fluctuations.

According to the IEC flickermeter standard, the input voltagesignal is first squared and then input to the bandpass filter inBlock 3. The squared input voltage signal in (1) becomes

v2(t) = [sin(2πf1t) + m sin(2πfIHt + θIH)]2

=12(1 + m2) − 1

2cos(2π · 2f1 · t)

− 12m2 cos(2π · 2fIH · t + 2θIH)

+ m cos (2π · (f1 − fIH) · t − θIH)

− m cos (2π · (f1 + fIH) · t + θIH) . (3)

As a result of the bandpass filter in Block 3, the dc term andfrequencies higher than the cutoff frequency (42 or 35 Hz) in(3) are filtered out. Therefore, only two frequency terms, i.e.,2fIH and f1 − fIH, can possibly remain after filtering.

To simplify the discussion, we will only consider 60-Hzsystems (f1 = 60) and the corresponding cutoff frequency(42 Hz) of the bandpass filter in Block 3. We also assume that

the filters in the flicker processing blocks are ideal. Dependingon the fIH value and the corresponding response of the IECflickermeter, the entire frequency range can be divided intothree regions, and we will look into the response of the IECflickermeter in each frequency region.

A. 0 < fIH ≤ 18

In this case, 0 < 2fIH ≤ 36, and f1 − fIH ≥ 42. Hence,only the 2fIH term remains after bandpass filtering. Thus,the 2fIH term represents the fluctuating term detected bythe flickermeter. However, the actual beat frequency associ-ated with this interharmonic frequency fIH is not 2fIH butf1 − fIH, according to (2), since the closest odd-order har-monic number h is 1.

This discrepancy in the actual beat frequency and that de-tected by the flickermeter can lead to inaccurate flicker eval-uation results. For example, the weighting filter following thebandpass filter in Block 3 simulates the frequency responseof lamp–eye reaction to light flicker. Therefore, if the input



Fig. 3. Compliance test result of the numerically implemented IEC flickermeter. Plot of the percentage deviation versus modulation frequencies.

frequency to the weighting filter is not matched with the actualbeat frequency, the resulting flicker-evaluation indexes canbecome erroneous with respect to actual flicker phenomena.

Fig. 2 illustrates the beat frequency discrepancy of the IECflickermeter in the presence of a low-frequency interharmonic.Fig. 2(a) and (b) presents a sample voltage input (f1 = 60 Hz,m = 0.2, and fIH = 9 Hz) and the discrete Fourier transform(DFT) magnitude of the Block 3 output of the IEC flickermeter,respectively. As discussed earlier, only the 16(= 2 × 8) − Hzterm remains after bandpass filtering in Block 3 of the IECflickermeter. Fig. 2(c) and (d) illustrates the half-cycle root-mean-square (RMS) fluctuations and the corresponding DFTmagnitude, and Fig. 2(e) and (f) demonstrates the absolutevalue peak fluctuations and the corresponding DFT magnitude.Since the input ac voltage is squared (e.g., incandescent lamps)or rectified (e.g., fluorescent lamps) to produce the light out-put, the fluctuations of the half-cycle RMS values and of theabsolute values should be considered. For both cases, the actualfluctuation terms are 52(= 60 − 8) Hz, as shown in Fig. 2(d)and (f), which is not consistent with the DFT magnitude resultof the IEC flickermeter shown in Fig. 2(b). Furthermore, themagnitude of the 16-Hz term in Fig. 2(b) is determined by thesquare of the interharmonic relative magnitude, namely, m2,which is very small since usually m � 1.

B. 18 < fIH < 102

This frequency range can again be divided into two cases, asfollows.

1) 18 < fIH ≤ 21: Both the 2fIH and f1 − fIH frequencyterms are left after filtering. However, the 2fIH term can beignored since the corresponding magnitude m2 is very small,as stated earlier. Therefore, the f1 − fIH term represents thebeating term.

2) 21 < fIH < 102: In this frequency range, only the f1 −fIH term in (3) remains after bandpass filtering. This frequencyis consistent with the actual beat frequency, as indicated in(2), with h = 1. Thus, the IEC flickermeter appears to prop-erly assess flicker caused by interharmonics in this frequencyregion.

C. 102 ≤ fIH

The deficiency of the IEC flickermeter in this frequencyrange has been discussed by other authors [4], [7], [8], [13].

