Available online at www.ciencedirect.com ScienceDirect Procedia Technology 1 (15 ) 589 595 SMART GRID Technologie, Augut 6-8, 15 An Optimization Algorithm for Voltage Flicer Analyi Athira S a*, Hariumar K a, Soman K P a, Prabaharan Poornachandran b a Center for excellence in Computational Engineering and etworing, Amrita Vihwa Vidyapeetham b Amrita Centre for Cyber Security ytem and etwor,amrita Vihwa Vidyapeetham Abtract Maintenance of power quality tandard i a critical iue in the electrical ditribution networ today. Proper quality can be maintained only by continuouly monitoring and analyzing the ignal of interet o a to identify the ource of ditortion. Thi paper preent an algorithm for analyi of voltage flicer in power ignal, a common phenomenon in mot of the networ involving heavy load. The low frequency component reponible for modulating the power ignal i extracted uing an optimization algorithm. Further the algorithm eparately identifie the ignal affecting the fundamental component and the harmonic preent in the power ignal. 15 Publihed The Author. by Elevier Publihed Ltd. by Thi Elevier i an open Ltd. acce article under the CC BY-C-D licene (http://creativecommon.org/licene/by-nc-nd/4./). Peer-review under reponibility of Amrita School of Engineering, Amrita Vihwa Vidyapeetham Univerity. Peer-review under reponibility of Amrita School of Engineering, Amrita Vihwa Vidyapeetham Univerity Keyword: Flicer; low freqency; envelope; optimization. 1. Introduction Power quality indicate the qualitative nature of power ignal in term of frequency, phae, amplitude and the harmonic preent [1]. In the preent day cenario, the load nature ha changed ignificantly and hence maintenance of power quality require better method for their analyi and maintenance []. With the advent of mart grid networ there i the preing need for maintenance of power quality throughout the networ; and hence efficient analyi methodologie that can identify the exact reaon and ugget compenation for the ame. A precie compenation for the diturbance can be introduced only by arriving at an etimate of the component involved after a detailed tudy of the ditortion introduced. Voltage flicer are one of the very common power quality diturbance frequently oberved in all networ, epecially in networ involving heavy load [3], [4]. * Correponding author. Tel.97944639 E-mail addre:athiraubhah@outloo.com 1-173 15 Publihed by Elevier Ltd. Thi i an open acce article under the CC BY-C-D licene (http://creativecommon.org/licene/by-nc-nd/4./). Peer-review under reponibility of Amrita School of Engineering, Amrita Vihwa Vidyapeetham Univerity doi:1.116/j.protcy.15.1.64
59 S. Athira et al. / Procedia Technology 1 ( 15 ) 589 595 Flicer refer to the low variation in voltage value of power ignal over time. Such unexpected variation produce evident diturbance in the ytem and damage device in the circuit. Flicer variation are typically reflected in the ignal a a low frequency envelope modulating the power ignal. Thu, the problem of flicer analyi can be conidered a the tudy of thi low frequency envelope to identify the ource for thi modulation. An algorithm that can effectively extract thi low frequency modulating portion of the ignal can facilitate the proper analyi of uch voltage flicer. Several uch method exit in literature, which vary by the method they adopt for the ditortion component etimation. One of the traight approache to the problem i the ue of Fourier tranform to etimate the low frequency component preent in the ignal and thu identify the potential ource for the ditortion. Thi method fail to deliver exact etimation owing to non-tationary nature of ignal and leaage effect [5]. Several method baed on Hilbert tranform [6] and wavelet tranform [7] were alo propoed for the analyi of thee power quality diturbance. Thee pectrum baed method cannot provide high time and frequency reolution at the ame time, and further are not capable of identifying the component with high reolution a i required for the exact analyi ofpower ignal. [8] dicue a tochatic load flow analyi to be applied for analyi of power ignal for heavy load uch a arc furnace. Thi paper propoe an algorithm for flicer analyi via an optimization framewor, which extract the low frequency component modulating the fundamental component of the power ignal a well a the harmonic preent in the ignal eparately. Although the optimization framewor would demand more computation a compared to the conventional method, the method offer better reult. The ret of thi paper i organized a follow. Section II decribe the optimization problem formulation and the Section III dicue the reult obtained for the algorithm applied to a ynthetic data et and finally the concluion i drawn in Section IV.. Problem Decription Conider an appropriately ampled power ignal x R taen a x [ x() x(1)... x( 1)] T coniting offundamental frequency f R Hz and it harmonic. Here we