RESEARCH ON RESPONSE OF ROGOWSKI COIL ELECTRONIC CURRENT TRANSFORMER TO TRANSIENT SIGNAL

Transcription

1 23 rd International Conferene on Eletriity Distriution Lyon, June 2015 RESEARCH ON RESPONSE OF ROGOWSKI COIL ELECTRONIC CURRENT TRANSFORMER TO TRANSIENT SIGNAL Jian fei JI Yuan Yu BO Qiang Sheng BU Jiangsu Provinial Power Company Eletri Power Researh Institute China ABSTRACT During the running proess, Rogowski oil ECT will ativate protetion malfuntion due to anormal waveform aused y transient signal whih is generated y swith operation or short-iruit fault. This paper uilds a Rogowski oil ECT model, analyzing signal transformation progress etween eah link in the transformer, and onduting a simulation experiment as well. Both theoretial analysis and simulation experiment results show that, transient signal waveform is distorted in the integration aumulation operation progress after differential and disrete sampling y Rogowski oil ECT. Beause transient disturane signal, whih aused y swith operation or short-iruit fault, ontains highfrequeny omponents, while the olletor sampling rate is low. Aording to this question, an improvement measure is proposed that using fourth order Bessel Filter to suppress high-frequeny signals, applying olletor to omplete software integration, and improving A/D sampling rate to 100kHz, in whih way to help improve running reliaility of Rogowski oil ECT. INTRODUCTION In intelligent sustation, eletromagneti disturane aused y swith operation or short-iruit fault an lead to anormal operation of Rogowski oil ECT, thus ausing inorret operation of relay protetion devie. And this eome the most diffiult prolem in atual operation and key fator affeting its widely appliation. It also eomes the fous of attention. The experiene of eletroni transformer operation still needs aumulated. Relevant system files and aliration standards need further optimized. While the operating experiene of onventional transformer in sustations with various voltage levels has already een mature. Before the eletroni transformer getting mature, all the new estalished sustations should apply onventional transformer with merging unit to ahieve on-site analog digital onversion, using fier upload, in whih way not only improve the antiinterferene and reliaility of signal transmission, ut also redue the onfiguration amount of transformer seondary winding. So as to redue the transformer volume and imporve its reliaility. stipulated in 2011 [58]th supplementary doument signed y Infrastruture Department of State Grid Corporation. Researhers all over the world have done a lot of studies [1]-[13] on Rogowski oil ECT. But, urrently, there are few analyses on Rogowski oil ECT anormal work aused y eletromagneti disturane. Ref[14] disusses aout the impat of integral part on transformer transient harateristis. Aove referenes do not study in detail on the transmission progress of Rogowski oil ECT etween eah link, neither point out the primary reason of generating anormal waveform. This paper will uild a Rogowski oil ECT model, analysing signal transmission etween eah link in transformer and influene mehanism of Rogowski oil ECT differential and integral proess on transient eletromagneti disturane signal, therey revealing the generating reason of anormal waveform in Rogowski oil ECT, and finally proposing an improvement measure of using fourth order Bessel Filter to suppress high-frequeny signals, applying olletor to omplete software integration, and improving A/D sampling rate to 100kHz, in whih way to help improve the anti-eletromagneti interferene aility of Rogowski oil ECT. ROGOWSKI COIL ELECTRONIC TRANSFORMER MODEL Power Soure I differential proess Rogowski oil ECT (eletroni urrent transformer) R1 Rogowski oil Colletor sampling proess Digital Integrator Integral proess Output Fig.1 Model of Rogowski oil ECT (eletroni urrent transformer) Comparing with onventional eletromagneti urrent transduers, the priniple of Rogowski oil eletroni urrent transformer is to measure urrent y differential and integral. As shown in Fig.1, Rogowski oil eletroni urrent transformer is onsisted of Rogowski oil (differential proessing), olletor (sampling proessing) and digital integrator (integral proessing). Differential Proess Rogowski oil output indued emf equation: it () e() t 0 N A (1) Where it () represents primary urrent, 0 represents vauum magneti permeaility, represents turns density, represents single turn area. As equation (1) shown, Rogowski oil transfer harateristis an e regarded as differential on primary signal. From frequeny domain point of view, this proess only hanges the amplitude and phase of the signal, ut no effet on the frequeny of the original signal. If the primary urrent ontains a large numer of higher harmoni, then Rogowski oil output also ontains a lot of higher harmoni, and the amplitude of high-frequeny omponent A N CIRED /5