Fig. 4. Minimum interharmonic magnitude mmin necessary to produceone PU of the IEC flickermeter response versus interharmonic frequenciesranging from 0 to 120 Hz.

Since the bandpass filter in Block 3 suppresses all interhar-monic terms whose frequencies are higher than 102 Hz (85 Hzfor 50-Hz systems), the IEC flickermeter cannot detect flickercaused by interharmonics in this frequency range, regardless oftheir magnitude.

To investigate the response of the IEC flickermeter to in-terharmonics, we numerically implemented the IEC standardflickermeter algorithm using MATLAB and experimented withinterharmonics ranging from 0 to 120 Hz. To ensure the com-pliance of the implemented flickermeter to the IEC standard,the compliance test based on the response of the flickermeter tosinusoidal amplitude modulation was carried out according tothe “Normalized flickermeter response for sinusoidal voltagefluctuations” table in the standard [2], and the results arepresented in Fig. 3. Fig. 3 is a plot of the deviation of theminimum voltage fluctuation level, which generates a unit valueof PU, from the reference values in the IEC standard tableversus different modulation frequencies. The IEC standardallows a maximum 5% deviation, and the test results in Fig. 3show that our numerical implementation is well within the 5%allowance.

Fig. 4 illustrates the sensitivity of the IEC flickermeter tointerharmonics ranging from 0 to 120 Hz. Each point onthe curve in Fig. 4 represents the minimum interharmonicmagnitude mmin, which results in a unit value of PU at the



Fig. 5. Proposed flicker peak detection blocks.

IEC flickermeter output at each interharmonic frequency. Thefrequency range 0–120 Hz is divided into three regions based onthe characteristics of the IEC flickermeter, as discussed earlierin this section.

In the frequency region A in Fig. 4, the beat frequency dueto an interharmonic is not properly represented by the IECflickermeter. Note that the local minima point in the A region isat 4.4 Hz, which is half of the maximum response frequency(8.8 Hz) of the weighting filter in Block 3. Furthermore, asdiscussed earlier, only the squared value of the relative inter-harmonic magnitude appears in this frequency region A, so thatthe influence of the interharmonics in this region is significantlyreduced.

The IEC flickermeter seems to properly work for inter-harmonics in region B, but as the interharmonic frequencybecomes greater than 102 Hz in region C, the interharmonicmagnitude required to generate perceptible flicker rapidlygrows, and it goes beyond the 10% of the fundamental fre-quency magnitude after about 110 Hz. However, according tothe real field case reported in [8], light flicker was observed byresidential customers when 180- to 200-Hz interharmonics with1% magnitude of the fundamental component were present,thereby confirming the deficiency of the IEC flickermeter stan-dard in the high-frequency range.

Based on our analysis and numerical experiments, the defi-ciency of the IEC flickermeter with regard to flicker caused byinterharmonics has been demonstrated. To address this prob-lem, we propose a new approach based on down-up samplingof the input voltage signal in the next section.

IV. DOWN-UP SAMPLING-BASED

PEAK DETECTION METHOD

With interharmonics present, there are two types of fluctu-ations due to the interharmonics (RMS and peak value fluc-tuations) [6]. According to the experimental study reportedin [4], the flicker behavior widely varies, depending on thetype of lamp. Analysis results in [6] suggest that certain typesof lamps (e.g., compact fluorescent lamps) are more sensitiveto peak value fluctuations of the input voltage rather thanRMS value fluctuations. Furthermore, even small magnitudeinterharmonics that do not cause significant RMS fluctuationscan result in considerable peak fluctuations. Therefore, for agiven interharmonic amplitude, a possible “worst” case might

occur if the customers’ lighting system consists mainly of peak-fluctuation-sensitive lamps.

Hence, in this paper, we propose alternative signal processingblocks to detect peak fluctuations for flicker evaluation. Fig. 5presents the signal processing blocks for our proposed flicker-detection method. As we have already mentioned, light flickerdue to peak value fluctuations is caused by modulation imposedon the absolute value of the voltage signal, which has thesecond harmonic of the fundamental frequency as the dominantfrequency term. Therefore, our basic idea is to examine thepeak fluctuation imposed only on the second harmonic of thefundamental frequency.

First, the input voltage signal is sampled by the samplingfrequency fS = Mf1, as schematically indicated in Block A inFig. 5. The sampling frequency should be an integer multiple Mof the fundamental frequency to prevent spectral leakagearound the fundamental term. The IEC standard allows eitheranalog or digital implementation of the recommended flicker-meter [2], and flickermeters are digitally implemented usingvarious approaches in the literature, including [14]–[17].