conider x to be um of two ignal x v w, n where v R i the fundamental component modulated with a low frequency ignal un ( ) Rto imulate the effect of flicer ditortion and i expreed a fn vn ( ) un ( )co, (1) f w R i the um of harmonic of f taen to be K 1 f n i wn ( ) co () i1 f n [,1,,..., 1] Z and f R i the ampling frequency. Our aim here i to extract the low frequency variation un ( ). A per the contruction of the power ignal decribed in (1) and () it original component arepure ingletone of nown frequencie, and hence the ample of the modulating component un ( ) can be conidered a the time varying multiplier for the original ingletonecomponent. Hence the ignal x can be repreented a K 1 xn ( ) ancn (, ) (, ) bnn (, ) (, ) (3) Where cn (, ) co( fn/ f) and n (, ) in( fn/ f). Thi repreentation i quite imilar to the decompoition of ignal uing K frequency bin a preented in [9]. An optimization frame wor that could aborb the flicer variation and the original power ignal variation into eparate component uing the model (3) i expreed a follow:
S. Athira et al. / Procedia Technology 1 ( 15 ) 589 595 591 min K 1 ab, o K 1 Da t.. xn ( ) ancn (, ) (, ) bnn (, ) (, ), Db (4) K where acb,,, R, 1 1 1 1 D (5). 1 1 Da and Db repreent the firt order difference of a and b where a, c, b, R an [1,,..., K 1] Z.. K The idea behind thi formulation i very imilar to the concept preented in [9], but here the objective ha been framed differently to achieve the optimization of interet. The formulation decribed in [9] minimize um of the 1 norm of Da and Db for all Z. Such an optimization would drive a and b to be piecewie contant ignal [9], a oppoed to our preent requirement of extracting mooth low frequency envelop and hence we minimize um of the norm of Da and Db for all Z.If the exact fundamental frequency f i nown, then c and can be choen properly o that a and b capture the flicer envelope. 3. Reult and Dicuion For the purpoe of analyi, a tet ignal of the form K 1 f n f n f n l xn ( ) co A co co (6) f f f i1 wa generated a per the model decribed by (1) and (), where f i the frequency of flicer component and l A Ri it correponding DC hift. Such a ignal of length 14 { f, f, f } {1,15, 5}, K 4 and n Z 1 3 i hown in Fig.1. taen at f 14 with f, f 5 Hz, f n l The flicer component co f and the fundamental component fn co are eparately hown in Fig.1 (a) f and (b) repectively. The ignal wa ubjected to algorithm (3) and correponding reult are hown in Fig. which ugget that the algorithm could exactly capture the flicer variation a the vector a. The recovered component are hown in Fig. 3.Precie election of frequency bin and the number of frequency bin i required for the good performance of the algorithm; in cae where the exact component of the original ignal i unnown a reliable frequency etimation technique need to be performed prior to analyi for deciding the frequency bin [1], [11], [1]. The algorithm wa alo found to be capable of removing different low frequency component affecting harmonic. Thi wa teted on an input ignal (different from (6)) of the form l
59 S. Athira et al. / Procedia Technology 1 ( 15 ) 589 595 Fig. 4(c) how uch a ignal of length 14 taen at f 14 with f, f 1, 1 K 1 f n f n f n fn f n l l1 1 xn ( ) co A co co A co co (7) 1 f f f f i f f n l { f, f, f, f } {5,1,15, 5}, K 4. The flicer component co 1 3 f and fn l 1 co are eparately f hown in Fig.4 (a)and (b)repectively and A, A R are their correponding DC hift. 1 l l Fig.1.(a) Flicer component (b) Original power ignal (c) Power ignal with flicer. Fig.. Flicer component eparated from (a) fundamental component (b) econd harmonic (c) third harmonic (d) fifth harmonic
S. Athira et al. / Procedia Technology 1 ( 15 ) 589 595 593 Fig. 3 Signal component eparated from (a) fundamental component (b) econd harmonic (c) third harmonic (d) fifth harmonic Fig. 4 Flicer component effecting the (a) fundamental component (b) econd harmonic (c) Input ignal
594 S. Athira et al. / Procedia Technology 1 ( 15 ) 589 595 Fig. 5 Flicer component recovered from (a) fundamental component (b) econd harmonic (c) Third harmonic (d) Fifth harmonic Fig. 6 Recovered ignal component correponding to (a) fundamental component (b) econd harmonic (c) Third harmonic (d) Fifth harmonic The reult flicer component and ignal component are hown in Fig. 5 and Fig. 6 repectively. The reult how that the algorithm wa capable of identifying thee cloe low frequency component affecting the harmonic. 4. Concluion The propoed algorithm i capable of effectively eparating the low frequency flicer component modulating the power ignal fundamental component and it harmonic. The flicer component effecting each harmonic i eparately available uch that with good frequency reolution technique the exact frequency reponible for the flicer can be preciely determined thu facilitating detailed analyi and identification of their ource in the electrical power ditribution networ..
