flicker emission of wind turbines contonuous operation

8/13/2019 Flicker Emission of Wind Turbines Contonuous Operation

1/5

114 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 17, NO. 1, MARCH

Flicker Emission of Wind Turbines DuringContinuous Operation

ke Larsson

Abstract This paper presents an analysis and the modelingof the flicker emission of wind turbines. Measurements comparedwith international standardsarediscussed. Thepaperconcentrateson the theoretical aspects of the flicker algorithm, wind turbinecharacteristics and the generation of flicker during continuous op-eration of wind turbines.

Index Terms Flicker, power fluctuations, wind turbine.

I. INTRODUCTION

AMONG utilities wind power is sometimes consideredto be a potential source for bad power quality. Uneven

power production and weak connections due to long feederlines are some of the factors behind this opinion. Not only theuneven power production but also other factors contribute to thepower quality of wind turbines. One of these factors is flicker.Electrical flicker is a measure of the voltage variation whichmay cause disturbance for the consumer. Flicker emissionsare not only produced during start-up, but also during thecontinuous operation of the wind turbine. The flicker emis-sion produced during normal operation is mainly caused byvariations in the produced power due to wind-speed variations,the wind gradient and the tower shadow effect. In areas wherewind power is an emerging technology, some actions have been

taken. One example is Germany, where power quality standardsfor grid connected wind turbines have been in use for someyears. The German standard for grid connected wind turbinesincludes rules on power fluctuations and flicker.

The International Electrotechnical Commission (IEC) iscurrently working on power quality requirements for gridconnected wind turbines. The work has resulted in a committeedraft designated IEC 61 400-21 [1]. This draft includes quan-tities to be specified for characterising the power quality of wind turbines and measurement procedures for quantifying thecharacteristics. Also wind turbine requirements with respectto power quality are given and methods for assessing windturbine impact on power quality are suggested. The proposed

standard pays particular attention to flicker. In addition togeneric standards, flicker may become a serious limitation towind power, at least in case of weak grids.

Flicker from wind turbines has become an important issue. Astudy of different types of wind turbines concluded that flickeremissionincertaincasesexceedslimitswhichareexpectedtobenormative in the future [2]. In order to predict flicker produced

Manuscript received August 25, 2000. This work was supported in part byELFORSK AB.

. Larsson is withthe DepartmentofElectrical Power Engineering, ChalmersUniversity of Technology, Gteborg, Sweden.

Publisher Item Identifier S 0885-8969(02)01509-7.

Fig. 1. Block diagram of the flicker meter.

by a wind turbine at the design stage, software tools are bedeveloped [3], [4]. For such software the physical dynamics o

the turbine, the wind turbulence and the electrical dynamicsthe generator and the network itself need to be modeled. Tpurpose of this paper is to provide a concise review of the anysis andmodelingof theflickeremissionofwind turbines,alowith measurements and a comparison with international stadards. This paper concentrates on the theoretical aspects of flickeralgorithm, windturbinecharacteristicsandflickerduricontinuous operation.

Section II in this paper provides a brief review of the flickmeter according to IEC Standard and the flicker algorithm. Stion IIIpresents thecharacteristicsof the turbine. Finally, flickduring continuous operation is described in Section IV.

II. DESCRIPTION OF THE FLICKER METERThe level of flicker is quantified by the short-term flick

severity value . The calculation of flicker severity takes inaccount the response of the light emission from incandesclights to voltage variations and also the response of the humeye and brain in perceiving variations in illumination. The fution and design of the flicker meter are specified in the StandIEC Publication 868 [5]. The block diagram shown in Fig. 1describes the flicker meter architecture. Although the block agram consists of five blocks, the flicker meter can be dividinto two main parts, each performing oneof thefollowing task1) a simulation of the response of the lampeyebrain chain 2) an on-line statistical analysis of the flicker signal and prestation of the results. Blocks 2, 3, and 4 perform the first tawhile the second task is accomplished by block 5.

