TECHNICAL DEVELOPMENTS AND PRACTICAL EXPERIENCE IN LARGE SCALE INTRODUCTION OF ON-LINE PD DIAGNOSIS

Transcription

1 ISBN Copyright 009 SAIEE, Innes House, Johannesburg TECHNICAL DEVELOPMENTS AND PRACTICAL EXPERIENCE IN LARGE SCALE INTRODUCTION OF ON-LINE PD DIAGNOSIS P.A.A.F. Wouters *, S. Mousavi Gargari, P. Wagenaars, I.J. Tighelaar, B.H.M.M. Simons, P.C.J.M. van der Wielen, E.F. Steennis, Eletrial Power Systems Group, Eindhoven University of Tehnology, The Netherlands KEMA, Arnhem, The Netherlands * Abstrat: On-line Partial Disharge (PD) detetion and loation systems for medium-voltage ables are at present being introdued in Duth utilities and worldwide. The tehnial hallenges now move from the development of the diagnosti tehnique itself to effiient implementation on a large sale. In this paper we disuss several implementation related hallenges and will propose adequate solutions. These hallenges inlude robust algorithms to determine time of arrivals of distorted PD waveforms, signal propagation along able types and onfigurations as three-ore and ross bonded ables, and effet of ring main units or substations on signal propagation. Algorithms based on signal energy and on phase angle in frequeny domain are preferred above e.g. threshold detetion to determine PD arrival times. By introduing effetive dieletri properties, able parameters for aurate fault loation as harateristi impedane and propagation veloity an be estimated also if data on semionduting layers are unavailable. Models are proposed for ross-bonded onnetions and for three-ore ables with ommon earth sreen. A pulse injetion iruit, already inluded in the PD equipment for time synhronisation, an be employed to extrat a model for PDs passing ring main units or even entire substations.. INTRODUCTION Partial disharge (PD) monitoring solutions have been introdued for HV able iruits ( 50 V) based on sensors per aessory. This approah is unpratial for MV able iruits ( 36 V), beause of the sheer number of omponents installed. Therefore, in ase of MV able onnetions a solution with only two sensors per able onnetion, one at eah able end, is preferred. Two sensors an over a able onnetion with several able segments, simply by measuring the differene in arrival time at both sensors. Suh an on-line PD measuring system wors for all types of MV ables and all types of aessories installed. To realise this approah the following problems have been addressed: - Synhronisation of the PD detetion units at both able ends - Coupling to the able for PD signal extration from the able - Reovering real PD signals from noise inherent to online measurement - Handling vast amount of data from ontinuous data streams At the CIRED 005 onferene, a prototype of the PD measuring system, alled PD-OL, was presented for the first time []. PD-OL stands for Partial Disharge monitoring On-line with Loation. In this system indutive signal oupling was preferred to allow for installation without need to swith off the able onnetion. The synhronisation between PD units at both able ends, whih is required for PD loation, was realised through inluding a pulse injetion unit whih sends a referene pulse to the other side. PD signals have to be distinguished from disturbanes from e.g. power equipment inside and outside the grid. A tehnique based on mathed filtering was adopted []. Finally, the arrival time and magnitude of the deteted waveforms, after being proessed loally at either able side, are ommuniated to the ontrol entre for further analysis and interpretation. Approximately 0 PD-OL systems have been put in operation sine it beame ommerially available in 007. One PD-OL system onsists of two separate units (Figure ), eah of them to be installed at one of the able iruit ends in either substation or ring main unit (RMU). Figure : PD-OL installation, with at eah able end a ontrol unit (PD-OL - CU) for signal proessing and ommuniation via internet and a sensor/injetor unit (PD-OL - SIU) for the atual measurement and pulse injetion. Eah measurement unit onsists of: - A sensor / injetor unit (PD-OL - SIU). This unit ontains both a sensor, to measure pulses from the able, and an injetion devie, to injet pulses into the able. This unit an be split in two parts and in this way lamped around the able or able earth onnetion, whih all an be done on-line. Pg. Paper D-0

