Untitled

Incipient Fault Detection in 33/11kV Power Transformers by Using Combined Dissolved Gas Analysis Technique and Acoustic Partial Discharge Measurement and Validated Through Untanking Mohd Raffi Samsudin Ahmad Qisti Ramli Ahmad Berhanuddin Young Zaidey Yang Researcher, High Voltage Tenaga Nasional Bhd Universiti Tenaga Tenaga Nasional Bhd TNB Research Sdn Bhd Selangor, Malaysia Kajang, Malaysia Kajang, Malaysia Abstract- Power transformer consists of components which
Electrical fault in power transformer can be categorized into are under consistent thermal and electrical stresses. The
two which are Partial Discharge and Arcing. major component which degrades under these stresses is
The Partial Discharge (PD) is the pre-breakdown of the the paper insulation of the power transformer. The
paper insulation. A PD in a transformer occurs when the electrical fault can develop into thermal fault such as
electric field in a localized area changed in such a way that localized insulation burning or hot-spots. Any fault in the
localized current stream is produced. This localized current transformer can be detected by using Dissolved Gas
will produce current pulses that are measurable at the output Analysis technique. In this paper, the detection of
of the transformer. PD can be classified into three categories electrical and thermal faults in 14 units of 33/11kV, 30
which are voids, coronas due to sharp edge or floating MVA and 15MVA transformers were done by using
components and surface tracking. However, the detection of Dissolved Gas Analysis (DGA). Then, the acoustic partial
the PD created by floating component and sharp edges did not discharge test was carried out to detect the activity and
yield any useful information about the insulation because their locate the source of the electrical fault. All the
appearance is not directly related to the condition of the transformers were untanked and the inspection was done.
insulation. Besides, the PD due to the floating components and From the inspection done, there were a few incipient faults
sharp edges will give rise to the Hydrogen and Methane detected such as overheating due to loose connection,
content in the insulation oil. This will eventually give a false sharp edges, insulation burning, choking effect due to
alarm on the insulation breakdown. Insulation breakdown moisture and surface tracking in On-Load Tap Changer
happened mainly due to PD in voids and small cracks. Voids (OLTC) compartment. As a conclusion, the combination
are defined as gaps in dielectric material which less dense than of the acoustic partial discharge technique and DGA
the dielectric material itself such as gas bubbles in oil that fills the transformer tank, or cracks in the paper insulation. The technique have proved to be a useful tool in detecting and
void region has a lower dielectric constant than the locating incipient faults in the power transformer.
surrounding material, which creates capacitance. Keywords: Partial Discharge, Arcing, Dissolved Gas A partial discharge can then occur when the electric field Analysis, On Load Tap Changer (OLTC) difference across the void if it exceeds minimum breakdown field strength. A PD will eventually develop into arcing. Any electrical fault in the transformer can be detected by using Dissolved Gas Analysis technique. The DGA can be used to Power transformer experiences consistent thermal and differentiate between the types of faults in the transformer. electrical stresses. The major component which degrades However, DGA alone is not conclusive in determining the under these stresses is the paper insulation of the power electrical fault in the transformer. As a complement, acoustic transformer. The insulation paper determines the life of the partial discharge technique was used to detect the electrical power transformer. The degradation of the paper insulation fault in the transformer. In this paper, the detection of will be accelerated with the presence of electrical fault. electrical and thermal faults in 14 units of 33/11kV, 30 MVA and 15MVA transformers were done by using Dissolved Gas Modern Electric Power Systems 2010, Wroclaw, Poland MEPS'10 - paper 14.6 Analysis (DGA). Then, the acoustic partial discharge test was carried out to detect the activity and locate the source of the electrical fault. Finally, the transformers were untanked for inspection and repairing. II. ACOUSTIC DETECTION OF PARTIAL DISCHARGE [1] The main advantage of the acoustic PD detection of partial Figure 2: Acoustic PD Detection Circuit [1] discharge is the location of the discharge occurrence. The principle of the acoustic PD detection is the detection of the The system frequency response (time constant) determines pressure waves generated by the discharge within the most system detection characteristics. The amplitude and insulation which appears as a small explosion. The explosion frequency characteristic of the signal that arrives at the sensor will excite a mechanical wave, and propagates through the and the ambient mechanical background noise determines