Block 1 performs the first step in the flicker meter. Thblock scales the input voltage to a reference level. Block squares the input voltage in order to simulate the behavior olamp. Block 3 is composed of a cascade of two filters whthe first filter eliminates the dc voltage and the double mafrequency. The second filter simulates the frequency responof voltage fluctuations of a light bulb combined with the humvisual system. Block 4 is composed of a squaring multipl

08858969/02$17.00 2002 IEEE
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


2/5

LARSSON: FLICKER EMISSION OF WIND TURBINES DURING CONTINUOUS OPERATION

and a first order low-pass filter with a time constant of 300ms. Together they form a nonlinear function. The output fromBlock4 represents theinstantaneousflicker level. Block5 isthe final flicker meter block, which makes an on-line statisticalanalysis of the instantaneous flicker level. The statistical anal-ysis can be divided into two parts. First, the cumulative proba-bility function of the instantaneous flicker level is established,and second, the short-term flicker severity value is calcu-lated using a multipoint method.

The cumulative probability function of the instantaneousflicker level from Block 4 gives percentages of observationtime for which flicker levels have been exceeded. The cu-mulative probability, , that the instantaneous flicker levelexceeds is defined as

(1)

where is the duration of timewhich the signal remains aboveand is thetotalobservation time.This method hasbeen termedtime at level classification. For practical purposes, only a lim-

ited number of curve points can be computed. The IEC 868states that the analysis is to be performed with at least 6 bits res-olution using at least 64 classes. The minimum sampling rate is50 samples per second. After this classification, the short-termflicker severity value iscalculated usinga multipoint methodexpressed by the equation

(2)

where the percentiles , , , and , are theflicker levels exceeded for 0,1%, 1%, 3%, 10%, and 50% of thetime during the observed period, i.e., the instantaneous flicker

levels exceeded for of the observed period. The suffix inthe equation indicates that the smoothed value should be used.These smoothed values are obtained by

(3)

(4)

(5)

(6)

The 300 ms time constant in Block 4 ensures that

cannot changeabruptly andno smoothingis neededfor this per-centile.According to IEC standards the short-term flicker severity

value is based on a 10-min period. The short-term flickerseverity evaluation is suitable for assessing disturbances causedby sources with a short duty-cycle. When flicker sources withlong and variable duty-cycles are under consideration, it is nec-essary to provide a criterion for the long-term flicker severity.For this purpose, the long-term flicker severity is derivedfrom the short-term severity values using the formula

(7)

where are consecutive readings of the short-term severit. The long-term flicker severity value is calculated f

, i.e., a 2-h period.The method for measuring instantaneous flicker and the a

gorithm required for calculating are rather complicated. general analytical method of calculating is not possible find. However, there are methods for determining the total suof flicker from a set of known flicker sources. In the StandaIEC Publication 61000-3-7 the following general relation short-term flicker severity caused by various loads is stated 6]

(8)

where is the individual level of flicker severity values frosource and is a coefficient which depends upon the charateristics of the main source of fluctuation. If the fluctuationcoincident stochastic noise should be used.

III. TURBINE CHARACTERISTICS

Wind turbines have some kind of control for regulating tpower from rated wind-speed up to the shutdown wind-speToday, two different types of regulation principles are mainused; stall-regulation and pitch-control. Regardless of the reglationprincipleused, thepowerwill fluctuatedueto wind-spevariations, the wind gradient and the tower shadow effect. If turbine has three blades, a power drop will appear three timper revolution. This frequency is normally referred to as 3- two-blade and a three-blade wind turbine have been studied[7].Bothturbinesarepitch-controlledandoperateatfixedspeedFor both wind turbines studied, the greatest power pulsatioccurs at rated power at the highest wind-speeds. According

[8], wind turbines equipped with inductiongenerators operatiatfixedspeedgeneratepowerpulsationsupto 20%of theaverapower.

Pitch-controlled turbines will also have power fluctuatiocaused by the limited bandwidth of the pitch mechanism in adition to fluctuations caused by the tower shadow. The powof pitch-controlled wind turbines is controlled by the anglethe blades. This means that the steady-state value of the powoutput should be kept close to the rated power of the generaat high wind-speeds, normally between 1214 m/s to the cut-wind-speed at 2025 m/s. This is achieved by means of pitchthe blades. The steady-state value of the power (solid line) a function of wind-speed is shown in Fig. 2. The steady-st

value of the power is, as illustrated in Fig. 2, kept equal rated power at wind-speeds above 12 m/s. However, pitchithe blades implies that the power curve is transferred. This islustrated in Fig. 2 where the dotted line shows the instantaneopower curve when the blades are pitched for rated power awind-speed of 15 m/s.

