MEASUREMENT AND EVALUATION OF THE FLICKER EMISSION LEVEL FROM A
PARTICULAR FLUCTUATING LOAD
2nd DRAFT
Prepared on request of CIGRE / CIRED WG CC02 (Voltage Quality)
Emmanuel De Jaeger, Laborelec, Belgium
June 2000
1. BRIEF RECALL OF THE UIE - IEC FLICKER MEASURING METHOD .....................................................2
2. EMISSION LEVEL EVALUATION WITH SWITCHING OF THE CONSIDERED LOAD ..........................2
2.1 STATISTICAL APPROACH BASED ON NORMALITY ASSUMPTION ..............................................................................3
2.2 APPLICATION OF THE GENERAL SUMMATION LAW ................................................................................................3
2.2.1 Summation effects........................................................................................................................................3
2.2.2 Application to comparative measurements .................................................................................................4
3. STATISTICAL APPROACH FROM SIMULTANEOUS MEASUREMENT OF FLICKER AND POWER
CONSUMPTION ..............................................................................................................................................................5
3.1 ESTIMATING THE LOCATION OF FLICKER SOURCES................................................................................................5
3.2 FULL STATISTICAL ASSESSMENT OF EMISSION LEVELS IN THE PARTICULAR CASE OF ARC FURNACES ....................6
4. DIRECT ON-LINE ASSESSMENT, BASED ON SIMULTANEOUS VOLTAGE AND CURRENT
MEASUREMENTS ..........................................................................................................................................................8
5. DIRECT ON-LINE MEASUREMENT OF FLICKER EMISSION LEVEL......................................................8
5.1 ORIGINAL METHOD: FLICKER EMISSION LEVEL RELATED TO A REFERENCE IMPEDANCE ........................................8
5.2 FLICKER EMISSION LEVEL RELATED TO THE ACTUAL OR CONTRACTUAL SHORT-CIRCUIT LEVEL ...........................9
5.3 DERIVED METHOD ..............................................................................................................................................11
2
1. Brief recall of the UIE - IEC flicker measuring method
All flicker measurements discussed in this report are based on the internationally agreed UIE - IEC method.
The basics of this method have been presented in the UIE publication [ 1] and agreed by IEC [ 2], [ 3].
The aim of flicker measurements is to define and assess a criterion expressing directly the degree of irritation of human
beings. The UIE-IEC flickermeter was developed in order to achieve this task through the measurement of voltage
fluctuations and correct simulation of the lamp-eye-brain behaviour. A simplified signal flow chart of the UIE-IEC
flickermeter is given in
Figure 1.
Figure 1 - Signal flow chart of the UIE - IEC flickermeter
The first block (voltage-input adapter) controls the 1-minute average of the voltages to a constant output so that the
fluctuations are always related to the average voltage level of the last minute.
The voltage fluctuations are then demodulated by squaring and band-pass filtering. These operations simulate the
physical behaviour of an incandescent lamp.
The combined influence of incandescent lamp and eye is simulated by a 4th order band-pass (weighting filter, having a
transfer function with a maximum amplitude at a frequency around 8.8 Hz).
The following blocks (squaring and smoothing through a first order low-pass filter) represent the brain reaction
(Rashbass model).
The obtained output signal is thus proportional to the voltage variations, weighted by the sensitivity of the combination
lamp-eye-brain : this is a measure of the momentary flicker sensation level. The absolute value of this signal must be
divided by the corresponding value of the perceptibility threshold in order to be expressed in terms of units of
perceptibility.
The final statistical evaluation calculates the so-called flicker severity level Pst which directly expresses the degree of
irritation, with Pst = 1 p.u. meaning the limit of disturbance (threshold of irritability). One Pst value is based on a short-
term observation period (usually ten minutes) and is calculated as the root sum of some weighted percentiles of the
momentary flicker sensation level.
Long-term flicker severity Plt is calculated as the cubic mean of twelve consecutive Pst values (i.e. over a time window of
two hours).
