Power Quality Disturbance Data Compression Using Wavelets Transform

Transcription

1 Internional Conference on Computer, Sytem and Electronic Engineering (ICSCEE') April 6, Johanneburg (South Africa) Power Quality Diturbance Da Compreion Uing Wavelet Tranform S. Mohammadzadeh, S. Gh. Seifoad, and M. Ahmadzadeh Abtract---In thi paper we preent a wavelet compreion technique for power quality diturbance da. Power quality tranient da are compreed and tored for analyi and claificion purpoe. Original da can be recontructed and then analyzed uing wavelet tranform. Different power quality diturbance including voltage ag, well, ag with harmonic and well with harmonic are ued to tet the performance the propoed approach. The file ize the compreed da i only one-ixth to one-third th the original da. Therefore, Da torage and tranmitting cot are reduced ignificantly. Keyword----Power quality diturbance, da compreion, decompoition, ag, well. I. INTRODUCTION OWER Quality (PQ) i one the mot important iue P in the electricity indutry today. The f expanion in ue power electronic device led to a wide diffuion nonlinear, time-variant load in the power ditribution networ, which caue maive eriou power quality problem. Electromagnetic diturbance caue big economic loe for indutry and reidential uer. Therefore, monitoring power quality (PQ) diturbance electric power i fundamental in order to fer olution epecially for utilitie and indutrial cutomer. Fourier tranform i ten ued in teady te diturbance monitoring. However, due to the preence tranient diturbance i not adeque. The Wavelet Tranform (WT) proceing technique ha been propoed for power quality monitoring. Thi tranformion i ued becaue timefrequency multi-reolution analyi property. [] Since tranient diturbance occur in the order microecond, the diturbance recorded uing thi type monitoring ytem produce megabyte da. A a reult, the volume the recorded da increae ignificantly. Therefore, it eem neceary to develop technique for da compreion. [], []. S.Mohammadzadeh, Department Electrical Engineering, Mahhahr branch, Ilamic azad univerity, mahhahr, Iran. (correponding author to provide phone: +9899,.mohammadzadeh8@gmail.com) S. Gh. Seifoad, Electrical Engineering Department Shahid Chamran Univerity, Ahvaz, Iran. (eifoad@yahoo.com) M.Ahmadzadeh, Department Of Electrical Engineering, Mahhahr branch, Ilamic azad univerity, mahhahr, Iran. (motafa.ahmadzadeh@yahoo.com) The purpoe thi paper i to introduce wavelet tranform method a a tool for da compreion. And Power quality diturbance will be compreed. Waveletbaed compreion cheme have been utilized epecially in the image proceing area []. II. WAVELET TRANSFORM The dicrete wavelet tranform (DWT) move a time domain ignal into it correponding wavelet domain. Thi i done through a proce called ub-band codificion, which i done through digital filter technique. []. A given diturbance waveform i tranformed into the time cale domain uing multi-reolution ignal decompoition (MSD). The bai function thi tranform i called the mother wavelet ψ(, which can be Stretched, and hifted to produce the daughter (or baby) wavelet [8]: t b j, ( a b By varying the value the dyadic caling parameter a = j, different frequency band and, hence, different decompoition j can be obtained. To illutre the tranform, let be the time domain ignal to be decompoed or analyzed. The dyadic wavelet tranform i then defined a [7], m m/ * t n DWT m, n) ( ) dt () m In principle, a recorded digitized time ignal c (n) i decompoed into it ed d (n) and moothed c (n) ignal uing filter g(n) and h(n), repectively. Filter g(n) ha a bandpa filter repone. Thu, the filtered ignal d l (n) i a ed verion c (n) and contain higher frequency component (i.e., harp edge, tranition, and jump in the original power diturbance ignal) than the moothed ignal c (n), becaue filter h(n) ha a lowpa frequency filter repone. The decompoition c (n) into c (n) and d (n) i a firt-cale decompoition. Mhemically, they are defined a follow: c ( n) d ( n) h( n) c g( n) c ( ) ( ) () () 78

