ONLINE MONITORING OF TRANSFORMER HEALTH USING FUZZY LOGIC APPROACH

Transcription

1 ONLINE MONITORING OF TRANSFORMER EALT USING FUZZY LOGIC APPROAC SRIYA SA Project Fellow, Surface Engineering and Tribology, Central Mechanical Engineering Research Institute Durgapur, India Abstract Diagnosis of transformer is carried out using Dissolved Gas Analysis (DGA). There exists various diagnosis methods based on DGA such as Key gas method, IEC, Roger Ratio, Dornenburg ratio etc. But these methods are unable to detect the presence of more than one incipient fault in the transformer. Furthermore, it is difficult to achieve an accurate diagnosis without an expert human operator. Therefore, to simplify the incipient fault detection technique and make it autonomous, the use of Artificial Intelligence tool, Fuzzy logic, has been suggested in this paper. Many experiments have been carried out to illustrate the performance of the proposed fuzzy logic model. The result of various experimental data has been analyzed using Fuzzy logic technique and also compared with the empirical test. The test results indicate significant improvement in the fault detection capability of the system when using Fuzzy logic model. Keywords DGA; Incipient fault; Power Transformer; Expert System; Fuzzy logic. I. INTRODUCTION II. TRANSFORMER FAULTS The power transformer is a major apparatus in a power system. Its well being is very important for the proper functioning of an electrical power system. Wide varieties of thermal and electrical stresses often age the transformer and subject them to incipient faults []. These faults deteriorate the working condition of the transformer and may even cause fatal breakdown. Being one of the most expensive and important element, power transformer is a highly essential component, whose damage and failure may cause the outage of a power system. Devices as, Buchholz relays or differential relays, have evolved to monitor the serviceability of power transformers. But, they respond only to a severe power failure requiring immediate removal of the transformer from service. Thus, preventive techniques for early detection of faults would avoid outage and thereby reduce the losses incurred. One such preventive technique is the Dissolved Gas Analysis (DGA) [-]. Power Transformers are subjected to electrical and thermal stress. This gives rise to formation of gases which can be found by analyzing the insulation oil of the transformer using Gas Chromatography []. DGA results can be interpreted by various methods such as Key Gas analysis, Roger s ratio method, Doernenburg Ratio method and Duval s triangle model [5]. The main drawback of using these methods for fault analysis in its present form is that, many DGA results falls outside the proposed code for fault detection. In this paper, an online monitoring technique has been suggested for incipient fault analysis in Power transformers. Artificial Intelligence tools such as Fuzzy logic has been implemented. The result obtained from the Fuzzy method has been compared to that obtained using traditional ratio method. IEC Publication 6599 [6] provides a coded list of faults detectable by dissolved gas analysis (DGA):. Partial discharge (PD): PD occurs in the gas phase of voids or gas bubbles. It is usually easily detectable by DGA, however, because it is produced over very long periods of time and within large volumes of paper insulation. It often generates large amounts of hydrogen.. Low energy discharge (D): Low energy discharge is caused by tracking, small arcs, and uninterrupted sparking discharges. 3. igh energy discharge (D): igh Energy Discharge gives rise to extensive carbonization, metal fusion and possible tripping of the equipment.. Thermal faults T <3 C (T): T evidenced when insulation paper turns brownish. 5. Thermal faults 3 <T< 7 ºC (T) : T evidenced when paper carbonizes. 6. Thermal faults T > 7 ºC (T3) : T3 evidenced by oil carbonization, metal coloration or fusion. III. DISSOLVED GAS ANALYSIS The two principal causes of gas formation within an operating transformer are electrical disturbances and thermal decomposition. All transformers generate gases to some extent at normal operating temperatures. But under faulty condition, the concentration of some gases like hydrogen (), methane (C), acetylene (C), Ethylene (C), and ethane (C6) increases abnormally. Furthermore, when cellulose insulation is involved, thermal decomposition or electric faults produce methane (C), hydrogen (), carbon di-oxide 6