Unfortunately, the wind-speed is not constant but varies the time. Hence, instantaneous power will fluctuate around rated mean value of the power due to gusts and the speed the pitch mechanism (i.e., limited bandwidth). As can be sein Fig. 2, variations in wind-speed of 1 m/s gives power flutuations having a magnitude of 20 . Fig. 3 shows the powfrom a stall-regulated turbine under the same conditions as t
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


3/5


Fig. 2. Power as function of wind-speed from a pitch-controlled wind turbine.

Fig. 3. Power as function of wind-speed from a stall-regulated wind turbine.

pitch-controlled turbine in Fig. 2. Variations in the wind-speedof the stall-regulated turbine also cause power fluctuations butthey are small in comparison with a pitch-controlled turbine.

Fig. 4 shows the measured power of a pitch-controlled fixed-speed wind turbine with a rated power of 225 kW under highwind-speed conditions. In the figure, variations in the powerproduced by the wind turbine is shown. As previously men-tioned, fixed-speedwind turbines produce power pulsations dueto wind speed gradients and the tower shadow.

Thefrequencyofthepowerpulsationsisequaltothenumberof blades multiplied by the rotational speed of the turbine, e.g., the3- frequency. The figure also indicates the power fluctuationscausedby wind gusts and the speed of the pitch mechanism.

IV. FLICKER DURING CONTINUOUS OPERATIONThecommitteedraftof IEC61400-21suggests that theflicker

from a single source should not be determined from voltagemeasurements, in order to avoid disturbance due to the back-groundflicker on thegrid.Themethodproposedforovercomingthis problem is based on measurements of current and voltage.The short-term flicker from the wind turbine should be calcu-lated using a reference grid where the measured current is theonly load on the grid. This procedure is performed in two steps.First, the measured time-series of the current are used to calcu-late the time-series of voltage variations on the fictitious grid bythe following equation:

(9)

Fig. 4. Measured power during normal operation of a pitch-controlfixed-speed wind turbine (solid line). The steady-state power is plotted (doline).

where is an ideal voltage source, and is the resistance and the inductance of the fictitious grid, respective

is the measured instantaneous current. The ideal voltasource shall be given by

(10)

where is the rms value of thenominal voltage of thegrid ais the electric angle of the fundamental of the measur

voltage. Second, the voltage variation is used as an inpto the flicker algorithm in compliance with the Standard IEPublication 868.

Fig. 5 shows the short-term flicker at different cut-off frquencies of a fixed-speed wind turbine and a variable-spewind turbine. Different cut-off frequencies have been achiev

by means of filtering the measured timeseries in an eighth ordButterworth filter at different cut-off frequencies.For both wturbines a short circuit ratio (SCR) of 20 and a phase angle45 has been used. The SCR is defined as the ratio betweenshort circuit power of the grid and the rated power of the wturbine at the point of common connection (PCC). The phangle is the tangent of ratio between the network reactance athe resistance. The short-term flicker has been calculated usa PC-program developed by Ris National Laboratory [9]. Thisprogram uses the Standard IEC Publication 868 with amenment 1 to calculate [10]. The input to the program is a timeseriesofactive andreactive power, shortcircuit powerandphaangle of the grid.

As shown in Fig. 5, the calculated short-term flicker considerably higher for the fixed-speed wind turbine as for variable-speed wind turbine. At high cut-off frequencies tshort-term flicker is 0.163 for the fixed-speed wind turbine a0.051 for the variable-speed wind turbine. It is worth noting tapproximately 30% of the total flicker emission produced the fixed-speed wind turbine depends on power variations wa frequency above the 3- frequency (2.15 Hz) of the turbiThis flicker contribution over the 3- frequency emanates frthe mechanical properties of the wind turbine, most likely dynamics of the induction generator. This is due to the flickcurve and the dynamics of the induction generator. The flickcurve is most sensitive at 8.8 Hz and the dynamics of t
http://-/?-http://-/?-http://-/?-http://-/?-


4/5

LARSSON: FLICKER EMISSION OF WIND TURBINES DURING CONTINUOUS OPERATION

Fig. 5. Short-term flicker from a fixed-speedwind turbine anda variable-speedwind turbine calculated at different cut-off frequencies. The cut-off frequencyhas a logarithmic scale.

