Comparison of metal oxide surge arrester models in overvoltage studies

Transcription

1 MultiCraft International Journal of Engineering, Science and Technology Vol. 3, No. 11, 2011, pp INTERNATIONAL JOURNAL OF ENGINEERING, SCIENCE AND TECHNOLOGY MultiCraft Limited. All rights reserved Comparison of metal oxide surge arrester models in overvoltage studies André Meister 1, Rafael Amaral Shayani 2, Marco Aurélio Gonçalves de Oliveira 2 1 Agência Nacional de Energia Elétrica (ANEEL) 2 Universidade de Brasília - Faculdade de Tecnologia - Departamento de Engenharia Elétrica Campus, Universitário Darcy Ribeiro, Caixa Postal 4386, Brasília DF, BRAZIL * Corresponding Author: mago@unb.br, Tel: ; fax: , andremeister@aneel.gov.br(andré Meister), shayani@unb.br (Rafael Amaral Shayani) Abstract During switching or lightning overvoltages, surge arresters play an important role in limiting voltage levels and protecting substation equipment, by conducting the excess of current in the system, which would otherwise damage the equipment. This paper aimed to conduct a technical comparison of the models developed to represent the frequency dependent characteristic of metal oxide surge arresters. The scope of this paper also included outlining a methodology to choose which one would be more appropriate. The major contribution of this article is the validation of the models in a typical 500 kv substation insulation coordination study. Keywords: Overvoltages, Insulation Coordination, Surge Arrester, Surge Arrester Modeling, ATP program DOI: 1. Introduction The substitution of silicon carbide arresters for metal oxide surge arresters has brought benefits to overvoltage protection. However, there is the demand for the development of a model to be used in transient overvoltage and switching surge studies. The ATP Alternative Transients Program program allows the modeling of this non-linear resistance through the ZnO Fitter routine and the Type 92 card (ATPRB, 1997). Laboratory test data of metal oxide arrester discharge voltage and current have indicated that the arrester has dynamic characteristics that are significant for studies involving fast front surges, which are not well represented by the ATP model previously mentioned. Technical data show that for fast front surges, with rise time less than 8µs, the voltage waveform peak occurs before the current waveform peak and the residual voltage across the arrester increases as the time to crest of the arrester discharge current decreases. The increase could reach approximately 6% when the front time of the discharge is reduced from 8µs to 1.3µs. According to the IEEE Working Group (1992), this peak can reach up to 12%. It may be pointed out that the voltage across the arrester is not only a function of the magnitude of the discharge current, but is also dependent on the rate of increase. This fact is particularly important in lightning studies. Several models, at different voltage levels, have been proposed to represent the frequency dependant characteristic of metal oxide surge arresters. The model proposed by the IEEE Working Group, although having the purpose of finding a mathematical model that adequately reproduces these effects without requiring excessive computing time, uses a trial and error procedure. Besides, it is necessary to have physical parameters (e.g. overall height, block diameter, column numbers), which makes modeling more difficult. Other models have been developed, but adjustment of parameters, in some cases, requires iterative procedures. Furthermore, the technical data necessary for other models are not always easily obtained through catalogues. The purpose of this study was to conduct a comparison of surge arrester models. The results show that all models have similar performance when subjected to fast front surges. Conclusions are drawn that will help the selection of which model would be more appropriate for each kind of study. Lastly, the advantages, disadvantages, and errors associated with the use of these models are mentioned. The main innovation of this paper lies in the technical comparison of the models with the results of an insulation coordination study in a typical 500 kv substation subjected to fast-front surges. To be represented in the ATP, each piece of