2 23 rd International Conferene on Eletriity Distriution Lyon, June 2015 is muh larger than that of power-frequeny omponent. Sampling Proess To get all information of the signal and restore it in susequent proessing, sampling proess should meet the Nyquist Sampling Theory, whih means sampling frequeny should e greater than twie the highest frequeny omponent of the signal, otherwise, frequeny aliasing will our, resulting in inorret restoration of signal. Atually, eletromagneti disturane aused y swith operation or short-iruit fault ontains a lot of high-frequeny omponents whose frequeny is more than half the sampling frequeny, whih auses the frequeny aliasing in sampling proess, generating a lot of DC omponents and tail urrent phenomenon. For illustration, the onept of Sampling Instant Offset is introdued to show the influene of low sampling frequeny on signal transmission, and remark it as, as Fig.2 shown. Fig.2 Sketh map for sampling point offset In Fig.2, the solid lines and dotted lines represent two sampling instants for one signal. Although two sampling frequeny are the same, the sampling instants on the time axis are different. And this fixed time interval etween two instants is alled sampling instants offset, marked as. When sampling frequeny meets the Nyquist Sampling Theory, the onept of sampling instant offset will no longer have pratial signifiane, as the signal an e orretly transmitted and finally restored, no matter whih signal instant is sampled. Whereas, when sampling frequeny doesn t meet the Nyquist Sampling Theory, the onept of sampling instant offset does have pratial signifiane. As frequeny aliasing ours, the generated signals in different sampling instants will e in ig differene. So for the same primary urrrent, the final output urrent waveform will e tremdously different with the hange of t. To orretly analyze the signal hange in transmission proess in transformer, the influene of sampling instant offset an not e ignored. This paper uilds the final model as Fig.3 for analysis with adding time delay link efore sampling to simulate this phenomenon. With this model, when sampling frequeny is onstant, the orrespondent sampling instant is fixed. In order to simulating sampling instant offset, the signal waveform from a will e translated on the time axis y time delay, so that to get different point in sampling link y hanging time delay. Unless otherwise stated, the signal in susequent analysis in this setion refers to single frequeny. Primary urrent Rogowski Coil a Time delay Sampling Integration Output Fig.3 Model for analysis Firstly, with figure 3, the transfer harateristis of time delay link on signal waveform will e analyzed. As mentioned aove, time delay only translate signal waveform in part a on the time axis, and it does not hange signal frequeny and amplitude. But it hanges the signal phase for the following sampling proess. Supposed that signal frequeny in part a is, the signal translation time in time delay link is sampling instant offset, marked as, the signal waveform phase in part a and respetively laelled as and, then it an e given y: 1 θ = θa + fa 2π = θa + 2πf aδt (2) Δt Then, under the ondition that sampling proess does not meet the Nyquist Sampling Theory, the transfer harateristis of sampling proess on signal waveform will e analyzed with Fig.3. Different from time delay, sampling proess has little effet on signal amplitude, ut it an hange signal frequeny and phase. Supposed signal frequeny in part as, sampling frequeny as SF, the signal frequeny after sampling proess in part as f, then f an e otained y formula (3) (4): Suppose m as aliasing times: f f θ a f a θ n Int( 0.5) (3) SF f = f -nsf (4) f 2 f m Int( ) Int( ) SF 2 SF (5) Where Int() represents round numers reservation only, that is to say all figures after the deimal point should e disarded. Unit:Hz f Unit:1000Hz Fig.4 Diagram for frequeny alias Supposed signal phase from part in Fig.3 as θ, sampling frequeny as SF, sampled signal phase from part as θ : f CIRED /5

3 23 rd International Conferene on Eletriity Distriution Lyon, June 2015 m m 1 ( 1) ( 1) (6) 2 2 For example, SF=4000Hz, then we an figure out that all signal frequeny after sampling is not more than 2000Hz and signal frequeny aliasing is regular periodi. (as Fig.4) As shown in Fig.4, six setions on horizontal axis, whih are divided y dotted line, are used to explain the phase variation of signals with different frequeny. The orrespondent aliasing times m are 0~5 when the signals from setions (0)~(5)generate frequeny aliasing. Aording to formula (6), when figure is in the range of setions (0)(2)(4), the signal phase from and d are equal. When figure is in the range of setions (1)(3)(5), f the sum of the signal phase from and d is π. Finally, ased on the aove analysis, onsidering signal from part a in