Online detection and localization system of low voltage winding radial deformation in power transformers using electromagnetic waves

Abstract

This patent involves a monitoring system for online detection and localization of radial deformation errors in the low voltage windings of transformers. By employing this system within power plants, as well as transmission and distribution stations, continuous and real-time monitoring of the transformer's condition, particularly the low voltage coil, is achievable. Upon identifying a defect and determining its location, if the defect is deemed critical, the transformer can be strategically removed from the circuit at a predetermined time for necessary repairs, thereby mitigating the risk of severe damage. The insights provided by this approach to operators, including the identification and localization of radial deformation in low voltage windings, facilitate a more efficient and expedited repair process.

Description

Online Detection and Localization System of Low Voltage Winding Radial Deformation in Power Transformers Using Electromagnetic Waves

[0001] The presented plan continuously monitors the power transformer. If any radial deformation defect occurs in the low voltage winding of the transformer, it is identified. Therefore, it is possible to remove the transformer from the circuit at the first opportunity and prevent the occurrence of a short circuit in the coil. This patent provides a new method for detecting and locating the radial deformation on low voltage winding in power transformers using electromagnetic waves in the frequency band of 9 to 13 GHz, which is established by setting up equipment in the microwave laboratory. Moreover, by performing multiple simulations in CST software, the results were validated. This design has the advantage over existing transformation function and frequency response methods that it can be implemented online. And its other important advantage over other electromagnetic methods is the detection of radial error on low voltage windings in the transformer, which was not done in the previous methods.

[0002] G01S 13 / 00 – G01R 31 / 62 – G01B 7 / 16

[0003] US20170356733

[0004] DETECTION OF RADIAL DEFORMATIONS OF TRANSFORMERS

[0005] A method for detecting radial deformation in a winding of a transformer may include synthetic aperture radar (SAR) imaging of the winding using ultra high frequency (UHF) electromagnetic signals in a first instance of the winding to obtain a first image of the winding; SAR imaging of the winding using UHF electromagnetic signals in a second instance of the winding to obtain a second image of the winding; and comparing the first image of the winding and the second image of the winding to detect a radial deformation in the winding. The UHF electromagnetic signals may be transmitted as a plurality of successive sinusoidal signals, where frequencies of the successive sinusoidal signals gradually change from a first frequency to a second frequency.

[0006] This mentioned patent, similar to our claimed one, is a transformer deformation detection method that uses electromagnetic signals. However, their processes differ in that this one relies on SAR imaging while ours involves antenna installation and data mining techniques.

[0007] IN1700 / MUM / 2009

[0008] AN ON-LINE DIAGNOSTIC METHOD FOR HEALTH MONITORING OF A TRANSFORMER

[0009] An on-line diagnostic method for health monitoring of a transformer. In the case of a single phase or three phase star connected transformer deformations in the winding are determined by representing the transformer winding as a lumped parameter circuit and dividing the winding into at least two sections. A first set of fingerprint values are generated to determine the location of the deformed section of the winding and the type of deformation. A second set of finger print values are generated to determine the extent of deformation of the deformed section. The location and extent of radial or axial deformation or combination of both radial and axial deformation in the winding are then determined. The change in the capacitance of the bushing of the transformer connected at the line end of the winding is also determined. The state of the insulation system of the transformer is determined by detecting partial discharge pulses in the transformer winding. The change in the dielectric characteristics of the insulation system of the transformer is detected on the basis of phase angle difference.

[0010] This patent, like our claimed system, involves online health monitoring of transformers. While both patents rely on fingerprint values to locate the deformation, and there are similarities in their methods, ours focuses on low-voltage winding monitoring via data mining but this one makes no mention of this specification.

