Assessment of Wind Power Quality: Implementation of IEC Procedures

Transcription

1 Assessmet of Wid Power Quality: Implemetatio of IEC614-1 Procedures A. Morales 1, X. Robe ad J.C. Mau 1 1 Departmet of Electrical Egieerig. CP 165/5 Uiversité Libre de Bruxelles Campus of olbosch 15, Brussels (Belgium) Phoe/Fax umber: /6 5, amorales@ulb.ac.be, jcmau@ulb.ac.be L Air Liquide Luxembourg,.A. L-481 Rodage (Luxembourg) Phoe/Fax umber: /6 6 18, Xavier.Robe@airliquide.com Abstract. The purpose of this paper is to preset the implemetatio of power quality procedures meetig the IEC stadard. The system s architecture is detailed i the paper. A experimetal laboratory setup which represets a fixed speed geeratio system is described. Measuremets are carried out to test the proposed algorithm accordig to the measuremet procedure for cotiuous operatio of wid turbies. The assessmet procedure is applied to ivestigate flicker emissio. Key words Power quality, flicker, wid power. 1. Itroductio The IEC stadard provides recommedatios for preparig the measuremets ad assessmet of power quality characteristics of wid turbies [1]. Oe importat questio is raised whe tryig to assess the power quality: who is resposible of flicker emissios? Geerally, the ifluece of a wid turbie o the voltage quality of the grid depeds ot oly o the turbie itself but also o the grid where it is coected. The Europea Wid Turbie Testig Procedure report ad other publicatios have illustrated that the grid impedace agle ad short circuit power have a importat ifluece o the procedures proposed by the stadard [, ]. We are the facig a very complex problem: how to measure the harmoic source impedace of a power etwork ad to assess the emissio level of a particular distorted geeratio [4].. IEC-614-1: The measuremet procedure The measuremet procedures described i the IEC stadard are valid to test the power quality characteristic parameters for the full operatioal rage of a wid turbie, coected to a MV or HV etwork with fixed frequecy withi ±1Hz, sufficiet active ad reactive power regulatio capabilities ad sufficiet load to absorb the wid power productio. The measuremet procedures are desiged to be as osite-specific as possible, so that power quality characteristics measured at for example a test site ca be cosidered valid also at other sites. The stadard specifies a method that uses curret ad voltage-time series, i m ( ad u m (, measured at the wid turbie termials distributed over the wid speed iterval from cut-i to 15 m/s, to simulate the voltage fluctuatios o a fictitious grid with o source of voltage fluctuatios other tha the wid turbie. This fictitious grid is preseted i Fig. 1 by a ideal phase to eutral rated voltage source u ( with a appropriate apparet short-circuit power k,fic ad four grid impedace agles k (,5,7 ad 85 ). Fig. 1: Fictitious grid as proposed by IEC The wid turbie is represeted by a curret geerator i m (, which is the measured istataeous value of the phase curret. Thus, the fluctuatig voltage that will be iput to the voltage flicker algorithm i compliace with IEC to geerate the flicker emissio value P st,fic is give as: RE&PQJ, Vol. 1, No., March 5

2 u fic d ( = u( + R fic im( + L fic im( (1) dt The fudametal voltage has to be determied from the measured voltage at the wid turbie termials. Oce this measuremet ad simulatio procedure is doe, flicker coefficiets c( k,v a ) are reported for cotiuous operatio after ormalizatio ad weightig processes depedig o the etwork impedace phase agle k ad the aual average wid speed v a. Flicker step factors k f ( k ) ad voltage chage factors k U ( k ) are geerated together i the case of switchig operatios. These factors are ormally measured ad provided by maufacturers, who tred to overestimate them. Durig the assessmet procedure, i order to evaluate the power quality expected from a wid turbie whe deployed at a specific site, the followig equatio applies at cotiuous operatio: where U is the rms value of the omial voltage of the grid, ad the electrical agle m ( of the measured voltage varies with frequecy f( withi a measuremet widow. At the begiig of the timeseries, the