VFT insulation coordination study of a 400 kV GIS

DLO, DANIEL LEO OLASONHEF

DenmarkTEB, THOMAS EBDRUP

Energinet.dkDenmark

FFS, FILIPE FARIA DA SILVAAalborg University

DenmarkCLB, CLAUS LETH BAK

Aalborg UniversityDenmark

daniel.olason@gmail.comtebdrup2@gmail.com

21, rue d’Artois, F-75008 PARIShttp : //www.cigre.org

INNOVATION FOR SECURE AND EFFICIENT TRANSMISSION GRIDS

CIGRÉ Belgium ConferenceCrowne-Plaza – Le Palace

Brussels, Belgium | March 12 - 14, 2014

SUMMARY

In November 2008 the Danish government decided that all overhead lines below 400 kV should be replaced by underground cables. This is due to a demand of reducing the overall visibility of the transmission system, sometimes referred to as the beautification of the transmission system. The agreement furthermore included to reinforce some of the existing 400 kV transmission lines. This is due to both increasing wind penetration and power flow between Scandinavia, Germany and possible future connections. As a part of reinforcing the 400 kV transmission system in Jutland, Denmark, the Danish TSO (Energinet.dk) is in the process of constructing a new gas insulated substation (GIS) in Revsing. As a part of this process, new Eagle pylons will replace some of the existing Donau pylons. The new Eagle type pylon is meant to reduce the visual impact of transmission lines. The reliability of the substation in Revsing is of great importance as it is part of the 400 kV systems backbone between Sweden, Norway, Germany and the offshore windfarms in Horns Rev. The design of the insulation coordination for GIS must therefore be studied carefully. During a disconnector operation in GIS, very fast transient (VFT) may generate overvoltages (VFTO) inside the enclosure. Because the gas insulated system must be viewed as non-self-restoring, it is important to ensure that the voltage inside the GIS does not exceed the insulation strength. This must therefore be accounted for, when conducting an insulation coordination study of a GIS.This article describes how the VFT phenomenon occurs inside the GIS and how it may generate overvoltages. This includes an explanation of how it is generated, what causes it and why it is so fast. A schematic consisting of the surge impedances from the manufacturer of the GIS is simulated and compared to the same model with a number of capacitances added (representing the corresponding component). These added capacitances were not modelled by the manufacturer, but were added in order to further increase the level of detail. This is important as VFT may also be generated by circuit breaker (BRK) operations, a ground switch or due to a fault [1]. A detailed model is more likely to detect a VFT generated by e.g. a fault than a simplified model. It is shown via simulations in a EMTP software, how the level of modelling detail affects the results. As may be seen from the simulation results, there is a significant difference between the voltage characteristics when simulating the GIS with and without the added components. The difference is approximately 1.5 p.u., and is apparent 35 ns from when the disconnector operation is performed. It is of interest to investigate what the primary cause of this difference is. A further analysis of this difference, lead to a closer look at how the breaker is represented in the model from the manufacturer and the model with the added components. A comparison of the two models revealed the importance of detailed modelling, especially all of the capacitances which are present in the GIS for e.g. circuit breaker, disconnector (DS) and spacers.

KEYWORDS

Very fast transients, GIS modelling, EMTP modelling, VFTO, VFT modelling.

2

IntroductionThe insulation material used in GIS is Sulphur hexauoride gas (SF6) which greatly increases the insulation strength from 27 kV/cm∙bar for air, to 89 kV/ cm∙bar [2, p.348]. The GIS is pressurized and the operating pressure of the GIS systems is 4.5 to 5.3 bar. The high insulating strength of SF6 has however some disadvantages as well. The type of gas used (SF6) is one of the factors responsible for the generation of very fast transients (VFT) inside the GIS. The overvoltage generated by VFT may range from 1.7 - 2 p.u. according to IEC 60071-4. There are however other sources which indicate higher overvoltages e.g. as high as 2.5 p.u. [3, p.1]. Even though the overvoltage is only reported to reach levels below 3.0 p.u. it must be investigated, especially for higher voltage systems. This is due to the fact that as the voltage level of the system rises, the ratio between the switching withstand levels and the system voltage decreases as shown in table 1. The switching impulse withstand level (SWIL) is used as a reference of the overvoltage [5], as VFT often originate from a switching event. According to [4, p.612], a 20% safety margin should be applied in insulation coordination studies of gas insulated substations.

Table 1: Switching impulse withstand level (SWIL) in comparison to the system voltage IEC60071-1.

Highest voltage for equipment Um [kVRMS]

Switching impulse withstand level (SIWL) [kVPEAK]

Ratio [p.u.] Ratio [p.u.] with 20 % safety margin.

