Y.R. Sood. N.C. Joshi Department of Electrical Engineering, National Institute of Technology, Hamirpur , INDIA

Transcription

1 MIT International Journal of Electrical and Instrumentation Engineering, Vol. 2, No. 2, Aug. 2012, pp. (77-81) 77 Transformer Internal Winding Faults Diagnosis Methods: A Review N.C. Joshi navinjoshi.gbpec@gmail.com R.K. Jarial larji2001@yahoo.co.in Y.R. Sood yrsood@gmail.com Rakesh Thapliyal National Institute of Technology, Hamirpur , INDIA ABSTRACT A transformer is a vital component in any power system network and its catastrophic failure may cause disturbance in the proper operation of electrical power system network. Therefore, it requires continuous monitoring and diagnostics of its operation. This paper presents the several on-line and off-line techniques related with internal winding fault detection in transformers. This paper also reviews and compares the different diagnostics methods based on their advantages and limitations. This highlights a further scope in decision-making for utility engineers for monitoring and diagnostic of the internal winding faults. Keywords: Diagnostics methods, internal winding faults; monitoring and transformer. I. INTRODUCTION The faults occur in a transformer are classified in two types: external and internal faults. External faults are those that occur outside of the transformer: overvoltage, over-fluxing, under frequency, and external system short circuits. Internal faults are those that occur inside of the transformer: winding turn-to-turn, turn-to-ground, over-fluxing [1]. From the last few decades, a continuous growth has observed in the power system and the progress will continue in the upcoming years. A transformer, being an integrated part of the power system, is an important link between a generating power station and a point of power utilization. Due to various kinds of intricate loads and their control systems, transformer is prone to faults. Internal winding faults in transformers can cause huge damages in a very short time, and in some cases the damages are repairable[2], and also about 70%-80% of transformer failures are caused by internal faults[3]. Among these faults, winding turn to turn fault is challenging to monitor and detect, especially at lower magnitude of the fault current. So locating fault is a necessary work for repairing the faulted transformers, which can be repaired and put back in service again. There are many different techniques available for detecting and/or locating these kinds of faults such as: off-line techniques, high frequency analysis [3], Frequency Response Analysis (FRA)[4, 5], Artificial Neural Networks (ANNs) and winding transfer functions[6], finite element analysis[7], online diagnostics techniques park vector approach[8] combination of discrete wavelet transforms and backpropagation neural networks[9], experimental studies [10], [11] etc. II. METHODS IN PRACTICE Dissolved Gas Analysis (DGA) DGA is one of the earliest tool for the diagnoses of internal winding faults in transformers, which had been in practiced 92 years ago. This method is based on analysis of dissolved gases inside the oil of transformer. DGA is one of the popular diagnostic methods, which stands for Dissolved Gas Analysis. Hydrocarbon oils are used in transformer as insulating and coolant fluids because of their high dielectric strength, better heat transfer capabilities. Insulating material is generally decomposed under electrical, thermal, mechanical stresses. The decomposition generates various gaseous products, which gets dissolved in mineral oil and reduces its dielectric strength. The main gases that are generated and dissolved in oil of the transformer are acetylene (C 2 H 2 ), ethylene (C 2 H 4 ), methane (CH 4 ), hydrogen (H 2 ) and ethane (C 2 H 6 ). The ratio of these

2 MIT International Journal of Electrical and Instrumentation Engineering, Vol. 2, No. 2, Aug. 2012, pp. (77-81) 78 gases is used to distinguish between the normal and the faulty operation condition of the transformer. There are two generally used conventional gas ratio methods namely Doernenburg ratios method and Rogers ratios method [12]. These two methods have been used to detect the internal winding faults in transformer. The collected DGA data information provides detail about the health of the transformer and provides early warning of developing faults. Table 2(a): Roger s Ratio Codes (a) Doernenburg Ratios Method This method utilizes the gas concentration from ratio of CH 4 /H 2, C 2 H 2 /CH 4, C 2 H 4 /C 2 H 6 and C 2 H 2 /C 2 H 4.This method is used to detect three types of faults (1) Thermal fault (2) Corona fault(low energy partial discharge) (3) Arcing (high intensity partial discharge)[12]. The value of the gases at first must exceed the concentration A1 to indicate whether there is really a problem with the unit or not and then we check the ratio of generated gases for indication of type of fault [13]. Table 1(a) shows the key gases and their concentration A1 and 1(b) shows