2. Emission level evaluation with switching of the considered load
Simple voltage measurements provide only the global effect of all the disturbers present in the network. However,
comparative measurements with and without the concerned disturbing load can be useful in order to assess its emission
level, providing that some assumptions are made.
3
2.1 Statistical approach based on normality assumption
If the flicker distributions (i.e. distributions of the measured Pst values) respectively with and without the considered
fluctuating load can be considered as normal (gaussian), the emission level distribution is also normal and its
characteristic parameters can be deduced as follows [4] :
µ i = µ with − µ without
(1)
σi = σ 2with − σ 2without
where µ represents the mean and σ, the standard deviation of the distribution.
Figure 2 shows an example of such a calculation of the flicker emission level of a DC arc furnace.
Figure 2 - Statistical deduction of the emission level of a DC arc furnace
The major disadvantages of the method are :
- The normality assumption is generally valid for the “background” flicker level, i.e. the flicker level resulting from
various slightly fluctuating loads (central limit theorem), or in certain situations with only one important disturbing
consumer. However, the assumption may not be valid in situations where two or three dominating loads are
operating simultaneously (this situation leading usually to multimodal distributions).
- Even if the normality assumption is globally acceptable, the actual 95 % and 99 % percentiles of the Psti distribution
(which are the interesting parameters for emission levels assessments [ 5 ] ) may differ significantly from those
obtained with this method. If the actual distribution does not fit perfectly the normal one, even small discrepancies
lead unfortunately to significant differences for the 95 % or 99 % of the cumulative function, due to the very low
slope of the curve in the high probability region.
- Measurements are not performed simultaneously and the result is influenced by possible changes in the operating
conditions and / or the network topology.
2.2 Application of the general summation law
2.2.1 Summation effects
A general combination relationship for short-term flicker severity caused by various loads has been found in the
following form, where Pst i are the various individual levels of flicker severity to be combined [ 1 ], [ 5 ] :
4
1m
Pst = ç m
Psti ÷ (2)
i
The value of the coefficient m depends upon the characteristics of the sources of fluctuation and is generally chosen
between 1 (where there is a very high occurrence of coincident voltage changes) and 4 (summation of voltage changes
due to arc furnaces specifically run to avoid coincident melts).
The value m = 3 is used for most types of voltage changes where the risk of coincident voltage changes is limited. The
vast majority of studies combining unrelated disturbances will fall into this category and the use of this value is
recommended whenever there is doubt over the magnitude of the risk of coincident voltage changes occurring.
In fact, recent empirical studies have shown that the summation law which best fits measurement results depends on the
Pst percentile which is chosen for the evaluation [ 6], [ 7]. In a two arc furnaces case, for example, it appeared that the
summation is practically linear up to a probability (P) of 50 %; it becomes quadratic for P ≅ 75 % and cubic for P ≅ 90
%. For P ≥ 95 % it is practically difficult to identify a summation law, the measured flicker level being almost entirely
caused by the most disturbing load (m ≥ 4).
In general, for voltage quality characterising purposes (95 % or 99 % probabilities), the cubic summation gives thus a
conservative evaluation but may be close to the truth. It seems justified to keep it as a general approach for
predetermination studies and emission limits assessment.
2.2.2 Application to comparative measurements
The general summation law based on exponent m = 3 can be applied to comparative measurements with and without the
concerned disturbing load in order to evaluate its emission level [ 8] :
( )
1 3
Psti = Pst
3
, with − P st , without
3 (3)
Evaluation resulting from this method is rather uncertain, mainly due to two reasons :
- measurements are not performed simultaneously and the result is influenced by possible changes in the operating
conditions of other disturbing loads and / or the network topology (influencing particularly the short-circuit level).
- the use of the exponent m=3 may not be adequate for each type of fluctuating load; this can be solved by using a
better fitted parameter, if known.