Then, the absolute values of the samples are generated, asshown in Block B in Fig. 5. Assuming that the voltage signalsare processed in the discrete-time sample domain, we can assessthe peak fluctuations of the absolute-valued voltage signals bydownsampling to 2f1 or downsampling by (2/M), as shownin Block C in Fig. 5. The downsampling process actuallycorresponds to taking one sample for each half-cycle of thefundamental. The process can be represented in an analyticalway using the signal model in (1).

First, the analytical expression of the absolute value ofthe voltage signal should be considered. We can consider theabsolute values of a sinusoidal signal as the multiplicationbetween the sinusoidal signal and a square-wave function withthe same period. Assuming that we have a sinusoidal signalsin(2πf1t), then the Fourier series representation of the cor-responding square-wave function s(t) is as follows:

s(t) =4π

∑n=1,3,...

1n

sin(2πnf1t) (4)

where n is an odd integer. Therefore, the absolute values of thesignal model in (1) are given in (5), shown at the bottom ofthe next page, where the vf (t) part indicates the terms relatedto the fundamental frequency, and the vIH(t) part indicates



the terms related to the interharmonics. Then, the discrete-timeversion of the vf (t) part can be represented as follows:

vf [k] = vf

(t0 +

k

2f1

)=

2π

∑n=1,3,...

1n

{cos

(2π(n − 1)f1

(t0 +

k

2f1

))

− cos(

2π(n + 1)f1

(t0 +

k

2f1

))}

=2π

∑n=1,3,...

1n

{cos (2π(n − 1)f1t0 + (n − 1)πk)

− cos (2π(n+1)f1t0+(n+1)πk)

}(6)

where k denotes the discrete-time index, and t0 is the initialsampling time. vf [k] becomes a dc term (a constant value)since the 2π(n ± 1)f1t0 terms in the cosine functions areindependent of k, and the other terms in the cosine functions,i.e., (n ± 1)πk, become 2lπk(l = 0, 1, 2, . . .) due to the factthat n is an odd integer.

According to (2), fIH can be represented as hf1 ± fB (h =2p + 1; p = 0, 1, 2, . . .), and then, the discrete-time version ofvIH in (6) becomes the following:

vIH[k] = vIH

(t0 +

k

2f1

)=

2m

π

∑n=1,3,...

1n

cos(

2π(hf1 ± fB − nf1)

×(

t0 +k

2f1

)+ θIH

)− 2m

π

∑n=1,3,...

1n

cos(

2π(hf1 ± fB + nf1)

×(

t0 +k

2f1

)+ θIH

)=

2m

π

∑n=1,3,...

1n

{cos

(2πlfB

k + θn−

IH

)

− cos(2πlfB

k + θn+

IH

) }(7)

where θn−IH = 2π((h − n)f1 ± fB)t0 + (h − n)πk + θIH, and

θn+

IH = 2π((h + n)f1 ± fB)t0 + (h + n)πk + θIH. Note thatlfB

(= ±(fB/2f1)) corresponds to the discrete frequency rep-resentation of the actual beat frequency in (2). Hence, vIH[k]represents the magnitude m of the actual beat term and itsfrequency fB .

Consequently, we can find a frequency term that representsthe actual beating effect via the downsampling process. How-ever, there are potential problems if only the downsamplingmethod is used. For example, the remaining parts of the flick-ermeter should be changed for the downsampled signal. More-over, to obtain the proper accuracy recommended by the IECspecifications, the sampling frequency should be sufficientlyhigh [14], [16]. Additionally, we want to replace those signalprocessing blocks in the IEC standard, which potentially causeproblems in detecting flicker caused by interharmonics andseamlessly combine the newly proposed blocks with the restof the flickermeter blocks, which comply with the current IECstandard.

Therefore, we up-sample the downsampled signal to theoriginal sampling frequency, as indicated in Block D in Fig. 5.The upsampling process is done by inserting zeros betweensamples, which is known as zero padding [18]. This down-up sampled signal carries the peak fluctuation information assideband terms around each harmonic of the fundamental term.