fig.3 ontains various frequeny omponents, we take fundamental frequeny as example to illustrate signal transfer proess from to. In Fig.3, signal ontained fundamental sequene from part an e onsidered as superposition of two parts. One part is the original ontained fundamental sequeny from part a, whih is less than half the sampling rate and undistorted after transferred to part. While the other part is distorted from high-frequeny signal, suh as frequeny in Fig.3 that all signal frequeny, like 3950Hz, 4050Hz and 7950Hz eome 50Hz after sampling proess. Integral Proess f () Fig.6 Signal waveform after sampling and output waveform of integrator As Fig.5 shown, supposed that high-frequeny disturane signal is added to primary urrent. The ontinuous urve in Fig.6 represent the differentiated waveform of primary urrent whih is superimposed with high-frequeny disturane signal, and the lak loks represent sampling instants of the olletor. It an e seen from Fig.6(a) that the output waveform of integrator is not oviously distorted, as relatively ig value of the highfrequeny disturane signal is not olleted. While from Fig.6(), it an e seen that the output waveform of integrator is oviously distorted, as ig value of the highfrequeny disturane signal is olleted. IMPROVING METHOD OF ROGOWSKI COIL ECT SOFTWARE INTEGRATION AND SAMPLING SYSTEM Primary urrent Rogowski Coil fourth order Bessel filter Sampling (FS=100kHz) Integration Output Fig.5 Superimposed waveform of primary urrent and high-frequeny pulse disturane signal (a) Fig.7 Shemati diagram for rogowski oil eletroni transformer (fourth order Bessel filter and Sampling rate FS=100kHz) To solve the frequeny aliasing prolem of the higher harmoni integration, the improvement measure is proposed that using fourth order Bessel Filter to suppress high-frequeny signals, applying olletor to omplete software integration, and improving A/D sampling rate to 100kHz, in whih way to help improve running reliaility of Rogowski oil ECT, as Fig.7 shown. Bessel filter Among the ommon three lowpass filters, Butterworth type, Cheyshev type and Bessel type, only Bessel filter an make least distortion for input and output waveform, as its phase is linear with frequeny in the speifi frequeny range. But Bessel s frequeny response is worse than Butterworth s and Cheyshev s of the same order. In order to get etter frequeny response, using high-order Bessel filter is neessary. Aording to eletroni transformer appliation requirement, the utoff frequeny of lowpass filter is hosen to e 2kHz, and the zero frequeny group delay of fourth order Bessel filter is less CIRED /5

4 23 rd International Conferene on Eletriity Distriution Lyon, June s than.the amplitude-frequeny harateristis of the fourth order Bessel filter is etter than that of the seond order RC lowpass filter. Within passand, the phase-frequeny harateristis is linear and the normalized group delay is onstant. Improving sampling rate to 100kHz The effiient signal frequeny looped A/D sampling is ontrolled within 15kHz after through fourth order Bessel filter. Aording to Shannon Sampling Theory, it is ovious that frequeny aliasing an e eliminated if sampling rate improved to 100kHz.(K=100/15 >2). There are three ommonly used methods for A/D onversion, suessive approximation, sampling and integration, while the latter two onverted speed is etween a few milliseonds and tens of milliseonds, unale to meet the requirement of relay protetion to ECT output harateristi. Thus, using suessive approximation A/D hip with high auray and fast onversion is a etter hoie. Aording to sampling rate alulation, the sampling interval for olletor is T s= s. Triggered timingly y FPGA miroproessor, A/D sampling is ativated to omplete interruption trigger, and inform FPGA to read A/D onversion data. Considering from the amplitudefrequeny harateristis, the approximation degree etween its real harateristi and ideal harateristi eomes higher along with the dereasing of T s value. While T s value is inversely proportional to sampling spot N in a unit time. That is to say, the sampling data eome more when the T s value is smaller, and it rings urden for susequent data alulation and transformation. Simulation experiment The three-phase line short iruit model is uilt (as shown in Fig.9. Three phase short-iruit urrent waveform is shown in Fig.10. At 0.15s, the A phase short-iruit ours. And after 100ms, the A-phase iruit reaker jumped. ()B phase ()C phase Fig. 10 The three phase short-iruit waveform and the output waveform of Rogowski oil ECT The three phase short-iruit waveform and the output waveform of Rogowski oil ECT are shown in Fig.10. The output waveform is distorted, whih is quite different from the original short-iruit urrent waveform. On shortiruit instant, the short-iruit urrent ontains a lot of high frequeny omponents. The sampling frequeny of the olletor is relatively low. After sampling and integrating, the output of Rogowski oil ECT is distorted. After using fourth order Bessel filter and improving sampling rate to 100kHz, the output waveform of Rogowski oil ECT is shown as Fig.11. Bessel filter will filter out the high-frequeny omponents efore signal enter the olletor, then improving sampling rate to 100kHz. And the original waveform will e well restored after integration. Fig.9 Three-pahse line short iruit model (a)a phase (a)a phase ()B phase CIRED /5

