Harmonic power measurements on discharge lamps

Transcription

1 Harmoic power measuremets o discharge lamps C.. Orfaos I.F. Goos F. V. opalis, Member IEEE atioal echical Uiversity of Athes, School of Electrical ad Computer Egieerig 9 Iroo Politechiou Str., 578 Athes, Greece, el: , Fax: , topalis@ieee.org ABSRAC: his paper presets a experimetal method to perform harmoic power measuremets o discharge lamps (metal halide, fluorescet etc). A virtual istrumetatio system has bee developed usig the LabView graphical programmig laguage. he system is actually a spectrum & power aalyser. It also icludes a arbitrary waveform geerator i order to perform measuremets with distorted supply voltages. his virtual istrumet has also bee developed i the LabView eviromet. he experimetal apparatus icludes voltage ad curret probes to tae the measuremets ad a specially desiged power amplifier to supply the loads with a user-defied waveform. he voltage ad curret sigals are recorded i real time ad the aalysed usig FF ad DF i order to determie their frequecy spectra. he magitude ad the phase agle of the harmoic compoets are computed as well as the HD. Moreover, the system performs power aalysis of the idividual harmoics ad of the recorded voltage ad curret sigals i the time domai. herefore their power parameters are determied (apparet, active, reactive ad distortio power). A importat advatage of the system is that the virtual istrumets allow simple modificatios by the user i order to perform additioal measuremets for ew quatities while the measuremet of the o-liear load is i progress. Keywords: Power quality, harmoics, harmoic aalyser, virtual istrumet, curret distortio, discharge lamps I. IRODUCIO Discharge lamps (fluorescet, high itesity discharge etc.) are characterised by a high lumious efficiecy ad provide sigificat eergy savig if compared to the icadescet lamps. hese lamps iject harmoic currets ito power systems as all o-liear loads. heir ballast ad their arc tube ca be a importat source of higher-order harmoic compoets of curret []. he harmoic currets pass through the impedace of the system ad cause a voltage drop for each harmoic. his results i voltage harmoics appearig at the load bus ad cosequetly it iflueces the quality of the supplied power as well as the electrical appliaces [, 3]. Surveys [4, 5] report that the 95% probability level for the 5 th harmoic voltage sustaied a steady icrease from 3% to 5% i.e..67% per te years. he compact fluorescet lamps (CFL) exhibit the highest harmoic distortio amog the discharge lamps. heir total harmoic distortio (HD) is usually higher tha %. he iteratioal stadard IEC 6-3- requires that the 3 rd ad the 5 th curret harmoic shall ot exceed 86% ad 6% of the fudametal respectively [6]. ASI defies a limit of 3% [7] as the maximum curret HD of lamps with electroic ballast. his stadard also specifies the limit of the magitude of all high-order harmoics to 3% of the fudametal magitude. he upper limit is defied as 7% of the fudametal for all higher tha the th order harmoics. he limit of the curret HD of electroic ballasts is %. Harmoic measuremets require istrumets that are desiged for spectral aalysis. hey shall have the capability to measure simultaeously voltage ad curret i order to obtai the harmoic power flow. For the same reaso it is required to measure both magitude ad phase agle of idividual harmoic compoets. Sychroizatio is also essetial ad fast samplig rate to obtai accurate measuremets of harmoic compoets at least up to the 39 th harmoic. his paper presets a experimetal apparatus with the required specificatios for the ivestigatio of the electrical performace of o-liear loads ad especially of discharge lamps. he cofiguratio of the system is compact ad user friedly. he cost for the developmet of such a system is very low compared with covetioal oes, which cosist of digital oscilloscopes, spectrum aalysers, ad several measurig istrumets. II. DEFIIIOS he power ad harmoics aalysis of this paper is performed usig the voltage ad curret waveforms of the discharge lamp. Both of these waveforms are recorded simultaeously i the time domai. heir effective (rms) value is calculated from the formula: = y ( t) dt where y(t) represets the waveform of voltage v(t) or curret i(t) ad the respective effective value V ad I. his formula is valid for every periodic sigal, siusoidal or distorted. If the sigal is distorted by harmoics the a FF (or DF) aalysis is required to obtai the frequecy spectrum ad cosequetly the effective values of the harmoic compoets. I this case the effective value of the sigal is calculated with harmoic represetatio as: = = where is the DC offset, the amplitude of the fudametal, the amplitude of the harmoic of th order ad the highest harmoic order. () ()

