A NEW APPROACH FOR AN EARTH FAULT DISTANCE LOCALISATION ALGORITHM IN COMPENSATED NETWORKS

Transcription

1 A NEW APPROACH FOR AN EARTH FAULT DISTANCE LOCALISATION ALGORITHM IN COMPENSATED NETWORKS Georg Achleitner, Clemens Obkircher, Lothar Fickert, Manfred Sakulin Graz University of Technology Inffeldgasse 8/, A-8 Graz, Austria Abstract Earth fault compensated networks improve power reliability, due to the reason that most of the earth faults extinguish without interferences of the grid operation, thus allowing uninterrupted power supply during the fault situation. However, this type of neutral treatment implicates problems in the localization of earth faults. Up to now distance protection relays which measure the distance between the point of their installation and fault location are only available for direct grounded networks. In previous investigations the authors found that the standard algorithm of these distance protection relays principally can be used also for compensated networks, - however the accuracy of the distance calculation strongly depends on the network conditions. In the paper presented the main influence parameters are studied and a new improved algorithm is developed. For this purpose an exact -phase mathematical simulation model of the investigated network is used. As factor with the highest impact on fault location the earth fault transition impedance is identified. Also errors of the so called k-factor (which considers the earth return path impedance) are of strong influence. The simulations show, that this improved distance calculation provides very good results up to earth fault transitions impedances of k Ohm. As further shown in the paper, this new algorithm increases the accuracy of earth fault distance protection at high ohmic earth faults also in the standard application in direct grounded networks. At the end of this paper the simulation results are validated with real test data to verify the usability of this improved algorithm. Keywords: Earth fault, distance protection, earth fault compensated networks, fault location, localization of earth faults, distance protection, high ohmic earth faults switched off manually or automatically. Therefore, a selective earth fault distance protection is necessary. Nowadays it is common to know the faulted line; however the exact location of the faulty line sector is mostly unknown. In this paper a detailed exact -phase network computation for a given grid circuitry and given fault location and fault conditions is used in order to calculate the voltages and currents in the grid nodes and branches, thus simulating the input signals of the investigated protection devices with their different incorporated algorithms. For fault distance calculation it is necessary to measure the line impedance between the protection relay position and the fault location. The standard distance calculation algorithm was designed for direct grounded networks and is usually used only in those. It deliveres satisfying results in the case of low ohmic faults, but it is not reliable in case of higher fault impedances. The purpose of the following investigations is to analyse if this standard algorithm can be used also for other types of grids (e.g. compensated networks) and to find out the aberrations occuring in case of higher fault impedances and/or earthing impedances. Furtheron it s the aim to improve the algorithm to gain higher accuracy in the above mentioned conditions. Results from field tests in medium and high voltage networks prove the possibility of using common distance protection relays for earth fault distance protection in compensated grids. However, common relays for neutral earthed networks suppress actions in case of small currents (in the range of rated currents) in order to avoid wrong operation. INTRODUCTION Most of the medium and high voltage networks in Central Europe are operated with a compensated neutral point (arc suppressing coil, Petersen coil). This method improves the power quality in these networks due to the reason that most of the earth faults extinguish without interferences of the grid operation, thus, allowing uninterrupted power supply during single phase-to-ground faults. For safety reasons earth faults have to be localized as fast as possible. Earth faults in compensated networks, which do not disappear automatically, have to be SIMULATION MODEL Figure : Earth fault detection - basic scheme of the simulation model th PSCC, Glasgow, Scotland, July -8, 8 Page

2 Z pet Z add Z E Z F earth fault compensation coil (Petersencoil) additional impedance in parallel to the Petersencoil grounding impedance fault transition impedance Figure shows the single line scheme of the earth fault detection simulation model. Figure shows the equivalent circuitry in symmetrical components including the fault impedance Z F and the grounding impedance Z E of the measurement station as well as all capacitances. Normally the latter impedance is of minor influence. However, in case of an earth fault, the residual earth current causes a voltage drop over this grounding impedance which leads to inaccurate line-to-earth-voltage measurement results. Figure : Simulation model in symmetrical components (for exact calculation) EARTH FAULT DISTANCE CALCULATION ALGORITHM In the following the standard distance algorithm as it is used in common distance protection relays is explained. For the derivation of the algorithm the circuitry in Figure is used, however the capacitances and Z E being neglected. The location of the distance protection relay in Figure and Figure is symbolized by U meas. The model equations are based on symmetrical components. As already shown in [] and [] the following basic equations can be developed. At the fault point (U F ) the following basic equations can be derived: F F F U + U + U = U F A symmetrical network (no phase preferences) is assumed: Z Z = and with meas meas meas LE U + U + U = U and L + L + L = L I I I I LE F ZE = IL I L + IL + I + I F TR Z U Z Z Z Z Z Z with k = and I L = Z The k -factor contains the ratio of the earth return path to the (positive sequence) line impedance. ULE ZF ZE = I L + IΣ k + IF + I TR Z Z Z meas L + meas L + meas L TR E = F F U I Z U I Z U I Z I Z I Z th PSCC, Glasgow, Scotland, July -8, 8 Page