5 23 rd International Conferene on Eletriity Distriution Lyon, June 2015 ()C phase Fig.11 The three phase short-iruit waveform and the output waveform of Rogowski oil ECT (using fourth order Bessel filter and improving sampling rate to 100kHz) CONCLUSION In this paper, Rogowski oil ECT model is uilt, and signal transformation etween eah link in transformer is analyzed, simulated analysing the impat mehanism of Rogowski oil ECT differential and integral proess on transient eletromagneti disturane signal. Besides, solutions are proposed finally, and the main onlusions are as follows: 1. The transient eletromagneti disturane signal aused y swith operation or short-iruit fault is a signifiant reason for anormal waveform generated y Rogowski oil ECT. 2. Reduing ut-off frequeny of the filter an derease the amplitude of transferred signal high-frequeny omponents, reduing frequeny aliasing extent after sampling proess, so as to eliminate anormal waveform. 3. Improving sampling rate in sampling proess an enlarge the right transferred signal frequeny range, reduing signal distortion degree, and the output signal tends to e ideal. At the same time, it an e onsidered as filter mathing with sampling. As the simulation experimental results show in this paper, tail urrent an e eliminated when filter ut-off frequeny is 8 harmoni and sampling rate is 10kHz. When the ut-off frequeny eomes 60 harmoni, the sampling rate should e improved to more than 100kHz. 4. Signal delay time on the sene is unontrollale, and its oinident value is a neessary ut not suffiient ondition for generating anormal waveform. When the filter ut-off frequeny mathes relatively well with the sampling frequeny, the variety of signal delay time will have little effet on the final results. 5. It is a feasile improvement method y using fourth order Bessel filter to suppress high-frequeny signal, improving A/D sampling rate to 100kHz and finishing software integration y olletor. And it will help improve the running reliaility of Rogowski oil ETC. Aknowledgments This projet is supported y Natural Siene Foundation of Jiangsu Provine (BK ), National Natural Siene Foundation of China ( ), Natural Siene Foundation of Jiangsu Provine (BK ), the Fundamental Researh Funds for the Central Universities. REFERENCES [1]Li J, Zheng Y, Gu S, et al, 2007, Appliation of Eletroni Instrument Transformer in Digital Sustation, Automation of Eletri Power Systems, vol.37, No.7, [2]Liao Jingsheng, Guo Xiaohua, Zhu Mingjun,et al, 2003, ROGOWSKI COIL CURRENT TRANSDUCER FOR LOW AMPLITUDE CURRENT MEASUREMENT, Automation of Eletri Power Systems, vol.27, No.2, [3]Luo C, Zhang G, Wang P, 2007, Eletroni Type Instrument Transformer and Its Present Tehnial Development Manner, Eletrial Equipment, vol.8, No.1,20-24 [5]Liu B, Ye G, Guo K, et al. Quality Test and Prolem Analysis of Eletroni Transformer[J]. Hign Voltage Engineering. 2012, 38(11): [6]WANG Hongxing, ZHANG Guoqing, GUO Zhizhong, 2009, Eletroni transformer and its appliation in digital sustations, Eletri Power Automation Equipment, Vol.29, No.9, [7]ZHANG Ming-ming,ZHANG Yan,LI Hong-in, 2004, Tehnology of designing integrator of Rogowski urrent transduer, Highvoltage Engineering, vol.30, No.9, [8]GUO Le,SHEN Di-qiu,LU Jia-li, 2010, Aomparative study of intergral methods in eletroni instrument transformer, Power System Protetion and Control, vol.38, No.8, [9]YIN Ming, TIAN Zhi-guo, ZHOU Shui-in, 2010, Researh on pratiaility of digital integrator ased on Rogowski oil, Power System Protetion and Control,vol.38, No.16, [10]WU Bin-in, SHEN Biao, DU Xiao-ying, 2010, Digital Integrator with Low Drift Based on Field Programmale Gate Array, Computer Measurement&Control, vol.18, No.7, [11]YIN Ming, TIAN Zhi-guo, ZHOU Shui-in, 2010, Researh on pratiaility of digital integrator ased on Rogowshi oi, Power System Protetion and Control, vol.38, Np.16, [12]Xie B, Yin X, Zhang Z, 2007, Tehnology in designing integrator of eletroni urrent transduer ased on Rogowski oil, RELAY, vol.35, No.3, [13]Shenglin W, Xin C, Lei C,2013, Study of ECT ased on Rogowski oil used in smart sustation, IEEE 7th International Power Engineering and Optimization Conferene (PEOCO), Langkawi, [14]Li W, Yin X, Chen D, 2008, Transient harateristis of Rogowski oil-ased urrent sensor, Eletri Power Automation Equipment, vol.28, No.10, CIRED /5