iitial agle of the fudametal voltage is. This ideal voltage agle calculatio esures that the phase agle betwee u fic ( ad i m ( is correct, uder the assumptio that the voltage drop at the fictitious source impedace is much smaller tha the ideal voltage amplitude. As illustrated i Figure, the error o the estimated phase agle icreases with the source impedace value ad agle. A appropriate apparet short circuit power of the fictitious grid k,fic should be at least 5 times, as recommeded by the stadard. P = P = c ) st lt (ψ k, va () k where is the rated apparet power of the wid turbie ad k is the apparet short-circuit power at the PCC. Other equatios are applied to assess the total cotributio from several wid turbies. As reported i [], flicker results are foud to be strogly depedet o the grid impedace value, ad the flicker summatio from idividual wid turbies gives too low total flicker values i case of limited supply stregth.. Proposed implemetatio of the stadard IEC The proposed system s architecture for implemetatio of power quality stadards for wid turbies is detailed i this sectio. Oly the measuremet procedure is specified i the case of cotiuous operatio. The implemetatio for switchig operatio is straightforward. Devices proposig basic power quality moitorig fuctios are already o the market meetig the IEC 61 stadards, but the IEC stadard icludes a simulatio stage of voltage fluctuatios o a fictitious grid. This simulatio is ot very complicated except for the geeratio of the ideal voltage source u ( because it should fulfill the followig properties [1]: 1) The ideal voltage should ot display ay fluctuatios (zero flicker level). ) It should have the same electrical agle m ( as the fudametal of the measured voltage. The ideal voltage is defied as: t u ( = U si( α ( ) = U si(π ( ) + α ) f t dt () m Fig. : Phase displacemet of u fic ( versus apparet short circuit power of the fictitious grid. The obtaied voltage time-series u fic ( calculated usig Equatio (1) is the iput to the flicker algorithm i compliace with IEC to give a flicker emissio value P st,fic o the fictitious grid. The agle of the ideal voltage is very importat durig the simulatio phase sice voltage fluctuatios i u fic ( are very sesitive to the agle value. If a error is committed o the fudametal voltage phase, the ideal voltage will be give by: t = U si(π f ( dt + α ± ε ( α m u ) (4) where m represets the error or deviatio from the measured voltage phase. Ay deviatio from the real electrical agle m ( of the measured voltage has a impact o the amplitude of the calculated voltage u fic (, what could brig the obtaied simulated values outside the valid rage of the flicker algorithm. This is illustrated with a umerical example i Figure, where the amplitude variatio of the fictitious fudametal voltage u fic ( expressed i per cet is plotted versus a variatio of m RE&PQJ, Vol. 1, No., March 5

3 The acquisitio fuctio is built aroud oe loop coutig the umber of data widows that have bee acquired. Oe data widow cotais C N poits where C is the umber of cycles per widow ad N is the desired umber of poits per cycle. (elected values are C = cycles/widow ad N = 18 poits/cycle.) The while-loop couts the umber of samples i a output buffer. Whe this buffer is full, a ew data widow is acquired. Fig. : Variatio o u fic ( relative amplitude versus deviatios from the measured grid agle ( m ). Whe implemetig the full stadard, special attetio has to be paid to the calculatio of the ideal fudametal voltage u (. Nowadays, the way to achieve the extractio of the fudametal frequecy is by meas of phase lock loop (PLL) equipmet. The samplig frequecy is calculated with a delay of 4 ms ( cycles). The phase shift betwee widows due to this FFT error may be corrected ad compesated. But the system frequecy variatios remai a importat source of error whe usig preset techology, ad this is traslated i erroeous phasor calculatio. Figure 4 illustrates the complex system proposed i this paper. It uses a costat samplig rate ad thaks to a software resamplig algorithm, measuremets are rebuilt at a costat umber of poits per period. This