Computer Assisted Detection of Interturn Short-Circuits in Rotor Windings

Transcription

1 Computer Assisted Detection of Interturn Short-Circuits in Rotor Windings José Tomás Ramírez Niño and Alberth Pascacio de los Santos Paper originally presented at the Computer Application in Power, IEEE, october Vol. 14, num. 4 Resumen El problema de detectar y localizar cortocircuitos en devanados de rotores para turbogeneradores ha representado siempre todo un reto tecnológico. Estos devanados están en serie y completamente embebidos en el acero de la flecha. El aislamiento entre vueltas de los devanados está sujeto a grandes esfuerzos mecánicos producidos por la velocidad de rotación. Por tanto, es común que los cortocircuitos se presentan solamente cuando el generador está en operación. La técnica propuesta por la Gerencia de Equipos Eléctricos del Instituto de Investigaciones Eléctricas permite evaluar los rotores en operación. Además de proporcionar datos de la localización de los cortos en el rotor, proporciona información del número de vueltas en cortocircuito. Con esta técnica se propone incrementar la confiabilidad en la detección de problemas en rotores mediante el procesamiento estadístico de las señales medidas por medio de una bobina exploradora instalada en el estator. 46

2 Artículo de investigación Introduction The integration of measurement systems based on different types of sensors is now possible and computers play an important role in it. These systems rely on monitoring many variables of electrical machines and on sophisticated data analysis and databases. In recent years, the Electric Equipment Department of Mexico s Instituto de Investigaciones Eléctricas (IIE) has been working on the development of equipment that integrates electrical measurement systems, combining the sensors output of the electric machines and features that computers presently provide. The role of the computer is determinant because only with its help can the statistical analysis needed to obtain the desired results be completed. Comisión Federal de Electricidad, one of the two Mexican electric power utilities, and Petroleos Mexicanos, the Mexican oil company, have been using this IIE s development. The quality of service in the electric industry depends on the reliability of its equipment. The current tendency in the industry is to continuously monitor all critical variables of power equipment that are essential for highquality service. The electric power industry has special problems regarding instrumentation. It requires control and monitoring of a great number of variables of the electric equipment in a high electromagnetic noise environment. Particularly, power generators are critical electrical equipment and much effort developing diagnostic techniques has been made in order to improve their reliability. Short circuits in rotor winding Short-circuited turns in power generator rotor windings produce operational problems such as high vibration levels and therefore its early detection is convenient. Usually, interturn short-circuits appear only during operating conditions, due to mechanical, electromagnetic or thermal stress conditions. Therefore, it is necessary to use on-line monitoring in order to detect these short circuits. Normally, the resistance of the windings on opposite poles is identical, for example the winding A1 and B1 in Figure 1 are identical. The heat produced by the Joule s effect is distributed symmetrically about the quadrature axis. If the interturn insulation is damaged so that two or more turns of the winding become short-circuited, then the resistance of the damaged coil diminishes and less heat is generated 47

3 than in the symmetrical coil in the opposite pole. The rotor body thus experiences asymmetric heating, which produces a thermal bow in the rotor body, causing vibration. The unbalanced magnetic forces on the rotor, produced by the change in the mmf (magneto-mechanical force) from the winding, contribute also to increase the vibration. If a shorted turn occurs in a coil close to the pole face, the increase in the bearing vibration may exceed the maximum level recommended for continuous operation. Therefore, this operation condition forces to limit the excitation rotor current, thus restricting the unit s flexibility and operability. The problem of detecting and locating shorted turns in rotor windings has always been a difficult technological challenge. Since the winding is completely embedded in the steel and the turns insulation is not physically loaded until it is subjected to the centrifugal forces produced above certain speed, the problem of detecting shorted turns appears to be very cumbersome. Interturn short circuit detection The comparative measurement of the magnetic flux density produced by each winding is one of the current techniques available for short circuit detection in the rotor s winding. Placing a measuring coil tangentially and near to the rotor does this measurement. The schematic set-up for a typical rotor with six windings in each pole is shown in figure 1. This technique can be applied solely for non-salient pole machines and either when the rotor is under testing out of its stator or installed in a stator slot when the test is on line. The tangential orientation of the search coil is preferred to the radial orientation because in the latter the main magnetic flux density is larger than the magnetic flux density produced by the individual windings and masks their separate behaviour. Ideally, the search coil should be placed as close as possible to the rotor surface. In practice, the measurement is obtained from a stationary search coil placed near the rotor surface but in the stator slot, although the signals provided by the search coil are disturbed by the magnetic flux density in the stator. Figure 1. Schematic diagram of a rotor with three windings and the search coil. Figure 2. Detail of the search coil in the installation process. Pictures of the search coil installation in the stator slot are shown in figure 2. The detection technique is based on the construction symmetry of the rotor, which considers that the first winding of one pole is equal to the first winding of the opposite pole. Therefore, if there is an interturn short circuit in one of these windings, a voltage difference on the induced at the probing coil will be produced. This method is able to determine the magnitude or severity of the failure, that is, the method could determine how many turns within the winding have a short circuit condition. For every complete turn of the rotor, the magnetic field produced by each winding induces a voltage at the 48