2 ISBN Copyright 009 SAIEE, Innes House, Johannesburg x = m AIC = ln( σ, ) + ( N ) ln( σ +, N ) thres P n EC xi i= = P x t oa, g = N t x P = N = t x P N n N n = Figure : Example of reorded pulse with four time-of-arrival methods applied. Time axes are in μs. Bla line: reorded PD signal, grey line: threshold (a) / normalised and shifted AIC urve (b) / normalised EC urve (), *: timeof-arrival, and dashed line: end of reord used for AIC analysis. - A ontroller unit (PD-OL - CU). This unit is onneted to the SIU by means of an optial fibre. It ontrols the measurement sequene, the data olletion and the signal proessing. It has also ommuniation failities on board in order to upload the resulting data via the internet to the ontrol entre for further interpretation. Furthermore, the PD-OL units an be remotely reahed via the internet for diagnosti purposes and updates. In atual MV grids many ompliating fators arise. In order to extend the appliation range of PD-OL even further, PD-OL has to be apable to ope with fators impeding the signal waveform. PD signals are affeted by the able types and able onnetions used. Further, PD signals are distorted by the effet of the omponents in and the size of substations or RMUs. Corret timing of the deteted PD waveforms may be impeded by these fators as well. In this paper the following aspets are analysed: - Loation auray: To obtain high loation auray, algorithms are developed whih are robust with respet to noise level, signal distortion during propagation, refletions on aessories in the MV iruit. - Cross bonding joints: In long onnetions the able sreens an be ross bonded, whih disturbs PD signals during propagation. - Cable parameters: Models for several types of able are enountered in the field inluding three-ore ables with ommon earth sreen and ables from whih not all relevant speifiations for signal propagation are available. - Connetion with RMUs: Diagnosing several able onnetions interrupted by a RMU with a single set of PD-OL units requires models for signal propagation through these RMUs and even through an entire substation.. LOCATION ACCURACY The auray of defet loation depends on the auray of the time-of-arrival t oa estimation of eah PD pulse. The PD signals in on-line appliations are distorted by other omponents. Also additional refletions an onfuse measured waveforms. Further, the expeted noise level is higher, due to ontribution from the onneted grid. Different tehniques to lin a time of arrival to a signal waveform are ompared. These tehniques are illustrated in Figures and 3, inluding their mathematial formulation [3],[4]: t oa, p ( ω) X e = τ h ω jωτ h Figure 3: Signal before and after resolving the phase ambiguity. The signal is plotted in time domain and the phase in frequeny domain. A value of τ h = 3.05 μs was iteratively found. Bla line: signal before iteration, grey line: signal after iteration, and *: time-of-arrival. Table : Summary of strong and wea points of five t oa - methods. Threshold AIC EC Gabor Phase Noise Pulse shape - / 0 a - / 0 a - / 0 a - / 0 a + / + a Signal refletions Loation auray / + b a) by inorporating the hannel impulse response b) if load impedanes are aurately nown - Trigger threshold: The instane that a signal exeeds some threshold level x thres, whih equals the average noise power P n multiplied by some hosen fator m, defines the t oa. Pg. Paper D-0

3 ISBN Copyright 009 SAIEE, Innes House, Johannesburg - Aaie information riterion (AIC): The AIC method detets the signal by giving suitable weights to the umulative variane before and after eah sample point. This results in a urve with a global minimum, whih position orresponds to the t oa. - Energy riterion (EC): The EC is determines the umulative energy of the signal up to a sample point orreted for the energy based on the average power P x of the omplete signal. Similarly as for the AIC method this results in a global minimum orresponding to the t oa. - Gabor entroid: Gabor defines the t oa aording to a weighted average based on the sample energy. A orretion is made for the noise power. - Phase method: From the onverted signal in frequeny domain X a narrow band around ω is hosen. The t oa in time domain orresponds to the phase angle in frequeny domain, up to an ambiguity with πm. To remove the ambiguity, X(ω) is