Voltage Flicker Monitoring Based on Virtual Instrument and Wavelet

Size: px

Start display at page:

Download "Voltage Flicker Monitoring Based on Virtual Instrument and Wavelet"

Emory Strickland
5 years ago
Views:

1 Voltage Flicer Moitorig Baed o Virtual Itrumet ad Wavelet Pacet Traform School of Electric ad Electroic Egieerig, Shadog iverity of Techology, Zibo, 5509, Chia,lzm@dut.edu.c Abtract Thi paper propoe a ew method of moitorig ad aalyzig voltage flicer baed o wavelet pacet traform ad virtual itrumet, ad etablihe a arithmetic model of the ytem. Firt, through the wavelet pacage traform, we ca extract the amplitude modulatio waveform, get the accurate time whe the voltage flicer occur ad top, ad cout the frequecy ad amplitude of voltage fluctuatio. The, the eatio level weighted filter, quare multiplier ad firt order low pa filter ca be deiged. Fially, the evaluatio idice of hort-term flicer everity ca be obtaied by uig tatitic ortig algorithm. Experimet reult how that thi method i very applicable for the detectio of o-tatioary voltage fluctuatig ad flicer igal. Keyword: Wavelet Pacet Traform, Flicer, Fluctuatio, Amplitude Modulatio Waveform, Frequecy 1. Itroductio With the developmet of idutrial techology, the impact load icreaed greatly i the power etwor, cauig power load chagig rapidly, ad leadig to the voltage fluctuatig of electrical equipmet everely[1].there are may method of moitorig the voltage fluctuatio ad flicer, uch a half-wave virtual value, full-wave rectificatio circuit, quare demodulatio[]. But thee method are ut uitable for aalogy meauremet itrumet, ot uitable for the time-variable voltage flicerig igal detectio ad time-frequecy aalyi. FFT[3] ca be applied to voltage flicerig detectio, but the mai drawbac of FFT i that the igal i uppoed to be table. Therefore, FFT ca't attai qualified accuracy i voltage flicer aalyi. A tatic LAV tate etimatio algorithm[]wa propoed. The method ca moitor the voltage flicer compoet with high accuracy, but the flicer frequecy mut be determied i advace, ad fat algorithm i lacig. Nowaday, Kalma filterig algorithm ca be applied to o-tatioary flicer igal etimatio[5-6], but the parameter electio mut be very precie, ot eay to achieve. Thi paper put forward a ew method for meaurig voltage fluctuatio ad flicer baed o wavelet pacage traform, which ca ot oly extract the feature of the voltage flicer ad examie the frequecy ad the amplitude of the voltage fluctuatio, but alo detect the momet whe the voltage flicer occur..the Arithmetic of Moitorig Voltage Fluctuatio ad Flicer.1 The arithmetic model The flicer i the ubective feelig whe we tarig a icadecet lamp ad it ha ome relatiohip with the brai. The huma brai erve eed a limited memory time to perceive illumiatio fluctuatio. The viual feelig of huma eye ad brai to the illumiatio fluctuatio are (a tatitic): to the icadecet lamp with 30V, 60W, the frequecy rage perceivable ca ot urpa 0.05Hz to 35Hz. So we eed to et up a mathematic model that imulate the light-eye-brai proce, the cout ad deal with the waveform ut gotte. Here we cotruct a arithmetic model of moitorig voltage fluctuatio ad flicer baed o the wavelet pacage traform a how i Figure 1. Iteratioal Joural of Advacemet i Computig Techology(IJACT) Volume, Number3,December 01 doi: /iact.vol.iue