24 145 7.40 5.9252 250 5.89 4.71123 550 5.48 4.38145 650 5.49 4.39245 950 4.75 3.80420 1050 3.06 2.45765 1550 2.48 1.98

Even if the overvoltage does not reach the SIWL, the overvoltage may speed up the aging and degradation process of the GIS [5, p.1]. It is however questionable if it should be compared to the lightning impulse withstand level (LIWL) instead of SIWL. The reason for this is that the fast rise of the VFT should perhaps rather be placed in a protective category with lightings

Origin of very fast transient overvoltage (VFTO)A VFT is a result of an instantaneous change in the voltage inside the GIS. In most cases this change in voltage is due to the opening or closing operation of a disconnector (DS). VFT may however also be generated by circuit breaker (BRK) operations, a ground switch or due to a fault [3, p.1]. An example of a DS operation is shown in figure 1, where a part of the GIS in Revsing is illustrated.

Figure 1: During a DS operation a travelling wave is generated which may cause overvoltages in the GIS. The dashed line indicates the current path in this specific switching scenario.

VFT has two main characteristics

3

1. Are in the highest frequency range in power systems: 1 to 50 MHz [3, p.1].The reason for the high frequency is the overall compactness and construction of the GIS. This means that from a modelling perspective it may be considered as several short sections of transmission lines in series, each with its own surge impedance. An example is shown in figure 2, where a closed disconnector is modelled according to [6]. This results in a vast number of discontinuities. In Revsing there are 8 bays with 4 DS for each bay, resulting in a total number of 32 DS, again resulting in a total number of 224 surge impedance to be modelled, only with respect to the DS. This results in many reflections and refractions of the travelling wave occurring at the points of discontinuity, which may superimpose each other. As a result, high frequency overvoltages will appear in the GIS [7].

2. Have a rise time of 4 to 100 ns.The reason for the fast rise time of the VFT is due to several factors, which are further explained later in this paper.

Trapped charge and its influence on VFTODuring a DS operation e.g. as shown in figure 1, numerous discharges (pre- or re-strikes) occur due to the relative slow speed of the moving contacts [12]. Figure 3 shows possible voltage restrikes during an opening sequence of a DS from an ideal capacitive floating section of the system. The floating section is the section denoted as load side DS in figure 6. The disconnection of capacitive loads results in trapped charges, which influence the amplitude of the VFTO, when closing the DS. The exact amount of trapped charges depends on the disconnection and it may be explained by the following sequence for figure 3:1. A disconnection occurs and the voltage in the load side remains constant while the voltage in the

source continues oscillating at power frequency.2. As time passes, the electrical potential between the two increases.3. If the voltage breakdown level is exceeded, sparking occurs.4. The current flowing through the spark will charge the capacitance on the load side to a voltage

equal to the supply voltage.5. During the charging process the insulation strength between the two contacts increases as the

contact distance increases and the spark will eventually extinguish.6. The process may repeat itself, resulting in a staircase type waveform.

Figure 3: Voltage on each side of a DS during the opening sequence, which can lead to a trapped charge on the load side

The amplitude of the trapped charge will have consequences on the magnitude of the travelling wave which is transmitted during sparking, when the DS is closed again. The worst case scenario would be if the voltage potential on the floating section would be 1 p.u. and the voltage on the other side of the DS would be -1 p.u. or vice-versa. This would result in a 2 p.u. between the DS contacts. There are however limits to the amplitude of the trapped charge. That is to say that the voltage amplitude will according to various sources never reach 1 p.u. e.g. IEC 60071-4 states that the maximum trapped

4

Figure 2: Modelling of a closed disconnector in GIS [6].

charge will reach 0.5 p.u., but according to [5] and generally throughout this study a worst case scenario of 1 p.u. should and will be investigated.Origin of VFT, why is it very fast?The breakdown field strength (E/p)0 of an insulating gas is dependent on the difference between the ionization coefficient α and the attachment coefficient η. These coefficients are defined as follows:

Ionization coefficient α: Considering a swarm of electrons moving in a gas under a constant field, the growth of ionization rate is defined in terms of the number of ionizing collisions per electron per cm travel in the gas parallel to the applied field [9, p.39].

Attachment coefficient η: The removal of electrons from the swarm is determined by an attachment coefficient [9, p.39].

In other words α describes how fast free electrons are created, whereas η is describes the gas ability to absorb free electrons. Shown in figure 4 is the relationship between the effective ionization and the breakdown field strength. Figure 4 indicates that SF6 is a “brittle” gas, as a slight increase in electrical field will increase the rate of ionisation much faster than e.g. for air. That means that the breakdown process for SF6 is faster than the breakdown process for e.g. air. The rise time of the VFT may be determined by the following equation [6, p.1].

t z=13.3kT

( Ep )

0∙ p ∙ηh

[ns]

Where:kT = Toepler spark constant = 0.5∙10-2 [V∙sec/m] , for: air, N2 and SF6.(E/p)0 = Breakdown field strength [V/m∙bar].ηh = Field efficiency factor (1 for a uniform field and 0 for radius of curvature approaching zero) [2, p.203].p = Gas pressure [bar].