2 Internional Conference on Computer, Sytem and Electronic Engineering (ICSCEE') April 6, Johanneburg (South Africa) The above equion can be hown a Fig.. At each decompoition, the length the decompoed ignal i half the length the ignal in the previou tage. Higherorder decompoition are then performed in a imilar manner. The recontruction c (n) can be performed by a pyramidal cheme imilar to decompoition. Fig. how the recontruction c (n) from decompoed ignal c (n) and d (n). Mhemically, thi can be expreed a: c( n) c( ) h( n ) d( ) g( n ) a) b) Fig. a) decompoition and b) recontruction proce () Higher-cale recontruction i performed in a imilar manner. III. DATA COMPRESSION METHOD In thi ection, a imple but effective method for da compreion uing wavelet decompoition and recontruction are preented. Wavelet tranform are able to effectively detect and localize the power quality diturbance. To do thi, wavelet decompoition technique are ued. The magnitude wavelet tranform coefficient aocied with diturbance event i much larger than th "diturbance-free coefficient." Thi fact lead to an effective method for da compreion power quality diturbance. [] The compreion i carried out in the wavelet domain. Wavelet tranform coefficient aocied with diturbance event are ept and all other diturbance-free coefficient are dicarded. The mot-moothed verion the original recorded ignal c (n) can be ept for recontruction purpoe. [] In order to recontruct the, wavelet recontruction technique are ued. Since Compreion proce elimine many the wavelet tranform coefficient, ome informion will be lot. Since many thi informion are reled to noie, the quality the recontructed diturbance ignal i very high, and mot diturbance event are preerved nicely ince the wavelet tranform coefficient aocied with the diturbance are aved. In the other word, to ome extent, the recontructed ignal i actually a better ignal ince it contain le electrical noie compared to the. []. Threholding wavelet tranform coefficient (WTC) can be performed in everal way. One the implet method i to dicard wavelet tranform coefficient below a certain threhold q, whoe value may vary from one cale to another cale. The threhold i baed on the abolute maximum value the wavelet tranform coefficient the aocied cale d (n) which Can be expreed a follow: [] ( p)max{ d( n)} () Where η i the threhold value, p lie in the range <p< and d (n) i the wavelet tranform coefficient the aocied cale and i given by: ^ d ( n) d ^ d ( n) ( n) for d ( n) for d ( n). (6) After threholding, it i clear th d (n) are maller than η are dicarded and thoe th are larger are tored. Compreion Rio i the rio original file ize to the compreed file ize. By chooing a different threhold value, a different compreion rio can be obtained. Compreion Rio i defined a: [] original file ize Compreionrio compreed file ize (7) In the next ection, we how compreion reult uing three-cale decompoition and recontruction to further enhance the compreion rio. IV. DATA GENERATION The imulion da wa genered in MATLAB area baed on the model in reference [] and []. The IEEE Std9 i conidered a performance meaure. five different PQ diturbance, named well, ag, harmonic, ag with harmonic and well with harmonic, were conidered. Table I give the ignal generion model and their control parameter. Starting and ending time, amplitude ag and well and quantity frequency component are genered randomly in MATLAB. V. SIMULATION The propoed da compreion technique i applied to different power quality diturbance. All the da are genered uing the MATLAB code a ampling re Hz. To demontre the efficacy the propoed technique ome tet cae are preented below. For all imulion tudy a pure inuoidal ignal Hz and p.u. amplitude i conidered. We apply a three-cale wavelet decompreion cheme to everal example. Different PQ ignal are conidered to ee the performance the approach. 79

3 Internional Conference on Computer, Sytem and Electronic Engineering (ICSCEE') April 6, Johanneburg (South Africa) TABLE I MODEL OF POWER QUALITY DISTURBANCES PQ diturbance MODEL Parameter normal in( - Swell, t A( ( t t ) t t )))in(, t t,..8, T t t 9T, t Sag A( ( t t ) t )) in(..9, T t t 9T Harmonic A( in( in( in( 7 in(7t )).,, 7., i A(( t t ) t t Swell with Harmonic ( in( in( in( t ) A(( t t ) t t Sag with Harmonic ( in( in( in( t ) ))) )))..9, T t t.,.,..9, T t t.,., 9T i 9T i A. Voltage ag Voltage ag i a udden reduction ( 9%) in the voltage magnitude, ling for. cycle to everal econd. Such a phenomenon may be caued by witching operion aocied with temporary diconnection upply, flow heavy current aocied with the tarting large motor load or the flow fault current. [8] Fig (a). how a voltage ag. Actual tarting time i. and actual ending time i.9. a i mentioned in [], -. wavelet tranform can be ued in detection and localizion power quality diturbance and firt i the bet original ignal criteria for thi purpoe. Fig (b) how. firt. By. uing thi figure, etimed tarting.. time i. and etimed ending time i.89 th i approximely - -. acceptable and repreent mall error -..