CONDUCTOR CORONA NOISE PREDICTION ON HIGH VOLTAGE AC LINES

Transcription

1 CONDUCTOR CORONA NOISE PREDICTION ON HIGH VOLTAGE AC LINES R.G. Urban*, H.C. Reader*, J.P. Holtzhausen*, K.R. Hubbard**, A.C. Britten** & D.C. Smith** * Department of EE Engineering, University of Stellenbosch, ** ESKOM Abstract: The conductor corona noise of four local 400 kv AC transmission lines is measured and compared to predictions. The prediction software developed is based on the EPRI semi-analytical approach using excitation functions and the theory of natural modes. The nature and accuracy of the prediction software is discussed along with relevant corona noise metrology issues encountered. Keywords: Conductor corona noise, PLC, excitation function, corona cage, theory of natural modes. 1 INTRODUCTION Significant electric fields at the surface of phase conductor bundles cause self-sustaining ionizing discharges in the surrounding air. These conductor corona discharges are responsible for many macroscopic phenomena on the line, including conducted electromagnetic noise and radiated electromagnetic fields. This noise causes interference to neighbouring communication systems. Line designers are therefore concerned that the noise is within limits [1]. Power line carrier (PLC) communication systems are installed on AC transmission lines in South Africa principally for teleprotection signalling for line protection. Conductor corona is a significant source of the noise in the PLC band on high voltage lines. A reliable calculation of the line noise, and consequently the resulting signal-to-noise ratio in the teleprotection channel, is an essential part of the system design process. A software package has been developed [2] for the prediction of conductor corona noise of high voltage AC lines. The prediction method is based on the EPRI semi-analytical approach [3], which utilizes excitation functions [4] and the theory of natural modes [5]. To evaluate and verify the accuracy of the software, the conductor corona noise of four 400 kv AC lines was measured. The excitation functions used in the simulations were measured in ESKOM s Megawatt Park corona cage facility. Previous papers ([6], [7]) described important aspects of the research such as the use of the theory of natural modes and the PLC system models used. Preliminary results were shown for two 400 kv lines. This paper reports on the operational line measurements, corona cage data extraction and the various correction factors used in the software for accurate predictions. A comparison between measured and predicted noise levels for the four lines and conclusions on the accuracy and reliability of the prediction software are presented. 2 THE PREDICTION SOFTWARE The fourth revision of the software, Corona Generated Radio Noise Modelling [8], is used for predictions presented in this paper. The corona noise performance of a line is defined by three measurements [1]. These are: the conducted corona noise across the terminals of the PLC receiver (i.e. the radio influence voltage), the radiated noise 20 m from the outer phase conductor of the line (i.e. the radio interference), and a lateral profile of the radiated noise in the vicinity of the line. The software is thus written to output these variables across the desired frequency band. The modal attenuation of a line is also a measurable quantity and can be used to check the software s use of the theory of natural modes. This is integral to the corona noise prediction process and can be considered as a verification of the software. To begin the prediction process, the line geometry is entered and displayed with the GUI. The software determines the average maximum surface gradient on the line s subconductors. The result is displayed on the GUI next to each phase conductor. The surface gradient is calculated using the method of successive images [9] to within 1% of the actual gradient. The excitation function for the conductor bundle used and the calculated surface gradient is obtained from corona cage measurements or calculated using empirical excitation functions. The selection and the accuracy of an empirical