3 U I Z I Z = = Z LE F F TR E IL + I k Z Z Z Z E U LE I L I TR measured positive sequence impedance of the fault distance positive sequence line-impedance related to one length unit grounding impedance measured phase-ground voltage at the location of the relay measured phase current at the location of the relay measured residual current at the neutral point of the transformer l l () measured residual current at the location of the relay distance to the fault location For very low fault transition impedances and grounding impedances, Z F ~, Z E ~ equation () can be simplified to the well known equation () which is used in the common protection relay as standard algorithm. ULE Z = = Z l () I + I k L Formula () only contains measurable signals (U LE, I L, ) and computable factors (k ), thus the fault distance can be calculated directly. Up to now, this standard algorithm is used only in low-ohmic neutral earthed grids with high fault currents, assuming low fault impedances. However, as can be seen the neutral point impedances (Z pet and Z add ) do not exist in the above formula (), thus they can t have an adulterant influence on the distance computation. This means that the same algorithm can be used also in compensated grids. In [] and [] the usability of the standard algorithm for low ohmic earth faults is shown. Load currents raise the current levels (I L ), so that (also in case of current-transformer errors) a more accurate distance determination is possible. To establish more precise detection levels in practice, an additional resistance in parallel to the arc suppressing coil can be installed. The expected advantage of this method is the raise of the residual fault current to get a more accurate fault distance estimation []. LIMITATIONS OF THE STANDARD ALGORITHM In [8], simulations based on the exact simulation model Figure were used to investigate the limits of the standard algorithm (). The influence parameters are varied and the outcome of the algorithm (equ. ) is compared with the exact results. The simulation data are presented in Table. Network -kv overhead lines Transformer Z / Z j.9 / j.9 length km Z / Z per km [Ω].+j./.+j. cap C / C per km nf / nf Fault distance km Load Ω Petersen coil Ω+ j Ω Additional resistance Ω Capacitive current from 8 A the network Table : Data of the simulation for checking the standard algorithm. Influence of the grounding impedance Z E This impedance was chosen due to the reason that in practice at some measuring points grounding impedances can be high. Normally the measured voltages are line-to-earth voltages. If the grounding impedance is too high, and there is current over the grounding, for example, if the Petersen coil is placed in this substation, the voltage rise cannot be neglected because it would lead to a miscalculation of the fault distance. X without additional resistance X with additional resistance..... grounding impedance ZE in Ohm Figure : Standard algorithm: Variation of ZE Figure shows the influence of the impedance Z E. The values were chosen.+j. Ohm for the starting point and the resistive part was varied. The inductive part was chosen due to measurement, which has shown this inductive part []. It can be seen, that with an additional parallel resistance the aberration becomes smaller. th PSCC, Glasgow, Scotland, July -8, 8 Page