5
3. Statistical approach from simultaneous measurement of flicker and power consumption
3.1 Estimating the location of flicker sources
Simultaneous measurements of flicker and additional parameters such as currents, active or reactive powers can be used
in order to detect the main flicker sources, through correlation analysis.
The following example illustrates how such an analysis can be done [ 8 ], [ 9].
Figure 3 - Example of correlation analysis showing the determination of the dominating flicker source
Two different disturbing consumers A and B (steel plants with arc furnaces of 15 and 25 MW respectively) are causing
flicker at the substation C (110 kV). Correlation analysis has been performed in order to determine which of the
consumers was the main source of the flicker level measured in C. The flicker level (Pst) is plotted versus fluctuations of
reactive powers in A and B. A strong dependence of the flicker to plant A is observed (good correlation) while there is
virtually none to plant B (weak correlation).
6
3.2 Full statistical assessment of emission levels in the particular case of arc furnaces
If the flicker level is continuously recorded at the point of common coupling for a sufficient long period of time,
together with the power consumption of the different consumers, a clear distinction can be made between situations with
0, 1 or more furnaces (for instance, each furnace is considered as continuously working as long as no interruption longer
than 2h occurs; a longer interruption being considered as an OFF period). For each of the possible combinations, all the
Pst results are put together and statistics (such as Pst 95% or Pst 99%) are calculated.
An example of such measurements (one day) is shown in Figure 4, for a 3-furnaces case.
Figure 4 - One day flicker measurements in a HV network (individual Pst values and gliding Plt curve) in a 3-
furnaces case
........... furnace A (AC, 100 MVA)
______ furnace B (AC, 100 MVA)
|||||||||| furnace C (DC, 95 MVA)
Selecting the Pst values according to the energy demand record of the different arc furnaces seems to be an efficient
method to assess the contribution from each individual furnace. A drawback is that time is necessary before getting
reliable results. The measurement period leading to the equivalent of one-week results (± 1000 Pst values) may in fact be
very long. For instance, in a practical experience described in [ 6 ] and [ 7 ], two and a half months were necessary for a
given DC furnace.
Figure 5 gives the statistics of the flicker level running from the start of the measurements up to the considered point,
each point giving an additional period of 2 weeks : e.g. for the overall flicker level (Figure 5.a), the first point
corresponds to a period of 2 weeks and the 14th point to a period of 7 months (14 times 2 weeks). For furnace C (Figure
5.d), we have 13 points instead of 14 because there was one 2 weeks period where no Pst value was available for this
furnace alone.
The comparison between Figure 5.a and Figure 5.d may seem surprising: during roughly the first half of the
measurement period, the global flicker level was lower than the one from the most disturbing furnace. The explanation is
that the number of Pst values is much greater (about ten times) for the overall level, so that a greater number of high Pst
values are eliminated when assessing Pst 99%.
7
1.5
1.4
1.3
1.2
Pst, Plt (pu)
1.1
0.9
0.8
0.7
Pst99% Pst95% Plt99% Plt95%
0.6
0 5000 10000 15000 20000 25000 30000 35000
Number of Pst values
(a) Continuous measurement (overall flicker level)
1.5
Pst99% Pst95% Plt99% Plt95%
1.4
1.3
1.2
Pst, Plt (pu)
1.1
0.9
0.8
0.7
0.6
0 500 1000 1500 2000 2500 3000
Number of Pst values
(b) Discontinuous measurements, periods when furnace A (AC) was alone
1.5
Pst99% Pst95% Plt99% Plt95%
1.4
1.3
1.2
Pst, Plt (pu)
1.1
0.9
0.8
0.7
0.6
0 500 1000 1500 2000 2500 3000 3500
Number of Pst values
(c) Discontinuous measurements, periods when furnace B (AC) was alone
1.5
1.4
1.3
1.2
Pst, Plt (pu)
1.1
0.9
0.8
0.7
Pst99% Pst95% Plt99% Plt95%
0.6
0 500 1000 1500 2000 2500 3000 3500
Number of Pst values
(d) Discontinuous measurements, periods when furnace C (DC) was alone
Figure 5 - Flicker measurements in a HV network at the PCC of a DC furnace (95 MVA) not far from two other AC
furnaces (100 MVA each). Statistics for periods of 0.5, 1, 1.5, ... , 7 months
Another disadvantage of the method - linked to the rather long measurement period - is the risk of changes in network
topology, possibly leading to significantly different short-circuit levels. The obtained individual Psti values are then no
longer related to a constant short-circuit capacity.
