21, rue dArtois, F-75008 PARIS http : //www.cigre.

org

(107)

CIGR Canada Conference on Power Systems Vancouver, October 17- 19, 2010

FAULT LOCATION IN EXTRA LONG HVDC TRANSMISSION LINES USING DISCRETE WAVELET TRANSFORM
O.M.K.K. NANAYAKKARA1, A.D. RAJAPAKSE1, RANDY WACHAL2 University of Manitoba1 (CAN), Manitoba HVDC Research Centre 2(CAN)

SUMMARY In this paper, the accuracy of the two-terminal travelling wave based transmission line fault location using wavelet transform technique is investigated for a 2400km long overhead HVDC transmission line. The feasible of using the measured DC voltages and the surge capacitor currents to detect travelling wave arrival was studied. Simulations were carried out with PSCAD/EMTDC and the fault location algorithm was implemented in MATLAB. Results show that the wavelet based edge detection can be used in the fault identification in HVDC systems with extra long overhead lines. The influence of noise on the fault location accuracy is presented. The results also show that either DC terminal voltages or the surge capacitor currents can be used in line fault location.

KEYWORDS
Fault location, HVDC, Long transmission line, Travelling wave, Wavelet, DWT

[email protected]

1. INTRODUCTION The travelling wave based fault location principle, which utilizes the propagation times of the voltage and current travelling waves generated on a transmission line when a fault occurs, is well known [1]. Although application of this principle is challenging in the highly branched and meshed AC networks, it has been successfully applied to transmission line fault location [2] in the conventional HVDC systems, which have only two terminals. Depending on the signals used, two travelling wave based fault locations methods can be identified: twoterminal method and single terminal method. The two-terminal method is more reliable than the single-terminal method as it only makes use of the fault generated initial surges but not the secondary reflections. In the two-terminal method, the difference in the times of arrival of the first fault-generated waves at the two line terminals is used to determine the fault location, given that the propagation velocity of the surge is known. The widely used and more accurate two-terminal measurement method requires precise determination of wave front arrival times at both line terminals, and an easy means of bringing the measurements from the two terminals to a common point, so that the fault position can be determined. This method became a feasible solution with the developments of Global Positioning System (GPS) which provides time synchronization accuracies of better than 1 s over the entire surface of the Earth, 24 hrs per day [3][4] and with developments in telecommunication technology [4]. There are number of line fault locators based on this principle [2] are installed in HVDC systems worldwide. These fault locators can achieve fault location accuracy of less than 0.3 km depending on the length of the transmission line. As the demand for power grows in countries that a rapidly industrializing, many new HVDC transmission schemes are being built. There are several HVDC transmission systems with extra long overhead (O/H) lines such as the 2500 km long Porto Velho-So Paulo HVDC system [5] under consideration. Accurate fault location in such extra long transmission lines is a challenging task and when the transmission lines are longer than 1000 km, fault location has been achieved with the help of repeater stations. Installation of extra hardware at the repeater stations, which are required to locate line faults using the existing technology, increases the cost of these transmission projects. In this paper, an investigation of line fault location in a HVDC system with 2400km long O/H line using two-terminal measurement method is presented. Discrete Wavelet Transform (DWT) is used to detect the travelling waves arriving at the terminals. Merits of using terminal voltage measurements and terminal surge capacitor current measurements for detection of travelling waves is examined. Furthermore, the influence of noise in the fault location accuracy is studied. All simulations were carried out with PSCAD/EMTDC and the fault location algorithm was implemented in MATLAB. 2. LINE FAULT LOCATION USING SYNCHRONIZED DOUBE-ENDED MEASURMENTS A schematic diagram showing the arrangement for line fault location (LFL) using two-end synchronized measurements is shown in Figure 1. Arrival of the travelling waves at each end of the line is time stamped using clock signals obtained from GPS. Distance to the fault location from the rectifier end, xF is calculated by using the travelling wave principle as ( ( (1) 2 where L is the total O/H transmission line length, ti is the inverter terminal surge arrival time, tr is the rectifier terminal surge arrival time, V is the propagation velocity of the travelling = 1

surge. Synchronized two-end measurements allow for simple and accurate fault location for conventional HVDC system with two terminals. The possibility of detecting surge arrival time by measuring (i) the terminal voltages at two ends (VDCRp1 ,VDCIp1) and (ii) the surge capacitor currents at two ends (SCMRp1, SCMIp1) is considered in this paper.
Smoothing reactor Rectifier end
VDCRp1 GPS

xf Fault Surge capacitor

OH line

Smoothing reactor
VDCIp1

Inverter end

SCMIp1

Converter

DC filter

DC filter

Converter

Time stamped Time stamped Calculation of LFL inputs inputs Figure 1 - Schematic diagram of two-end synchronized LFL using GPS