2 A simple method to determie whether a sigal is distorted is to calculate the crest factor that is defied as the ratio of the pea value of the waveform to its rms amplitude: = V, I of voltage ad curret: pea CF (3) S = V I If this ratio is ot equal to the the waveform is distorted. Usually the harmoic compoets are expressed power: with respect to the effective value of the fudametal. his is ow as the idividual harmoic rate H of the S = V I + compoet. It is give by the followig formula accordig = = = to the CIGRE defiitio: H = % (4) For distorted sigals it is very useful to determie the effective value H of the harmoic cotet. he waveform should be cleared off the fudametal compoet ad the DC offset: H = = he magitude of the distortio is usually expressed with respect to the rms value of the fudametal or i some cases to the rms value of the sigal. Accordig to the IEEE defiitio the total harmoic distortio (HD) of a sigal is the ratio of the harmoic cotet to the magitude of the fudametal: HD = H % (5) (6) Apparet power is defied as the capacity of the power system required to deliver active power P. Accordig to the IEEE defiitio it is give as the product of the rms values (3) Substitutig the equivalet expressios of V ad I from Eq.() to Eq.(3) results the followig form for the apparet V I (4) Worig similarly with Eqs.(8), () the followig form for the active power is derived: P = ( V I cosϕ ) = + V I cosϕ (5) V I cosϕ = = + A similar formula is obtaied from Eqs.(), () for the reactive power: Q = ( V I siϕ ) = + V I siϕ (6) V I siϕ = = + he total active power that is produced by v(t) ad i(t) is the mea value of the istataeous power: shall be added i order to fulfil the equatio: P = v( t) i( t) dt (7) S = P + Q + D he active power of the idividual harmoics -icludig the fudametal- is determied from the effective amplitudes V, I of their voltage ad curret respectively: P = V I cosϕ i=,, 3,,. (8) where cosφ is the phase displacemet betwee them. his is the power factor of each idividual harmoic: PF = cosϕ (9) herefore, the sum of the active powers P of the harmoics -plus the DC power P - will give the total active power: P = P = Obviously the results from Eq.(7) ad () shall be equal. he reactive power of the harmoics is determied usig the same method as above: Q = V I siϕ i=,, 3,,. () As for the total reactive power, there is a disagreemet betwee the aalysts o how to defie it i the presece of harmoics. Most of them agree that the total reactive power Q is the sum of reactive powers Q at each frequecy: Q = Q = Observig Eqs.(4), (5) ad (6) becomes obvious that the sum P +Q is ot equal to S for distorted sigals v(t) ad i(t). his would be true oly i case of pure siusoidal voltage ad curret waveforms i.e. without harmoics. For distorted sigals the apparet power cosists ot oly of active ad reactive power but also of a ew quatity that (7) his quatity is equal to the sum of the cross products of the harmoics of differet order V I ( ). It is called distortio power D sice these harmoics do ot produce wor ad yield o average power. D may ot be referred as power sice it does ot flow through the system as power is assumed to do. D is calculated from the followig formula that is obtaied from Eqs.(4)-(7): D () = = = + ( V I cosϕ V I cosϕ ) + ( V I si V I siϕ ) = = + ϕ (8) Special attetio should also be paid to power factor. For o-distorted sigals it is equal to the phase displacemet factor cosϕ betwee v(t) ad i(t). However, this is ot true for distorted sigals. I this case the value of phase displacemet diverges from the value of the power factor. he divergece becomes high whe the sigals are cosiderably cotamiated. I ay case the power factor should always be determied as the ratio of the active power P to the apparet oe S: P PF = (9) () S