2 line geometry shield wire phase conductor PLC coupling Fig 1 The graphic user interface (GUI) for the corona noise prediction software [8]. excitation function is discussed in a companion paper at this conference. The excitation functions used for the corona noise predictions presented in this paper are measured in ESKOM s Megawatt Park corona cage. Lastly, the earth model (see [6]) and PLC system coupling (see [7]) are defined for the line. The line attenuation and corona noise levels are then extracted for the line, defined in the GUI, across a defined bandwidth. The output variables can then be plotted. The software s GUI, showing an entered line geometry and calculated values, is presented in figure 1. 3 CAGE EXCITATION FUNCTIONS A corona cage is described in [9] as a rectangular, grounded-mesh cage structure surrounding a centred conductor bundle, between 15 and 60 m long and energized by a single phase high voltage source. Cylindrical cages are more expensive to construct, but provide a uniform radial electric field along the length of the conductor bundle. The average maximum surface gradient of a phase conductor can be replicated in a coaxial corona cage geometry by placing the bundle in the centre of the cage and applying the correct voltage. The total corona current on the conductor bundle, I P, saturates at the level defined by the attenuation constant of the coaxial transmission line, α, if the separation distance, s, between individual corona sources, i g, is sufficiently small, i.e. there are enough corona sources on the line [9]: lim ig [ I ] 0 P = (1) 2 α s The corona cage is much shorter than a transmission line and the number of corona sources on the centre conductor of a corona cage may therefore not be sufficient to produce a saturated noise level. Corona activity increases under heavy rain conditions and the cage length required to ensure a saturated corona current level is reduced. A minimum cage length has not been derived in the literature, although it is suggested in [9] that the cage should be longer than 30 m for heavy rain tests. An advantage of corona cages is the ability to apply uniform rainfall of a controllable intensity along the entire length of the conductor bundle tested. The excitation functions of numerous conductor bundles, under various weather conditions, were made available for the noise predictions of the studied lines. These excitation functions were extracted from the Megawatt Park corona cage and are similar to the measurements used to calculate empirical excitation functions defined in the literature, since they were made in accordance with CISPR 18 specifications [1]. The empirical functions are however general functions defined for a wide range of conductor bundle geometries. By using the excitation function for the actual conductor bundles used on the

3 studied lines, a more accurate prediction of the line s noise levels should be obtained. Fig 2 The Megawatt Park corona cage The Megawatt Park corona cage is a screened cylindrical cage design located in Johannesburg at an altitude of 1750 m. A photograph of the cage is given in figure 2. The facility is equipped with meteorological equipment that is used to log ambient weather conditions during each measurement sequence. Numerous reactive elements are used at the facility to ensure a low harmonic content in the source voltage. Both fair and foul weather excitation functions can be measured in the cage. The L 5 wet excitation function is extracted by applying uniform rain at a rate of about 2 mm/hr. The L 50 wet level is next extracted by switching the artificial rain off and waiting 1 minute. After the minute has passed, the noise level is measured. This procedure is repeated, producing a measurement sequence of noise levels measured at one minute intervals until the conductor is completely dry and the noise level reaches the L 50 fair weather level. dry The three consecutive QP noise measurements, plotted in figure 3(a) below, are made under the same weather conditions in the Megawatt Park corona cage at 500 khz for a twin IEC- 800 conductor bundle. They show significant variation at all measured surface gradients. The excitation functions used in the simulations are therefore the average value of numerous noise measurements. The accuracy of the prediction will therefore be dependent upon the number of measurements used to calculate the excitation function. The QP plot in figure 3(b) below indicates that the relationship between the noise level and the surface gradient is a smooth curve above about 18 kv/cm. Peak and average detector levels were also measured with the QP levels. The differences between these levels, which are important to the metrology community, have been discussed in [2] and will be examined in a future paper. The variation of the noise level at a single surface gradient was investigated by making 50 consecutive measurements at regularly spaced time intervals over a 5 minute period. The result revealed that the variation of the measured noise is reduced from 8 % to 3 % with an increase in surface gradient from 14 kv/cm to 20 kv/cm. Empirical corrections for altitude, weather condition and wind speed are also included in the software. Two relationships are extracted from a comparison of various empirical Γ [dbµa] Γ [dbµa] (a) Fig 3 Fair weather QP excitation function measurements and (b) measured QP, peak and average excitation functions extracted from the Megawatt Park cage. (b)

4 weather correction terms in [2]: ( 17 3) 50 wet = L50dry + ± L (2) ( 24 1) L (3) 1 wet = L50dry + ± These equations are used to calculate fair weather corona noise levels from the more reliable heavy rain excitation function extracted from the short corona cage. The correction term for a change in altitude, A [km], used in the CIGRÉ, BPA and ENEL methods is used [9]: A RI A = (4) 0.3 The correction term derived for wind speed, WS [km/hr], by LaForest et al. is used [2]: ( WS ) 0. 3 RI ws = (5) 4 LINE MEASUREMENTS Corona noise measurements were made on four 400 kv lines. These were the Hydra- Droërivier, Aries-Kokerboom, Acacia- Koeberg and Kriel-Tutuka lines. The radiated noise measurements are greatly affected by two main issues: The high radiated noise levels present at short lateral distances from the line s outer phase conductor overloaded the input amplifier of the active loop antenna and caused it to saturate. Reflections and excessive radiation caused by line discontinuities near the measurement site resulted in misleading radiated noise measurements. The accuracy of the software prediction was therefore assessed by comparing the measured and predicted conducted noise levels measured across the terminals of the PLC receiver in the PLC band. (a) (b) (c) Fig 4 The measured and predicted conducted corona noise levels of (a) the Acacia- Koeberg, (b) Hydra-Droërivier, (c) Aries-Kokerboom and (d) Kriel-Tutuka 400 kv lines. (d)

