Tme-frequency Analyss Based State Dagnoss of Transformers Wndngs under the Short-Crcut Shock YUYING SHAO, ZHUSHI RAO School of Mechancal Engneerng ZHIJIAN JIN Hgh Voltage Lab Shangha Jao Tong Unversty 800 Dongchuan Road, 200240, PR. CHINA zsrao@sjtu.edu.cn Abstract: - How to effectvely dentfy the looseness or deformaton of wndngs s mportant to keep the safety of transformers operaton. The short-crcut experments have been done on the large power transformer of 50000KVA. The vbraton ntensty change s used to ntally detect the wndngs' state by onlne montorng the vbraton sgnal of the shocks. Then the tme-frequency analyss further processes these sgnals n order to recognze the fault nformaton contaned. Ths study fnds that the tme-frequency analyss can be used to more precsely detect the fault characterstcs of wndngs. Compared wth tradtonal off lne short crcut reactance method, the tme-frequency analyss s more effectve. Key-Words: - Transformer wndng, tme-frequency analyss, Vbraton 1 Introducton Transformers play an mportant role n the safe operaton of power system. Durng ts operaton, the transformer often suffers from the short-crcut shocks, whch causes the looseness of the wndngs clampng pressure, eventually leadng to collapse of the wndngs[1~ 4]. Therefore, t s most mportant to contnuously montor and assess the condton of the wndngs to ensure relablty of transformers operaton[5]. The tradtonal methods to dentfy the wndngs deformaton nclude the short-crcut reactance method[6] and frequency response analyss[2, 7], both of whch adopt electrcal measurement methods. Based on the dstrbuted parameter transfer model of transformer s wndngs, these methods dentfy the wndngs deformaton by the change of dstrbuted parameter or leakage reactance. However, when the wndng undergoes a small deformaton, especally when the axal compacton force decreases, whle the electrcal parameter doesn t change, these methods are not effectve n detectng the real status. The looseness or deformaton of wndngs essentally affects the mechancal characterstcs of transformers body[8], whch can be dentfed and analyzed by measurng the related vbraton sgnals. The short-crcut shock experments have been done on the large power transformer of 50000KVA. The vbraton ntensty and the tme-frequency analyss are used to detect whether wndng s n fault. Ths study fnds that these methods are more effectve to montor the state change of wndngs than the short crcut reactance method. 2 Short-crcut Shock Tests of Transformer Wndngs The expermented transformer s rated capacty s 50000KVA and rated voltage s 110KV. The short-crcut shock tests have been conducted accordng to the standards[9] of GB 1094.5-85 or IEC 76-5: 1976. To begn wth, the low-voltage sde of the transformer s short-crcuted. Then the short-crcut shock occurs three tme separately for every phase A, B, and C wndngs. Every tme, the exctng current keeps constant and the shock tme lasts 0.25s. Meanwhle, the vbraton sgnal caused by the shock s obtaned by the accelerometers whch are attached on the transformer tank of the low-voltage sde. The measurng ponts are located as shown n Fg. 1. ISSN: 1790-5117 283 ISBN: 978-960-474-054-3
C B A c b a o 3# 2# 1# Fg. 1 The locaton of measurng ponts Takng measurng ponts 3# for phase C as an example, Fg. 2 shows ts tme-doman sgnals under the three tmes shocks. From ths fgure, t can be seen that the vbraton response keeps a rather large vbraton level durng the load perod of short-crcut current, whle t gradually decays wth the dsappearance of the short-crcut current. However, t s dffcult to tell the dfferences of the three tmes shocks from ths fgure. Therefore, the further sgnal analyss s needed. Ampltude (m/s 2 ) 100 50 0-50 1st shock 2nd shock 3rd shock -100 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Tme (s) Fg 2. The tme-doman comparson of measurng ponts 3# of phase C under the three tmes shocks 3 Analyss of Expermental Data 3.1 Vbraton ntensty