2 (CO), and carbon mono-oxide (CO). These are the key gases under DGA. The concentration of these key gases is monitored and accordingly fault is predicted [7]. Generally advanced techniques such as key gas ratio methods are used to determine the potential problem within the transformer.the evaluation method applied for Doernenberg ratios and Rogers ratios [8] utilizes the following gas ratios: C/,C, C, C6 and C6. The ratio of COO is sometimes used as an indicator of the thermal decomposition of cellulose. If cellulose degradation is the problem, CO,, C, and C6 will also be increasing significantly. IV. NEED OF AN EXERT SYSTEM The ratio method for fault detection in transformer is powerful as well as highly adaptive. But one shortcoming of using these methods for fault analysis in its present form is that, many DGA results falls outside the proposed code for fault detection. Also, these methods are suitable in predicting only one fault. Another shortcoming is due to the structure of the IEC codes used. Based on the gas ratios, these codes are quantized to define the crisp boundaries of, and. But, in practice, these boundaries are often non-crisp (or fuzzy) especially under multiple faults. As a result, these codes could lead to errors and abrupt changes in the diagnosis result. To solve this problem, a Fuzzy DGA system is proposed. The system represents both IEC 599 [8] and critical key-gas levels for fault prediction. Fuzzy logic system allows processing of data in a manner similar to human mind. Also, this system can be later improvised to learn from its mistakes so that the diagnosis system can match up with specific transformers [9]. V. IEC STANDARD According to IEC 6599 [6], the extended Rogers method is used to produce a three digit code for fault determination. The code is determined based on three gas ratio C, C/ and, C6 as given in Table I. Table : IEC Code determination criteria Gas ratio Value Code X=C X<..<X<3 X>3 Y=C/ Y<..<Y< Y> Z=C6 Z< <Z<3 Z>3 A list of codes along with the faults designated by them is drawn from IEC 6599 [6]. Given below is a table that represents the faulty condition. Table 3: Fault diagnosis Fault X Y Z igh energy discharge(d) Low energy discharge (D) or or X or X Partial Discharge X (DP) Normal(NF) Thermal fault(t) 5⁰<T<7⁰ Thermal fault(t) T>7⁰ VI. or BASIC CONCEPT OF FUZZY LOGIC In the standard set, an element either belongs to or does not belong to a set {,}. In contrast, the fuzzy set enables the description of concepts where the boundary is not explicit [,]. The elements in Fuzzy set are allotted a degree of membership. A membership function (MF) is a curve that defines how each point in the input space is allotted a degree of membership between and. Different membership functions such as triangular, trapezoidal and Gaussian curves are available []. A chosen fuzzy inference system (FIS) is responsible for drawing conclusions from the knowledge-based fuzzy rule set of if-then linguistic statements. Fuzzy inference systems is of two type; Mamdani-type FIS and the Sugeno type FIS []. Mandani has expressive power of formalization and interpretability, while Sugeno FIS has the computational efficiency and robustness. Also, Sugeno FIS is more sensitive in applications where the input is imprecise such as our present case. Thus, we finally take Sugeno FIS for developing the expert system. VII. FUZZY EXPERT SYSTEM The development of fuzzy expert system involves 3 processes namely: fuzzification, fuzzy inference, and defuzzification. A. Fuzzification It converts a crisp gas ratio into a fuzzy input membership value. In the proposed fuzzy diagnosis technique, each crisp value of gas ratio C is represented by a Gaussian Bell fuzzy-membership 7

3 function illustrated in figure. The same follows for the other gas ratios C / and C 6. Screen shot : Membership function Y=C/ Figure : Membership of code, & for C6 B. Fuzzy inference system (FIS) It is responsible for drawing conclusions from the knowledge-based fuzzy rule set of if-then linguistic statements. ere in this paper, Sugeno Fuzzy inference system has been used in the MATLAB Fuzzy logic tool box. C. Defuzzification It then converts the fuzzy output values back into crisp output actions. This step is required to get the output result which predicts the incipient fault in the transformer. Six output functions are defined to predict different fault conditions with reference to table 3. Screen shot 3: Membership function Z=C6 Table : Output Fault code Output Code (U) Fault Type No fault (NF) DP D 3 D T 5 T A. FUZZY MEMBERSIP FUNCTION fuzzy membership functions are constructed for the three gas ratios C, C/ and, C6 and one for the output fault type. The Matlab screen shots of all the four membership functions are shown below. Screen shot : Output variable Fault B. FUZZY RULE BASE The set of three Gas ratios denoted by the membership functions form the premise for fuzzy logic analysis. A fuzzy rule set (linguistic if-then statements) is then used to form judgment on the fuzzy inputs. For example, the first fuzzy rule of figure reads: If C is Code and C / is Code or then fault type is D which maps the first fault listed in Table 3. The other 8 fuzzy rules also map the 6 other fault types and a screenshot of these rules is shown in MATLAB environment in below figure. Figure : Fuzzy rule base Screen shot : Membership function X=C 8