4 . Influence of the fault transition impedance Z F Principally, the fault impedance is unknown. In the simulation, this impedance was varied from to Ohm. X without additional resistance X with additional resistance 8 9 Figure : Standard algorithm: Variation of ZF. Influence of a deviating value of k Another problem is to find the appropriate value of k. Figure shows the effects of a wrong choice of k. In this simulation the angle and the value of k was varied. As can be seen considerable errors occur with wrong k choice. EXTENDED ALGORITHM As stated in the above performed sensitivity analysis, with higher fault transition impedance large aberrations of the calculated fault distance arise. Therefore it is necessary to adapt the algorithm in order to be able to take this influence into account. Going back from formula () to formula () one sees that the terms I F *Z F and I TR*Z E were omitted. However, to consider these terms one has to know the current at the fault point and the fault impedance which are both not measurable quantities, - at least not at the location the protection device is installed. In the following a new approach for estimating the fault impedance and the fault current in radial networks is presented.. Estimation of the fault impedance The fault impedance can be calculated by dividing the phase-to-earth voltage U LE of the faulted phase by the measured residual current and by using only the resistive part of the resulting calculated impedance. This estimation is acceptable because the line resistance is normally much smaller than the fault resistance. If the fault impedance is too small, there would be an aberration in the calculation because the whole loop resistance is calculated and not only the fault resistance. In these cases the use of the standard algorithm is preferable []. Due to experiences the fault impedance mostly is an electric arc and therefore predominantly ohmic []. R F U = real () LE I R F U LE Resistance at the fault point measured phase-ground voltage at the location of the relay measured residual current at the location of the relay The result of the simulations are presented in fig. Figure : Standard algorithm: Variation of k Comment: As a result of the above performed sensitivity analysis it can be stated that the standard algorithm leads to large aberrations, if the chosen parameters (k ) are not correct or if the neglected parameters become too large (Z F, Z E ). To receive higher accuracy of the calculated faulted line impedance, besides accurate choice of the parameters contained in the standard algorithm, above all the neglected parameters have to be taken into account. calculated transitions impedance RF in Ohm 9 8 RF calculated 8 9 Figure : Estimation of RF th PSCC, Glasgow, Scotland, July -8, 8 Page

5 It can be seen that the estimation is linear, even for high ohmic earth faults.. Estimation of the fault current As shown in Figure the fault current I F consists of the inductive current over the starpoint impedance and the capacitive currents from the faulted line and from the other feeders. The starpoint current plus the capacitive currents of the other feeders equal the current which is measured by the protection relay. The missing capacitive current is the capacitive current I cap of the faulted line (related to the system zero voltage). "standard" algorithm extended algorithm distance nominal..... grounding impedance ZE in Ohm Figure 8: Extended algorithm: Variation of ZE. Extended Algorithm applied to compensated networks The extended algorithm formula () now uses the simulated signals of the exact grid calculation. The results of the line impedances calculated by the extended algorithm show, that the accuracy is increased significantly and gets better with higher ohmic earth faults. In these simulations the additional resistance was put in parallel to the Petersen coil. Figure : Figure to explain the fault current I I ji U F = Σ + cap meas () Unom I F Fault current Umeas System zero voltage (the sum of the measured line-to-earth voltages by the protection device) measured residual current at the location of the relay U nom Nominal line-to-earth voltage Icap Nominal capacitive current contribution from feeder Equation () gives an estimation of the fault current at the earth fault point. As can be seen all variables can be measured by the protection relay at the point of its installation or can be calculated from nominal grid data.. Influence of the grounding impedance Z E on extended algorithm In chapter. it is shown, that ZE has a high influence on the calculated fault distance. Applying the extended algorithm on the same simulations give significantly better results as presented in figure 8. X with "old" algorithm X calculated with extended algorithm 8 9 Figure 9: Compensated networks with high ohmic fault impedance: Simulation of the extended algorithm and comparison with the old standard distance protection algorithm. Extended Algorithm applied to direct grounded networks The common opinion is, that in low ohmic grounded networks earth faults can be detected very easily with the standard algorithm. However high ohmic earth faults cause problems, because they cannot be detected correctly by formula (). Figure shows that the extended algorithm also can be used for these problems with much better accuracy. th PSCC, Glasgow, Scotland, July -8, 8 Page

6 X with "old" algorithm X calculated with extended algorithm 8 9 Figure : Low ohmic grounded networks with high ohmic fault impedance: Simulation of the extended algorithm and comparison with the old standard distance protection algorithm FIELD TESTS To prove the usability of the algorithm in practice, several field tests were done. Measurements were done in compensated networks as well as in low ohmic grounded networks. X distance X calculated R F R F calculated,8 Ω,9 Ω Ω,9 Ω,8 Ω, Ω Ω,9 Ω,8 Ω,9 Ω Ω 9,9 Ω,89 Ω, Ω Ω, Ω,89 Ω,8 Ω Ω, Ω,89 Ω,9 Ω Ω, Ω, Ω, Ω Ω 9, Ω 8, Ω, Ω Ω, Ω 9, Ω, Ω Ω, Ω Table : Results of the distance calculation of the field tests In a network earth fault tests were done. The extended algorithm was applied to the measurements and the results are presented in figure, which shows the aberration of the distance calculation algorithm in the different tests. The maximum fault transition impedance was relatively small (maximum Ohm). In these tests, the fault location was known exactly and therefore the error-analysis was possible. Z E was unknown and therefore not taken into account., Field Tests,. Tests in compensated networks The fault location was varied as well as the fault impedance. The tests were performed at different fault locations (X distance ) and with different transition impedances (R F ). The transition impedances were measured (R F ). The extended algorithm was applied to the measurement. The results are presented in table. Table and show field tests in a -kv network which is compensated. It has a short time low resistance grounding with around A. Type of network I cap I additional -kv / compensated A A -kv / compensated A A -kv / compensated A A -kv / compensated A A -kv / compensated A A -kv / compensated A A -kv / compensated A A 8 -kv / compensated A A 9 -kv / compensated A A Table : Data of the field tests With the extended algorithm the field tests were analyzed. The results are presented in table. Abberation in %,,,, -, -, -, 8 Distance in Ohm Figure : Results of earth fault tests Standard Algorithm Improved Algorithm As can be seen in Figure, larger aberrations only occur at small fault distances.. Tests in low ohmic grounded networks Table shows the results from earth fault tests in a low ohmic grounded network. It can be seen, that the algorithm detected earth faults up to a transition impedance of around Ohm. X distance X old X calc tpye of R F R F calculated, Ω, Ω, Ω tree 8 Ω wire on, Ω, Ω,9 Ω earth Ω good, Ω, Ω, Ω grounding Ω good, Ω, Ω, Ω grounding Ω Table : Results of the distance calculation of the field tests th PSCC, Glasgow, Scotland, July -8, 8 Page