2 36 equipment was modeled for high frequency studies. The substation was modeled with distributed parameters through their impedance surge, time travel and length. The transmission lines were modeled through their surge impedance. All other equipment was modeled for the phenomenon studied. Lastly, the surge arresters models were introduced in the study for analysis. 2. Surge arrester models 2.1 The ATP conventional model In the ATP Program, despite the existence of many types of surge arrester models, the exponential non-linear resistive device is the most widely used. The voltage-current characteristic is represented by several exponential segments, each one defined by eq. (1). q v i = p (1) Vref In this equation q is the exponent, p is a multiplier, and V ref is an arbitrary reference voltage that normalizes the equation and prevents numerical overflow during exponentiation. The first segment of the device is linear which speeds up the simulation. The second segment is defined by parameters p, q and a minimum voltage level. When the voltage across the surge arrester reaches a predefined minimum level, the algorithm tries to find a solution to the equation. The more exponential the model, the more precise are the results. The simulation of this model shows that for fast front surges, the peak voltage and current occur at the same time. Therefore it is not suitable to represent phenomena which are frequency dependent. 2.2 The Tominaga et al. model The aim of having a frequency dependent model, based on the desired behavior between voltage and current, led to a model with a varistor in series with an inductance (Figure 1) (Tominaga et al., 1979). Although the voltage across the device increases with a higher current level, this model is a good approximation to specific situations. For example, a chosen inductance for the model could produce accurate results for an 8µs surge front. But for a 2µs surge front, the results would not be satisfactory. 2.3 The Kim et al. Model Figure 1 The Tominaga et al. model. As in the previous model, a non-linear inductance was placed in series with the varistor (Figure 2) (Kim et al., 1996). The major problem with this model is that it is necessary to build a program to calculate the non-linear inductance. Furthermore, many voltage-current points are necessary to represent the surge characteristic.

3 37 Figure 2 The Kim et al. model. 2.4 The IEEE Model Due to the difficulties pointed out, it was necessary to study frequency dependent models. This was the task of the IEEE Working Group [2], that developed a model in which the non-linear V-I characteristic is represented by two sections of non-linear resistances designated by A0 and A1. These resistances are separated by a R-L filter (Figure 3). Figure 3 The IEEE Model. For slow front surges, the R-L filter has very little impedance and the two non-linear sections of the model are essentially in parallel. For fast front surges, the impedance of the R-L filter becomes more significant. This results in more current in the nonlinear section designated by A0, than in the section designated by A1. Since characteristic A0 has a higher voltage for a given current than A1, the result is that the arrester model generates a higher voltage. This model yields good results for arrester discharge voltages, when the discharge current has a time to crest within the range of 0.5 to 45µs. The major problem associated to it is to calculate its parameters. The IEEE Working Group suggests an iterative method, where adjustments are necessary to achieve better results. The initial parameters depend on electrical and physical data. 2.5 The Mardira and Saha model The IEEE model was simplified, the resistive devices were eliminated, and another way to define the parameters was chosen (Figure 4) (Mardira and Saha, 2011). The authors state that the model yields good results for a current discharge with a 8x20µs waveform, and does not require an iterative process. However, it does not work properly for a wide variety of waveforms.

4 38 Figure 4 The Mandira et al. model. 2.6 The Pinceti and Giannettoni model In this model, also derived from the IEEE Model, all necessary data are easily collected in datasheets, there is no need for an iterative correction of the parameters, and the model s performance is accurate (Figure 5) (Meister and Oliveira, 2005a). Besides, the capacitance is eliminated, and only electrical parameters are used. The two parallel resistances are substituted for only one, in order to avoid numerical overflow. 2.7 The Fernandez and Diaz model Figure 5 The Pinceti and Giannettoni model. The IEEE Model is also simplified in this model (Fernandez and Diaz, 2001). This time, both electrical and physical parameters are necessary. The two parallel resistances are substituted for only one, in parallel to a capacitance, and one of the inductive devices is eliminated (Figure 6). The relation between the A0 and A1 currents and their voltages must necessarily be 0.02 (Pinceti and Giannettoni, 1999). Figure 6 The Fernandez and Diaz model

5 39 3. Model implementation The Conventional, IEEE, Pinceti and Giannettoni, and Fernandez and Diaz models were compared using the ATP Program, version ATPDRAW 3.9. The other models were not considered, because their results were not accurate. The waveforms used in the simulations were developed with the ATP Source 15 Heidler Type. The use of Source 13 does not demonstrate the frequency dependent characteristic, where the current signal leads the voltage signal. The other sources available did not allow correct modeling of the desired waveform. To compare models, the ABB EXLIM P-E kv varistor was used (Table 1). The model parameters are shown in Table 2. Rated Voltage [kv] 4. Simulation results Nominal Discharge Current [kacrest] Table 1 Technical data for ABB EXLIM P-E kv Max. Residual Voltage (kvcrest) Height 30x60µs 8x20µs 1x(2-20)µs [mm] Table 2 Model parameters Model L1[µH] L0[µH] R1[Ω] R0[Ω] C[pF] IEEE 21,75 0,29 94, ,97 Pinceti and Giannettoni 0,487 0, Fernandez and Diaz model 0, ,3 Results for both the 8x20µs waveform with 10 ka, and the 1x2µs waveform with 10 ka are shown in Table 3, together with the error as compared to datasheet values (Table 1). Time to voltage crest for each model is presented in Table 4. The response of each model to waveforms 8x20µs with 10 ka, and 1x2µs with 10 ka are shown in Figures 7 and 8 respectively. Figures 9, 10, 11, and 12 present their dynamic behavior. Table 3 Models Comparison - Voltage results Model 8x20µs - 10 ka 1x2µs - 10 ka Voltage [kv] Error Voltage [kv] Error Conventional % % IEEE % % Pinceti and Giannettoni % % Fernandez and Diaz model % % Table 4 Models Comparison Time to voltage crest. Model Time to crest (µs) 8x20µs 10 ka 1x2µs 10 ka Conventional IEEE Pinceti and Giannettoni Fernandez and Diaz model