Accuracy of the travelling wave based LFL depends largely on the precise identification of the surge arrival time. For example if the precision of time measurement at a single terminal is only up to 10s, the approximate accuracy of fault location that can be achieved is 3 km (assuming the travelling wave propagation velocity is 300000 km/s). Typically, steep wave front is visible in terminal voltage or current of overhead line with a permanent solid fault. However due to noise and degradation of waveform during the travel, precise detection of the arrival of a travelling surge is harder using conventional edge detection methods such as Short-Time Fourier Transform (STFT) and Finite Impulse Response (FIR) filtering [8][4] etc.. However, wavelet transform can be used to identify this change in either voltage or current. Wavelet transform is a linear transformation similar to the Fourier transform [7]. However, it is different from Fourier transform because it allows time localization of different frequency components of a given signal [7]. In this paper, discrete wavelet transform (DWT) is used to detect incoming travelling wave as it is used by many novel fault location algorithms [5], [2], [7]. Wavelet transform uses basis functions that are known as mother wavelets and are defined as [4]: (2) ( = ( where * means complex conjugate; p is a scaling factor and is the shifting. Performance of the fault location by using family of Daubenchie (db) mother wavelet types was tested in this research. According to the results haar mother-wavelet (also called db2 mother wavelet), was found to be the most suitable. Shape of the haar mother-wavelet is shown in Figure 2 and it is considered the simplest mother wavelet type available. Therefore, it is expected to be computationally less demanding and requires less resources when implementing in hardware.
1 haar (t)

0.5 1

0.5

1.5

t
Figure 2 Haar mother wavelet

In DWT, the scale p changes as powers of 2, i.e.: p0=20=1, p1=21=2, etc. and generally: pj=2j [4] where j is referred to as the level of details. The value of also changes in a discrete 2

manner as powers of 2: at jth detail level, the time shifts are changes as 0=2j0=0, 1=2j1=2j, etc. and generally: n=2jn [4]. DWT of the sampled waveform x(k) can be derived as shown in equation (3): = ( ( (3)

(4) ( =2 2 where: and k is the sample number of x(t),M-Total number of signal samples of non zero ( for the time period. Moreover, after substituting the generalized values to values of and p to the mother wavelet function shown in equation (2); the equation (4) is derived [4]. The DWT, which is also considered as multi-resolution analysis, differs from continuous wavelet transform (CWT) with clear steps in time-frequency plane [4]. The DWT can be used to decompose the input signal in to multiple frequency bands and this can be implemented efficiently as a filter bank as shown in Figure 3 under DWT decomposition [4]. Only a single level is shown in Figure 3 but this can be extended for series of levels by substituting approximation value with the input signal of the next level.

Figure 3 - Structure of one-level DWT algorithm

This implementation is commonly known as Mallat tree algorithm and consists of series of low-pass filters (LPF) and their dual high pass filters (HPF). The circle with downward arrow behind 2 denotes down sampling by a factor of 2. The output xd(n) is called the detail wavelet coefficients while the output from the last low pass filter is referred to as the approximation wavelet coefficient. It is possible to obtain the original signal x(t) through wavelet series reconstruction. The reconstruction can also be carried out efficiently using a tree algorithm as shown in Figure 6 under DWT reconstruction. The filters HPF1 and LPF1 are the inverse filters of HPF and LPF respectively. In Figure 6, the circles with upward arrow behind 2 denotes up sampling by a factor of 2. 3. TEST NETWORK Modified version of the first Cigr benchmark HVDC scheme [9] is selected as the test HVDC system. This test network has 500kV as the nominal DC voltage and it is designed to deliver 1000MW of active power between two 50Hz AC networks. The simplified PI model representing a cable transmission scheme in the original Cigr model [9] was replaced with a frequency dependent distributed parameter model of a 2400km long overhead transmission line in the modified network. Furthermore, a bipolar HVDC configuration is used since most of the present day HVDC systems are built in bipolar configuration, instead of the mono-polar arrangement in the original reference [9]. Original test network does not contain a surge capacitor. Surge capacitor is used to protect converter station equipment from surges travelling alone the overhead transmission line because when a steep wave-front of surge travelling along the DC transmission line is first slowed down by 3