The following Block E in Fig. 5 is a bandpass filter thatremoves the dc term and unnecessary frequency terms resultingfrom the upsampling. In this paper, we used a first-order high-pass filter whose cutoff frequency is 0.05 Hz and a sixth-order,low-pass filter whose cutoff frequency is 60 Hz for 60-Hzsystems and 50 Hz for 50-Hz systems, which are half of thedownsampling frequency. Note that the high-frequency cutoffof this bandpass filter is different from that of the bandpass filterin the IEC flickermeter.

Blocks A–E in Fig. 5 comprise the proposed alternative sig-nal processing blocks. The proposed signal processing blocksreplace Block 2 and the first bandpass filter in Block 3 of theIEC standard flickermeter in Fig. 1. The rest of the proposedflickermeter, including the weighting filter in Fig. 5, is the sameas the IEC standard flickermeter.

|v(t)| = v(t) · s(t)

= (sin(2πf1t) + m sin(2πfIHt + θIH)) ·(

4π

∑n=1,3,...

1n

sin(2πnf1t)

)

=4π

∑n=1,3,...

1n

sin(2πf1t) sin(2πnf1t) +4π

∑n=1,3,...

m

nsin(2πfIHt + θIH) sin(2πnf1t)

=4π

∑n=1,3,...

{12n

cos (2π(n − 1)f1t) −12n

cos (2π(n + 1)f1t)}

︸︷︷︸vf (t)

+4π

∑n=1,3,...

{ m

2ncos (2π(fIH − nf1)t + θIH) − m

2ncos (2π(fIH + nf1)t + θIH)

}︸︷︷︸

vIH(t)

(5)



Fig. 6. Down-up sampling-based flicker detection with an 8-Hz interharmonic term. (a) Discrete time-domain signal after Block D in Fig. 5. (b) CorrespondingDFT magnitude spectrum.

Note that our method can be considered “conservative” inassessing the likelihood of flicker, particularly when it is usedfor assessing flicker of lighting systems, which consist mainlyof incandescent lamps. That is because our method evaluatesthe peak fluctuations of the voltage input, while incandescentlamps are sensitive to the RMS value fluctuations. As discussedin [6], peak values generally tend to fluctuate more severelythan RMS values do in the presence of the same amplitudeinterharmonic.

We implemented our new proposed algorithm usingMATLAB, and the experimental results based on the implemen-tation are presented in the next section.

V. EXPERIMENTAL RESULTS

To demonstrate the efficacy of the proposed method, wefirst show a proper flicker-detection example using the samesignal model in Section III, i.e., the 60-Hz voltage signal withan 8-Hz interharmonic. The signal is sampled at 1200 Hz(= 60 × 20), which also prevents spectral leakage around thefundamental term. Fig. 6 illustrates the signal processing resultof the proposed alternative blocks. Fig. 6(a) shows the discrete-time waveform after Block D in Fig. 5, which is the down-up sampled signal, and Fig. 6(b) presents the correspondingfrequency spectrum result.

Contrary to the misleading frequency spectrum resultin Fig. 2(b), which showed a 16-Hz term instead of52(= 60 − 8) Hz, we can successfully detect the 52-Hz fluc-tuation term in the DFT magnitude spectrum of Fig. 6(b). Theother higher frequency terms and the dc term will be filtered outby a bandpass filter in Block E in Fig. 5. The 52-Hz component

is then processed by statistical processing units to generateflicker-severity indexes.

An example of detecting flicker caused by a high-frequency interharmonic (172 Hz) is also presented in Fig. 7.As mentioned earlier, the current IEC standard flickerme-ter cannot detect the flicker caused by this high-frequencyinterharmonic since the higher frequency term is filtered outby a bandpass filter in Block 3 of the IEC flickermeter in Fig. 1.However, the down-up sampled signal presented in Fig. 7(a)and the corresponding DFT magnitude in Fig. 7(b) accuratelyshow the actual beating 8(= 3 × 60 − 172) − Hz term.

For a quantitative comparison, the sensitivity curve of theproposed method to interharmonics is presented in Fig. 8, alongwith that of the IEC flickermeter. These curves represent theminimum interharmonic magnitude mmin to produce one PU inthe flickermeter output. The sensitivity curves for both methodsappear identical in the frequency range from around 30 to 90 Hzin Fig. 4, where the IEC flickermeter properly works withinterharmonics.