The breakdown field strength (E/p)0 will according to [8, p.4], increase in proportion to the pressure of the insulation gas. Meaning that at e.g 5 bar the breakdown field strength becomes 89 kV/cm∙bar∙5 = 445 kV/cm. Given a GIS pressure of 5 bar and a uniform field, the rise time is equal to 14.9 ns. This correlates well with the definition of VFT which defines the rise time as between 4-100 ns.

Modelling componentsAll components are modelled in general according to IEC 60071-4. Detailed information regarding the GIS components from the manufacturer was not available. Due to this reason and the fact that IEC 60071-4 does not specify values for each component, the authors searched for other sources to obtain values for a similar system. Each component value and modelling method is shown in figure 6.

Generating the sparkThis process may be modelled according to [5] and [10] with an exponentially decaying resistance in series with a small resistance. This is based on a worst case assumption were a spark of maximum amplitude is considered. The spark resistance is shown in the following equation [10].

R=Rarc+Ropen ∙ e−t / τ [Ω]

Where:Rarc is the arcing resistance = 0.5 Ω.

5

Figure 4: Effective ionisation coefficients in air and SF6 [9, p.40].

Ropen is the resistance of the gap = 1012 Ω.

τ is a time constant = 0.6∙10-9 s.The implementation of the spark resistance equation in EMTP and simulation results from EMTP is shown in figure 5, where the variable resistance is a function of time. As can be seen from figure 5 the

value of the variable resistance decays from 1 TΩ to zero in app. 20 ns. This correlates well with the limits between 4 - 100 ns for VFT, more specifically the 14.9 ns as previously mentioned.

Overall GIS modelAs detailed geometrical information regarding the GIS was not available, the possibility to construct a detailed model of the GIS was limited. That is to say that including every component such as every section, elbow, spacer and so forth was not possible. What was available, was a schematic of the GISmodel constructed by the manufacturer. The problem was that this model is a somewhat simplified model, compared to the recommended level of detail, making it difficult for the authors to determine if it is sufficient in detail for the study of VFT. A custom model is therefore constructed. The main reason why the model from the manufacturer cannot be used to construct a detailed model, is that the only geometrical data available was the length of the ducts and for each length the surge impedance was given. This is in fact the only data available from the manufacturer regarding modelling of the GIS. It is however of interest to investigate among other, if the model should be constructed in greater detail. There are however some components which may be added to possible improve the level of detail. It is therefore of interest to implement these components, and to compare the simulation results. The following components are implemented as they were not accounted (or the value was unknown) for in the manufacturer model:

Capacitance for DS, circuit breaker (BRK), surge arresters, Transformer and Reactor. These components are documented by IEC60071 as standard components to model a GIS. Other components require more detailed geometrical information.

Shown in figure 6 is the overall custom model to be implemented in EMTP. The components colored in grey and the voltage transformers were the only available data from the manufacturer.

Simulation resultsThe following case study will be simulated for each model:A trapped charge of 1 p.u. will be simulated in order to account for the worst case scenario.

The following locations will be measured for each model:Load side DS, OHL terminal, Transformer terminal, Reactor terminal.

6

Figure 5:Simulations results from PSCAD/EMTP and the implemetation of the spark generator.

Figure 6: The overall custom model to be implemented in an EMTP software.

Shown in figure 7 is the simulation results for voltage simulated at the load side DS, for both the manufacturer and custom model. The maximum overvoltage at the DS is well below 2 p.u for both of the models. There is an apparent difference in the simulation results between the two types of models.

Figure 7 : Simulation results from PSCAD/EMTP, comparing the model constructed by the manufacturer and the custom model, measured at the DS terminal.

Shown in figure 8 is the simulation results for the voltage simulated the OHL terminal, for both the manufacturer and custom model.

7

Figure 8: Simulation results from PSCAD/EMTP, comparing the model constructed by the manufacturer and the custom model, measured at the DS terminal.

The maximum overvoltage at the OHL is well below 2 p.u for both of the models. There is an apparent difference in the simulation results between the two types of models. Shown in table 2 are all of the peak voltage p.u. measured for each model. Marked in red/bold is the maximum simulated overvoltage. Not only are the waveforms very different, but it is apparent that the overvoltage for the custom model is higher than for the manufacturer model. They are however both well below 2 p.u. and thereby well below the SIWL limit.

Table 2: P.u. values for the different measuring points for the manufacturer and custom.