2 u(t) Wavelet pacage Traform Seatio level weighted filter Square multiplier Firt order low pa filter Statitic aemet of flicer Figure 1. The arithmetic model of voltage fluctuatio ad flicer Sice the flicer i due to voltage fluctuatio, the firt tep of moitorig voltage flicer i to extract compoet of voltage fluctuatio from igal u(t). By aalyzig the amplitude modulatio waveform of igle frequecy to the modulatio power frequecy carrier wave. The itataeou value of a voltage ca be expreed a: u( t) A(1 m co t) co t (1) Where A i the amplitude of the power frequecy carrier wave ad ω it agle frequecy, m the amplitude of the amplitude modulatio waveform ad Ω it agle frequecy. By mea of the wavelet pacage traform, amplitude modulatio waveform i extracted from u(t) ad give by v( t) m co t () The eatio level weighted filter imulate the frequecy repoe of huma eye to the icadecece lamp iflueced by ie volute fluctuatio. The imulatio of huma eye frequecy chooe characteritic ca ue the trafer fuctio IEC recommeded K( ) ( )( ) (3) The quare multiplier imulate the o- liearity ee proce of eye brai. The firt order low pa filter ue itegral fuctio to average the moothe, ad imulate the oliear ad memory effect of huma brai erve to reflect viio. The output of firt order low pa filter i waveform of itataeou flicer eatio level S(t). Short-term flicer everity Pt i a tatitical calculatio reult which how whether the flicer i trog or wea. Pt ca be attaied through tatitical calculatio of S(t).. Wavelet pacet algorithm Let h 0,h 1 be the coefficiet of the filter of multi-reolutio aalyi, defie the followig recurio formula: w t) h w (t ) () ( 0 Z w 1( t) h1 w (t ) Z 0 ( t) ( t), w1 ( t) ( t, (t) w t) Z Whe =0, w ) i orthoomal calig fuctio, (t) i wavelet fuctio, the above defied fuctio aggregate ( i wavelet pacet that decided by (5) 67

3 w0 ( t) ( t), wavelet pacet w ( t) Z i the et of fuctio havig certai relatio icludig calig fuctio w ( ) ad mother wavelet w ( ) [7-9]. 0 t To dicu cotituet pace of the wavelet pacage, itroduce the ymbol 1 W, Z 1 t. Accordig to wavelet multi-reolutio aalyi, could get V V V Z (6) Expad the Eq.6 to the wavelet pacage, get 1 1, Z, Z (7) 1 ad 1 1 are ubpace of 1, 1 1 correpod to w, 1 correpod to w 1, i wavelet multi-reolutio aalyi, L ( R) W,Expre with W, the pace decompoitio of Z wavelet pacage i W (8) I wavelet multi-reolutio aalyi, ( R) Z L pace i decompoed ubpace Z, V ad W.I wavelet pacet decompoitio, accordig to dyadic form decompoitio, becaue =0 i correpodig to wavelet decompoitio, o oly coider =1,,,ad =1,,, W (9) Ad recurio o o. Wavelet pacet traform ca get more abudat iformatio about time-frequecy localizatio, it i more uitable for aalyi of umpig igal ad o-tatioary igal. For example, here i a three-level decompoitio, it wavelet pacet decompoitio i how a i Figure, A mea low frequecy, while D mea high frequecy, the lat equece umber expree decompoitio level of wavelet pacet. The decompoitio of wavelet pacage i S AAA3 DAA3 ADA3 DDA3 AAD3 DAD3 ADD3 DDD3 (10) S A1 D1 AA DA AD DD AAA DAA ADA DDA AAD DAD ADD DDD Figure. Wavelet pacet decompoitio After level wavelet pacet decompoitio, the volute igal ca get decompoitio ode ad correpodig frequet bad ubpace, the correpodig frequet bad are f {[ 0 ], ] 1 [ f f (, ] 1 [ -1) f f }, where 1 1 f i amplig frequecy. The frequecy bad of igal ca be divided evely by the wavelet pacet traform, but i fact, the actual frequecy bad ditributio i ot arraged accordig to the frequecy after the wavelet pacet traform. There are diorder i the frequecy bad diviio. For example, a i level, the 68

4 theoretical order of ode would be (, 0), (, 1), (, ), (, 3), However, the actual order of ode i (, 0), (, 1), (, 3), (, ). A a matter of fact, the frequecy bad of wavelet pacet i alo a regular arragemet. The relatiohip betwee the theoretical order ad actual order ca be expreed uig cyclic code. Such a 3 level, B=B B 1 B 0 expre theory biary umber, G=G G 1 G 0 expre actual biary umber. Gi Bi 1 Bi (11) Where G0 B1 B0, G1 B B1, G B3 B, B 3 =0. The relatiohip of theoretical ad actual ode equece i how i Table 1..3 Numerical imulatio Table 1. Relatiohip of theoretical ad actual ode equece Theoreti cal Theoreti cal Actual biary l Actua ode ode biary umbe equece equece umber r B B 1 B 0 G G 1 G (3,0) (3,0) (3,1) (3,1) (3,) (3,3) (3,3) (3,) (3,) (3,6) (3,5) (3,7) (3,6) (3,5) (3,7) (3,) 0 Figure 3. The imulatio reult. (a) igal u(t), (b) the amplitude modulatio waveform, (c) igal(amplitude beig 0V) i ub-bad 1.5Hz~5Hz,(d) igal with the amplitude beig 380V ad the frequecy beig 50Hz. Suppoe that the iput voltage igal u(t) i a ie waveform with the amplitude beig 380V ad the frequecy beig 50Hz ad voltage flicer with the amplitude ad the frequecy beig 38V ad 6Hz repectively occur at t=0.5 ad top at t=1.65, the amplig frequecy i 300Hz. ig the 69