%.. -. original. ignal (a) (b) Fig. a) Voltage ag b) firt The ha 7 ample.. point. The ignal ag i decompoed - by the DWT up to three and the reult are hown in Fig Each thee figure contain the following plot: (a): firt (6 poin (b): firt (6 poin (c): econd (8 poin (d): econd (8 poin (e): third (97 poin (f): third (97 poin Now, we recontruct the ignal by uing - ignal - recontruction technique. Recontruction proce i hown in fig. In thi figure: recontuction..ignal for reyag (a) firt (c) econd (d) econd (e) recontuction ignal third (f) third for reyag for reyag Fig. Decompoition voltage ag in three. (a): in econd recontructed -. ignal (8 poin - (b): in.6 firt recontructed.6.7 ignal for. reyag (8.. poin (c):. recontructed ignal (76 poin.. (a) in for reyag econd for reyag recontructed ignal (b)... firt (b) Approximion for reyag in firt for reyag. recontructed ignal (c) recontructed ignal Fig. Recontruction proce voltage ag. recontuction ignal a d approxima appr de o recontuction ignal ignal recontuction recontuction ignal ignal appro approximio.. approxima -.. 8

4 oximion In econd recontruction ignal we. ue third in decompoition ignal and a. zero vector. In ame way, in firt. recontruction ignal i ued the econd - recontructed ignal and a zero vector with the ame ize. In -. recontruction. original... ignal..6 we.7 ue econd recontruction ignal and firt recontuction. ignal. To tudy the accuracy recontruction, we decompoe thi ignal again - and tain tarting and ending - time. Approximion and firt decompoition i -. - hown in fig 6. Detail firt in fig (b) how.8 and.87 for tarting... and.. ending.6.7 time repectively th repreent accuracy imulion. for reyag - In recontruction ignal the third (97 poin and firt (8 poin i ued. Th mean to recontruct the, it i neceary to be.. - tored (8+97=) point f recontuction ignal tion.. for reyag..6.7 imion proximion for reyag for reyag (a) firt (b) firt Fig. Approximion and firt decompoition B. Swell In a imilar way, a well i imuled and i hown in fig 6(a). The ignal i decompoed in three. Detail and in each i hown in figure 6(b)-6(g). A i mentioned in previou ection, the ignal i recontructed and the reult are hown in 6(h)-6(j) (a) voltage well (c) firt recontuction ignal (e) recontuction recontuction econd ignal ignal for reyag for reyag for reyag (b) firt (d) econd (f) recontuction ignal third (i) for reyag in firt (j) recontructed ignal. recontructed ignal Fig. 6..Decompoition voltage.7.ag. in.three and recontuction ignal Recontruction proce recontuction ignal -. To tudy the accuracy method, the recontructed ignal i decompoed again and Approximion and firt decompoition i hown in fig 7. Table.. II. repreent reult thi.. cae recontuction ignal for reyag. for reyag (a) firt (b) firt Fig. 7 Approximion and firt decompoition Table II Reult well parameter time Actual tarting time.9. Actual ending time.96. Etimed tarting time before decompoition.6 Etimed ending time before decompoition.9 -. Etimed tarting time after decompoition Etimed ending time after decompoition.89 C. Sag with harmonic and well with harmonic On the bae table (), thee diturbance are imuled and are mapped to wavelet domain. recontuction ignal Decompoition diturbance and recontruction proce are in imilar way. Figure 8(a) and 8(b) have hown thee two diturbance. Detail firt decompoition are - hown if figure 8(c) and 8(d). Similar to wh wa ted in the previou ection,.. diturbance decompoed in three and recontructed again. for reyag (Fig 8(e) and 8(f)). To. verify the accuracy the method, firt are. how in figure 8(g) and 8(h). Our imulion reult are given in Table III. recontuction ignal for reyag a recontuction recontuction ignal ignal recontuction ignal recontuct leve recontuction ignal. Internional Conference on Computer, Sytem and Electronic Engineering (ICSCEE') April 6, Johanneburg (South Africa) (g).... third original.ignal.6.7. (h) for in reyag.7 econd fo for reyag for reyag. recontructed ignal. recontuction ignal for reyag recontuction ignal for reyag for reyag recontuction ignal for reyag for reyag