3 III. HARMOIC POWER AALZER For the purposes of this ivestigatio a harmoic power aalyser was developed which cosists of a computer, a data acquisitio multifuctio card PCI- of atioal Istrumets ad the software pacage LabView. he system also icludes a resistive divider for the measuremet of voltage ad a clamp meter with aalogue output for the measuremet of curret. he supply voltage ad the curret waveforms are recorded ad trasferred ito the computer through aalogue iputs of the card. Both waveforms are aalysed ad their harmoic compoets are computed. he computer cotrols the experimetal procedure usig Lab View software pacage. he eviromet of LabView is very friedly for developig programs usig graphical programmig laguage. It uses termiology, icos ad ideas familiar to scietists ad egieers ad relies o graphical symbols rather tha textual laguage to describe programmig actios. It has extesive galleries of fuctios ad subrouties for most programmig tass icludig data acquisitio. Virtual Istrumets (VIs) imitate actual istrumets. he developed VI simulates the pael of three physical istrumets: a oscilloscope, a spectrum aalyser ad a power aalyser. It cotais a iteractive user iterface that is the frot pael of the aalyser (Fig.). he experimetal procedure of this project is divided i a series of tass that ca be divided agai util the complicated applicatio becomes a series of simple subtass. he tass ad the subtass are performed by the user of the computer via several VIs that have especially bee developed for the purposes of the project. he DAQ board is a PCI- card with maximum samplig rate S/s. he card has output chaels ad 8 differetial (or 6 sigle-eded) iput chaels. he voltage-iput rage is software programmable for -V (uipolar) or ±5 V (bipolar). A software programmable gai amplifier has gai selectio,, 5,,, 5,. he PCI- has a -bit ADC with aalogue resolutio of.44 mv at a gai of. he card is samplig at 5 Hz o two iput chaels for the recordig of the applied voltage to the lamp ad of the curret of the lamp. his meas that the samplig frequecy is.5 Hz per chael. his frequecy is high eough to prevet aliasig, accordig to yquist theorem, sice the maximum harmoic frequecy of the measured sigals is 95 (39 th harmoic). herefore, a atialisig filter is ot required. he asychroous samplig o two iput chaels causes a time delay of /5 KS/s=.4 ms. his is ot a problem because from field measuremets harmoics up to 7 th order have bee detected i the utility voltage. I that case the period of 7 th harmoic is.86 ms. Hece, the itroduced asychroous samplig error i the phase displacemet betwee the voltage ad curret harmoics of 7 th order is oly 5 o. his meas that the maximum itroduced error i the measuremet of power factor does ot exceed 8.7%. he frot pael receives istructios from a bloc diagram that provides a pictorial solutio to the programmig problem. Samplig starts after a aalogue trigger from the software. A specified umber of samples of voltage ad curret waveforms are retrieved through two aalogue iput chaels. A triggered acquisitio ca start multiple time while is avoided the overhead of cofiguratio ad the buffer allocatio each time. his is a timed acquisitio, meaig that a hardware cloc is used to cotrol the acquisitio rate for fast ad accurate timig. It is also a buffered acquisitio because the data are stored i a itermediate memory buffer after they have bee acquired from the DAQ board. Sice a aalogue trigger is used from the software, data are cotiuously acquired while the program is checig whether the itermediate buffer for the aalogue trigger specificatios is beig met. Oce met the program retrieves ad displays data from that buffer after the appropriate amout of data have bee acquired before ad after the trigger. he program waits after each iteratio for the ext trigger the it reads the same amout of data agai from the same chaels at the same rate. he user ca set up the trigger chael, the polarity of the slope (positive or egative), the hysteresis aroud the trigger level withi which variatios i the aalogue sigal level do ot cause triggers. He ca also set the time delay for the trigger ad the umber of scas as well as the umber of pre-trigger scas i.e. the umber of scas before the trigger. he samples of voltage ad curret retur to a x array which is split ito two x arrays, oe for the samples of voltage waveform ad the other for the curret waveform. Each vector of samples is scaled i order to be i accordace with the real values. Whe usig FF or DF the digital sigal is modified to iteger umber of periods i order to avoid leaage of spectral iformatio. he umber of periods is at least four. Additioally, the user may apply smoothig widows such as Haig, Hammig, Blacma, Blacma-Harris, Exact Blacma, Flat op, Geeral Cosie (4 erm B-Harris, 7 erm B-Harris) to improve the spectral characteristics of the sampled