the line surge impedance and then by the surge capacitor. Since the use of surge capacitor current to detect surge arrival as in the existing fault locator installed at the Nelson River HVDC scheme [10] to be investigated, a 20 nF surge capacitor is added to the test network. The presence of a series DC reactor is essential in line commutated type HVDC scheme [11]. However, original test network does not contain a separate smoothing reactor because it is included in the simplified cable circuit. General value of the smoothing reactor is in the range of 0.5-1H [11]. Therefore, 0.5H smoothing reactor is placed in series with the transmission line at both ends. The terminal voltage and surge capacitor current measurements are monitored for large number of simulation cases with different fault locations. The input signals were conditioned assuming 2 MHz sampling rate, 16-bit Analog to Digital (A/D) conversion resolution, and 020 V range before use in the line fault location algorithm to understand the performance under realistic conditions. These A/D parameters are selected after careful testing and considering the availability of commercial A/D converters.
Terminal Voltage
20 18 Rectifier end Inverter end 20 18 Rectifier end Inverter end

Surge Cap.Current

Normalized voltage

14 12 10 8 6 4 2 0 2.402 2.404 2.406 2.408 2.41 2.412

Normalized current

16

16 14 12 10 8 6 4 2 0 2.402 2.404 2.406 2.408 2.41 2.412

Time (s)

Time (s)

Figure 4 - Monitored input signals conditioned by A/D

A sample of simulation results is shown in Figure 4. The waveforms shown are the measured terminal voltages and surge capacitor currents after normalization between 0-20V and quantization with a 16-bit resolution at 2MHz sampling rate during a short circuit fault at 400km from the rectifier end. The input vector limits are taken as 0-2pu for voltage and -1 kA to 1 kA for surge capacitor currents. Therefore values below zero are set to zero due to normalization effect as seen in voltage waveform in Figure 4. This does not have any impact in operation of the line fault locator since it only uses initial change in voltage. In the two-end synchronized measurement based fault location method proposed in this paper, the transient can be recognized by analyzing the DWT coefficients of the measured terminal voltage or surge capacitor current signals. Magnitude values of the DWT detail coefficients were obtained for the corresponding input signals for scale (j) values of level 2 to level 5 for each terminal. The DWT coefficients for the above sample waveforms are shown using voltage at the rectifier end (VDCRp1) and surge capacitor current at the rectifier end (SCMRp1) in Figure 5. Surge arrival point is clearly visible in both input signals with a sharp wave front for each level shown in Figure 5. If a proper threshold value is set then the initial surge arrival point can be easily detected.

Detail coef. mag.

Detail coef. mag.

Level 2 details
0.4 0.2 0 Using VDCRp1 Using SCMRp1

Level 4 details
1.5 1 0.5 0 2.27 2.28 2.29 2.3 2.31 x 10
4

Using VDCRp1 Using SCMRp1

2.27

2.28

2.29

2.3 x 10
4

Detail coef. mag.

0.8 0.6 0.4 0.2 0 2.27 2.28

Using VDCRp1 Using SCMRp1

Detail coef. mag.

samples Level 3 details

Samples Level 5 details

Using VDCRp1 Using SCMRp1

2.29

2.3 x 10
4

Samples

Samples

x 10

Figure 5 - Magnitude values of DWT level details applied to terminal voltage

4. LFL ALGORITHM Simplified fault location algorithm is presented in Figure 6. Initially either terminal voltage or surge capacitor current measurement, which is the input data, is stored to a buffer. DWT is applied to the input buffer and magnitude values of detail coefficients of level 2, 3, 4 and 5 are extracted.

