Transformer Condition Monitoring based on Neural Networks & Fuzzy Logic

Generation, transmission and distribution of electric energy are to be characterized by a very high degree of reliability. Reliability level is directly related to the quality of the equipment used and the periodic maintenance carried out on the system. In our country, the distribution system is the most neglected part, which is being given some importance only in the recent years. Distribution Transformer failure rate is very high due to poor workmanship, not adhering to quality at manufacturing site and lack of proper maintenance at site. This has lead to non availability of power to end-customers and also large cost involved in the in the repair of these equipments. Power sector is the prime mover of any country’s economic growth and hence the determinant of development. Hence, increasing plant load factor and cutting down transmission and distribution losses would be interesting. From a technical angle, any failure in the insulation system of transformer due to various types of faults namely, permanent ground fault, improper voltage distribution in winding. This will result in over stress to some parts of the winding. Identifying the type and location of fault is a critical factor. But, transformer maintenance is neglected area of power distribution management. Knowledge based systems like neural networks incorporated with fuzzy logic are one of the reliable approaches to diagnose the transformer faults. Hence, investigation in the proposed research is aimed to develop software, adopting Fuzzified Neural Networks (FNN) and on-line testing arrangements involving traveling wave technique for locating the transformer faults. The main purpose is to develop software along with the fabrication of on-line testing arrangements for transformer condition monitoring as a basis for preventive maintenance, ultimately leading to transmission and distribution loss reduction. One of key areas of concern identified in performance of power sector is the massive loss during transmission and distribution (covering technical and commercial losses). Enormity of the problem can be gauged when one looks at the vast expanse of the system, vast size of transformer population, the technical diversity, and geographical distribution. Virtually, the system is open ended and performance greatly depends on the honesty and commitment of the personnel in the field and extent of public co-operation. Identifying the type and location of fault in the transformer is a critical factor in this regard. Knowledge based systems like Neural Networks incorporated with Fuzzy logic is one of the reliable approaches to diagnose the failure. This will involve • Developing electrical equivalent circuit model of transformer • Simulating the fault under consideration in the model, and • Obtaining the training data by using simulation software. Neural Network software is proposed to be used to design, train and test appropriate Fuzzified Neural Networks (FNN). A FNN so trained will give the fault location with acceptable accuracy, after defuzzification. The proposed work will have the following objectives: • Developing software package for determining transformer fault for power distribution transformer of chosen specification. • Determining the transformer faults so that preventive maintenance of the transformer population can be carried out more effectively. • Laying the foundation of Transformer Condition Monitoring System so as to introduce preventive maintenance system for transformers. Any electrical apparatus (or device / instrument) invariably involves an electrical insulation system. Experience shows that more than 95% of all failures of electrical apparatus are due to failure of electrical insulations even when the conductor system is properly designed. It is therefore obvious that a long and satisfactory performance of electrical apparatus can be ensured by proper design, incorporation and maintenance of the electrical insulation. • In real life, it is essential to verify whether the insulation is properly designed • Properly incorporated by satisfactory manufacturing techniques under appropriate Conditions • Exhibiting expected performance during service. Such verification is carried out by • design tests by manufacturer • Acceptance and routine tests before installation at the manufacturer’s premises or other laboratories • Periodic tests during service at site or, if deemed essential, at other labs. Above tests are essentially high voltage tests, which have been established by long practice, and national/international standards exist for most of tests. Sometimes, the manufacturers/users may desire to carry out non-standard, but probing tests with specific aims. Obviously, such testing requires a reasonable good high voltage laboratory and skilled dedicated personnel to carry out the tests and interpret the results properly. All reputed electrical industries in the developed world have their own high voltage laboratories or use independent laboratories to ensure quality of the equipment manufactured by them. High voltage laboratories tend to be very expensive, need skilled dedicated personnel for proper operation and experts for interpretation of the results. Unfortunately, the middle and lower level electrical industries in our country cannot afford to own such laboratories. On the other hand, laboratories abroad have many disadvantages like • Indefiniteness about availability of laboratory facility at the desired time • Transportation & accident hazards of damage during transit • Could be very expensive for the lower & middle level industries. The result is that, substandard products have flooded the market and frequent avoidable and expensive failures are spoiling the electrical power scenario. Tests that can be carried out Following tests are carried out for condition monitoring of Transformer: • Power frequency voltage withstand and flashover test on insulators, Porcelain bushings, Condenser bushing, Isolators and Surge arrestors • Lighting impulse voltage withstand and flashover tests on Insulators, Porcelain bushings, Isolators, surge arresters, Potential transformers and Current transformers • Residual voltage test on Surge arresters • Capacitance and Tan measurements • Oil breakdown voltage measurements • Partial discharge and Radio noise measurements (Provided ambient noise permits measurement). Many other custom required tests within the capabilities of the equipment can also be thought of. With doorstep availability of above evaluation & testing facilities at a reasonable price (without very heavy investment), it is expected that the quality of electrical products will go up very significantly. It is proposed to identify the type and location of the faults in power transformer by traveling wave technique. Neural networks incorporated with fuzzy logic (FNN) can be effective to diagnose the fault. The procedure involve developing equivalent circuit model of transformer, simulating the fault under consideration in the model and obtaining the training data by traveling wave technique using simulation software. Neural Network software is proposed to be used to design, train and test appropriate Fuzzified neural networks (FNN). FNN so trained can give the fault location with acceptable accuracy after defuzzification. Working principle of ANN based Transformer condition monitoring is as follows: • Traveling wave is applied to the transformer winding at one end and is received on the opposite end. • Similarly, the experiment is repeated with traveling wave being propagated with the other winding. • Received wave forms are captured using the digital oscilloscope, stored in the PC and compared with the standard signal. • Neural Networks examine the difference and decide the fault and location of the faults. Advantages of condition monitoring are that one can identify the type and location of the fault in power transformer through on-line testing and interpretation of test data using suitable software that would be developed for the purpose. This can lead to reduction in system stress and thereby increase the life of all electrical equipment used by distribution companies. The proposed system will help to improve the life of transformers thereby increasing profitability of the organization (by reducing the replacement cost).Directly certain other cost like repair cost can be monitored more effectively and thereby controlled. This approach will help in the following ways: • Servicing parameters and level of servicing can be improved substantially by effective detection of monitoring system or transformer faults. • Further, same system can be utilized for product testing at purchase and stores system, so that good transformers without any fault can be purchased. • Software would be developed keeping in mind the managerial control system and can be upgraded to web based central controlled system for transformer fault detection and monitoring system. • System design will also keep in mind line loss minimization and control system. G.S.Sheshadri Professor&HOD. Dept.of E&EE CIT,NH206 Gubbi.Tumkur-572216