first shifted in time over τ h suh that the phase jumps (Figure 5, bottom) from -π to +π vanish. The simulated results with respet to noise, sensitivity to waveform distortion, presene of refletions and to loation auray are summarised in Table. The simulations are performed on a 000 m able having harateristis obtained from a real power able. The simulation is based on 000 runs with added Gaussian noise. Also the test waveforms are based on field experiene. Details are given in [4]. From Table it is onluded that the energy riterion and phase method perform best. However, the phase method is sensitive to the load seen at the able ends. A phase angle introdued there diretly translates into a loation error (ompare Figure 4a and Figure 4b). The PD-OL system enables, owing to its inluded pulse injetion unit, to determine this phase shift and ompensate for it by reating a proper model for the load experiened at the able ends. The reproduibility of the phase angle method and the energy riterion are exellent under all onditions as shown in Figure CROSS BONDING JOINTS In a ross bonding (CB) joint the earth sreen of eah phase is interrupted and onneted to the earth sreen of another phase. Usually, the atual ross-bonding ours in a CB box. The CB ables are either single ore or oaxial (as shown in Figure 5). CB joints at as refletion points for the signal. In addition, part of the pulse may start travelling between able sheaths, i.e. the intersheath mode. To simulate pulse propagation, measurements were performed on an artifiial CB system made up of 50 Ω measuring ables. Refleted and transmitted signal are measured upon a square pulse (3 V, 50 ns) injeted in able A. The signals shown in Figure 5 are the refleted signal in A, the transmitted signal in B, and a pulse entering C, one of the other ables. The signals D-F are similar to C (D and F have opposite polarity, E has same polarity). The alulated waveforms are based on the CB model shown left. Sine a omplete model as given in [5] involves many, in pratie unnown, parameters a simplified model is proposed [6]. The model onsists of a series impedane Z js whih represents the indutane of the loop in the CB onnetion, and a parallel impedane Z jp representing its apaitane and taing the intersheath mode into aount. a b Figure 4: Figure a: loation error if able load mathes the able impedane; figures b and : error and standard deviation in ase that the load impedane at the near end is an indutane of μh and the load impedane at the far end a apaitane of nf. Cross-bonding model Figure 5: Shemati drawing of CB joints, ables and box; left, model of eah joint; right refleted and transmitted signals upon injeted signal at point A (dashed measured, grey simulated waveforms). Pg. 3 Paper D-0

4 ISBN Copyright 009 SAIEE, Innes House, Johannesburg b a Figure 6a. The refletion at 3 μs ours at a regular joint. The refletions at the CB joints and at the far end are indiated within dashed retangles. They our both in the yellow phase and in the other phases. The existene of an intersheath mode is shown in Figure 6b. To this end the earth onnetions between two HV phase ables were diretly onneted and an indutive probe was plaed around it. The bla line is the normal mode signal also shown in Figure 6a, propagating with 85 m/μs. The grey line represents the intersheath mode after refletion on the first CB joint, travelling with 0 m/μs. Qualitatively the signals an be explained, but onsidering the signal amplitudes, deviations from the proposed model our. E.g. the refletions in the red and blue phases should be equal. The reason is related to the use of single phase CB ables whih ouple mutually depending on their atual positions. This problem does not arise for CB joints interonneted with oaxial CB ables. 