6 40 Figure 7 Response to waveform 8x20µs 10 ka Figure 8 Response to waveform 1x2µs 10 ka Figure 9 Dynamic behavior: Conventional Model.

7 41 Figure 10 Dynamic behavior: IEEE Model Figure 11 Dynamic behavior: Pinceti and Giannettoni Model Figure 12 Dynamic behavior: Fernandez and Diaz Model This study shows that the Conventional Model does not represent the dynamic behavior of the varistor, since the current and the voltage peak occur at the same time (Figures 7 and 8). Besides, the Conventional Model reaches the same voltage peak value for both waveforms (Table 3). The IEEE Model presents a relatively higher error for faster front surges (Table 3). It also requires physical data and an iterative method to determine the parameters. The occurrence of the voltage maximum before the current maximum was observed in all models, except for the Conventional (Figures 7 and 8). The results of the Pinceti and Giannettoni model are very accurate. Physical data are not needed, nor does it require the use of iterative methods, thus not needing much

8 42 computational effort. The Fernandez and Diaz model also presents good results. The major difficulty associated to using this model is determining the inductance value, which is related to curves between the waveform time to crest and the residual voltage. 5. Insulation coordination study In order to study energy absorption, models were compared through an insulation coordination study in a typical 500 kv substation (Figure 13), subjected to fast-front surges. The substation was modeled with distributed parameters through their impedance surge, time travel and length; and the transmission line through its surge impedance (the value of 300 Ω was adopted). The transformer 525/440/13,8 kv was modeled as in Figure 13. All other equipments were adequately modeled for the phenomena studied as described in Table 7. (Meister and Oliveira, 2005b). Figure 13 Substation diagram and Transformer model The line is protected by the ABB EXLIM P-E kv varistor (Table 1), while the transformer is protected by the ABB EXLIM T kv varistor (Table 5). Surge arrester models were introduced to represent both varistors and their parameters are shown in Table 6. The surge current was injected from the line using the ATP Source 15 and Table 8 presents the results, while Figures 14 to 16 present energy waveforms. Rated Voltage [kv] Nominal Discharge Current [kacrest] Table 5 Technical data for ABB EXLIM T- 420 kv Max. Residual Voltage (kvcrest) Height 30x60µs 8x20µs 1x(2-20)µs [mm] Table 6 Surge Arrester model parameters Model L1[µH] L0[µH] R1[Ω] R0[Ω] C[pF] IEEE 6,90 0, ,84 73,6 73,60 Pinceti and Giannettoni 1,30 0, Fernandez and Diaz model 0, ,6 Table 7 Other equipments parameters modeling Equipment C[nF] Potencial transformers 5,00 Circuit breakers 0,10 Switches 0,10 Current transformers 0,50 Reactors 4,00

9 43 Model Table 8 Insulation coordination study results. Line [kv] Line [J] Varistors Transformer [kv] Transformer [J] IEEE E E+05 Pinceti and Giannettoni E E+05 Fernandez and Diaz E E+05 Figure 14 IEEE Model Figure 15 Pinceti and Giannettoni Model

