New Method for Transformer Winding Fault Detection

Transcription

1 POSTER 2015, PRAGUE MAY 14 1 New Method for Transformer Winding Fault Detection Martin KNENICKY Department of Electrical Power Engineering, Faculty of Electrical Engineering, Czech Technical University in Prague, Technicka 2, Prague, Czech Republic knenimar@fel.cvut.cz Abstract. This paper contents a description of frequency diagnostic methods for detection of transformer winding condition. This group of considerate offline test methods is very important for a modern power transformer diagnostic. First, principles of often used methods the Low Voltage Impulse Method (LVIM) and the Sweep Frequency Response Analysis method (SFRA) are briefly explained and then the new method using a chirp signal is proposed. The test program is programmed in the software MATLAB. Theoretically, the new method using the chirp signal should provide better evaluation and application. The functionality and performance of these methods is verified by realization of real test measurements and the methods applicability compare to chirp signal method is discussed at the end. Keywords Power transformer diagnostic, transformer winding faults, frequency response analysis, chirp signal. 1. Introduction A power transformer is one of the most important part of power grids. A number of installed transformers in power grids are 5 and more times greater than generators. If some undetected defect is formed in a winding in the transformer, sooner or later a breakdown of this device will occurs. This accident usually totally destroys the transformer and brings very large economic losses caused by the destruction of very expensive device and by partial blackout consequences in the power grid. The possible economic losses are magnified with an increase in transferred electrical power. A high level of maintaining reliability and operability is necessary for transformers of the highest voltages and powers. This assures early detection of an arising defect. Currently, there is a good reason to increase the importance and development of new diagnostic methods in this field. Defects of transformer windings can occur after longterm employment or under an effect of a short circuit or extensive handling activities of the transformer. Correct determination of actual winding condition is a problem. The exact condition of the transformer windings is possible to detect only through opening of the transformer container. Usually it is not possible to perform associated works on the site of installation and special workplace is needed. Also the transportation of a large oil power transformers, which have weights about several tons and large dimensions, is very problematic. Furthermore, internal parts of windings are not visible for control though opening transformer container. This is the reason why various diagnostic methods for detection of winding conditions without opening the transformer container and handling the transformer from the operation site are developed in recent decades. Currently, frequency methods for transformer winding diagnostic are the most important group. Frequency methods can detect winding shifts and mechanical deformations, partial inter-turns short circuit of windings or connection of magnetic core sheets. Possible winding defects are very well described in [1, 2]. 2. Frequency Response Analysis 2.1 Introduction to Frequency Response Analysis A typical high voltage test was at the beginning of transformer winding testing. The first frequency test method was proposed and described in [3]. The frequency response methods are based on the fact that each transformer has own unique frequency response similarly as a human finger print. Fig. 1. A simple equivalent circuit of a transformer winding; Rin input resistance, Rout output resistance, Ls serious own inductance, Cs serious inter-turn capacitance, Cg grounding capacitance, Cb bushing capacitance