4. CABLE PARAMETERS Figure 6: Measured voltages after pulse injetion in the yellow phase; a: normal mode signals in all three phases; b: indiated in grey an indutively measured intersheath urrent. A field measurement was onduted on a 50 V able iruit under onstrution with a set of CB joints at 50 m interonneted with single-ore CB ables, followed by an open end 00 m further. A pulse (3.5 V, 0 ns) is injeted in the yellow phase. This signal and its refletion at the transition from injetion (50 Ω) to HV able (6 Ω) are trunated utmost left in Models of power ables require detailed nowledge of the dieletri properties of the applied materials. Espeially the omplex relative permittivity ε r of semionduting layers is often not available. Still, for PD-OL a able model is required to reliably estimate the PD magnitudes and time of arrival. Typial XLPE ables, both single-ore and three-ore able, whih an be enountered in the field are depited in Figure 7. A simplified model is presented, where the unnown parameters related to the semionduting layers of the dieletri material are omitted and effetive parameters are introdued. Clearly, signal attenuation an not be predited in suh a model, sine the losses are aused to a large extent by the semionduting able properties. The harateristi impedane and propagation veloity depend less ritially on these properties. An effetive relative permittivity, for a single-ore able, is introdued aording to [7],[8]: Figure 7: Left, shemati drawing of a typial single-ore XLPE able equivalent iruit and definition of effetive parameters to model the able harateristis; right, three-ore XLPE able type with semionduting layers around eah ondutor. Condutors and earth sreens are indiated in light grey; the semionduting layers in dar grey. Pg. 4 Paper D-0

5 ISBN Copyright 009 SAIEE, Innes House, Johannesburg a b Frequeny (MHz) aount aording to [8],[9]. Figure 8a shows the measured harateristi impedane and the estimate aording to equation. On MV level three-ore power ables are often applied. If eah ore has its own metalli sreen they an be treated as three independent single-ore ables. The able shown in Figure 7 (right) has only one ommon earth sreen. However, eah ore is surrounded by its own semionduting layer and swelling tape, whih restrits the eletri field but not the magneti field. In the shown trefoil symmetry, the able exhibits two distint propagation modes [0]. The behaviour of this able type an be approximated by numerial tehniques. Figures 8b and 8 show the math between the measured and estimated (numerially using a boundary element method) harateristi impedane of the two distint propagation hannels. Detailed information, also on omparison of other able harateristis, an be found in [9]. Figure 8: Measured and estimated harateristi impedane of a single-ore XLPE able (a), and of the two distint propagation modes of a three-ore XLPE able (b,). r s ln ( ) ', ( ) r ε r (), eff ω ε r insu ω r ln s tis r + ts The symbols are defined in Figure 7. The effetive relative permittivity is defined suh that the standard equation for the apaitane of a oaxial struture holds. The harateristi impedane and the propagation veloity are now obtained from: μ 0 r s Z ( ω ) = ln ; v p () π ε 0ε r, eff r = μ0ε 0ε r, eff If the able earth sreen has a helial struture, the redution of the propagation veloity an be taen into 5. CONNECTION VIA RMU The effiieny of PD-OL implementation an be greatly improved if a single set of PD-OL units is apable to diagnose multiple onseutive able setions inluding RMUs. A hypothetial situation is shown in Figure 9. In fat the situation shown is part of a real MV grid from whih the measurements are taen. In order to model suh a system the transfer funtion of able setions, RMUs and substation have to be nown. The transfer through the RMU halfway hardly affets passing PD signals. This an be expeted, sine a PD signal, either oming from the right or from the left, experienes a load formed by the outgoing able in parallel to the distribution transformer. The latter impedane, apaitive in nature for frequeny omponents in the sub and low MHz range [0], is typially an order of magnitude larger than the able impedane. If inoming and outgoing able impedanes are equal, the signal Cable Cable PD-OL unit PD-OL unit Cable 3 Cable 4 Tr. Tr. Tr. RMU RMU Substation Cable 5 Figure 9: Typial onfiguration of able setion with two RMUs and a substation. Indutive signals deteted in the substation on different in- and outgoing ables are measured to determine the substation transfer funtions. Pg. 5 Paper D-0