fault according to gases ratio. The limitation of this method is, it can detect only three types of faults. Table 2(b): Classification of faults based on Roger s Codes Table 1(a): Concentration level for Doernenburg Ratio Table 1(b): Fault diagnosis for Doernenburg Ratio (b) Roger s Ratios Method The Roger s method makes use of four gas ratios: CH 4 /H 2, C 2 H 6 /CH 4, C 2 H 4 /C 2 H 6 and C 2 H 2 /C 2 H 4. This method indicate five different types of faults such as (i) low temperature fault, (ii) medium temperature thermal fault, (iii) high temperature thermal fault, (iv) partial discharges and (v) high energy arcing. Diagnosis of faults is done with a coding scheme based on ranges of the ratios as shown in table 2 below[14]. Advantage of Roger s method over Doernenburg method is that it can indicate more number of faults. The different combination of the coding provides 12 different types internal winding faults in transformer. The type of faults based on the code is shown in Table 2(a) and Table 2(b) below[14]. Frequency Response Analysis (FRA) Frequency Response Analysis technique involves basically measuring the impedance of the windings of the transformer with a low voltage sine input varying in a wide frequency range and it is one of the most accurate in detecting winding displacements. Dick and Erven initially demonstrated this method in 1978 [15]. In FRA, the impedance is consider as a strong function of frequency and called transfer function (TF). The impedance is measured over a wide range of frequencies and the obtained results are than compared with the reference set of results. If any deviation occurs in these two types of results indicate the winding fault inside transformer. How a fault detected by FRA is shown in Figure 1. Figure 1: Graphical Differentiation of FRA Measurements

3 MIT International Journal of Electrical and Instrumentation Engineering, Vol. 2, No. 2, Aug. 2012, pp. (77-81) 79 Although the fault detection method is easy, the FRA technique could show two disadvantages: It is difficult to find exactly which part of the transformer has failed. No limits are fixed in order to know how much difference is required to represents a real fault or only a discrepancy between measurements. One of the main reasons of the disadvantages in the diagnose procedure is the complex connections among the different parts inside the transformer which are reflected in the frequency response and make difficult to distinguish the different parts that make up the transformer (core and windings). One another limitation in this method is the impedance measurement, mainly below 10 khz, is observed to be significantly dependant on DC magnetization and demagnetization [16]. ANN based fault detection in transformer Artificial Neural Network (ANN) is an advance fault detection method used in a transformer. It is good tool for the applications such as pattern classification, function approximation, system identification[17], [18] due to inherent characteristics of learning. A multi-layer feed-forward neural network is used to detect the fault and distinguish between them[18]. In multilayer feed-forward network gas ratios considered as a input and the output is different types of internal faults Figure 2. Training of a multilayer network is done with the help of apply input. A multi layer artificial neural network is trained for the various differences in the winding functions and then, it acts as a decision making tool to identify fault in the winding [19]. One of the advantages of ANN method for detecting internal faults inside the transformer is that it can deal with the complex situations such as missing data. In order to use ANN in detecting the inter-turn fault, large number of fault cases data needs to train the neurons [20]. Disadvantage of this advanced method is computation and requirement of large memory and hence, require large number of processors and instruments. Figure 2: Multi-layer feed-forward network Wavelet transforms based fault detection Wavelet transform is a mathematical tool for analysis of signal. it is the extension of Fourier analysis and can determine the exact time at which particular frequency occur in the signal. Which cannot be determining with the help of Fourier analysis? In fact, a signal can be representing in terms of finite length or fast decaying oscillating waveform. In wavelet transform we can decompose the signal to obtain time history of different frequency band. Wavelet analysis uses various popular mother wavelets such as Daabechies wavelet, Haar wavelet, Sym, Bior and Coiflet for analysis of signal into a set of approximate and detailed coefficients. The signal analysis with wavelets is found in many applications related to Engineering, Applied Mathematics, Physics and other Sciences [21],[22]. One of the advantages of wavelet analysis is that it used as a online fault detection technique. It uses multiresolution analysis (MRA) approach and that was introduced by Mallat in [23]. Multiresolution analysis (MRA) signal decomposition is effective for extracting the information of the signals. The MRA is obtained through a filter bank build up of band-pass filters. In conventional MRA every down sampling gives as a result a detail through a high-pass filter, while the low-pass filtered portion gives rise to an approximation Figure 3. This latter is successively passed through another set of two band-pass filters, what leads to another detail at a higher level of resolution plus a new approximation. In order to attain a proper solution the filters must have some well known properties, which are summarized in the known as mother wavelets and scale functions. Figure 3: Wavelet decomposition tree RVM (Reverse Voltage Measurement) Technique It is an offline diagnostic tool for detecting the moisture content, ageing and overheating of insulation material. It has been known for years that charging and discharging characteristic of insulation in influence by material impurity, in presence of particular moisture content. The paper insulation in a transformer is of crucial for continuous and proper operation of the transformer. Therefore RVM technique has been developed to assess the condition of insulation paper without opening the transformer. Recovery Voltage Measurement (RVM) is a technique based on measurement of voltage maxima across the paper insulation of transformer after a charge discharge cyclic process. This technique gives good spread information about the moisture content and aging process in the transformer insulation. The concept of RVM originated in the mid 1970 s from the Budapest Technical University in Hungary [24]. In this technique, a specialized instrument is used to generate waveforms of the type shown in Figure 4. This voltage

4 MIT International Journal of Electrical and Instrumentation Engineering, Vol. 2, No. 2, Aug. 2012, pp. (77-81) 80 cycle is used to apply on insulation of transformer. First we apply DC step voltage and kept it for a charging period of Tc. This time period is called charging period for insulation of transformer. After that some time delay period (Td) is provide and in this period the insulation is discharged by shorting it. After the short circuit released, the charge bounded by the polarization will turn into free charges (i.e. voltage build-up between the electrodes on the dielectric). This phenomenon is called the return voltage [24]. To overcome this problem park negative sequence method is used [26]. It is one of the effective techniques for diagnosing the occurrence of internal winding faults in the windings of operating transformer. In this technique primary and secondary phase currents are measured and negative sequence differential current is calculated. The magnitude of this negative-sequence current signifies whether the fault is internal or external. However, with this approach, it is difficult to discriminate between unbalanced loads and winding faults. Also this method unable to detect the fault when the transformer is unloaded. Table 3: Outlines of the methods Figure 4: RVM waveform The instrument measures this voltage as short circuit releases and note down the maximum value of the voltage up to which it build and the time required for it to reach maximum value [24], [25]. This cycle is repeated for many times and the maximum recovery voltage and the time to obtain. Once this cyclic process is completed, the various measured maximum voltages are plotted against the respective charging times (Tc). The time at which the maxima of plotted curve occur is indication of the moisture content and aging of the insulation. This maxima is strong function of moisture content in the insulation. The more the moisture content for a given temperature, the more the maxima will shift to the left. Thus an early maxima indicates high moisture content and late maxima indicates low moisture content. Park Negative Sequence Method Protective devices are a crucial part for detecting fault conditions in a power system. The appropriate protection scheme must be selected to ensure the safety of power apparatus and reliability of the system. One that type of protecting scheme is the differential protection scheme which is aimed at detecting internal winding faults in transformer. Differential protection is one of the most used methods for protecting transformers against internal faults. The technique is based on comparison of currents both primary and secondary side of transformer. Taking into account voltage ratio and vector group adjustment, the related relay trips whenever the difference of currents magnitude in both sides crosses the limit. Although, this protection scheme most accurate one, is subjected to false operation in some special cases. Such as problems related to a mismatching between the transformer ratio and the CTs ratios, magnetizing inrush current. III. CONCLUSION This paper presents different diagnostics methods for internal winding fault analysis in transformers, which are employed in practice. This paper