However, should a reliable value of the short-circuit level be known for each data, the results could be re-processed in
order to be related to a fixed reference short-circuit capacity (e.g. the contractual value).
8
4. Direct on-line assessment, based on simultaneous voltage and current measurements
Using simultaneous voltage and current measurements at the point of coupling of a fluctuating load allows the
determination of the network’s source impedance [10]. This relies upon the assumption that the network actually has a
linear behaviour and may be represented by the classical Thevenin equivalent, i.e. a constant voltage source behind an
impedance.
The determination of the network’s impedance has to be done by processing the simultaneous voltage and current
fluctuations known to be caused by the considered fluctuating load only This process is illustrated in Figure 6 where
the module of the impedance Z is given by the slope of the least-square regression straight line passing through the
points.
125
124
Voltage (V)
123
122
121
120
-3.5 -3 -2.5 -2 -1.5
Current (I)
Figure 6 - Plot of voltage vs. current for a single event known to be caused by a given fluctuating load
Knowing the system’s impedance Z, the emission level of the considered load will be evaluated as the flicker caused by
V emission = E0 − Z Iload (4)
where E0 is the constant voltage source of the network’s Thevenin equivalent circuit and Iload is the fluctuating load
current.
5. Direct on-line measurement of flicker emission level
5.1 Original method: flicker emission level related to a reference impedance
An original direct on-line measurement method has been developed by Dr M. Sakulin, (University of Technology, Graz,
Austria) [ 8 ], [11], [ 9 ].
The flicker measurement unit Pst is homogeneous to relative voltage changes. Accordingly, rather than looking at the
summation result of Pst values, the method relies upon the superposition principle, applied to voltage fluctuations
sources. The typical following configuration is considered (Figure 7) :
Figure 7 - Typical configuration for direct on-line assessment of flicker emission level of consumer A
9
There must be a known impedance ZAB between the measuring points A (consumer) and B (point of common coupling
with other possible disturbing loads). For instance, this impedance may be the HV-MV transformer feeding the
considered load.
The method consists in measuring the flicker related to the differential voltage fluctuation defined as :
Vdiff = ∆ VA − ∆ V B (5)
Applying the superposition principle, it can be shown that in general situations, where fluctuating loads are
simultaneously present at A and B, the differential voltage is given by
Vdiff = Vdiff ( A ) + Vdiff ( B) = Vdiff ( A ) (6)
In this equation, Vdiff (A) must be understood as the differential voltage fluctuation due to the disturbing load A, while
Vdiff (B) is the differential voltage fluctuation caused by other disturbing loads connected in B.
Equation ( 5 ) results from the fact that when the disturbing load in A is switched off, voltage fluctuations in A are
exclusively caused by the other fluctuating loads acting in B :
∆ V A ( B ) = ∆ V B( B )
(7)
Vdiff ( B) = ∆ VA ( B) − ∆ V B( B) = 0
On the other hand, it can also be shown that Vdiff is related to the impedance between A and B :
Vdiff = Vdiff ( A ) = ZAB ∆ I (8)
( ∆I being the current fluctuation in the impedance ZAB )
It is thus established that the flicker emission of the load Psti (A) is equal to the Pst calculated from the difference of the
fluctuation signals in A and B.
The result, Pst emiss, is related to the impedance ZAB , i.e. to a defined reference short-circuit level Ssc emiss = V2 / ZAB .