Fig. 6. Short term flicker from a fixed-speed anda variable-speedwindturbineat different power. The flicker emission increases at higher wind-speeds dueto higher turbulence in the wind.

induction generator have a resonance frequency of approxi-mately 10 Hz. The flicker contribution from the variable-speedwind turbine is quite different. The variable-speed system hasthe ability to reduce 3- pulsations from the turbine but themechanical properties of the wind turbine seem to contributeto a higher flicker level at frequencies of approximately 10 Hz.As can be seen in Fig. 5, the cut-off frequency of the measuredtime series from a wind turbine must exceed 50 Hz in order toachieve a good result.

Fig. 6 shows the short-term flicker emission from a fixed-speed and a variable-speed wind turbine at different power. Theflicker is calculated using an SCR of 20 and a grid angle of 45 ,i.e., thesame conditions as thewind turbines presentedinFig. 5.

As can be seen in Fig. 6, the flicker emissions increase at higherwind-speeds due to higher turbulence in the wind.In the case of the fixed-speed wind turbine, the flicker in-

creases around three times from lower to higher wind-speeds.Even in the case of the variable-speed wind turbine, the flickerincreases with an increase in wind-speed. Except for the flickerlevel, there is one fundamental difference between the fixed-speed and the variable-speed wind turbines. The flicker levelincreases at increasing wind-speed for the fixed-speed turbinewhile the flicker level decreases at rated wind-speed for thevariable-speed wind turbine. As the wind turbine reaches ratedpower, the variable-speed system will smooth out the powerfluctuations and, thereby, limit the flicker. However, the flicker

level should be based on an annual wind speed distributioHence, flicker must be measured at all wind-speeds.

A. Flicker Coefficient

Accordingto thecommittee draft of IEC61400-21 theflickcoefficient from wind turbines is to be determined by measuments and simulations. The three instantaneous phase currenand the three instantaneous phase-to-neutral voltages are to measured at the wind turbine terminals. The cut-off frequenof the voltage andcurrent measurement must be at least 400 HMeasurements are to be taken so that at least thirty ten-minutime-series of instantaneous voltage and current measuremearecollectedforeach 1 m/s wind-speed binbetween cut-inwinspeed to a wind speed of 2 m/s above reference wind speeThe voltage time-series for each set of 10-min measurvoltage and current time-series are then to be calculated usi(9) and (10). The voltagetime-series isto beused asinputotheflicker algorithm togive one flicker value onthefictitious grid for each 10-min time-series. The flicker coefficiis to be determined for each of the calculated flicker values applying

(11)

where is the flicker coefficient and is the rated apparent power of the wind turbine. is the flicker level caculated at the short-circuit power level of a fictitious referengrid defined as

(12)

The phase angle, of the fictitious grid is defined as(13)

The flicker emission produced by a wind turbine connectto a grid with an arbitrary short-circuit power may, then, recalculated by

(14)

Fig. 7 shows the short-term flicker from a wind turbine cculated at different SCRs. The short-term flicker in Fig

has been calculated in two different ways: 1) by measuremeand (9) on a reference grid using different SCRs (this methodcalled in Fig. 7), 2) by using (11) and (14) according to tcommittee draft of IEC 61400-21 (called IEC in Fig. 7). As cbe seen in Fig. 7, the two different methods for calculatingvalues agree well with each other.

B. Summation of Flicker

Wind turbines are often placed close to each other, e.g., wind parks. Wind turbines located in a wind park will expeence approximately thesamemean valueof thewind-speed.Tmean value of the power from the wind turbines will, therefobe correlated.


5/5


Fig. 7. Short-term flicker emission from a wind turbine calculated at differentSCRs. The dots are calculated directly by using measurements and the IECdots by using equations according to the committee draft of IEC 61400-21.

Fig. 8. Calculatedshort-termflickerfromthirty differentmeasuredtime seriesof two fixed-speed wind turbines (cross) and ten different measured time seriesof two variable-speed wind turbines (circle).