4 Artículo de investigación search coil. Two values from each winding are measured and compared with the corresponding winding at the opposite pole. The winding A1 in the pole A from figure 1 for example, and its symmetrical winding, winding B1 in the pole B, pass near the search coil twice, A1a, A1b and B1a, B1b (first and second pass voltage for the pole A and B respectively). Thus, it is possible to compare the induced voltage between symmetrical windings at each rotation cycle. However, several factors could introduce errors in the measurement, which may lead to a mistaken electrical failure diagnosis. Changes in the excitation current produced by electrical noise are one of the main factors that directly affect the magnetic flux density generated. The excitation systems, mainly static excitation systems, generate electrical noise that always causes excitation current variations and changes on the induced voltage over the search coil. On the other hand, the rotor could have magnetic irregularities that could produce abnormalities or magnetic asymmetries, which produce a waveform distortion, also detected by the search coil. The amount of turns for each winding is a decisive factor affecting the resolution of this technique. The more turns the windings have, the more difficult to detect one turn in short circuit conditions inside the winding. Data acquisition set up The measurement set up is shown in figure 3. The electrical signal provided by the search coil is first filtered to reduce the first and the third harmonic components. A high pass filter is the first stage in the signal processing and this reduces the frequency components that are produced by the main magnetic field at 60 Hz and 180 Hz, when the rotor speed is 3600 rpm, for example. Then, a low-pass filter is used in order to reduce the aliasing error in the digitalizing process. Both filters make a band pass filter shown in figure 3. This signal process is performed by an external hardware module, which conditions the signal to be acquired by the computer. A commercial data acquisition hardware module is used for the analogue to digital conversion with a sampling rate of 100k samples/seconds and with a 12 bit resolution. Figure 3. Measurement set up arrangement. Computer application A graphical programming language was used to build this application. The program runs in the Windows platform that controls the acquisition process and the statistical signal analysis. In order to illustrate the program structure, a programming block diagram of the main menu control is shown in figure 4. The application software was designed to monitor on-line power generators and it runs in a desktop or laptop computer. This application is able to evaluate short-circuits in the rotor periodically and store the measurements in a database. Additionally, the system can send the information via TCP/IP or modem to a maintenance control office upon request. A typical waveform from the signal provided by the search coil is shown in figure 5. The voltage induced on the search coil by each winding and the effect of the poles when they pass over the search coil is shown in figure 5. The first computer application task is to extract, from this waveform, the value of the induced voltage provided 49

5 by each coil and store them in vector form in order to be statistically processed. Statistical analysis Often electrical noise and interference produce errors in the measured data and therefore, produce errors in the failure diagnosis. Statistical analysis is a very useful tool for determining windings asymmetry obtained from real measurements. Figure 4. Graphic programming block diagram. The technique proposed by the IIE provides the data, at nominal rotor speed, to locate the shortcircuited windings and the number of their short-circuited turns. This technique improves the sensitivity and dependability by means of a statistical process applied to the signal provided by the search coil. When the first and second passes of each winding are plotted as the abscissa and ordinate respectively, a plot, as shown in figure 6, is obtained. Usually, there are two clusters of points associated with the two poles and the statistical distribution of the measured data can be observed. The clusters separation is a measure of the windings asymmetry. Figure 5. Typical waveform provide from the search coil. The dispersion produced by noise in the measurement and imperfections in the construction of windings can be observed, as well as the differences produced by interturn short circuits in the winding. The mean of each set of data is obtained for each pole separately. Figure 6. Two clusters of points of the measured data. 50