presents an ample review of all these methods that can help the diagnosis of deferent type of internal faults. Every method have their advantages and disadvantage, which is summarized in Table 3. This table can help the diagnostic industries, which method is suitable for their product. REFERENCES [1] IEEE Std C , IEEE Guide for Protective Relay Applications to Power Transformers. [2] J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering, Transformer Protection, John Wiley & Sons, Inc1999. [3] M.R. Barzegaran and M. Mirzaie, Detecting the position of winding short circuit faults in transformer using high frequency analysis, European Journal of Scientific Research, Vol. 23, 2008, pp [4] A. Shintemirov, W.J. Tang, W.H. Tang, and Q.H. Wu, Improved modelling of power transformer winding using bacterial swarming algorithm and frequency response analysis, Electric Power Systems Research, Vol. 80, 2010, pp

5 MIT International Journal of Electrical and Instrumentation Engineering, Vol. 2, No. 2, Aug. 2012, pp. (77-81) 81 [5] M.A. Abdul Rahman, H. Hashim and P.S. Ghosh, Frequency response analysis of a power transformer, Electrical Engineering Department, College of Engineering, University Tenaga National. [6] M. Faridi, M. Kharezi, E. Rahimpour, H.R. Mirzaei and A. Akbari, Localization of turn-to-turn fault in transformers using artificial neural networks and winding transfer function, International Conference on Solid Dielectrics, Potsdam, Germany, July 4-9, [7] H. Wang and K.L. Butler, Finite element analysis of internal winding faults in distribution transformers, IEEE Transactions on Power Delivery, Vol. 16, July [8] L.M.R. Oliveira and A.J. Marques Cardoso, On-line diagnostics of transformer winding insulation failures by Park s vector approach, Proceedings of the 9th International Electrical Insulation Conference, pp , Berlin, Germany, June 18-20, [9] A. Nagopitakkul and A. Kunakorn, Internal fault classification in transformer windings using combination of discrete wavelet transforms and back-propagation neural networks, International Journal of Control, Automation, and Systems, Vol. 4, No. 3, pp , June [10] L. Satish and Subrat K. Sahoo, Locating faults in a transformer winding: an experimental study, Electric Power Systems Research, Vol. 79, pp , [11] P. Palmer-Buckle, K.L. Butler, and N.D.R. Sarma, Characteristics of transformer parameters during internal winding faults based one experimental measurements, IEEE Transmission and Distribution Conference, [12] IEEE guide for the Interpretation of Gases Generated in Oil- Immersed Transformers, ANSI/IEEE std. (1991) C [13] C , I., IEEE Guide for Interpretation of Gases Generated in Oil-Immersed Transformer, I. The Institute of Electrical and Electronic Engineers, Editor. 1992, The Institute of Electrical and Electronic Engineers, Inc p. 27. [14] Siva Sarma, D.V.S.S. and G.N.S. Kalyani, ANN Approach for Condition Monitoring of Power Transformers using DGA IEEE Region 10 Conference, TENCON C: p [15] E.P. Dick and C.C. Erven, Transformer diagnostic testing by frequency response analysis, IEEE Trans. Pwr. App. Syst, vol. PAS-97, No. 6, nov/dec. 1978, pp [16] N. Abeywickrama,Y.V. Serdyuk, and Stanislaw M. Gubanski, Effect of Core Magnetization on Frequency Response Analysis (FRA) of Power Transformers, IEEE Trans. Pwr. Deliv. vol. 23, no. 3, Jul. 2008, pp [17] R.P. Lippmann, An introduction to computing with neural nets, IEEE ASSP Mag. 4 (2) (1987) [18] N. Yadaiah, B.L. Deekshatulu, L. Sivakumar, V. Sri Hari Rao, Neural network algorithm for identification of parameters of a dynamical systems involving time delays, Int. J. Appl. Soft Comput, Jul 2007, pp [19] H. Firoozi, M. Kharezi, and H. Bakhshi, Turn-to-Turn Fault Localization of Power Transformers Using Neural Network Techniques, Int. Conf. Pro. Appl. Diel. Mat. B-11, Jul 2009, pp [20] M.R. Rao and B.P. Singh, Detection and localization of interturn fault in the HV winding of a power transformer using wavelets, IEEE Tras. Dielec. and Elec. Ins., vol. 8, no. 4, aug 2001, pp [21] R.M. Rao, A.S. Boparidkar, Wavelet Transforms An Introduction to Theory and Application, Pearson Education, India, [22] K.L. Butler-Purry, M. Bagriyanik, Identifying transformer incipient events for maintaining distribution system reliability, in: Proc. of the 36th Hawali Intl. Conf. on System Sciences, [23] S.G. Mallat, A Theory for multiresolution signal decomposition The wavelet representation, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 11, No. 7, pp SC Computer Science, [24] Giuseppe M. Urbani, Roger S. Brooks, Using the Recovery Voltage Method to Evaluate Aging in oil-paper Insulation, International Conference on Conduction and Breakdown in Solid Dielectrics, June 22-25, [25] Gustav Csepes, Istvan Hamos, Roger Brooks, Volker Karius, Practical Foundations of RVM (Recovery Voltage Method for OiVPaper Insulation Diagnosis), CEIPDP8. [26] Z. Gajic, Method and device for fault detection in an n-winding three phase power transformer, US Patent, US 2009/ A1, Mar