2 2 M. KNENICKY, NEW METHOD FOR TRANSFORMER WINDING FAULT DETECTION All frequency methods are based on applying of specific voltage to the input terminals of a transformer winding. The response signal is measured on output terminals of the transformer winding at the same time. The measuring signal is attenuated by influence of winding parasite and major passive components (resistivity, capacity and inductivity) and the measured signal can be shifted at time. The winding as a passive component system is utilized. A simple equivalent circuit of a transformer winding is shown in Fig. 1. The attenuation (in decibels) of signals with variable frequency is given by a relationship: Vin( j ) H( j) 20log Vout ( j ) where V in is input voltage and V out is output voltage of a measured transformer winding. The attenuation is usually plotted as dependence on frequency. Passive components in the equivalent circuit are modified when a fault of winding occurs. The winding frequency response is modified as well. A condition for successfully diagnostic is presence of a reference measurement for comparison of new and old frequency characteristics. Attention is necessarily paid to ambient conditions of the measurement, which can influence results of measurement. For the example, ambient conditions can be remanent magnetization of transformer core, length and type of measurement cables, status of tap changer, temperature of oil, configuration of all unmeasured transformer terminations (grounding, short circuiting or open circuiting), etc. These conditions must be recorded and taken into account at evaluation. For a good quality testing, the application of more measurements is recommended. These measurements should be at least three different connections of unmeasured terminals for a measured terminal. However, a number of measurements are limited by time consumption and some compromise is necessary to find. The suitable selection of measured connections is presented in [4] for various types of transformers. If a reference measurement is not available, then the evaluation of a measurement is performed by comparison with similar transformers or just comparison between windings of the transformer. Small inaccuracies should be ignored at evaluation in these cases. Frequency responses of outer windings are usually the same in contrast to the central winding of the three phase transformer. These differences are caused by an asymmetry position of the windings on the magnetic core [3-5]. A main problem of frequency response methods is a subjectivity of final interpretation. An experienced specialist is needed to evaluate measurements and identify possible fault. Some evaluation procedures are described in [6]. On the opposite the high sensitivity and easy applicability are large counterbalance to the problems of evaluation. (1) 2.2 Low Voltage Impulse Method (LVIM) The LVIM is a first frequency response analysis method and was successfully used for a long time. A lot of experiences and practical aspects of the LVIM are described in [7]. The fundamental principle of this method can be described as follows. A suitable steep voltage impulse is applied to input transformer terminals and frequency response is measured on output terminals. Measured data are filtered from signal disturbance and noise by digital filters. The filtered signals are then transformed to the frequency domain by Fast Fourier Transform (FFT). The attenuation H is calculated from the frequency spectrum using the relationship (1). There are two ways how to interpret results. The first way is a graph plotting of admittance dependency on frequency. The other way is a graph plotting of attenuation dependency on frequency. Results do not depend on input impulse shape. The impulse shape influences just frequency range of the measurement. The LVIM measurement is very fast. These are advantages of the LVIM. Disadvantages are relatively low sensitivity, frequency limit (less than 1 MHz) and application of complicated mathematical operation FFT. In addition, the LVIM has worse repeatability and stability in comparison with the SFRA. The very well detailed description of the LVIM is in [4-5, 7]. 2.3 Sweep Frequency Response Analysis (SFRA) Currently, the SFRA is a more effective and more used type of FRA methods. The reason is a direct measurement in frequency domain. Principle of this method is that individual points of a frequency characteristic are successively determined by successively applying of sine harmonic voltage signals with various frequencies. The signals are step by step applied to input terminals of a transformer winding. Response signals are measured on the output terminals of the winding. Input signals have a constant value of voltage and their frequency is changed in a chosen step for desired frequency range. A signal with changed amplitude and phase shift is measured on the output terminals of the transformer winding. The attenuation H is evaluated using the relationship (1). The interpretation of results is the same as for the LVIM. An increase of a quantization error and a problem with harmonics amplitude are advantages. This is the reason, why measurements are more accurate and more sensitive. Theoretically, the frequency range is unlimited. Practically, the frequency range is to about 10 MHz. Less instrument demand is a significant advantage also. The disadvantage of the SFRA is high time consumption. Many measurements are performed for the entire frequency range. The recommended maximal measure step is 2 % of frequency range. Then measurement of frequency