6 ISBN Copyright 009 SAIEE, Innes House, Johannesburg passes the RMU virtually undisturbed. At a substation many ables are onneted and inoming pulses reflet at and transmit into these ables. Moreover, the size of a substation an not be negleted with respet to wavelengths orresponding to the frequeny ontents of the PD signals. In Figure 9 the signals are shown, measured at the inoming able and three of the four outgoing ables. From these signal the transfer funtions in the substation from the inoming able to the outgoing ables are determined. The dashed lines in Figure 0 are obtained from a substation model whih inludes impedanes to tae into aount the effet of the indutanes from the onnetions, the resistive (radiation) losses and the apaitanes to simulate the MV/LV transformer, its onneting able and a measurement transformer. The parameters are optimised for a best least square fit on the measured transfer funtions ombined with impedane measurements as desribed in []. Figure 0: Measured (solid) and simulated (dashed) transfer funtions from one inoming able to three outgoing ables in a substation. 6. DISCUSSION AND CONCLUSION Appliation of the PD-OL system to monitor extended able systems inluding RMUs or substation seems feasible. In order to maintain the sensitivity and the high loation auray realised for single able setions, models have to be designed to aount for the influene of all omponents on the PD waveform. This paper presented models for ross-bonded ables, ables for whih the data on the semionduting layers are unavailable and on how to treat multi-ondutor ables. Also the effets of substation omponents ould be adequately modelled. Sine the PD waveform is affeted by all transfer funtions a robust time of arrival algorithm is required. The energy riterion and phase methods (if an adequate model for the omplete able lin is present) satisfy this requirement. Reognition of PD signals is presently done on basis of a predefined mathed filter ban. However, in omplex able onnetions this ban may not be optimal. Present and future wor is direted to the design of an adaptive filter set, whih adapts automatially to the waveforms of real PDs enountered for a speifi onnetion. 7. REFERENCES [] P.C.J.M. van der Wielen, J. Veen, P.A.A.F. Wouters, E.F. Steennis: On-line partial disharge detetion of MV ables with defet loalisation (PDOL) based on two time synhronised sensors, 8 th International Conferene on Eletriity Distribution, Session, Turin, June 005. [] J. Veen and P.C.J.M. van der Wielen: The appliation of mathed filters to PD detetion and loalization, IEEE Eletrial Insulation Magazine, vol.9, no.5, pp.0-6, 003. [3] C. Herold, T. Leibfried, S. Maralous and I. Quint: Algorithms for automated arrival time estimation of partial disharge signals in power ables, Proeedings of the 5 th International Symposium of High Voltage Engineering, Ljubljana, paper T7-33, Aug [4] P. Wagenaars, P.A.A.F. Wouters, P.C.J.M. van der Wielen and E.F. Steennis: Aurate estimation of the time-of-arrival of partial disharge pulses in able systems in servie, IEEE Transations on Dieletris and Eletrial Insulation, vol.5, no.4, pp.90-99, 008. [5] R.J. Jason, A. Wilson and D.B. Giesner: Partial disharges in power-able joints: their propagation along a rossbonded iruit and methods for their detetion, IEE Proeedings C, vol.7, no.6, pp.40-49, 980. [6] P. Wagenaars, I.J. Tighelaar, P.A.A.F. Wouters, P.C.J.M. van der Wielen and E.F. Steennis: Partial disharge propagation through able systems with ross-bonding joints, Proeedings of the 0 th Nordi Insulation Symposium, Lyngby, pp.9-, June 007. [7] G. Mugala, R. Erisson and P. Pettersson: Dependene of XLPE insulated power able wave propagation harateristis on design parameters, IEEE Transations on Dieletris and Eletrial Insulation, vol.4, no., pp , 007. [8] P. Wagenaars, P.A.A.F. Wouters, P.C.J.M. van der Wielen and E.F. Steennis: Approximation of transmission line parameters of single-ore and three-ore XLPE ables, submitted to IEEE Transations on Dieletris and Eletrial Insulation, 009. [9] D.A. Hill and J.R. Wait: Propagation along a oaxial able with a helial shield, IEEE Transations on Mirowave Theory and Tehnology, vol.8, no., pp.84-89, 980. [0] P.A.A.F. Wouters: On-line alibration of highfrequeny partial disharge signals in three-phase belted power ables, IEE Proeedings on Siene, Measurement and Tehnology, vol.5, no., pp.79-86, 005. [] Bart Simons, Paul Wagenaars, Peter Wouters and Fred Steennis: Online impedane measurement method for RMUs and substations, submitted to st Nordi Insulation Symposium, Gothenburg, June 009. Pg. 6 Paper D-0