[0011] CN109884459B

[0012] Intelligent online diagnosis and positioning method for winding deformation of power transformer

[0013] The invention discloses an intelligent online diagnosis method for winding deformation of a power transformer. After the power transformer is impacted by short circuit or collided during transportation, the characteristics of local distortion, bulging and the like of a winding can occur under the action of electrodynamic force or mechanical force, the winding deformation is called, and huge hidden dangers are buried for safe operation of a power network. The common winding deformation diagnosis methods are off-line diagnosis methods, and have the defects of needing the transformer to be stopped, having high requirements on professional skills of operators and the like. The invention provides an intelligent online diagnosis method for winding deformation by combining information entropy and a support vector machine, which utilizes permutation entropy and wavelet entropy to extract characteristics of current and voltage signals, integrates variation conditions of monitoring indexes of a power transformer in the aspects of complexity, time-frequency domain and the like, automatically learns diagnosis logic from fault characteristics through a machine learning algorithm, realizes intelligent diagnosis of winding deformation, and thus reduces labor cost and improves diagnosis efficiency.

[0014] This patent resembles our claimed system in overall function and they both provide online monitoring of transformers but their methods and componants are different. For instance, this one uses a vectory and entropy parameters while ours focuses on low-voltage winding monitoring and uses antennas, signal transmission and data analysis of the network.

[0015] United States Patent 11500037

[0016] Method and a system of detecting winding fault under online operation of an electrical machine

[0017] A method and a system of detecting winding fault during online operation of an electrical machine, said method comprising, acquiring a set of signals from the machine over a period, said set of signals comprising two or more magnetic flux signals, each flux signal obtained from a respective flux sensor positioned on the external surface of the machine; extracting an internal winding fault indicator from the set of signals; comparing the internal winding fault indicator with a baseline indicator; and determining the internal winding fault when the internal winding fault indicator deviates from the baseline indicator by a threshold.

[0018] This patent operates in a similar way to our claimed system in that they both establish a threshold and detect the error based on deviation from the baseline. However, they use different methods and techniques of data collection and data mining.

[0019] United States Patent Application 20200200813

[0020] ONLINE DIAGNOSIS METHOD FOR DEFORMATION POSITION ON TRASNFORMATION WINDING

[0021] The invention discloses an online diagnosis method for transformer winding deformation position, including: (1) collecting a transformer with known winding state and decomposing into several position sub-samples; (2) performing feature extraction on each position sub-sample with information entropy, adding with label indicating deformation and inputting into support vector machine to train diagnosis model; (3) decomposing a transformer under diagnosis into 9 position subsamples in the way of step (1), performing feature extraction of step (2) and inputting into the diagnostic model trained in step (2); (4) outputting diagnosis result from the support vector machine about whether the position sub-samples of the transformer is deformed. The invention can achieve intelligent diagnosis of winding deformation by comprehensively considering variations of monitoring indicators of the transformer in complexity, time-frequency domain and other aspects and automatically learning diagnostic logic from fault features through machine learning algorithms, thereby reducing labor costs and improving diagnostic efficiency.

[0022] This invention utilizes machine learning techniques and provides online fault localization so the function resembles our claimed system. However, their operating procedures and componants are different, for instance ours installs antennas on the windings and focuses on low-voltage type transformers while this one does not specify such qualities.

[0023] US20230014981

[0024] EQUIPMENT AND METHOD FOR IDENTIFYING A FAULT IN THE WINDINGS OF A DISTRIBUTION TRANSFORMER

[0025] An apparatus for identifying a fault in the windings of a distribution transformer, a transformer, and an associated method, said device comprising: a first Rogowski current sensor at a high-voltage incoming current terminal, and a second Rogowski current sensor in tandem at a low-voltage outgoing current terminal and at a low-voltage incoming current terminal; a first conductor of the low-voltage outgoing current terminal, passed through in one direction through the second sensor, and a second conductor of the low-voltage incoming current terminal, passed through in the opposite direction through the second sensor; the first and second sensors generate output signals indicating the primary current and the secondary current; both signals are integrated, generating output signals proportional to the primary current and the secondary current, obtaining a transformation ratio, which is compared with a threshold, and sending a fault signal if said threshold is exceeded.

[0026] This mentioned inventions bears a resemblance to our designed system as they both provide monitoring methods of transformers and utilize threshold deviations. However, they use different equipment, for instance this one relies on specific sensors while ours uses antennas. Also, ours offers online monitoring specifically for low-voltage systems, while this one makes no mention of this characteristic.