Model DS OHL Transformer ReactorManufacturer 1.447 1.32

31.032 1.032

Custom 1.699 1.758

1.045 1.050

A closer look at the simulation previously shown in figure 7 reveals a special area of interest where the main difference is shown. Further analysis of the cause of this difference leads to a closer look at the termination of the breaker in the two types of models. Shown in figure 9 is the breaker termination, used as the breaker model in the manufacturer model. The model is terminated by an open end, representing an open circuit breaker. The graph shown in figure 9 is a simulation of the manufacturer model and the custom model, this time using the open circuit breaker model shown in figure 9, for the custom model as well (replacing the one shown in figure 6). This was performed in order to see if the main difference between the manufacturer model and the custom model shown in figure 7 would disappear.

Figure 9 : Simulation results from PSCAD/EMTP, comparing the model constructed by the manufacturer and the custom model, measured at the DS terminal using the open circuit breaker model (shown on the right side).

8

As can be seen from the simulation in figure 9 the main difference disappeared. This shows the importance of simulating, using the correct values for the capacitors, when representing a breaker in GIS. For this study the voltage at the DS will result in a higher value with capacitors representing the breaker, instead of simulating with an ideal open end.

Conclusion As may be seen from table 2 that the maximum voltage will never reach the switching limits of 2.45 p.u. As was mentioned earlier, it is however questionable if it should be compared to the switching voltage limit, were VFT should perhaps rather be placed in a protective category with lightings. This would increase the allowable p.u. overvoltage to be as high as 3.32 p.u. placing it even further away from the limit.

Modelling the GIS should be done in great detail. Especially modelling the capacitance of components is very important as components capacitances greatly influence the simulation result. Modelling each small detail may however be troublesome and difficult to simulate. If the geometrical data is available a simplification, as shown in figure 10, would however be possible.

There is a need to model all of the capacitances in detail and as is shown in this paper, the capacitance of the breaker is important as it greatly influences the results. This is evident as modelling with the correct components (capacitances) resulted in a VFTO of 1.758 p.u. compared to 1.323 p.u. without the corresponding components.

REFERENCES

[1] P. Bolin, H. Koch, Gas insulated substation GIS, in: Power Engineering Society General Meeting, 2006. IEEE, 2006, p. 3.

[2] E. Kuffel, W. Zaengl, J. Kuffel, High voltage engineering: fundamentals Electronics & Electrical, Newnes, 2000.

[3] Modeling guidelines for fast front transients, Power Delivery, IEEE Transactions on 11 (1) (1996) 493 - 506. doi:10.1109/61.484134.

[4] A. R. Hileman, Insulation Coordination For Power Systems, CRC Press, 1999.[5] M. F. Tomasz Kuczek, Modeling of overvoltages in gas insulated substations, "Przeglad

Elektrotechniczny" (Electrical Review) 04a/2012, ABB Corporate Research Center in Krakow, Poland.

[6] D. Povh, H. Schmitt, O. Volcker, R. Witzmann, P. Chewdhuri, A. E. Imece, R. Iravani, J. A. Martinez, A. Keri, A. Sarshar, Modeling and analysis guidelines for very fast transients, Power Engineering Review, IEEE 17 (13) (1996) 71. doi:10.1109/MPER.1996.4311053.

[7] D. Pinches, M. Al-Tai, Very fast transient overvoltages generated by gas insulated substations, in: Universities Power Engineering Conference, 2008. UPEC 2008. 43rd International, 2008, pp. 1 - 5. doi:10.1109/UPEC.2008.4651627.

[8] S. Boggs, F. Chu, N. Fujimoto, A. Krenicky, A. Plessl, D. Schlicht, Disconnect switch induced transients and trapped charge in gas-insulated substations, Power Apparatus and Systems, IEEE Transactions on PAS-101 (10) (1982) 3593 - 3602. doi:10.1109/TPAS.1982.317032.

[9] A. Haddad, D. Warne, I. of Electrical Engineers, Advances in High Voltage Engineering, IEE Power and Energy Series, Institution of Electrical Engineers, 2004.

[10] M. Stosur, M. Szewczyk, W. Piasecki, M. Florkowski, M. Fulczyk, Gis disconnector switching operation; vfto study, in: Modern Electric PowerSystems (MEPS), 2010 Proceedings of the International Symposium, 2010, pp. 1 - 5.

[11] ABB, Gas-insulated Swithcgear ELK-3, product brochure (2010)

9

Figure 10: The figure on the left side is the original illustration of the GIS and the figure on the right is the possible equivalent representation of the GIS [11].

[12] J. A. Martinez, P. Chowdhuri, R. Iravani, A. Keri, D. Povh, Modeling guidelines for very fast transients in gas insulated substations, Report Prepared by the Very Fast Transients Task Forceof the IEEE Working Group on Modeling and Analysis of System Transients

10