5 .. Internional Conference on Computer, Sytem and Electronic Engineering (ICSCEE') April 6,. Johanneburg (South Africa) ximion original ignal etail recontuction ignal ion for reyag (a)voltage ag with harmonic (c) firt original. ignal (e) recontruction voltage ag.7.7 with harmonic for reyag.. -. recontuction recontuction ignal ignal - for reyag (b) voltage well with harmonic (d) firt (f) recontruction voltage well with harmonic (g) firt (h) firt Fig. 8. reult voltage ag and voltage well with harmonic TABLE III RESULT OF SAG WITH HARMONIC AND SWELL WITH HARMONIC parameter SAG & HARMONIC SWELL & HARMONIC Actual tarting time.. Actual ending time.99.7 Etimed tarting time before decompoition.. Etimed ending time before decompoition.98.7 Etimed tarting time after decompoition..9 Etimed ending time after decompoition A can be een from three cae, the accuracy thi method i verified by imulion. Thi imulion i repeed everal time and preciion method i confirmed in repetition imulion. VI. CONCLUSION In thi paper power quality diturbance da compreion i handled. The technique are baed on wavelet theory and multi-reolution analyi. Any ditorted event are decompoed into three reolution. By uing da compreion technique, power quality diturbance are recontructed. It can be een th wavelet theory can be ued to reduce ample point da. ACKNOWLEDGEMENTS recontuction.. ignal Thi wor wa upported by Ilamic Azad Univerity, mahhahr Branch, Iran original ignal for reyag recontuction ignal for reyag REFERENCES [] P. K. Dah, B. K. Panigrahi, D. K. Sahoo, and G. Panda, Power Quality Diturbance Da Compreion, Detection, and Claificion Uing Integred Spline Wavelet and S-Tranform, IEEE Tranaction on - Power Delivery, VOL. 8, NO., APRIL recontuction ignal [] S.K. Meher, A.K. Pradhan, G. Panda, An integred da compreion cheme for power quality event uing pline wavelet and neural networ, Electric Power Sytem Reearch 69 () recontuction recontuction ignal ignal [] S. Santoo, E.J Power, W.M Grady, Power quality Diturbance Da Compreion uing Wavelet Tranform Method, IEEE for reyag for reyag Tranaction on Power. Delivery, Vol., No., July for reyag. [] Haibo He, Januz A. Starzy, A Self-Organizing Learning Array Sytem for Power Quality Claificion Baed on Wavelet Tranform, IEEE TRANSACTIONS -. ON POWER DELIVERY, VOL., NO., JANUARY [] Walid G. Mori, M. E. El-Hawary, A New Perpective for the IEEE Standard 9- Via Stionary Wavelet Tranform in the Preence Nontionary Power Quality Diturbance, IEEE TRANSACTIONS ON POWER DELIVERY, VOL., NO., OCTOBER 8 [6] Xin-ling wen, Yu chen, meaurement voltage fluctuion and flicher in electric power ytem baed wavelet tranform, Proceeding the 7 Internional Conference on Wavelet Analyi and Ptern Recognition, Beijing, China, Nov. 7 [7] C. S. Burru, R. A. Gopinh, and H. Guo, Introduction to Wavelet and Wavelet Tranform, A Primer. Englewood Cliff, NJ: Prentice, 998. [8] Mh H.J.Bollen, Undertanding Power Quality Problem Voltage Sag and Interruption, IEEE Pre Serie on Power Engineering [9] S. Santoo, E.J Power, W.M Gray, P. Hmann, Power aement via wavelet tranform analyi, IEEE Tranaction on power Delivery Vol., No, Apr.996 Saeed Mohammadzadeh wa born in 98 in Lahijan, Iran. He received B.Sc. degree in power electrical engineering from Guilan Univerity, Iran, in 6, and M,,Sc. degree in power electrical engineering from Shahid Chamran Univerity, Iran, in 9. He i currently with department Electrical Engineering, Mahhahr branch, Ilamic azad univerity, mahhahr, Iran a lecturer. Hi current reearch interet include power quality Detection and Control. S. Gh. Seifoad wa born in Ahvaz, Iran, in 96. He received the B.S.c degree in electrical engineering from the Iran Univerity Science and Technology (IUST), Tehran, Iran in 989 and the M.S.c degree in electrical engineering from Ferdoi Univerity Mahhad, Mahhad, Iran, in 99 and the PH.D. degree from IUST, Tehran (Iran) in 6. Currently, he i with the department electrical engineering Shahid Chamran Univerity Ahvaz, where he ha been there icnce 99. Hi reearche interet are power electronic, protection relay, power quality and electric machinery. Motafa ahmadzadeh wa born in Shiraz, Iran, in 98. He received the B.S. degree from yazd univerity,yazd, iran in 6 and M.Sc. degree from chamran univerity, ahvaz, iran in power engineering in 9. He i currently with department Electrical Engineering, Mahhahr branch, Ilamic azad univerity, mahhahr, Iran a lecturer. hi reearch interet include tability power ytem, deign FACTS device and power quality. 8