However, the sensitivity curves for both methods exhibitdifferent behaviors in the lower frequency range below 30 Hzand the higher frequency range above 90 Hz. Contrary to the in-accurate response of the IEC flickermeter in the low-frequencyrange (0–18 Hz), as discussed earlier, the curve correspondingto the newly proposed method shows that the proposed methodsensitively reacts to the low-frequency interharmonics in that asmaller interharmonic magnitude is required to produce one PU.

In the high-frequency range (90–120 Hz), according to thecurve corresponding to the standard method, large magnitudesof the interharmonic are required to produce one PU in theflickermeter output, which is contrary to the field case reported


Ezio

Realce

Ezio

Realce

Ezio

Realce


Fig. 7. Down-up sampling-based flicker detection with a 172-Hz interharmonic term. (a) Discrete time-domain signal after Block D in Fig. 5. (b) CorrespondingDFT magnitude spectrum.

Fig. 8. Minimum interharmonic magnitude mmin necessary to produceone PU for the IEC flickermeter response and the proposed method versusinterharmonic frequencies.

in [8]. On the other hand, the curve corresponding to the pro-posed method is well extended into the high-frequency rangeand shows that relatively small magnitudes of the interharmoniccan produce one PU.

The numerical experimental results presented in this sectionindicate that our proposed flicker processing blocks are able toproperly detect flickers caused by interharmonics, regardless oftheir frequencies.

VI. CONCLUSION

In this paper, we have presented a deficiency of the currentIEC flickermeter with regard to low-frequency interharmonics

and propose a new approach based on down-up sampling ofthe voltage fluctuation signal to address this deficiency. Usingthe down-up sampling-based method, peak value fluctuationsof the input voltage were examined to assess interharmonic-caused flicker.

Based on our analysis and numerical experiments, it wasdemonstrated that our proposed method correctly representsthe actual beat frequency associated with the interharmonicsin any frequency range. As a result, the proposed method canextend the interharmonic-related flicker-detection range beyondthe current limited range of the IEC flickermeter.

Since our method is based on evaluating peak value fluctu-ations, it is particularly suitable for assessing flicker of peak-fluctuation-sensitive lamps, e.g., compact fluorescent lamps. Inthis sense, our method may be considered to be complementaryto the IEC standard that is focused on assessing flicker of incan-descent lamps, which are sensitive to RMS value fluctuations.Furthermore, our proposed signal processing blocks can beimmediately combined with the current flickermeter standardto address the interharmonic-related limitations.

ACKNOWLEDGMENT

The authors would like to thank the anonymous reviewers fortheir constructive suggestions.

REFERENCES

[1] S. Clippingdale and H. Isono, “Photosensitivity, broadcast guidelines andvideo monitoring,” in Proc. IEEE Int. Conf. Syst., Man, Cybern., Tokyo,Japan, Oct. 1999, pp. 22–27.

[2] Electromagnetic Compatibility (EMC). Part 4: Testing and measurementtechniques. Section 15: Flickermeter. Functional and Design Specifica-tions, IEC Std. 61 000_4_15, 2003.


Ezio

Realce


[3] IEEE Recommended Practice for Measurement and Limits of VoltageFluctuations and Associated Light Flicker on AC Power Systems, IEEEStd. 1453_2004, 2004.

[4] M. D. Koster, E. D. Jaiger, and W. Vancoistem, Light Flicker Causedby Interharmonics. [Online]. Available: http://grouper.ieee.org/groups/harmonics/iharm/docs/ihflicker.pdf

[5] T. Keppler, N. R. Watson, J. Arrillaga, and C. Shiun, “Theoretical assess-ment of light flicker caused by sub- and interharmonic frequencies,” IEEETrans. Power Del., vol. 18, no. 1, pp. 329–333, Jan. 2003.

[6] T. Tayjasanant, W. Wang, and W. Xu, “Interharmonic–flicker curves,”IEEE Trans. Power Del., vol. 20, no. 2, pp. 1017–1024, Apr. 2005.

[7] W. Xu, “Deficiency of the IEC flicker meter for measuring interharmonic-caused voltage flickers,” in Proc. IEEE Power Eng. Soc. Gen. Meeting,Jun. 2005, pp. 2285–2288.

[8] T. Tayjasanant and W. Xu, “A case study of flicker/interharmonic prob-lems caused by a variable frequency drive,” in Proc. 11th Int. Conf.Harmonics Quality Power, Sep. 2004, pp. 72–76.