the insulation. The speed of the acoustic wave propagation sensitivity and signal-to-noise ratio. Acoustic signals generally depends on the surrounding medium [1]. The acoustic wave is decrease at high frequencies due to absorption [1]. This effect shown in Figure 1. The acoustic features such as duration, rise is more for apparatus with large distances between PD sources time, counts, energy and amplitude are used in partial discharge detection. In acoustic PD detection, one need to consider the reflection and refraction, geometrical spreading On the other hand, the acoustic noise increases at lower of the wave and absorption in the materials which will lead to frequencies. The system is optimized through a tradeoff changes of sound propagation. between bandwidth, signal, and noise (Figure 3). Absorption often limits the sensitivity and the thermal noise of amplifiers increases as the square root of the bandwidth. This thermal noise can influence the signal to noise ratio. Sometimes, one discharge often results in multiple signals that propagate along different paths to the sensor. The frequency response of the system also determines which frequency components are detected. As the speed of sound and propagation path vary with wave type and frequency, the choice of sensor and bandwidth determines the appearance of the signals. Figure 1: Acoustic Wave [1] The PD impulses has a short duration resulting compression wave has frequencies in the ultrasonic region [2]. The frequency range is between 10 Hz and 300 kHz. In air and gases, microphones are usually used as sensors. On the other Figure 3: Signal and Noise vs. Frequency, Indicating the Basis for hand, piezoelectric transducers as acoustic emission (AE) Determining the Acoustic Detection Bandwidth that Provides Optimum Signal to Noise Ratio [1] sensors offer the best sensitivity for detection of ultrasonic waves in the enclosure [2]. B. Location of Discharges [3] A. Design of Acoustic PD Instrumentation [1] The possibility of PD activity location is one of the major features of acoustic discharge detection. Location can be Acoustic partial discharge detection apparatus consists of a based on either measurement of the signal arrival time at a sensor, filter, preamplifier, and some type of data acquisition sensor (Figure 4(a) and 4(b)) or on measurement of signal instrument as shown in Figure 2. level. The intensity of the wave decreases as a function of distance from the source when a wave propagates through a becomes complicated. If the electrical signal cannot be detected, a triangulation can be carried out as a simultaneous measurement with several acoustic sensors. In a locus, the source must be located on a hyperboloid between the two sensors. This can be determined from analyses of time lag of the signal. All locations on a hyperboloid have the same time lag between the signal arrivals at the two sensors. If the signal is repetitive, one of two sensors can be moved until the acoustic pulses arrive simultaneously at the two sensors where the location of the PD source is in between them. The simple rule is the signal magnitude will be high if it is close to the source. Then, only one sensor is required and is moved around until the position for maximum signal is located. The frequency of the signal will be higher as it reaches closer to C. Triangular Method for Location [1] Figure 4 (a): Triangulation of Source Location Based on Time of Flight Measurements Based on Measurement of both Electric and Acoustic Methods The triangulation method is time consuming and the result of the PD detection may diverge from it location [1]. As an alternative, the location calculation is derived from the time- distance relationship implied by the velocity of the sound wave. The absolute arrival time, t, of a hit in an event can combine with the velocity, v, of the sound wave to calculate the distance, d, from the sensor to the source [1] as in equation d = v × t (1) The distance between two points depends on the dimension of the object. The majority of the location modes are a variation of 2 dimensional source locations in a plane. In many cases the 2D plane will wrap around a 3 dimensional object. For two points in 2D, the distance equation is just the Pythagorean Theorem expressed in Cartesian coordinates as: d = (x − x ) + ( y − y ) (2) This calculation is complicated because the exact time the event origination is unknown. To overcome this problem, all Figure 4 (b): Triangulation of Source Location Based on Time of Flight the times are considered relative to the first hit in the event. Measurements Based on Acoustic Methods [1] Each arrival time difference for each sensor is referred This results from several mechanisms including geometrical relatively to the distance from the first hit sensor. For the spreading of the acoustic wave, acoustic absorption second hit sensor as relative to the first hit sensor, a difference (conversion of acoustic energy to heat), and scattering of the equation can be written as: wavefront. These phenomena result in a reduction