algorithm is a combiatio of a iterpolatio (by a oversamplig factor L) ad a liear approximatio. The mai beefit is that the filter does ot have to be computed for each widow. It has bee show that its performaces are good eough to meet the power quality stadards, eve for harmoics up to the 4 th [5]. The mai idea behid this desig is, startig from a costat samplig rate at 1 or khz, to subdivide data ito multiple streams at differet samplig rates allowig flicker calculatio ad also performig accurate harmoic moitorig fuctios. Accurate frequecy trackig is oe of the ew improvemets to the process. I this way, two phasors are obtaied per widow of (4 ms ± legth, with t the variable widow legth depedig o the detected frequecy. This allows a more accurate phasor calculatio. 4. Algorithm descriptio A. Acquisitio fuctio The umber of cycles per widow, parameter C, has bee selected to obtai the highest accuracy whe computig u ( from the resampled measured voltage V ew (. The voltage frequecy is cosidered statioary withi a widow, ad a widow may be composed of te, five or two cycles. Therefore, a shorter data widow will provide the highest accuracy i frequecy estimatio ad phasor calculatio. Normally, stadard power quality moitorig fuctios perform phasor calculatios o a 1-cycle width basis. B. Iterpolatio ad liear approximatio The first elemet of this algorithm is the low-pass FIR filter. It is a -D array of coefficiets. Each colum of the filter allows computig oe sample at L f s (f s iitial samplig frequecy) avoidig multiplicatios by zero. The filter is built computig a Blackma widow with the followig parameters: - L is the oversamplig factor. - M is the dowsamplig or decimatio factor (set to 1 whe usig the liear approximatio method). - taps is the umber of rows i the matrix. It correspods to the umber of samples at f s that will be used to compute oe sample at L f s. - cutoff is the cut-off frequecy i per-uit (1 represets f s /). Oce the filter has bee built, the ext step is to compute the samples at L f s that surroud the desired oe. The two sample idexes are a ad a+1. The sample, at f sd (desired ew frequecy, equal to N f V where f V is the voltage frequecy i the preset widow) is the give by the liear iterpolatio betwee a ad a+1. Whe C N poits are i the output buffer, a -cycle widow has bee acquired, every couter should be reset ad a ew resamplig frequecy should be computed, esurig the cotiuity betwee every - cycle widow. First, FFT, rms, or other computatios o the samples are made RE&PQJ, Vol. 1, No., March 5

4 ACQUIITION U m (,I m ( at f s =1 khz 1 mi legth Buffer # Widows C= cycles/widow New acquisitio Widow Couter> Ed While Liear approximatio at variable frequecy f' sd =N f computed =18 (5± f)= =6.4±18 f khz V ew,i ew at N poits/cycle Upsamplig Widow Couter< Ed Liear approximatio at costat frequecy f sd =N f=18 5=6.4kHz i m ( at costat rate Accurate phasor calculatio: fft(v ew )/cycle, fft(i ew )/cycle V ew_rms /cycle,i ew_rms /cycle PhasorV m /cycle,phasori m /cycle α m =mea(agle(phasorv m (1:C))) α m ideal voltage creatio at a costat frequecy samplig u ( ) = si(π f + ) VWidow αm for = 1: C N f V Widow u fic ( computatio ad dowsamplig Compute frequecy of the voltage f V Widow Compute harmoics IEC <49th Flickermeter IEC P st,fic Fig. 4: Flowchart of the proposed algorithm. Normalizatio ad weightig Report c(ψ k,va) for cotiuous operatio C. Frequecy computatio The estimated frequecy is calculated i the data widow usig the DFT method. The first step is the extractio of the fudametal phasors by the DFT ad phase correctio (oe for each cycle of the sigal). The algorithm performs a liear regressio, usig the least square method, betwee these phases. The slope of the lie is the frequecy i the data widow. A additioal feature of this algorithm is the rejectio of the out-of-rage measuremets: if a phase is too far from the regressio lie, the program deletes it from the array of phases ad recomputes a ew regressio lie util all the remaiig phases match it. If there are ot eough remaiig poits, o frequecy is computed. Harmoics are computed to the 49 th compoet. The rms value of the