The flicker produced by each wind turbine during contin-

uous operation emanates, as mentioned earlier, mainly from thetower shadow effect and wind turbulence. Since the wind tur-bines are not located on exactly the same place, but close toeach other, they will not experience the same wind-speeds onthe rotor disks. The variations in the wind-speeds and the po-sition of the rotors are, therefore, not correlated. The powerfluctuations and the flicker of two or more wind turbines areexpected to be uncorrelated stochastic noise. Hence, accordingto the Standard IEC Publication 61 000-3-7 (8) should be usedto sum the total flicker during normal operation. Fig. 8 showsthe calculated short-term flicker from thirty different measuredtime series of two fixed-speed wind turbines and ten differentmeasured time series of two variable-speed wind turbines. The

short-term flicker is calculated in two different ways. First, di-rectly by using the sum of the measurements from each windturbine, called in Fig. 8. Second, by using the short-termflicker from each wind turbine andthe summation law accordingto (8) with , called in Fig. 8.

As can be seen in Fig. 8, the short-term flicker varies dueto variations in the wind, i.e., turbulence. The mismatch of thetwodifferentways of calculating theflicker (i.e., deviation fromthe dotted line), however, is small. The correlation coefficient,

.According to the committee draft of IEC 61400-21 the fol-

lowing equation is to be applied in determining the flicker con-tribution from a number of wind turbines connected to a

common point

(15)

where is the flicker emission from a single wind turbin

V. CONCLUSION

Flicker emissions are produced during the continuous opation of wind turbines. The flicker is caused by power flucttions which mainly emanate from variations in the wind-spethetowershadoweffectandthemechanicalpropertiesofthewiturbine. Pitch-controlled turbines have in addition power flucationscausedby thelimited bandwidthof thepitch mechanism

Themethodformeasuring instantaneous flicker andthealgrithm required for calculating is rather complicated. A geeral analytical method of calculation for determining the shterm flicker from a set of arbitrarily chosen voltage distubances is not possible.

The committee draft of IEC 61400-21 suggests that thflicker emission from a single wind turbine should be dtermined by measurements. The measurements should nbe based on voltage measurements only, in order to avothe measurements from being disturbed by the backgrouflicker on the grid. The method proposed for overcoming tproblem is based on measurements of current and voltage. Tshort-term flicker emission from the wind turbine should calculated on a reference grid using the measured currentthe only load on the grid.

Finally, a flicker coefficient is introduced. The use of tflicker coefficient makes it possible to calculate the flicker pduced by the wind turbine connected to a grid with an arbitrshort-circuit power.

REFERENCES[1] Power Quality Requirements for Grid Connected Wind Turbines , Com-

mittee Draft, IEC Std., Publ. 61 400-21, 1998.[2] P. Srensen et al. , Flicker emission levels from wind turbines, Wind

Eng. , vol. 20, no. 1, 1996.[3] E. Bossanyi, Z. Saad-Saoud, and N. Jenkins, Prediction of flicker

duced by wind turbines, Wind Energy , vol. 1, no. 1, pp. 3551, 1998.[4] A. Feijo and J. Cidrs, Analysis of mechanical power fluctuation

asynchronous WECs, IEEE Trans. Energy Conv. , vol. 14, pp. 284291,Sept. 1999.

[5] Flicker MeterFunctional and Design Specifications , IEC Std., Publ.868, 1990.

[6] Electromagnetic Compatibility, Limitation of Voltage Fluctuations and Flicker Equipment Connected to Medium and High Voltage Power Supply Systems , IEC Std., Publ. 61000-3-7, 1995.

[7] P. Gardner, Flicker fromwind farms, inProc. BWEA/SERCRAL Work-shop Wind Energy Penetration Weak Electr. Netw. , Rutherford, U.K,June 1993, pp. 2737.

[8] G. Gerdes and F. Santjer, Power quality of wind turbines and theirteractionwith thegrid, inProc. Euro. Wind Energy Conf. , Thessaloniki,Greece, Oct. 1014, 1994, pp. 11121115.

[9] P. Srensen, Methods for calculation or the flicker contribution fwind turbines, Ris Nat. Lab., Roskilde, Denmark, Ris-I-939, 199

[10] Flicker MeterFunctional and Design Specifications , IEC Std.,Amendment 1 to Publ. 868, 1990.

ke Larsson received the M.Sc. degree in electrical engineering from tChalmers University of Technology, Gteborg, Sweden, in 1994. Hecurrently pursuing the Ph.D. degree at Chalmers University of Technologthe area of electrical machines and power electronic. His research interinclude power quality from wind turbines.

flicker emission of wind turbines contonuous operation

Documents