5 5 RESULTS The measured and predicted conducted corona noise levels at the Acacia, Droërivier, Aries and Tutuka substations are plotted in figure 4. Only altitude, wind and weather correction factors are added to the calculated excitation function predictions. It is evident in all four comparisons that the conducted corona noise performance is best predicted by the measured heavy artificial rain excitation function. The empirical weather relations defined in equation 3 are used to calculate the fair weather corona noise from the stable foul weather data. Two possible reasons for this are the reduced sensitivity of the corona cage measurements to changes in environmental conditions under heavy rain conditions and secondly, the decrease in the separation distance between pulses along the line, which causes the RMS summation of current pulses to saturate to a long line current level. The comparison between the corona noise predictions in figure 4 revealed that an appropriately chosen semi-analytical method produced a more accurate result for all measured lines except the Hydra-Droërivier line. This is attributed to the number of measurements used to derive both the measured cage and empirical excitations functions. Since more measurements are used for the empirical excitation function, the statistical average will have a narrower confidence interval. The choice of which empirical excitation function to use for a specific line is dealt with in a companion paper. The line measurements were made under favourable conditions and only a few correction factors needed to be included. The conducted corona noise predictions for these lines could therefore be made with a high level of confidence and are within 3 db for all lines (c.f. figure 4 at 500 khz). 6 CONCLUSIONS Software was developed to predict the corona noise performance of transmission lines. Comparisons between the measured and predicted conducted noise levels suggest that the predicted conducted noise levels are within 3 db of the measured levels at 500 khz. To obtain a high level of confidence in the predicted noise level under various conditions, appropriate correction factors defined in this paper must be applied and a heavy rain excitation function obtained from numerous measurements must be used. 7 ACKNOWLEDGEMENTS ESKOM ENTERPRISES are acknowledged for funding the research. The DTI is also acknowledged for its funding received through THRIP. Ulrich Buttner and Petrus Pieterse are thanked for their laboratory contributions. 8 REFERENCES [1] CISPR 18. Radio Interference Characteristics of Overhead Power Lines and High-Voltage Equipment, IEC, Bureau Central de la Commission Electrotechnique Internationale, 3 rue de Varembe, Geneve, Switzerland, [2] Urban, R.G. : Power Line Corona Noise Prediction from Small Cage Measurement, PhD Dissertation, Stellenbosch University [3] Comber, M.G. and Nigbor, R.J.: Transmission Line Reference Book 345 kv and above, Chapter 5: Radio Noise, EPRI, 2 nd Edition, Revised 1987, pp [4] Gary, C.H.: The Theory of the Excitation Function: A Demonstration of its Physical Meaning IEEE Transactions on Power Apparatus and Systems, Vol. PAS-91, January 1972, pp [5] Wedepohl, L.M.: The Theory of Natural Modes in Multi-Conductor Transmission Systems, unpublished lecture notes, Westband, British Columbia, Canada, January [6] Urban, R.G., Reader, H.C., Holtzhausen, J.P., Britten, A.C. & Hubbard, K.R.,: Calculating Conductor Corona Noise in Power Line Carrier Systems on AC Transmission Lines, Proceedings of the South African Universities Power Engineering Conference, SAUPEC 2003, pp [7] Urban, R.G., Reader, H.C., Holtzhausen, J.P., Britten, A.C. & Hubbard, K.R.,: Preliminary Predictions of Corona Noise on High Voltage AC Lines using Cage Data, Proceedings of the South African Universities Power Engineering Conference, SAUPEC 2004, Stellenbosch. [8] Urban, R.G.: Corona Generated Radio Noise Modelling, Version 4, Users Manual, Department EE Engineering, Stellenbosch University, [9] P. Sarma Maruvada, Corona Performance of High-Voltage Transmission Lines, Research Studies Press Ltd, 2000.