of the t − t = (d − d ) / v (3) intensity of the wave as it moves away from the source. In practical situations, a location based on a time-of-flight The distance equation (2) can be combined with the measurement requires two or more simultaneous difference equation (3) to get: measurements in order to use triangulation to determine the source location. Additional option which will ease the PD t −t = [ (x x)2 (y y)2 x )2 (y detection is to measure the electrical signal simultaneously with the acoustic signal. If the acoustic propagation velocity is known, then calculation of the source location will become s and ys are the unknown coordinates of the source. The equation contains two unknown and cannot be solved by itself. To get a second equation with the same 2 unknowns, a 3rd hit is added to the event to get: However, different wave components travel along different paths in a structure which makes the location determination

t −t = [ x( x)2 (y y)2 x )2 (y y )2 / These simultaneous equations can then be solved for xs and ys .The problem with this approach is that it gives more than one source location per event and if there is any error in the timing values, the source location can be wildly incorrect [1]. A better approach would be to average the data to produce a single location. The equation given in (5) can be solved by using multiple regressions. However, it does not actually average the results of multiple 3 hit calculations directly. If more hits are considered, the equation (4) and (5) and can be Figure 6: Software layout [6] t is the time of arrival for the other sensors. IV. RESULTS AND DISCUSSION A field test was conducted on 14 units of transformer The acoustic partial discharge test was done on 14 units of suspected experiencing partial discharge activity. These transformer for 24hrs. Oil samples were taken and sent to the transformers have been selected based on its hydrogen and lab. The DGA interpretation was done by using Roger Ratio, methane level, which are more than 100 ppm and 50 ppm IEC Ratio, IEEE ratio, Duval Triangle, Key Gas Analysis and respectively [4]. The equipment and the software layout for Doernenburg ratio. In addition, the transformers were the acoustic PD testing are shown in Figure 5 and Figure 6 untanked to perform the inspection internally. Table I summarizes the oil test results and the untanking findings. TABLE I: DGA INTERPRETATION AND UNTANKING FINDINGS. The location of the sensors should be similar with the layout. Oil DGA Interpretation
After the sensors has been attach to the transformer tank, Overheating and Partial Loose Connection at On Load Tap automatic sensor test (AST) was performed. This is to check Changer (OLTC) termination the operation of the sensors and the cabling connections to the Partial Discharge Sharp Edge at OLTC termination sensors. The test was run for 24 hours so that it can capture Partial Discharge and OLTC moving contact tracking the whole day loading cycle of the transformer. At the same Partial Discharge and OLTC moving contact tracking time, oil sample was taken to capture the condition of the oil during testing. After the testing completed, AST was Burned Insulation performed to check the condition of the sensors. A threshold Partial Discharge Wrong Cable lug sizing at OLTC of 45dB was used for the acoustic testing. Partial Discharge Wrong Cable lug sizing at OLTC Partial Discharge High Moisture and Insulation Partial Discharge High Moisture and Insulation Partial Discharge Wrong Cable lug sizing at OLTC Partial Discharge Wrong Cable lug sizing at OLTC Burned Insulation Partial Discharge Sharp Edge at OLTC termination Partial Discharge Sharp Edge at OLTC termination Figure 5: The Acoustic PD equipment and sensors T1 faced some problem with its OLTC termination. The OLTC termination was loose thus created some discharge and hotspot (Figure 7).

Loose connection causes Overheating Figure 9: Some tracking observed at the moving contact (T3, T4) Figure 7: Loose OLTC termination (T1) For the transformer T2, T13 and T14, partial discharge was detected due to sharp edges at the OLTC termination (Figure 8). This type of partial discharge is called corona and will give rise to the hydrogen and methane content in the insulating oil. However, no secondary damage was detected due to this For the transformer T3 and T4, some electrical discharge signals were picked-up by the acoustic partial discharge equipment from On-Load Tap Changer (OLTC) tank. The OLTC compartments were opened and some tracking observed at the moving contact. The tracking is due to the contaminated OLTC oil which becomes conductive (Figure For the transformers T6, T7, T10 and T11, some discharges were observed and distributed around the OLTC compartment. The untanking findings revealed that the Figure 10: Cable Lug with Gap (T6, T10, T11, T7) discharges are due to wrong cable lug sizes used at the termination. The wrongly sized cable lugs created air gap at the termination (Figure 10). Sharp edge caused Corona.
Note: The edge should be
bent instead of extruded out.