h th harmoic compoet, deoted by I h(rm), is obtaied i per cet o the rated curret basis. Total harmoic distortio is give by: THD(%) = D. Fast fourier trasform u ( ) = si(π fv Widow + α m ) = 1,..., CN (6) The FFT is computed o a fixed umber of poits per cycle after performig the sigal resamplig. The Whe usig a variable samplig rate, the measured kowledge of fudametal values ad phase curret is also resampled ad special care is take to displacemet betwee measured voltage ad curret geerate a variable time-step vector to perform the allows the computatio of active ad reactive power, ad cos h= I I h( RM) 1( RM) 1 (5) where I 1(RM) is the rms value of the fudametal compoet. E. imulatio of voltage fluctuatios u fic ( The ideal voltage u ( is obtaied every two cycles at a variable or fixed samplig rate, accordig to: 47 RE&PQJ, Vol. 1, No., March 5

5 derivative of the curret i equatio 1. The fiite differece approximatio is: di( dt i = +1 i t (7) where t varies every two cycles due to the variable frequecy samplig. The fictitious voltage is the obtaied accordig to equatio 1 at a variable samplig rate. Thus, iterpolatio ad dowsamplig are performed for u fic ( to be iput to the voltage flicker algorithm to geerate the flicker emissio value. Oversamplig ratios should be differet every -cycle widow to get a fial fixed samplig rate. The system architecture becomes more complicated. Therefore, we propose to geerate u ( at a fixed samplig rate (6.4kHz is the selected rate i our implemetatio) ad apply a classical dowsamplig to get a ew data stream suitable for the flicker algorithm ( 4 Hz i our implemetatio). 5. Laboratory test To test our implemetatio of power quality procedures meetig the IEC stadard, two laboratory setups have bee used. Results are show for the first test. It cosists of a 11 V wouded-rotor iductio motor coupled to a sychroous geerator via a elastic torque trasducer as show i Figure 5. The sychroous geerator supplies the ecessary load to simulate the power curve of a wid turbie. Fig. 5: chematic of laboratory setup. haft asymmetries esure the presece of modulated currets at termials as it is show i Figure 6. The geeratio of the ideal voltage u ( is correctly performed ad it fulfills the desired properties: same phase as the measured voltage ad o perturbatios. A detail is plotted i Figure 7. Fig. 7: igals superpositio, u ( ad V ew ( After ormalizatio ad weightig processes, ad depedig o the etwork impedace phase agle k ad the aual average wid speed v a, the 99% percetile of the values i the wid speed iterval from cut-i to 15m/s are reported as a output of the algorithm. Flicker coefficiets c( k,v a ) are show i Table 1 for cotiuous operatio of the asychroous motor. Table 1: Flicker coefficiets. v a 6 m/s 7.5 m/s 8.5 m/s 1 m/s k The assessmet that follows is a fairly simple procedure ad it specifies how the reported flicker coefficiets ca be used to estimate the flicker emissio from a sigle wid turbie or a group of wid turbies operatig cotiuously o ay specific site. Equatio () applies. P st values are calculated for the short circuit power of the istallatio. hort circuit impedace value ad agle are obtaied from the power trasformer feedig the asychroous motor: TFO MA = 5kVA = 11.4 kva pu TFO pu MA pu CC = 1pu =.pu = 1.5 = pu (8) Fig. 6: I( i per-uit at a wid speed of 5.5m/s. The short circuit ratio is give by : = (9) RE&PQJ, Vol. 1, No., March 5 CCréseau MA

6 The obtaied results are show i figures 8 ad 9, where the measured P st cotributio is plotted versus wid speed set poits (Figure 8) ad versus active power (Figure 9). The harmoic cotet of the curret is also computed ad preseted i Figure 1, where the mea values i per cet at rated curret is plotted. 