When the wndngs are n workng order, ther vbraton responses have no bg dfference under the shock of the same short-crcut current. However, when there exsts the clampng pressure looseness or dstorton of the wndngs, the change of ther structure characterstcs would be shown n the vbraton response[10]. Therefore, the changes of wndngs vbraton level can be used to detect the state of the wndngs by the on-lne montorng of the wndngs. To state the vbraton of the transformer under the shock of short-crcut more clearly, we ntroduce vbraton ntensty as follows: k m n rms rms rms ax ay az = 1 2 = 1 2 = 1 2 a d = ( ) + ( ) + ( ) k m n Where, a, a yrms, a zrms the vbraton level of xrms measurng pont n the drecton of x, y, z K, m, n the number of measurng ponts n the drecton of x, y, z From the above formula, the vbraton ntensty could be derved, as shown n Fg. 3 and Table 1. These comparsons show that the vbraton ntenstes of phase B change small under the three tmes short-crcuted shock, whch means the wndng of phase B has suffcent mechancal strength to wthstand the short-crcut shock. However, the vbraton ntenstes of phase A and C change obvously. Especally when they suffer from the second shock, the vbraton ntenstes ncrease more than 14%. Ths tells that the states of low-voltage wndngs of phase A and C change under the shock, wth the ncreasng tendency. vbraton ntensty 16 14 12 10 8 6 4 2 0 1st shock 2nd shock 3rd shock shock for phase A shock for phase B shock for phase C Fg. 3 The comparson of vbraton ntensty Table 1 The comparson of vbraton ntensty Phase A Increasng rato Phase B Increasng Phase C Increasng rato rato Shock 1 9.5566 6.6516 12.4995 Shock 2 10.9223 14.29% 6.7752 1.86% 14.6642 17.32% Shock 3 11.64 21.80% 6.7297 1.17% 15.2617 22.10% Accordng to the requrement of GB 1094.5-85 or IEC 76-5: 1976, the reactance of every wndng s measured after every shock. The detaled results are as shown n Table 2. ISSN: 1790-5117 284 ISBN: 978-960-474-054-3
Table 2 The comparson of reactance change Phase A Phase B Phase C Gap of Gap of Gap of Before Experment 46.94 37.08 28.8 Shock 1 47.01 0.15% 37.09 0.03% 28.79-0.03% Shock 2 47.01 0.15% 37.09 0.03% 28.67-0.45% Shock 3 47.01 0.15% 37.09 0.03% 28.56-0.83% Note: The gap of reactance s the devaton percentage of the reactance after shock from the value before the shock. The comparson of reactance shows that the reactance of phase C decreases gradually after the three tmes shocks and the gap approaches -1%, whch means the state of the wndng of phase C changes after the shocks. However, the reactance of wndngs of phase A and B doesn t change after the shocks. From the opnon of the short-crcut reactance method, these two wndngs are normal. The short-crcut method and vbraton ntensty method have reached the same concluson on the states of phase B and C, not on phase A. Therefore, the tme-frequency analyss s used to analyze the vbraton sgnal of phase A and C, n order to study the reason of the dfference and the wndngs fault characterstcs. 3.2 Tme-frequency analyss for the short-crcut shock sgnal Because the nsulaton cardboard of the wndngs s nonlnear materal, the vbraton response caused by the exctng current of 50Hz s manly conssted of harmonc components of 100Hz and 50Hz. Therefore, the requrement for the frequency resoluton precson s not hgh durng the tme-frequency analyss. Then Gabor transform s employed for the sgnal process. Fg. 4~7 show the tme-frequency dagrams of phase A and C. Fg 4 The tme-frequency dagrams of Measurng pont 1# of phase A for the short-crcut shocks ISSN: 1790-5117 285 ISBN: 978-960-474-054-3
Fg 6 The tme-frequency dagrams of Measurng pont 1# of phase A for the short-crcut shocks (50Hz) Fg 5 The tme-frequency dagrams of Measurng pont 3# of phase C for the short-crcut shocks Fg 7 The tme-frequency dagrams of Measurng pont 3# of phase C for the short-crcut shocks (500Hz) From the above dagrams, we can see that the vbraton level of phase A maly changes n the ISSN: 1790-5117 286 ISBN: 978-960-474-054-3