Figure 6 - Simplified DWT based LFL algorithm

A threshold to identify the surge arrival point is set about 10% above the maximum value of the corresponding input signal under the normal conditions. The safety margins are required to allow for the noise. Different threshold values are found for each level of the detail DWT coefficient considered in the algorithm. The time when the magnitude of the considered DWT coefficient rises above the threshold is recognized as the time of arrival of a surge at the terminal. From the measurements at the other end of the transmission line, the time of arrival of the surge in that terminal is received via telecommunication channel. Fault location is calculated by using the travelling wave principle according to equation (1). As different DWT coefficients represent different frequency bands in the signal, the velocity of propagation at each of these frequency bands could slightly differ. These velocities can be found and the algorithm can be calibrated by using test data for a known fault. The algorithm attempts to find an arrival of surge in the current data buffer, and if it did not find an edge, then the buffer window is shifted and the procedure is repeated.

5. SIMULATION RESULTS PSCAD/EMTDC is used to carry out the simulations with 0.1s simulation time step. Voltage at the pole 1 rectifier end (VDCRp1) and the inverter end (VDCIp1) together with surge capacitor currents at both ends (SCMRp1 and SCMIp1) were recorded for many different fault locations at a 2 MHz sampling frequency (0.5s sampling time step). For each case, permanent ground fault was applied between pole 1 and the ground with 0.01 fault resistance. Voltages are represented in per unit value based on 500kV voltage and the surge capacitor currents are represented in kilo amperes. PSCAD/EMTDC output data are imported to MATLAB environment and the fault location algorithm was implemented using MATLAB functions. The respective input signals were scaled and quantized before using in the fault location algorithm to simulate the actual signal acquisition steps. A/D conversion limits are assumed as 0-20 V and for scaling, the input vector limits are taken as 0-2pu for voltage and -1 kA to 1 kA for surge capacitor currents. After careful analysis, 16-bit resolution is selected as the A/D bit resolution. 6. FAULT LOCATION ACCURACIES Prediction accuracies were obtained for different fault locations using voltage and surge capacitor current inputs. Some selected results are shown in Table 1 using terminal voltage and Table 2 using surge capacitor current. Predicted fault locations were measured from the rectifier end is presented in kilometers. Fault locations were estimated using detail DWT coefficients of different levels (level 2, 3, 4 and 5 coefficients). In most cases, the accuracy is varying in the order of level 2, level 3, level 4 and level 5. Lower detail levels give higher prediction accuracy compared with higher levels. Overall accuracy of the system is within 500m limit. According to the results, surge capacitor current can also be used in fault location prediction. These results are also lie within 500 m accuracy limit.
Table 1 - LFL accuracies for different fault locations using terminal voltage without noise as the measurement
FL (km) Level 2 Level 3 Level 4 Level 5 10.00 10.01 9.71 10.60 12.37 100.00 99.96 99.36 100.19 101.83 850.00 850.06 850.06 850.32 852.03 1323.00 1323.12 1323.42 1323.33 1324.34 2040.00 2040.04 2040.34 2040.01 2039.96 2240.00 2239.97 2240.27 2238.90 2238.54

Table 2 - LFL accuracies for different fault locations using surge capacitor current without noise as the measurement
FL (km) Level 2 Level 3 Level 4 Level 5 10.00 11.51 9.71 10.60 12.37 100.00 99.96 99.96 101.38 101.83 850.00 850.06 850.06 850.32 852.03 1323.00 1323.12 1322.82 1322.14 1321.95 2040.00 2040.04 2040.34 2040.01 2039.96 2240.00 2239.97 2240.27 2238.90 2238.54

7. FAULT LOCATION WITH NOISY SIGNALS Performance of the fault location algorithm is tested with input signals contaminated with white noise. As an example, Figure 7 shows surge capacitor current contaminated with 0.001kA noise and terminal voltage contaminated with 0.01pu noise respectively.

LFL using DWT levels

LFL using DWT levels

Overhead line surge cap. currents

Overhead line voltages 0.01pu noise VDCRp1 Original VDCRp1

0
1.12

-0.005

1.1 1.08

Current (kA)

Voltage (pu)

-0.01 0.001kA noise SCMRp1 Original SCMRp1 -0.015

1.06 1.04 1.02

-0.02

1 0.98

-0.025 2.401 2.402 2.403 2.404 2.405 2.406 Time(s) 2.407 2.408 2.409
2.404 2.4045 2.405 2.4055 Time(s) 2.406 2.4065 2.407