10 44 Figure 16 Fernandez and Diaz Model The results show that in the voltage level studies, the highest difference was approximately 2% between the models. In the energy absorption studies, the Fernandez and Diaz as well as Pinceti and Giannettoni models seemed to yield more conservative results. 6. Conclusions In this study, a simulation of the dynamic behavior and energy absorption of metal oxide surge arrester models was conducted. Model implementation was performed with the Alternative Transients Program. All data necessary for the models was found in manufacturers catalogues, although this was not always a simple task. Regarding dynamic behavior, the Conventional model was the only one that did not work properly, with the current and voltage peaks occurring at the same time. The comparison with manufacturers data showed that the IEEE, Pinceti and Giannettoni as well as Fernandez and Diaz models acceptably reproduced the peak voltages. In the insulation coordination study the Pinceti and Giannettoni as well as Fernandez and Diaz models presented more conservative results regarding the energy absorption, but the difference between these results was insignificant. For the overvoltage coordination studies, all three models were appropriate. References Alternative Transient Program Rule Book (ATPRB), Can/Am EMTP User Group,USA. Fernandez F., Diaz R., Metal-oxide surge arrester model for fast transient simulations, paper 144, International Conference On Power System Transients, IPST 01, June. IEEE Working Group , Modeling of metal oxide surge arresters, IEEE Transactions on Power Delivery, Vol. 7, No. 1, pp Kim, I., FunabashI, T., Sasaki, H., HagiwarA, T. Kobayashi, M Study of ZnO arrester model for steep front wave, IEEE Transactions on Power Delivery, Vol. 11, No. 2, pp Mardira, K. P., Saha, T. K., A simplified lightning model for metal oxide surge arrester, The University of Queensland Austrália, Downloaded September Meister, A., Oliveira, M.A.G. 2005a. Modeling ZnO arrestes in ATP for surge protection - Comparação da representação de varistores de Óxido de Zinco no programa ATP para proteção contra sobretensões, XVIII SNPTEE, Curitiba, Paraná, Brasil, Outubro. Meister, A., Oliveira, M.A.G. 2005b. Modeling ZnO arrestes in ATP for inslulation coordination studies Modelagem de varistores de Óxido de Zinco para estudos de coordenação de isolamento. Dissertação de Mestrado em sistemas elétricos de potência Universidade de Brasília. PPGENE.DM-245A/05. Pinceti, P., Giannettoni, M., A simplified model for zinc oxide surge arresters, IEEE Transactions on Power Delivery, Vol. 14, No. 2, pp Tominaga, S., Azumi, K., Shibuya, Y., Imataki, M., Fujiwara, Y., Nichida, S., Protective performance of metal oxide surge arrester based on the dynamic v-i characteristics, IEEE Trans. Power App. Syst., Vol. PAS-98, pp Biographical notes Andre Meister was born in Brasilia, Brazil, on October 8, He received the Bachelor of Electrical Engineering degree and the M.Sc. degree from the University of Brasília, Brasília, Brazil, in 2000 and 2005, respectively. His professional experience includes Marte Engenharia (Brasil), where he was involved with load flow, stability, and electromagnetic transient studies. He joined the Agência Nacional de Energia Elétrica (National Regulatory Agency) in 2005, where he works with electrical transmission systems. His areas of interest include transmission systems and electromagnetic transient studies.

11 45 Rafael Anaral Shayani was born in São Paulo, Brazil, on March 25, He received the Bachelor of Electrical Engineering degree from the Polytechnical School of the University of São Paulo, São Paulo, Brazil, in 1998 and the M.Sc. and Ph.D. degrees from the University of Brasília, Brasília, Brazil, in 2006 and 2010, respectively. His professional experience includes the Centro de Gestão e Estudos Estratégicos (Brazil), where he worked as a consultant in the Prospective Study on Photovoltaic Energy, and Johnson Controls, where he supervised electric power and air conditioning systems. He joined the University of Brasilia in 2009, where he teaches a course on energy conversion. His areas of interest include electrical power systems, renewable energy sources, power quality, and energy efficiency. Mr. Shayani received an award at the 8th Brazilian Energy Congress in 1999, presenting a comparison on overall costs between energy production from hydroelectric power plants and from natural gas plants. Marco Aurélio Gonçalves de Oliveira was born on December 20, 1958, in Rio de Janeiro, Brazil. He received the Bachelor of Electrical Engineering degree from the University of Brasília, Brasília, Brazil, in 1982 and the M.Sc. and Ph.D. degrees from the University of Paris, Paris, France, in 1989 and 1994, respectively. From 1982 to 1988, he was with the Operation Division of Eletronorte (Brazil) where he was involved with load flow, stability, and electromagnetic transient studies. He joined the University of Brasilia in 1994, where he was the Head of the Department of Electrical Engineering from 2006 to His research interests include power electronics, power quality, renewables, and energy efficiency. He has published over 50 papers in those fields. Dr. de Oliveira was Chair of the IEEE Brasilia Section and IEEE Brazil Council. Received September 2011 Accepted October 2011 Final acceptance in revised form October 2012