5 Daubechie5 wavelet 9 to decompoe u(t), electig eve level decompoitio, each frequecy bad pace i 1.5Hz, frequecy bad of igal equece i ub-bad:(0~1.5hz), (1.5Hz~5Hz), (5Hz ~37.5Hz), (37.5Hz~50 Hz). Figure 3 how the imulatio reult, Figure 3 (a) how igal u(t), Figure 3 (b) how igal i ub-bad 0~1.5Hz (the amplitude modulatio waveform), Figure 3 (c) how igal(amplitude beig 0V) i ub-bad 1.5Hz~5Hz, Figure 3(d) how a ie waveform with the amplitude beig 380V ad the frequecy beig 50Hz. It i very clear that the voltage flicer occur at t=0.5 ad top at t=1.65, the frequecy ad of the amplitude of amplitude modulatio waveform are 6Hz ad 38V repectively. Table how imulatio reult of differet frequecy modulatio waveform. Table. Experimet rult of differet frequecy modulatio waveform Frequecy Theoretical Meaured (Hz) virtual virtual Error value(v) value(v) The Hardware ad Software Deig of the Sytem 3.1The hardware deig of the ytem The hardware the ytem coit of eor, igal coditioig circuit, trafer board, data acquiitio (DAQ) ad computer. The eor are HYH-SB-13- voltage eor ad HYH-SB-13-I curret eor. The igal emedatio circuit i compoed by LM3, a i-phae proportio amplifier, which revie error of eor caued by idividual differece. The trafer board i made by u. The data acquiitio i PCI-603E. a product of NI compay, it ha 16 chael igle, 8 chael differece aalogy iput, 1 bit preciio, 00KS/ amplig frequecy, ±0.05V to ±10V iput rage. 3.The oftware deig of the ytem (1) Wavelet pacet algorithm The Wavelet pacet algorithm ca be realized by mixed programmig of MATLAB ad LabVIEW, i developmet eviromet of LabVIEW, we li MATLAB cript program by uig MATLAB ode, realize call of wavelet fuctio. () The deig of eatio level weighted filter Coverig Eq.3 to z domai ytem fuctio. The traform Eq.1 to differece equatio a b z 0 H ( z) (1) 1 a z 0 1 y ( ) b x( ) a y( ) (13) Where a 1 = , a =.7158, a 3 = , a = , b 0 = , b 1 =0.0039, b = , b 3 = , b = The ubprogram of eatio level weighted filter wa completed i LabVIEW accordig to differece equatio (Eq.13). 1 (3) The deig of quare multiplier ad firt order low pa filter 70

6 Square multiplier i Square arithmetic for output of eatio level weightig filter. The trafer fuctio of firt order low pa filter i K ( ) (1) Coverig Eq.1 to z domai ytem fuctio ) 0.671(1 z H ( z) z (15) The traform it to differece equatio a y ( ) 0.671x( ) 0.671x( 1) y( 1) (16) Firt order low pa filter ca be achieve uig IIR Filter.vi i LabVIEW. We ca get waveform of S(t) from above-metioed arithmetic. () Short-term flicer everity Pt The formulatio of Pt a follow: P t 0.031P P P3 0.8P P50 (17) Where P 0.1,P 1,P 3,P 10 ad P 50 are detectio uit that the itataeou flicer eatio level S(t) exceed 0.1% 1% 3% 10% ad50% i 10 miute. The program of calculatig Pt wa deiged by uig tatitical ortig algorithm, becaue S(t) i dicrete equece i equality time iterval, if the um of the time period (where the S i o le tha a certai S(x)) i x percet of all the time period, that mea S i the highet probability value withi(100-x)% period of time. So the P value, P 0.1 P 1 P 3 P 10 P 50 which correpod to 0.1% 1% 3% 10% ad 50% ordiate of CPF curve are the highet probability value withi 99.9% 99% 97% 90% 50% of time period S(). We ut eed to realig S() from malle to bige to fid out correpodig probability highet value, the apply them to the formula metioed above. (5) The flow chart of program Figure how the flow chart of program. 71