3 POSTER 2015, PRAGUE MAY 14 3 characteristic takes about 10 min for frequency range 1 khz to 10 MHz. The well detailed description of the SFRA is in [4-6, 8-9]. 3. New Frequency Response Analysis Using Chirp Signal 3.1 Chirp Signal A chirp signal is a harmonic sine signal with variable frequency. Frequency can be varied linearly, quadratically, exponentially, logarithmically or otherwise according to a predefined function in time. A chirp signal is described by a mathematical relationship: x( t ) Asin 2 f ( t ) t (2) where A is signal amplitude, t is time and f(t) is frequency dependence on the time. Relationships of basic frequency dependences on the time for a chirp signal are given by mathematical notations: linear dependence: f f (3) m 0 f ( t ) f0 t Tch quadratic dependence: fm f f ( t ) f t (4) 2 Tch exponential dependence: The above mentioned relationships include initial frequency f 0, end frequency f m, period of the chirp signal T ch (time of frequency change f 0 to f m ) and time t. An example of quadratic chirp signal is shown in Fig Design of the Measuring System The aim of the method using a chirp signal is to imitate the method SFRA without its the greatest weak point of large time demands. A chirp signal is generated by a function waveform generator and that is amplified by a high power amplifier. This signal is applied to an input of a transformer winding. The input and output signals are recorded by an oscilloscope card placed in PC. The synchronization of the oscilloscope card and the function waveform generator is necessary for the successful measurement. The measured data are processed by software. The winding frequency response characteristic is evaluated by use of Fast Fourier Transform (FFT) or Hilbert Transform (HT) or Finding of Maxims (FM). The proposed measure circuit is shown in Fig. 3. Specific used devices for the measure circuit are described in [10]. 0 m 0 t f ( t ) f f f Tch (5) logarithm dependence: fm t f0 Tch f ( t ) f0 10 (6) Fig. 3. Proposed circuit for measurement of winding frequency response characteristic by method using chirp signal The voltage level of an input signal is limited by used waveform generator to several volts. Afterwards, the voltage signal is amplified to several hundreds of volts for feasibility measurements. That leads to use of oscilloscope voltage probes which can influence the frequency response. The maximal frequency independency of measurement should be assured. The frequency independent components of measuring system are required. Fig. 2. Example of chirp signal with quadratic change of frequency 3.3 Verification Measurements An automatic measuring system was created in software MatLab for verification of the method with a chirp signal. It is possible to set all parameters of a chirp signal in control panels (Fig. 4) on the left. These

4 4 M. KNENICKY, NEW METHOD FOR TRANSFORMER WINDING FAULT DETECTION Fig. 4. Control panel of MatLab program for measurement frequency response characteristic by method using chirp signal parameters are initial frequency, end frequency, time of the chirp signal cycle and type of the chirp signal (dependence of frequency change on the time). A lot of measure parameters are possible set also. Measured signals and a frequency response characteristic are plotted on the right side of the control window after the measurement process. A code of the MatLab program is presented and described in [10]. The example of measured input and output signals is shown in Fig. 4. The blue waveform is the measured input voltage signal. The green waveform is the measured output response signal. The red curve is the frequency response characteristic of the both measured signals. It was discovered that the time period of a chirp signal influences resulting characteristics. A waveform of frequency response for time period 1 ms is very different from other time periods about frequency value 150 khz (Fig. 5). Frequency response characteristics are smoother for time periods 5 and 10 s of chirp signals then for time period 1 s. The frequency response characteristics have also shifted their maxims. The problem is caused by a signal reflection in long time chirp signals. This effect is not evident in higher frequencies (Fig. 6). Very time short chirp signals are required to use for an accurate measurement. Fig. 5. Frequency response analyses characteristics with various period time of chirp signal (frequency range 1 to 300 khz) Fig. 6. Frequency response analyses characteristics with various period time of chirp signal (frequency range 300 to 1000 khz) Fig. 7. Frequency response analyses characteristics with various sampling frequency (frequency range 1 to 300 khz) Fig. 8. Frequency response analyses characteristics with various sampling frequency (frequency range 300 to 1000 khz)