6. Future work Fig. 8: P st level o the 11V grid from data acquisitio system vs. wid speed for,5,7 ad 85 grid impedace agle. The secod laboratory setup aims at evaluatig the flicker emissio whe several wid turbies are coected at the PCC. Three motor baks of similar characteristics as the oe utilized durig the first laboratory test are available. We wish to evaluate the ifluece of simulated wid speed, ad umber of motors operatig i parallel o the flicker emissio level. Previous work by the authors comprised the developmet of a prototype for measurig the harmoic source impedace see from the PCC of a Thevei equivalet [6]. The etwork equivalet is represeted i Figure 11 as see from the couplig poit of a wid farm. The source impedace value Z sh gives us a measure of the true harmoic emissio level, sice this impedace represets a accurate image of the etwork characteristics. The product Z sh by geerated wid curret I h costitutes the real emissio from the wid power productio. The resposibility of the etwork is estimated from the ope-circuit voltage E sh. Fig. 9 P st level o the 11V grid from data acquisitio system vs. active power for,5,7 ad 85 grid impedace agle. Fig. 11: Network Thevei equivalet. It is ot the aim of this work to discuss the actual methodology i use for coefficiet estimatio durig the measuremet procedure proposed by the stadard, but to propose a improvemet regardig the secod phase of power quality assessmet oce the wid turbie is coected at the PCC by the realizatio of a o-lie source impedace measuremet. Fig. 1: Mea values of harmoic currets i percet of rated curret. Nowadays, grid agle depedability is cosidered by selectig the right coefficiet from maufacturer s data. Geerally, the short-circuit power is calculated. The ature ad the accuracy of data to be provided for modellig the etwork are ot always evaluated i the same way. Besides, the source impedace is a timedepedet parameter, varyig with a chage of etwork topology, as for example the coectio or discoectio of a secod parallel trasmissio lie or a secod trasformer at the bus bar. The developmet of dispersed geeratio will lead to importat shortcircuit level variatios as well RE&PQJ, Vol. 1, No., March 5

7 We propose to cosider apparet short-circuit power variatios of the real grid, by evaluatig both the fudametal ad harmoic source impedace value ad its agle. This possibility gives rise to ew challeges o the voltage fluctuatio assessmet, such as more accurate grid-depedet o-lie estimatio of flicker emissio levels. The ifluece of other fluctuatig geerators or loads that may cause sigificat voltage variatios at the wid turbie termials could the be idetified ad resposibilities assiged. The fial step i this project is to compare both methods of evaluatig the flicker cotributio. The assessmet of the grid impedace will allow ot oly assigig resposibilities but will also facilitate the desig of equipmet, etwork compoets, tuig of cotrol systems ad protectio cofiguratio. The developmet of dispersed geeratio will be positively iflueced by the capacity of measurig the etwork stregth. 7. Coclusios IEC defies a method based o voltage ad curret measuremet to evaluate power quality characteristics, idepedet of the grid where the wid turbie is coected. The proposed algorithm has bee implemeted ad tested, ad it performs well. A high accuracy is obtaied durig the simulatio procedure thaks to a software resamplig algorithm i spite of the costat samplig rate origially set durig data acquisitio. The measured power quality characteristics ca thus be applied to calculate the ifluece o the voltage quality o aother site give by its short circuit power ad its impedace agle. Refereces [1] IEC-614-1, Wid Turbies, Part 1: Measuremet ad assessmet of power quality characteristics of grid coected wid turbies. [] P. ørese ad al, Europea Wid Turbie Testig Procedure Developmets. Task : Power Quality, Riso Natioal Laboratory, May 1. [] T. Thiriger, T. Petru ad. Ludberg: Flicker cotributio from wid turbie istallatios, IEEE Trasactios o Eergy Coversio, Vol. 19, No. 1, March 4. [4] O. Gobeau, L. Berthet, J. L. Javerzac, Method to determie cotributio of the customer ad the power system to the harmoic disturbace, CIRED, Barceloa, pai, 1-15 May. [5] X. Robe, Itroducig power quality moitorig fuctios i digital protective relays, DEA at the ULB, Electrical Egieerig Departmet,. [6] A. Morales ad J.C. Mau: Power quality resposibilities by grid impedace assessmet at a wid power productio, CIRED, Barceloa, pai, 1-15 May RE&PQJ, Vol. 1, No., March 5