Figure 7 Surge capacitor current and terminal voltage with added white noise

LFL accuracies for some test cases using terminal voltage measurement with 0.001pu white noise are listed in Table 3. LFL accuracy is sensitive to noise level in the input signal.
Table 3 - LFL accuracies for different fault locations using terminal voltage with 0.001pu noise as the measurement
FL (km) Level 2 Level 3 Level 4 Level 5 10.00 10.00 10.30 10.60 12.37 100.00 99.89 100.19 100.19 101.83 850.00 849.72 849.72 850.32 852.03 1323.00 1323.03 1323.03 1323.33 1324.34 2040.00 2040.31 2040.31 2040.01 2039.96 2240.00 2240.09 2240.09 2240.09 2238.54

LFL using

Similarly, LFL accuracies using surge capacitor current with 0.001kA white noise are listed in Table 4. In general, surge capacitor current is more sensitive to noise level than the terminal voltage. Performance of this method is poor for very close faults. In general, a better accuracy is achieved with lower levels DWT coefficients.
Table 4- LFL accuracies for different fault locations using surge capacitor current with 0.001kA noise as the measurement
FL (km) Level 2 Level 3 Level 4 Level 5 10.00 11.21 9.71 10.60 12.37 100.00 99.96 99.96 101.38 101.83 850.00 850.06 850.06 850.32 852.03 1323.00 1322.82 1322.82 1323.33 1324.34 2040.00 2040.04 2040.34 2040.01 2039.96 2240.00 2239.97 2240.27 2238.90 2238.54

8. CONCLUSION This paper has investigated the performance of two-ended travelling wave based line fault location algorithm with DWT based surge detection for a 2400km long HVDC transmission line. The applicability of both the terminal voltages and the surge capacitor currents as input signals was examined. The simulation result shows that both inputs signals would produce similar accuracy levels: in the range of 500m under normal conditions for permanent line to ground faults. The analysis considered the scaling and quantization effects of the A/D conversion. Increasing levels of noise degrades the fault location accuracy, especially in the case of faults very close to one of the ends. Generally, higher fault location accuracy is obtained with lower levels DWT coefficients (Levels 2 and 3) of the input signals. However, when the noise is present, use of higher levels (Levels 4 and 5) DWT coefficients may be desirable to achieve robust results.

LFL using DWT levels

levels

DWT

BIBLIOGRAPHY [1]L. Shang, G. Herold, J. Jaeger, R. Krebs and A. Kumar, Analysis and identification of HVDC system faults using wavelet modulus maxima, in Proc. 7th International Conference on AC-DC Power Transmission, Atlanta, Nov. 2001.pp. 315-320. [2]P. Chen, B. Xu and J. Li, A Traveling Wave Based Fault Locating System for HVDC Transmission Lines, in Proc. PowerCon 2006, International Conference on Power System Technology, 2006, pp. 1-4. [3]IEEE Std C37.114-2004, IEEE Guide for Determining Fault Location on AC Transmission and Distribution Lines, 2005 [4]M.M. Saha, J. Izykowski, E. Rosolowski, Fault location on power networks, Springer, 2010. [5]B. Backwel. (2009, July). ABB wins contract for world's longest HVDC link [Online]. Available:http://www.rechargenews.com/energy/wave_tidal_hydro/article184870.ece [6]C. H. Lee, Y. J. Wang, and W.L. Huang, A Literature Survey of Wavelets in Power Engineering Applications, in Proc. of Science Council, ROC (A), Vol. 24, No.4, pp 249-258, 2000. [7]F.H. Magnago and A. Abur, A Fault location using wavelets, Power Delivery, Vol. 13, Oct. 1998 pp.1475 1480. [8]N. Perera, A.D. Rajapakse, A.M. Gole, Wavelet-based relay agent for isolating faulty sections in distribution grids with distributed generators, in Proc. The 8th IEE International Conference on AC and DC Power Transmission, pp. 162-166, March 2006. [9]M. Szechtman, T. Wess, and C. V. Thio, First benchmark model for HVDC control studies, Electra, no. 135, pp. 5467, Apr. 1991. [10]T.W. Radford, HVDC line fault locator upgrade, in Proc. HVDC operating conf., 1987, pp. 189-200. [11]E.W. Kimbark, Direct Current Transmission, Vol. 1, John Wiley & Sons, 1971.