Figure 8: Sharp edges at the OLTC termination (T2, T13, and T14) Figure 11 : Burned Insulation [2] M. Muhr, R. Schwarz, "Partial Discharge Measurement as a Diagnostic The remaining transformers (T5,T8,T9,T12) had Tool for HV-Equipments", Institute of High Voltage Engineering and experienced overheating due to PD and some insulation System Management, Graz University of Technology, Inffeldgasse 18, 8010 Graz, Austria, IEEE 2006, Pg 195-198 burnings were detected. These transformers have been [3] DISP with AEWIN User's Manual Rev. 3 November 2005, PAC Part# : proposed to be repaired in the factory (Figure 11 and Figure 6320-1001, Sound Technology for Safety & the Environment, Acoustics Corporation, Princeton Junction, NJ, Copyright@2005. [4] IEC 60599,"Guide to the Interpretation of Dissolved and Free Gas Analysis"1999-03. [5] N.A. Muhamad, B.T. Phung, T.R Blackburn, K.X. Lai, "Comparative Study and Analysis Of DGA Methods for Transformer Mineral Oil" ,Power Tech, 2007 IEEE Lausanne 1-5 July 2007 [6] I. J. Kemp, "Partial Discharge Plant-Monitoring Technology: Present and Future Developments," IEE Proc.-Sci. Meas. Technology, Vol. 142, [7] IEEE Std C57.104-1991,"IEEE Guide for the Interpretation of Gasses Generated In Oil-Immersed Transformers"'1991. [8] F.H.Kreuger,"Discharge Detection In High Voltage Equipment", Temple Press Book, 1964. [9] Mistras Holding Group,"Acoustic PD Measurement Manual", [10] L.E Lundgaard, "Partial Discharge XIV, Acoustic Partial Discharge Detection-Practical Application", IEEE Electrical Insulation Magazine, Vol. 8, No. 5, September/October 1992, pp. 34-43. [11] Yasmin H. Md Thayoob, M.R Samsudin, P S Ghosh and Ahmad B. Abd. Ghani, "Analysis of Partial Discharge Signal Pattern in XLPE Cable under Various Soil Conditions using Self-Organizing Map", IEEE International Conference on Power and Energy (PECon), Dec. 2008, [12] M.J. Mousavi and K.L. Butler-Purry, "A Characterization Methodology for Distribution System Abnormalities Using Wavelet Packets and Self- Figure 12: Burned Insulation Organizing Map Neural Networks", Proc. of the 13th Int. Conf. Intelligent Systems Application to Power Systems, 6-10 Nov. 2005, pp. From the untanking and internal inspection, it has been observed that most of the PD activities in the transformers are [13] M.L Chai, Yasmin H. Md Thayoob, P S Ghosh Ahmad Zuri Sha'ameri due to the manufacturing defect. The defects shown in Figure and Mohd Aizam Talib, "Identification of Different Types of Partial Discharge Sources from Acoustic Emission Signals in the Time- 7, Figure 8 and Figure 10 can be avoided during the Frequency Representation", IEEE International Conference on Power manufacturing process. These kinds of defects will give a and Energy (PECon), Dec. 2006, Malaysia. false alarm on the occurrence of PD in the transformer. [14] C.M. Lee, A.Q. Ramli, P.S. Ghosh, Y.H.M. Thayoob and Z. Wang, Furthermore, the PD activity due to the latter can degrade the "The Effect of Different Partial Discharge Sources on Acoustic Waves Propagation in an Experimental Tank", Proceedings of the XIVth oil thus will reduce the insulation integrity of the transformer. International Symposium on High Voltage Engineering, Tsinghua University, Beijing< China, Aug. 25-29, 2005. [15] Wen-Yeau Chang and Hong-Tzer Yang, "Partial Discharge Pattern Recognition of Molded Type Transformers Using Self Organizing Map", 8th International Conf. on Properties and Applications of Based on the internal inspection done on 14 transformers, it Dielectric Materials, June 2006, pp. 246-249. has been found that most incipient faults can be prevented [16] Shie Qian, "Introduction to Time-Frequency and Wavelet Transform", during the manufacturing process. Typical defect observed National Instrument Corporation, Prentice Hall PTR, 2002. during the untanking were sharp edges, loose contacts and [17] Berkant Tacer and Patrick J. Loughlin, "Instantaneous Frequency and Time-Frequency Distributions", International Conference on Acoustic, wrong cable lug sizing. These kinds of defects will give a Speech and Signal Processing (ICASSP-95), Volume 2, Pages 1013- false alarm on the occurrence of PD in the transformer. On the 1016, 9-12 May 1995. other hand, the lack of OLTC maintenance caused internal [18] Richard G.Lyons, "Understanding Digital Signal Processing", Addison- tracking due to degraded insulating oil. Wesley Publishing Company, 1997. [19] FITST3-31, Facilities Instructions, Standards and Techniques in Transformer Diagnostics. 2003, Bureau of Reclamation Hydroelectric The acoustic partial discharge technique proves to be a Research and Technical Services Group Denver 2003. p. 5-13. useful tool in confirming the presence of partial discharges in [20] Md. Amanullah, et al., Analyses of Electro-Chemical Characteristics of the power transformer. This is supported with the detection of Vegetable oils as an Alternative Source to Mineral Oil-based Dielectric Fluid. 2005 IEEE International Conference on Dielectric Liquids, 2005. incipient faults which have been validated through physical ICDL 2005. , 2005: p. 397 - 400 inspection. The input from the acoustic partial discharge [21] IEC61294, Insulating Liquids - Determination of The Partial Discharge measurement served as additional information in diagnosing Inception Voltage (PDIV) Test Procedure,, E. I.T. Committees, Editor. the transformer incipient faults. 1993, International Electrotechnical Commission: Geneva, Switzerland. [1] L. E. Lundgaard, "Partial Discharge - Part XIII: Acoustic Partial Discharge Detection, 2005

Source: http://meps10.pwr.wroc.pl/submission/data/papers/14.6.pdf