5.2 Flicker emission level related to the actual or contractual short-circuit level
As explained above, the original Sakulin’s method produces a flicker emission level measurement related to the chosen
reference impedance. This result has then to be transposed to the actual or contractual HV short-circuit level, as usually
specified in contractual applications [ 5 ].
This can be achieved as explained in the following example, dealing with fluctuating loads connected to a HV grid
through a HV-MV transformer (e.g. big arc furnaces installations). In such cases, the reference impedance may be the
HV-MV transformer feeding the load: ZAB = Xtfo.
When there is no other significant flicker source in the neighbouring network, Sakulin’s method leads to the following
relationship between the measured Pst emiss, Pst(HV) and Pst(MV):
Pst (MV ) = Pst (HV ) + Pst emiss (9)
Moreover, there is usually a good linear correlation between Pst (MV) and Pst emiss :
Pst (MV ) = k Pst emiss ( 10 )
so that we also have, in that particular situation, a good correlation between Pst (HV) and Pst emiss:
Pst (HV ) = ( k − 1 ) Pst emiss ( 11 )
10
In a most general situation where other flicker sources are present in the network, there is no obvious correlation
between the HV flicker and the flicker emission level as given by Sakulin. However, at the MV side of the transformer,
the influence of other flicker sources is negligible with respect to the flicker produced by the load of which we want to
measure the emission and equation ( 9 ) is still valid - at least for values satisfying Pst(MV) > Pst(HV).
The measured MV flicker plotted versus the emission level according to Sakulin looks like in Figure 8.
Figure 8 - Measured MV flicker vs emission level as given by Sakulin’s method
On the other hand, the measured HV flicker plotted versus the emission level as given by Sakulin’s method looks like in
Figure 9. The straight line indicates the lowest measured levels in HV, corresponding to situations where no other
significant disturbing load is connected to the PCC.
Figure 9 - Measured HV flicker vs emission level as given by Sakulin’s method
The slope of this line is equal to k-1 and if the network impedances were perfectly voltage independent, it should
theoretically be equal to the ratio Xsc / Xtfo (where Xsc is the standardised short-circuit impedance of the HV network and
Xtfo, the transformer reactance).
11
In fact, it may be significantly different, due to the voltage dependence of the loads in the network. As explained in [12],
the influence of the various loads (present in the network) results in a virtual internal impedance of the network X*sc ,
different from the standardised short-circuit impedance.
Knowing this, the flicker emission level at HV is deduced from the emission level resulting from Sakulin’s method by:
X*sc
Pst emiss ( HV ) = Pst emiss ( 12 )
X tfo
where X*sc is the empirical value of the virtual network reactance, identified from Figure 9 (X*sc/Xtfo = k-1) or - more
easily - from Figure 8 (where k is given by the slope of the least square straight line). This approach relies upon the
assumption that this impedance is not too much varying (it is in fact depending on the loads) and that it makes sense to
use an average value.
A further step is still necessary in order to relate this result to the reference short-circuit power (if this is different from
the actual network conditions) as usually specified in contracts.
Therefore, it is also suggested in [12] to use, in first approximation, the following formula:
X*sc
P st contract ( HV ) = P st emiss
Ssc
( 13 )
X tfo Ssc contract
In this formula, Ssc represents the standard calculated short-circuit power, according to IEC 909, taking the actual
network configuration and operating conditions into account. Ssc contract is the standard calculated short-circuit power,
also according to IEC 909, for the contractual network configuration, which may be different than the actual one.
Finally, situations may also be found where the correlation between Pst emiss and Pst (MV) is not as good as pictured in
Figure 8, leading to uncertainty in the determination of the coefficient k and, consequently, the identification of X*sc. For
such cases, it is suggested in [12] to use Xsc rather than X*sc, for the evaluation of the flicker emission level using
equations (12) and (13), especially in critical cases where it not obvious whether an emission limit is fulfilled or not.