[0027] This designed system uses scattering parameters to diagnose radial deformation of low voltage winding in real-time. Electromagnetic waves are sent to the transformer winding via transmitter antennas, and return waves are received by the same antennas to obtain the scattering parameter value, stored in a database for healthy and defective states. By comparing the states, any winding changes can be easily detected. Various tests of radial errors with different intensities and locations on the low voltage bobbin are conducted to estimate possible deformation locations, with results recorded as a reference. Reference status results are then compared with online test results using identification algorithms to determine error location. Scattering parameters reflect energy distribution between input and output ports, providing a fingerprint for the transformer's healthy state in the database, enabling fault localization in abnormal conditions.

[0028] Transformers are one of the main and expensive components of energy production, transmission and distribution networks. The performance of the transformer at different levels has a vital and effective role in maintaining the stability and improving the reliability of the power network, therefore, its protection is one of the most important and vital issues. Many users are interested in knowing the inner workings of power transformers. By using online monitoring methods, in addition to preventing serious damage to the transformer, it is possible to ensure the continuity of electric energy transmission with timely notification. These equipments may suffer from various defects such as the radial deformation of the winding which may occur due to electrodynamic forces at the moment of short circuit. Over time this will weaken the insulation characteristics of the coils and cause short circuit and damage to the transformer.

[0029] Various methods have been introduced to diagnose radial deformation defects in transformer coils, some of which need to take the transformer out of circuit, and some of them are implemented online. Recently, the effectiveness of methods for detecting mechanical defects of coils using electromagnetic waves has been proven online. The use of electromagnetic waves makes it possible to diagnose the defect of the transformer winding protrusion and depression online by a device that transmits and receives electromagnetic waves.

[0030] However, all the electromagnetic methods that have been proposed so far have been able to monitor high voltage windings, while low voltage windings which are more prone to error have been monitored less due to their location and difficult physical and visual access. Among the existing methods that use electromagnetic waves, the primary focus is on monitoring high voltage transformers. The presented design is a system that has been implemented for the first time to detect the error of radial deformation in low voltage windings of transformers. By using this system in power plants, transmission and distribution stations, the status of the transformer, especially the low voltage coil, can be reached permanently and online. After detecting the defect and estimating its location, if the defect is significant, the transformer is removed from the circuit at a suitable and pre-planned time and is prepared for repairs to prevent serious damage. The information provided by this method to the operators (such as the presence and location of the radial deformation on low voltage windings) helps to repair the winding more easily and quickly.Solution of Problem

[0031] Among the main challenges of the existing electromagnetic methods, we can mention the need to change the structure of the transformer tank to create a dielectric window, problems in changing the position of the antennas by electric motors in order to scan the entire winding, the existence of obstacles such as tap changers and cooling connections and the lack of monitoring of low voltage windings.

[0032] In addition to the mentioned problems, on the one hand, the low voltage winding has been subject to more electromagnetic forces and mechanical errors due to the high passage of current, and on the other hand, it has been monitored less due to its location and difficult physical and visual access. Therefore, it is necessary to have a system that can detect the radial deformation error of the low voltage winding in the transformer.

[0033] In this design, scattering parameters are used for the real-time diagnosis of the radial deformation of the low voltage winding. The electromagnetic waves are radiated to the transformer winding by two transmitter antennas, and the return waves are received by the same antennas, then the scattering parameter value is obtained and stored in the database for healthy and defective states. By comparing the healthy and defective states, any change in the winding can be easily detected. Additionally, in order to estimate the possible location of radial deformation, various tests of radial errors with different intensities and in different places on the low voltage bobbin are performed and their results are recorded as a reference. Lastly, via the online tests, the results with the reference status are run through identification algorithms to determine the estimated location of error.

[0034] In this method, the transformer bobbins can be considered as plates or rectangular waveguides, which is a function of the physical surface inside the channel, the distance to the emission source, the wave emission medium, and other parameters. The transformer is a static system and by considering the emission and wave as constant, its behavior can be predicted and any small change in the level of the emission channel will be a function of the deformation of the transformer bobbins and will directly affect the scattering parameters and cause them to change.