S. Athira et al. / Procedia Technology 1 ( 15 ) 589 595 595 Acnowledgement The author would lie to than all at the Center for Excellence in Computational Engineering and etworing for their direct and indirect upport for thi wor. Reference [1] M. H. Bollen, Undertanding power quality problem. IEEE pre ew Yor,, vol. 3. [] V. Muolino, L. Piegari, E. Tironi, L. Martini, A. Brambilla, and C. Tornelli, Simulation and field tet reult for potential application of lv dc ditribution networ to reduce flicer effect, in Harmonicand Quality of Power (ICHQP), 1 14th International Conference on. IEEE, 1, p. 1 6. [3] Mettam GR, Adam LB.How to prepare an electronic verion of your article. In: Jone BS, Smith RZ, editor. Introduction to the electronic age. ew Yor: E-Publihing Inc; 1999. p. 81-34. [4] A. Roin and T. Tehnia uliool, Control, Superviion and Operation Diagnotic of Light Rail Electric Tranport. Tallinn Univerity of Technology Pre, 5. [5] A. A. Girgi, J. W. Stephen, and E. B. Maram, Meaurement and prediction of voltage flicer magnitude and frequency, Power Delivery, IEEE Tranaction on, vol. 1, no. 3, pp. 16 165, 1995. [6] R. C. Dugan, M. F. McGranaghan, and H. W. Beaty, Electrical power ytem quality, ew Yor, Y: McGraw-Hill, c1996, vol. 1, 1996. [7] F. Mui no, M. Carabajal, M. Morvidone, C. D Attelli, and M. Fabbro, Detection of voltage flicer in noiy ignal. [8] M. Manjula and A. Sarma, Comparion of empirical mode decompoition and wavelet baed claification of power quality event, Energy Procedia, vol. 14, pp. 1156 116, 1. [9] C. Chen, H. Chuang, C. Hu, and S. Teng, Stochatic voltage flicer analyi and it mitigation for teel indutrial power ytem, in PowerTech Proceeding, 1 IEEE Porto, vol. 1. IEEE, 1, pp. 5 pp. [1] Y. Ding and I. W. Selenic, Spare frequency analyi with pare derivative intantaneou amplitude and phae function, arxiv preprint arxiv:13.653, 13. [11] F. Anderon, M. Carlon, J.-Y. Tourneret, and H. Wendt, A new frequency etimation method for equally and unequally paced data, 14. [1] J. F. Hauer, C. Demeure, and L. Scharf, Initial reult in prony analyi of power ytem repone ignal, Power Sytem, IEEE Tranactionon, vol. 5, no. 1, pp. 8 89, 199. [13] T. Lobo, Z. Leonowicz, J. Rezmer, and P. Schegner, High-reolution pectrum-etimation method for ignal analyi in power ytem, Intrumentation and Meaurement, IEEE Tranaction on, vol. 55, no. 1, pp. 19 5, 6.