7 SUMMARY In this paper the mathematical derivation of an improved, extended algorithm for earth fault detection by common distance protection relays is presented. The usability of this method for detection of earth faults in compensated networks is shown. The parameters, which have the largest influence on the calculated distances, are identified and the standard algorithm, as it is used in existing distance protection relays is extended and improved. Simulations show, that this improved distance calculation provides very satisfying results up to earth fault transition impedances of Ohms. Field tests with single phase earth faults with fault impedances of up to Ohms were performed and show very promising results. In all cases significantly higher accuracy than with the standard algorithm could be received. For compensated networks an injected additional current is recommended because it increases the trigger level and the fault-to-load current factor [], even though the simulations also worked without additional currents. Further research has to be done concerning the influence of unsymmetrical networks on the accuracy. REFERENCES [] Achleitner G., Obkircher C., Fickert L., SakulinM.: A new approach for earth fault localization in compensated networks, PSP, Bled, Slovenia [] Achleitner G., Obkircher C., Fickert L., Sakulin M.: Earth fault localization in compensated grids - a new way of fault distance computation, th International Conference Electric Power Quality and Supply Reliability,, Tallin, Estonia [] Achleitner Georg, Fickert Lothar: Verfahren zur Entfernungsortung von Erdschlüssen, Patent Application, A / [] Frei Johann, Ermittlung der elektromagnetischen Felder von Fahrleitungssystemen, Dissertation, University of Technology Graz, [] Eberl Gerit, Einsatz eines Deltaverfahrens zur Berechnung der Erdschlussdistanz in kompensiert betriebenen Energienetzen, Dissertation, Dresden, [] Hänninen, Lehtonen, Earth fault distance computation with fundamental frequency signals based on measurements in substation supply bay, Espoo, VTT Tiedotteita [] Imriš Peter, Transient Based Earth Fault Location kv Subtransmission Networks, Dissertation, Helsinki, [8] Achleitner G., Obkircher C., Fickert L., Sakulin M., Impedanzmessung an Leitungen, e&i, Volume, Number /, Wien BIOGRAPHIES Georg Achleitner was born in 9 in Linz, Austria. He received the Diploma Engineer degree from the University of Graz, Austria in. Since he is a science assistant at the Department of Electrical Power Systems, Graz university of technology. His research activities are resonant grounded nets, protection, grounding, measurement of earth- and line impedances and faults. Clemens Obkircher was born 9 in Austria. He received the Diploma Engineer degree from the University of Graz, Austria in. Since he is a science staff member of the Department of Electrical Power Systems, Graz university of technology. His research activities are resonant grounded nets, power losses, protection, grounding and faults. Lothar Fickert was born in 99. After graduation as Diploma Engineer and receiving the PhD in 9 he worked for years in power industry. In 998 he became professor and head of the Institute of Electrical Power Systems, Graz University of Technology. His main research activities are protection, power quality and grid design and operation. Manfred Sakulin was born in Bad Ischl, Austria, in 9. He retired as professor and senior researcher at Graz University of Technology in, but is still working as lecturer and consultant. His main activities are in the fields of energy efficiency, renewable energy sources, power quality and energy economics. th PSCC, Glasgow, Scotland, July -8, 8 Page