5.3 Derived method
If the direct on-line emission measurement method mentioned in the preceding paragraphs is not available, a good
indication of the emission level is given by the flicker measured at the MV side of the step-down transformer. As already
stated above, at that voltage level, the flicker contribution of the other fluctuating loads, connected to the HV PCC, is
practically negligible. The measured MV flicker is thus a good indicator of the flicker emission level of the considered
load, provided that it can be transposed to the HV level, where emission limits are usually specified. For this reason,
measurements must still be carried out simultaneously at both sides of the step-down transformer and processed as
follows [13].
In the more general situation where other flicker sources are present in the network, the measured HV flicker plotted
versus the measured MV flicker looks like shown in Figure 10.
In this graphic, the straight line indicating the lowest measured levels in HValso gives the flicker emission level,
corresponding to situations where no other significant disturbing load is connected to the PCC. The slope of this line is
equal to (k-1)/k so that we have :
X*sc
Pst emiss ( HV ) = Pst ( MV ) ( 14)
X*sc + X tfo
where X*sc is the empirical value of the virtual network impedance. In a similar way as explained in the preceding
paragraph, the contractual emission level can be calculated by:
X*sc
Pst contract ( HV ) = Pst ( MV )
Ssc
( 15 )
* +X
Xsc tfo Ssc contract
12
1.5
0.5
0
0 1 2 3 4 5 6 7 8 9 10
Pst MV
Figure 10 - Measured HV vs. MV Flicker (general situation, with other disturbing loads connected)
REFERENCES
[ 1] UIE Working Group Disturbances, “Flicker measurement and evaluation”, UIE 2nd revised edition, 1991.
[ 2] International Electrotechnical Commission, “Flickermeter - Functional and design specifications”, IEC Report
868, 1st edition, 1986.
[ 3] International Electrotechnical Commission, “Amendment 1 to publication 868 (1986) : Flickermeter - Functional
and design specifications”, 1990.
[ 4] T. Gustafsson, “Evaluation of summation coefficient”, Danieli Centro Met, Technical Report, 96-02-06
[ 5] International Electrotechnical Commission, “Assessment of emission limits for fluctuating loads in MV and HV
power systems”, IEC Technical Report 61000-3-7, 1st edition, 1996.
[ 6] A. Robert, M. Couvreur, “Recent Experience of Connection of Big Arc Furnaces with Reference to Flicker
Level”, CIGRE 1994, paper 36 - 305 (CC02 IWD 9516)
[ 7] A. Robert, M. Couvreur, “Arc Furnace Flicker Assessment and Mitigation”, PQA 94, Amsterdam, paper B-1.08
[ 8] M. Sakulin, “Flicker, State of the Art, Measuring Techniques and Evaluation”, Jornada sobre la calidad de la
onda eléctrica, ACIO - AEIC - ASINEL, Barcelona, 13/11/1996
[ 9] M. Sakulin, H. Renner, “Field Experience with the Austrian UIE / IEC Flicker Analysis System”, Proceedings of
the XIIth UIE Congress, Electrotech 92, Montréal Canada, 1992, pp. 842 - 851
[10] J. Bruce Neilson, M. Brent Hughes (Powertech Labs Inc., Vancouver, B.C., Canada), « Enhanced Flickermeter »,
(Power Point) Presentation for IEEE Task Force on Light Flicker, January 1999 (CC02 IWD 9924)
[11] M. Sakulin, T. Key, « UIE/IEC Flicker Standard for Use in North America. Measuring Techniques and Practical
Applications », Proceedings of the PQA ‘97 North America Conference, Columbus OH, USA, March 1997
[12] M. Couvreur, E. De Jaeger, A. Robert, « Voltage Fluctuations and the Concept of Short-Circuit Power », CIGRE
2000, paper 13/14/36-08
[13] M. Couvreur, E. De Jaeger, A. Robert, « Short-circuit level and voltage fluctuations », Proceedings of the PQA
’98 Conference, Cape Town, South Africa, November 1998.