7 Set meauremet parameter Read voltage igal u(t) Call Wavelet pacet algorithm Call eatio level weighted filter Call quare multiplie Call firt order low pa filter Call tatitic ortig algorithm of Pt Diplay u(t),v(t), S(t),Pt. Experimet Figure. The flow chart of program.1emulatio reult of the eitivity coefficiet K(f) A the eitivity coefficiet K(f) ca repreet the frequecy characteritic of light-eye-brai. Table 3 how repectively the K(f) obtaied by IEC amplig ad from the actual meauremet, it ca be ee that the error i very lower. Freque cy (Hz) Table 3. Emulatio reult of K (f) Voltage K (f) of K (f) of actual fluctuatio IEC meauremet (%) Emulatio reult of itataeou flicer eatio level S(t) Table idicate the imulatio reult of itataeou flicer eatio level S(t), the error i very lower. 7

8 .3 Experimet reult Table. Emulatio reult of itataeou flicer eatio level S(t) Freque Voltage cy fluctuatio(%) (Hz) S(t) Error The voltage flicer moitorig ytem wa applied ito two typical ubtatio. The experimet reult of voltage fluctuatio ad flicer i how i Table 5, ad it proved that the tetig ytem had ice iteroperability, while it operatio wa table ad the cocluio of the tet were reliable. 5. Cocluio Table 5. Experimet reult Tetig time S(t) Pt :10: :11: :1: :13: :1: :15: :18: :: The voltage fluctuatio ad flicer moitorig ytem baed o Wavelet pacage traform ad virtual itrumet coit of eor, igal coditioig circuit, data acquiitio ad computer. I the ytem, the wavelet pacage traform wa applied to virtual itrumet by uig the MATLAB cript ode i LabVIEW, Seatio level weighted filter, Square multiplier, firt order low pa filter ad the evaluatio idice of hort-term flicer everity were deiged. Experimet Reult idicated that the amplitude modulatio waveform ca be extracted, ad the accurate time whe the voltage flicer occurred ad topped ca be detected, the error of the eitivity coefficiet K(f), itataeou flicer eatio level S(t) ad hort-term flicer everity Pt were very mall. 6.Acowledgmet Thi reearch wa upported by the Natioal Natural Sciece Foudatio of Chia ( ) ad the Natural Sciece Foudatio of Shadog Provice (ZR010FM005). 7. Referece 73

9 [1] A. M. Aladari, S. A. Solima, Meauremet of a power ytem omial voltage, frequecy ad voltage flicer parameter, Iteratioal Joural of Electrical Power ad Eergy Sytem, vol.31,.7-8, pp , 009. [] Bihop M.T., Sedzic V., Witte J.F., Sharma D., Adoptig the IEC voltage flicer methodology i the ited State, Iro ad Steelmaer (I ad SM), vol,9,.8, pp.65-70, 00. [3] C. J. Wu, T. H. Fu, Effective voltage flicer calculatio algorithm uig idirect demodulatio method, IEE Proceedig: Geeratio, Tramiio ad Ditributio, vol.150,, pp , 003. [] S.A. Solima; M.E. El-Hawary. Meauremet of power ytem voltage ad flicer level for power quality aalyi: a tatic LAV tate etimatio baed algorithm. Iteratioal Joural of Electrical Power & Eergy Sytem, vol., o.6, pp.7-50,000. [5] Zhou Shagbo, Hu Peg, Li Ku, Liu, Yuiog, A ew target tracig cheme baed o improved mea hift ad adaptive alma filter, Iteratioal Joural of Advacemet i Computig Techology, vol., o., pp , 01. [6] Yag Li, A ovel fuzzy votig deig baed o adaptive exteded alma filter, Iteratioal Joural of Advacemet i Computig Techology, vol. 3, o. 6, pp , 011. [7] Wog Ko Max; Wu Jiagfeg; Davido Tim N. Wavelet pacet diviio multiplexig ad wavelet pacet deig uder timig error effect. IEEE Traactio o Sigal Proceig, vol. 5, o. 1, pp , [8] Zhou Rui, Bao We, Li Nig, Dare Mechaical equipmet fault diagoi baed o redudat ecod geeratio wavelet pacet traform, Digital Sigal Proceig: A Review Joural, vol. 0, o. 1, pp , 010. [9] I. Daubechie, W. Swelde, Factorig Wavelet Traform ito Liftig Step, J. Fourier Aalyi Applicatio, vol., o. 3, pp. 5-67,

LP10 INFERENTIAL STATISTICS - Confidence intervals.

LP10 INFERENTIAL STATISTICS - Cofidece iterval. Objective: - how to determie the cofidece iterval for the mea of a ample - Determiig Sample Size for a Specified Width Cofidece Iterval Theoretical coideratio