sigals [8]. After that, both of the digital sigals are aalysed i the time ad frequecy domai. If the umber of samples is a valid power of, the the subroutie performs FF to reduce the umber of calculatios to 3 log. he maximum legth of FF is the legth of the buffer. he buffer legth is set up to 496 samples. FF is performed by applyig the Split-Radix algorithm, which is similar to the Radix-4 algorithms i additio to the efficiecy of Radix-8 algorithms. he Split- Radix algorithm requires the least umber of multiplicatios amog the Radix-, Radix-4 ad Mixed- Radix algorithms [8]. If the umber of samples is ot a valid power of, the the program computes the DF callig a efficiet DF routie. However, DF is slower ad uses more memory tha FF. he measured sigals ad their harmoic ad power aalysis are displayed o the mai widow of the frot pael of the istrumet (Fig.) while several other widows are available for the represetatio of various parameters of the measured quatities. IV. EXPERIMES he experimetal results which, are preseted below, demostrate the capabilities of the developed virtual istrumet for harmoic aalysis ad power measuremets. he selected lamps for the experimets represet two

4 categories of discharge lightig: compact fluorescet with electroic gear (CFL) ad metal halide (MH). A more extesive ivestigatio of other types lie mercury ad sodium vapor is out of the scope of this paper. Actually it is the subject of a special paper dealig exclusively with the electrical characteristics of discharge lamps. Such a ivestigatio is curretly i progress ad the results will be preseted i the ear future. he CFLs of this ivestigatio are fitted with electroic ballast which is built ito the lamp fixture. heir mout is E7 screw. All of them are desiged for the 3 V, 5 Hz electric utility systems. hree groups of such CFLs have bee tested from three differet brads. Each brad with wattages 5 W, W ad 3 W. he bulb of the MH lamps is a two-pi compact type. hey require exteral magetic ballast ad starter ad operate at 3 V, 5 Hz supply voltage. wo groups have bee tested with wattages 7 W ad 5 W. he curret waveform ad the harmoic spectrum of a typical W CFL is preseted i Fig.. A sigificat distortio is evidet i the curret of the lamp. All the ivestigated CFLs exhibit a highly distorted curret. It is very importat to observe i able that the 3 rd harmoic i six out of ie CFLs exceeds the limit of 86% that is set by IEC 6-3- [6]. O the other had the 5 th harmoic of these lamps also exceeds the limit of 6% that is set by IEC Fig. Harmoic power aalysis of a 7W metal halide lamp (with ballast) o the frot pael of the aalyser.

5 A,8,6,4,, -, -,4 -,6 -,8,,,,3 % 8 6 4,4 sec fudametal oe ad i some cases almost equal to it (able ). As expected the HD value is quite high especially the oe of CFLs that exceeds %. he crest factor also diverges from the ideal value of.4. Lamp able Amplitude of distortio. otal Harmoic HD curret curret (A) (A) (%) Crest factor CFL5W/a CFL5W/b CFL5W/c CFLW/a CFLW/b CFLW/c CFL3W/a CFL3W/b CFL3W/c MH7W/b MH5W/b Harmoic Order Fig. Curret waveform ad harmoic spectrum of a W CFL. able Amplitude of curret harmoics (% of the fudametal). he letters a, b, c deote the three differet brads Harmoic order Lamp CFL5W/a CFL5W/b CFL5W/c CFLW/a CFLW/b CFLW/c CFL3W/a CFL3W/b CFL3W/c MH7W/b MH5W/b For a distorted sigal it is importat to determie its harmoic cotet. he waveform should be cleared off the fudametal compoet ad the DC offset. his results to a sigal i the time domai that is purely the distortio of the waveform (Fig.3). he curret of all tested lamps is heavily distorted. he harmoic curret is comparable to the A,4,,, -, -, -, 4,,,,3 Fig.3 Waveform of the total harmoic cotet of a 5W CFL.,4 sec Oe of the features of the virtual aalyser is the determiatio of the phase displacemet of each harmoic with respect to a referece poit. Geerally this poit is the start of the voltage waveform i order to get the phase displacemet betwee supply voltage ad curret harmoics. he widow with the phase agles of the curret harmoics of a 5W CFL is displayed i Fig.4. Degrees Fig.4 Phase agle of idividual curret harmoics of a 5W CFL. It may be observed i Fig.4 that the phase agle of the fudametal curret is quite low. his remar cofirms the cojecture that all of the commercially available electroic CFLs are fitted with a compesatio circuit, which corrects the phase displacemet betwee the supply voltage ad the fudametal curret. Actually, all of the