[9] D. Gallo, R. Langella, and A. Testa, “Light flicker prediction basedon voltage spectral analysis,” in Proc. IEEE Porto Power Tech Conf.,Sep. 2001, pp. 10–13.

[10] D. Gallo, R. Langella, and A. Testa, “Toward a new flickermeter basedon voltage spectral analysis,” in Proc. IEEE Int. Symp. Ind. Electron.,Jul. 2002, pp. 573–578.

[11] E. W. Gunther, “Interharmonics in power systems,” in Proc. IEEE PowerEng. Soc. Summer Meeting, Jul. 2001, vol. 2, pp. 813–817.

[12] F. D. Rosa, R. Langella, A. Sollazzo, and A. Testa, “On the interharmoniccomponents generated by adjustable speed drives,” in Proc. IEEE 10thInt. Conf. Harmonics Quality Power, Oct. 2002, vol. 1, pp. 183–188.

[13] T. Kim, A. Wang, E. J. Powers, W. M. Grady, and A. Arapostathis,“Detection of flicker caused by high-frequency interharmonics,” in Proc.IEEE Instrum. Meas. Technol. Conf., Warsaw, Poland, May 2007, pp. 1–5.

[14] J. de Sousa Rocha and S. Deckmann, “Digital flickermeter implemen-tation,” in Proc. 38th Midwest Symp. Circuits Syst., Aug. 1995, vol. 2,pp. 757–760.

[15] M. Chen, “Digital algorithms for measurement of voltage flicker,” Proc.Inst. Elect. Eng.—Generation, Transm. Distrib., vol. 144, no. 2, pp. 175–180, Mar. 1997.

[16] S. Caldara, S. Nuccio, and C. Spataro, “A virtual instrument for measure-ment of flicker,” IEEE Trans. Instrum. Meas., vol. 47, no. 5, pp. 1155–1158, Oct. 1998.

[17] A. Bertola, G. Lazaroiu, M. Roscia, and D. Zaninelli, “A Matlab–Simulinkflickermeter model for power quality studies,” in Proc. 11th IEEE Int.Conf. Harmonics Quality Power, Sep. 2004, pp. 734–738.

[18] A. V. Oppenheim and R. W. Schafer, Discrete-Time Signal Processing.Englewood Cliffs, NJ: Prentice-Hall, 1999.

Taekhyun Kim (S’03) received the B.S. degree fromSeoul National University, Seoul, Korea, and theM.S. degree from The University of Texas at Austin,where he is currently working toward the Ph.D.degree in electrical and computer engineering.

His research interests include the theory and al-gorithm development of advanced signal processingtechniques for power system analysis.

Edward J. Powers (F’83) received the B.S. degreefrom Tufts University, Medford, MA, the M.S. de-gree from the Massachusetts Institute of Technology,Cambridge, and the Ph.D. degree from Stanford Uni-versity, Stanford, CA, all in electrical engineering.

He is the Texas Atomic Energy Research Foun-dation Professor of Engineering and a Professor ofelectrical and computer engineering with The Uni-versity of Texas at Austin. His primary professionalinterests are in the innovative application of digitalhigher order statistical signal processing in the analy-

sis, interpretation, and modeling of time series data characterizing nonlinearphysical phenomena, e.g., interharmonics and health monitoring of rotatingmachines, and the utilization of the wavelet transform and time–frequencyanalysis to detect and identify transient events in various physical systems,including electric power systems.

W. Mack Grady (F’00) received the Ph.D. degreefrom Purdue University, West Lafayette, IN.

He is a Professor of electrical and computer engi-neering with The University of Texas at Austin. Heworked for six years as a System Planning Engineerfor Texas Utilities, Dallas.

Prof. Grady is a Chairman of the IEEE PowerEngineering Society (PES) Transmission and Distri-bution Committee and was the Technical ProgramChairman for the 2003 IEEE PES Transmission andDistribution Conference and Exposition.

Ari Arapostathis (F’07) received the B.S. de-gree from Massachusetts Institute of Technology,Cambridge, and the Ph.D. degree from the Universityof California, Berkeley.

He has been a Faculty Member with The Uni-versity of Texas at Austin since 1982. His researchinterests include analysis and estimation techniquesfor stochastic systems, the application of differentialgeometric methods to the design and analysis ofcontrol systems, stability properties of large-scaleinterconnected power systems, and stochastic and

adaptive control theory.


detection of flicker caused by inter harmonics

Documents