4 C. FUZZY INFERENCE SYSTEM (FIS) The Fuzzy rules appear as strictly defined and similar to the conventional, logic. But the fuzzy membership function allows to interpret the results flexibly and thereby classifying borderline fault cases under two different fault types with individual probabilities of occurrences attached to both of them. Fuzzy Inference System involves the operations between input fuzzy sets, as illustrated graphically in figure 3. The Sugeno FIS derives output fuzzy sets from judging all the fuzzy rules. The solution is arrived by taking the weighted average of the fuzzy output rules. The output window shows the probability of occurrence of one fault or the other. Figure 3: Fuzzy Inference System CONCLUSION The limitations of the conventional DGA methods such as non detection and false alarm are addressed by the Fuzzy Logic based Online monitoring techniques of transformer incipient fault detection. In this paper, an atomization technique of fault prediction has been proposed and implemented successfully. The results have been compared with the actual fault present in the transformer. Comparison of the expert system result with the actual fault justifies the high efficiency and accurate fault finding capability of the proposed system. Sugeno-type FIS has an advantage that it can be integrated with optimization techniques such as neural networks and genetic algorithm so that the expert system can adapt to individual transformer on case by case basis by making the system self learning. REFERENCES D. MATLAB DGA TEST RESULT On testing the fuzzy system with a number of real life data obtained from South Eastern Railways transformers, we obtain the false alarm rate and non detection rate as 6.67% i.e. /3 cases. A sample screen shot of the output results is give below. Screen shot 5: Output Result Window The whole table of data and the inference drawn is given in Appendix A. []. Yoshida, Y. Ishioka, T. Suzuki, T. Yanari and T. Teranishi, Degradation of insulating materials of transformers, IEEE Trans. Electrical Insulation, vol. EI-, No. 6, pp , 987. [] M. Duval, Dissolved gas analysis: It can save your transformer, IEEE Electrical Insulation Magazine, vol. 5, no. 6, pp. -6, 989. [3] T. Kawamura et al, Dissolved Gas Analysis. Its Use forthe Maintenance of Transformers, CZGRE Paper -5,986. [] J. Singh, Y. R. Sood, and R. K. Jarial, Condition monitoring of power transformers Bibliography survey, IEEE Elect. Insul. Mag., vol., no. 3, pp. 5, May/Jun. 8. [5] T. K. Saha, Review of modern diagnostic techniques for assessing insulation condition in aged transformers, IEEE Trans.Dielectr. Elect.Insul., vol., no. 5, pp , Oct. 3. [6] IEC Publication 6599, Mineral oil-impregnated electrical equipment in service Guide to the interpretation of dissolved and free gases analysis, March 999. [7] R.R. Roger, IEEE and IEC Codes to Interpret Incipient Faults in Transformers, Using Gas-In-Oil Analysis, ZEEE Tram. Elect. Znsul., Vol. EI-3, No. 5, pp , 978 [8] IEC Publication 599, Interpretation of the Analysis of the Gases in Transformers and Other Oil-Filled Electical Equipment in Service, 978 [9] J.L.Naredo, P. Moreno, C.R. Fuerte, A comparative study of neural network efficiency in power transformer diagnosis using dissolved gas analysis, IEEE Trans. Power Delivery, vol. 6, pp , Oct.. [] J. Yen and R. Langari, Fuzzy Logic. Pearson Education,. [] A. aman, N. D. Geogranas, Comparison of Mamdani and Sugeno Fuzzy Inference Systems for Evaluating the Quality of Experienceof apto-audio-visual Applications, AVE 8 IEEE International Workshop on aptic Audio Visual Environments and their Applications, 8 [] ong -Tzer Yang; Chiung-Chou Liao, Adaptive fuzzy diagnosis system for dissolved gas analysis of power transformers, IEEE Trans. Power Delivery, vol., pp. 3-35, Oct [3] C.E. Riese, J.D. Stuart, TOGA-An Expert System for Transformer Fault Diagnosis, Artificial Intelligence Applications in Chemistry, American Chemical Society Ed. 9

5 Sl. No 9. 3 APPENDIX C C6 C C X=C/ C Y=C/ Z=C 6 Actual Fault Fuzzy Logic Interpr etation T T D D D D D D NF D T T NF NF T T NF D T T T T T T D D T T NF DP