tested CFLs are characterised by a quite high cosϕ at the fudametal frequecy, usually i the rage of (able 3). he presece of reactive power at all frequecies is justified due to the phase agle betwee supply voltage ad curret harmoics. he reactive power of some harmoics is egative due to their capacitive phase displacemet. For o-distorted supply voltage there is o active power at frequecies higher tha the fudametal, sice the mea value of the istataeous power is zero i that case. However, the slight distortio of voltage causes active power cosumptio at the frequecies of the respective harmoics. Further to the harmoic aalysis, the istrumet also performs a power aalysis of the recorded waveforms. It computes the istataeous active power ad the itegral of it over a specified umber of periods (mea value) that defies the active power (Fig.5). he other power parameters (reactive, apparet ad distortio power) are also calculated usig the methodology described i II Hz

6 (able 4). It seems that the CFLs operate at very low power factor as it was expected while the power factor of the MHs is much better. Watts 3 able 3 Power characteristics of lower order compoets. Lamp Harmoic P Q cosϕ order (W) (Var) CFL 5W/a CFL 5W/b CFL 5W/c CFL W/a CFLW/b CFLW/c CFL3W/a CFL3W/b CFL3W/c MH7W/b MH5W/b 6, this applicatio. he cost of the system is very low compared with a covetioal oe cosistig of a digital oscilloscope, a spectrum aalyser ad several measurig istrumets such as voltmeters, ammeters or power aalysers. he system cofiguratio is compact ad user friedly. It has bee developed to perform measuremets o discharge lamps but it ca also be used for similar experimets o other low voltage equipmet. he system facilitates the performace evaluatio of various appliaces for distorted supply voltages. able 4 Power aalysis of the ivestigated lamps. Power Power Active Reactive Distortio Apparet factor Lamp (W) (Var) (VA) (VA) CFL5W/a CFL5W/b CFL5W/c CFLW/a CFLW/b CFLW/c CFL3W/a CFL3W/b CFL3W/c MH7W/b MH5W/b VI. REFERECES [] F.V. opalis, Efficiecy of eergy savig lamps ad harmoic distortio i distributio systems, IEEE ras. o Power Delivery, Vol. 8, r. 4, 993, pp [] R.R. Verderber, O.C. Morse, W.R. Allig, Harmoics from compact fluorescet lamps, IEEE ras. o Idustry Applicatios, Vol. 9, r. 3, 993, pp [3] D.J. Pileggi, E.M. Gulachesi, C.E. Root,.J. Getile, A.E. Emauel, he effect of moder compact fluorescet lights o voltage distortio, IEEE ras. o Power Delivery, Vol. 8, r. 3, 993, pp [4] H.J. Koster, F. Weiel, Voltage ad curret harmoics i low ad medium voltage etwors, Eletrizitatwirschaft, Vol. 88, 989, pp [5] I.M. ejdawi, A.E. Emauel, D.J. Pileggi, M.J. Corridori, R.D. Archambeault, Harmoics tred i E USA: A prelimiary survey, IEEE ras. o Power Delivery, Vol. 4, o. 4, 999, pp [6] IEC, Electromagetic compatibility (EMC): Limits - Limits for harmoic curret emissios (equipmet iput curret 6 A per phase) - Defiitios, IEC 6-3- (-8), Geeve, Switzerlad,. [7] ASI, America atioal stadard for lamp ballasts: High frequecy fluorescet lamp ballasts, ASI C8. 993, ew or. [8] atioal Istrumets Corporatio, LabView Aalysis VI Referece Maual, 996, Austi ,,,,3 Fig.5 Istataeous power ad active power (straight lie) of a 3W CFL V. COCLUSIOS he developed virtual istrumetatio system (hardware ad software tools) facilitates experimets o power ad harmoic aalysis. he user supervises the tests through the virtual istrumets that have bee developed especially for,4 sec VII. BIOGRAPHIES Charis Orfaos was bor i axos, Greece, i 953. He received his diploma i Electrical Egieerig from Uiversity of Patras i 976 ad the Msc. Degree i Electroic Automatio from Uiversity of Athes. He is Ph.D. caditate at the School of Electrical ad Computer Egieerig of atioal echical Uiversity of Athes. His research iterests iclude lightig, power quality ad harmoics. Ioais F. Goos was bor o May 8, 97 i Artemisio, Arcadia, Greece. He received his diploma i Electrical Egieerig i 993 ad his Ph.D. i from the atioal echical Uiversity of Athes. Fragisos V. opalis was bor i Mitilii, Greece, o March 3, 955. He received the diploma i Mechaical ad Electrical Egieerig ad the Ph.D. degree from the atioal echical Uiversity of Athes (UA) i 979 ad 99 respectively. He is the head of the Laboratory of Photometry of UA (School of Electrical & Computer Egieerig) where he teaches the course of photometry ad lightig. His research iterests cocer photometry, lightig, ratioal use of eergy, power quality ad harmoics.