frequency of 100 Hz, where the component ncreases obvously wth every shock (as shown n Fg. 6). However, there s nearly no change for phase C n terms of the components of 50Hz and 100 Hz. The dfference exsts maly n the scope of 450Hz~ 600Hz, especally n the frequency of 500Hz. Wth every shock, the component of 500Hz not only changes obvously, but ts peak shft obvously (as shown n Fg. 7). For example, the maxmal response peak of the thrd shock appears n the 0.05s after the short-crcuted current, whch s dfferent wth the normal state after the shock and means the wndng of phase C s n wrong state. And the reactance method also detects ths problem. In addton, from the above fgures t can be seen that the fault characterstcs of phase A and C are not the same. 100 Hz s the exctng frequency of electromagnetc force. When the short-crcut current keeps constant, the response ampltude of 100Hz keeps no change for the normal wndngs. However, the response ampltude of 100Hz for phase A ncreases gradually, whch means that the wndng of phase A goes wrong [11]. The fault may be because of the looseness of the clampng pressure, whch causes the shft of the natural frequency, fnally resultng n the change of response ampltude of 100Hz. Therefore, these experments tell that the onlne vbraton ntensty method can show the change of wndngs state, more precsely than the reactance method. Meanwhle, based on ths ntal judgment, the tme-frequency analyss can further precsely show the fault characterstcs. 4 Concluson From the study of the short-crcut shock test for ths large power transformer of 50000KVA, we can get the followng conclusons: 1. Vbraton ntensty can be used to on-lne montor the wndngs states. Compared wth the off-lne short-crcut reactance method, ths method s more effectve and tmely to show the change of the wndngs states. 2. Based on the judgment of vbraton ntensty, tme-frequency analyss can be used to more precsely show the fault characterstcs and the reasons of wndngs states change by analyzng the vbraton sgnals. n power transformers, Proc. 68th Annu. Int. Conf. Doble, 2001. p. Clents: Sec 8\\\ 3. [2] Wang, M. Vandermaar, A.J. Srvastava, K.D., Transformer wndng movement montorng n servce - key factors affectng FRA measurements, Electrcal Insulaton Magazne, IEEE, Vol. 20, No. 5, 2004, pp. 5-12 [3] A.C. Frankln, D.P. Frankln, The J & P Transformer Book, Butterworths, 1983 [4] Wang, M. Vandermaar, A.J. Srvastava, K.D., Revew of condton assessment of power transformers n servce, Electrcal Insulaton Magazne, IEEE, Vol. 18, No. 6, 2002, pp. 12-25 [5] C. Booth* and J. R. McDonald, The use of artfcal neural networks for condton montorng of electrcal power transformers, Neurocomputng, Vol. 23, No. 1-3, 2004, pp. 97-109 [6] A. Babare et al., Ennel-dagnosss of on- and off-lne large transformers, Proc. Cgré Symp., Berln, 1993, pp. 110-04. [7] Dck, E.P. Erven, C.C., Transformer Dagnostc Testng by Frequency Response Analyss, IEEE Transactons on Power Apparatus and Systems, Vol. PAS-97, No. 6, 1978, pp. 2144 2153 [8] Patel, M.R., Dynamc response of power transformers under axal short crcut forces PartII: Wndngs and clamps as a combned system, IEEE Transactons on Power Apparatus and Systems, Vol. PAS-92, No. 5, 1973, pp. 1567-1576 [9] CHEN Ku, Short-Crcut Test method of transformer and ts development, TRANSFORMER, Vol. 37, No. 1, 2000, pp. 49~53 [10] C. K MECHEFSKE, Correlaton power transformer tank vbraton characterstcs to wndng looseness, Insght- J. Non Destruct. Test. Cond. Montor, Vol. 37, No. 8, 1995, pp. 599-604 [11] Garca, B. Burgos, J.C. Alonso, A.M., Transformer tank vbraton modelng as a method of detectng wndng deformatons-part I: theoretcal foundaton, Power Delvery, IEEE Transactons on, Vol. 21, No. 1, 2006, pp. 157-163 References: [1] V. Sokolov and B. Vann, Experence wth detecton and dentfcaton of wndng bucklng ISSN: 1790-5117 287 ISBN: 978-960-474-054-3