6 Artículo de investigación Then, different phenomena might be observed. The asymmetry between windings of the same pole and between the opposite poles are shown in figure 7. The distance between the means, as the statistical parameter, determines the presence of a short circuit in a particular winding. The means An and Bn are shown schematically in figure 8 as two points. A probability concept is often used as a way to determine whether or not a hypothesis is true. This concept of probability is a form of inductive reasoning that may be very useful in the diagnosis of electrical power equipment. In assuming that the measured data have a normal probability distribution, a statistical hypothesis is made in order to assess whether the hypothesis itself is true or false. The hypothesis test selected assumes that there is no difference between the data obtained from windings of pole A and pole B. The analysis is perform on the done means separation as in shown in Figure 8; the probability about the hypothesis is calculated. If a short circuit was found in some winding with a probability lower than 0.7 for example, then, the hypotesis is probably false. However, if the probability calculated is greater than 0.95, then, a true short circuit was found. The number of turns on short circuit that has been detected in a particular Figure 7. Statistical distribution plot that shows the windings asymmetry. Figure 8. Schematic diagram of the statistical distribution of measured data. winding is given as the ratio of the normalized distance between means and the number of turns from the winding. The software output, applying the proposed method, is shown in figure 9. The generator characteristics data, the test rotation speed, the short circuit turns from each winding and its probability are displayed. Field application The proposed method was tested in a 40 MBA Siemens generator located in a power plant at Tula, Hidalgo Mexico; the data shown in figures 6 and 9, were taken from that application. 51

7 is carried out, by considering random noise, periodic interference and the probability distribution differences in the measurements. Details of the electronic set-up and software are available from the authors, upon request. Acknowledgment The authors thank Dr. Roberto Canales for his valuable technical and editorial support. Figure 9. Statistical analysis results from a field measurement. For Further Reading In figure 9, the measure data from the coil # 4 are shown. In this particular case, two interturn short circuits, on the winding Nº 6, at the B pole were found with a probability of 92%. The interturn short circuits were produced by coal pollution and cleaning the windings eliminated this failure. The integrity of the insulation has not been compromised. The use of a probing coil for sensing the tangential magnetic flux density of the rotor is a non-invasive and inexpensive technique, which may be used when the rotor is rotating at nominal speed. The results obtained illustrate the effectiveness of the statistical analysis proposed as well as demonstrate the computer as a powerful informationprocessing tool. This technique allows acquiring the probability with which the diagnosis Kennedy R., Rickey D., Monitoring and Control of Industrial Power Systems. IEEE Computer Applications in Power Volume 2, Number 4, October 1989, pages Ramirez-Niño J., García A., Robles E.and Castaño V., A mathematical method for improving the detecting of interturn short circuits in rotor windings of power generators. Measurement Science and Technology, Vol 12, Nº 2, February 2001, pages Conolly H., Lodge L., Jackson R.J. and Roberts I. (1985). Detection of interturns shortcircuits in generator rotor windings using air-gap search coils. 2 nd. Conf. On Electrical Machines Design and Applications, pp 11-15, Albright D. (1971). Interturn short-circuit detector for turbine-generator rotor windings. IEEE Trans. Power Appar. Syst Currículum vítae José Tomás Ramírez Niño [jtrn@iie.org.mx] Was born in México D.F. He holds BS and MS degrees in Electrical Engineering from the Universidad Nacional Autónoma de México (UNAM) and PhD degree at the Universidad Autónoma del Estado de Morelos (UAEM). He has experience in optical communication links, signal analysis, and electronic instrumentation. He has been researcher at the Instituto de Investigaciones Eléctricas since 1981 and is currently working in the electrical equipment area. Alberth Pascacio de los Santos [pascacio@iie.org.mx] Was born in Tuxtla Gutiérrez, México. He holds as an example a BS degree in Electrical Engineering from the Instituto Tecnológico de Tuxtla Gutiérrez. He has experience in electronic instrumentation and graphical programming language for virtual instrumentation. He has been a researcher at the Instituto de Investigaciones Eléctricas since