5 POSTER 2015, PRAGUE MAY 14 5 Fig. 9. Frequency response analyses characteristics with various type of chirp signal (frequency range 1 to 300 khz) Used different values of sampling frequency did not influence the measurement results (Fig. 7, 8). The reason is reality that the lowest used sampling frequency (5 MHz) is large enough to record quality signal. Recorded measured signals can be unsuitably modified when lower sampling frequency is used. Evaluated characteristics can be been inaccurate and simpler. This effect is unwanted. The type of chirp signal is not important for a correct measurement when time period of a chirp signal is sufficiently short (Fig. 9, 10). The repeatability of measurement was verified by multiple measurements of frequency response characteristic for each setting of measured chirp signal parameters. The verification of measurements was performed for all above mentioned cases. 4. Conclusion The investigated diagnostic method using a chirp signal is very fast, simple and sensitive. These properties are advantages compared with existing frequency response analysis methods. However, it has also some restrictions. Use of a quality high power amplifier is necessary. Electromagnetic disturbance greatly influences measurements. The noise filter of measured signals is complicated. Experienced evaluations are required as well as other frequency response analysis methods. From the measurements some important parameters of chirp signal are discovered. The value of sampling frequency should be at least about 5 MHz. Time period of chirp should be maximal about 1 ms. Shape of a chirp signal is unimportant. Some components of the measuring system must be adapted according to the needs of this diagnostic method. A sufficiently powerful and stable testing voltage source is needed for quality and sensitive measurements. The same stable oscilloscope probes are needed. The maximal Fig. 10. Frequency response analyses characteristics with various type of chirp signal (frequency range 300 to 1000 khz) restriction of electromagnetic disturbance must be provided for measurements. Good measuring devices are necessary for fast measurements. This method is good for future investigation. A lot of other influences to measurement are possible to test. Theoretically, this method may be suitable for online diagnostic for its immediate measurement. But large problems could be great operating noise and creation of special workplaces on an electric grid. Acknowledgements The author wish to thank Ing. Radek Prochazka, Ph.D., Department of Electrical Power Engineering, FEE CTU in Prague, for his supervision and constructive discussion about research described in the paper. References [1] JEZIERSKI, Eugeniusz. Transformátory: Teoretické základy. Praha: Československá akademie věd, p. [2] PETROV, G. N. Elektrické stroje 1: Úvod - Transformátory. Praha: Academia, p. [3] DICK, E.P.; ERVEN, C.C. Transformer Diagnostic Testing by Frequency Response Analysis. In IEEE Transactions on Power Apparatus and Systems, p , IEEE Power & Energy Society, Nov [4] CIGRE Standard WG A2.26. Mechanical-Condition Assessment of Transformer Windings Using Frequency Response Analysis (FRA). P. Picher. CIGRE, 2008., 60 p. [5] PROCHÁZKA, Radek. Diagnostika poruch vinutí výkonových transformátorů. Praha, p. Doctoral thesis. Faculty of Electrical Engineering, Czech Technical University in Prague. [6] SINGH, Jashandeep, et al. Novel method for detection of transformer winding faults using Sweep Frequency Response Analysis. In Power Engineering Society General Meeting 2007, pp. 1 9, IEEE Power & Energy Society, June 2007.

6 6 M. KNENICKY, NEW METHOD FOR TRANSFORMER WINDING FAULT DETECTION [7] DROBYSHEVSKI, Alexandr A. Assessment of Transformer Winding Mechanical Condition by Low-Voltage Impulse Method. In Power Tech Conference Proceedings, 2003 IEEE Bologna, on p. 6 pp., IEEE Power & Energy Society, June [8] HANIQUE, E.; VAESSEN, P.T.M. A new frequency response analysis method for power transformers. In IEEE Transactions on Power Delivery, pp , IEEE Power & Energy Society, Jan [9] KRAETGE, A., et al. Aspects of the Practical Application of Sweep Frequency Response Analysis (SFRA) on Power Transformers. 6'th Southern Africa Regional Conference CIGRE, [10] KNĚNICKÝ, Martin. Detekce poruch vinutí transformátorů. Praha, p. Bachelor thesis. Faculty of Electrical Engineering, Czech Technical University in Prague. About Author Martin KNENICKY was born in Benesov u Prahy, Czech Republic, in He received the B.S. and M.S. degrees in electrical power engineering from the Faculty of Electrical Engineering, Czech Technical University in Prague, Czech Republic, in 2011 and Currently, he works towards the Ph.D. degree in High Voltage Laboratory at the Department of Electrical Power Engineering, FEE CTU in Prague. His research interests are high-voltage testing and diagnostic methods for high voltage cable systems and power transformers.