[0035] Scattering parameters describe how the energy is distributed between two input and output ports. These parameters in the healthy state of the transformer can be stored in the database as a fingerprint, and also in different radial error conditions, this test is repeated and the status of this parameter was recorded in the database. Then, after any transformer movement, short circuit test or high current critical conditions, abnormal results occurrence in periodic tests, earthquakes and other conditions, this test is repeated and the result is compared with the results in the database to perform fault localization. The schematic and hardware of the proposed design are given in.

[0036] In the laboratory prototype, the specifications of a transformer of 20.4 kV and 1.6 megavolt ampere with Dyn11 connection and 50 Hz were obtained from the manufacturer, (Table 1) shows the specifications of the dimensions of the transformer and the built model. In order to simplify the implementation of radial error models, a model of one phase of the studied transformer was made from copper and according to the real dimensions. And all the tests on it were done in the microwave laboratory, using an Agilent 8720ES model network analyzer.

[0037] Due to the selection of a frequency of at least 9 GHz, the wavelength of the transmitted signal is approximately equal to 30 mm, while the distance between the disks of the high voltage winding is 6 mm in real mode, so the waves do not pass through the distance between the disks and the emission channel is like a waveguide. They travel from the beginning to the end of the bobbin. Choosing a wide bandwidth at high frequency has many advantages, including increased penetration, accurate estimation of the fault location, high speed of transmission, small size of equipment and lower power consumption.

[0038] With the selected wavelength, the set of low and high voltage windings and the core are respectively seen as three metal cylinders inside each other, and there is no need to model other components.

[0039] In electromagnetic methods, the antennas were generally located in front of the coils or windings in order to have a complete scan of the high voltage winding, so they did not have the ability to monitor the condition of the low voltage one which is inside the high voltage coil and around the core. We used a 12mm wide channel between the low and high voltage windings, and for complete coverage of the channel, we used 4 antennas (two transmitting antennas and two receiving antennas), each pair facing each other and placed at the beginning and the end of the channel. The location of the transmitter antenna is shown at the entrance of the channel; the receiver antennas are located exactly in line with the end of the channel.

[0040] Patch antennas were used in the laboratory model due to having a suitable bandwidth. This antenna was designed in Antenna Magus software and then assembled with high precision by CNC machine on Rogers RO4003 board with a thickness of 1 mm. The gain of this antenna at 12 GHz frequency is 6 dB and the area of each antenna is nearly 4 square cm. The antennas are connected to the body or cover of the transformer through N or SMA connectors and through the coaxial cable, therefore it is possible to provide online monitoring.

[0041] In the experiments, the environment of air diffusion is considered, the transformer tank is not modeled, the temperature and humidity conditions are considered constant. After calibrating the network analyzer in a wide frequency band of 9 to 13 GHz, the first test was performed in the healthy state of the transformer (RD0) and the results (parameter S11) were recorded in this test, then the same test was repeated for the simulated transformer in the CST software. So, the development of the model in the software requires entering the exact dimensions of the transformer and defining the materials for the coils and core, and then defining the boundary conditions and other necessary parameters.

[0042] In order to apply the radial deformation in the laboratory sample, first-degree deformations at zero degree angle (in the direction of the antenna) and 70 mm steps were implemented on the low voltage winding from the laboratory model. We named these errors as RD1-7. With the same conditions, the experiments were repeated for the simulated transformer in CST software. In addition, due to the great similarity between the simulated results and the practical results, the second and third degree deformations were applied only on the simulated model and in the half cylinder in front of the antenna at the angles of 0, +90, -90 and named as RD8, RD9. In order to check the first-order error more precisely, we placed it at a distance of 210 mm from the top of the low voltage coil at angles with steps of 30 degrees in the negative and positive direction with respect to the direction of the antenna and up to an angle of ±90 degrees, and we named it RD10-RD15. All the radial changes are of the protrusion type and the dimensions are 5 x 10 x 50 mm. These tests were performed in the frequency band of 9 to 13 GHz and steps of 40 MHz by the network analyzer device. Therefore, the number of received samples for each situation is equal to 101.

[0043] Using the scattering parameter (S11), a comparison of the healthy state in the practical and simulated state is done. It can be seen that the simulation can predict the real change process. The measurement and simulation results of cases RD1, RD3 and RD6 are shown inand. Cases RD8 and RD9 are shown inand cases RD10, RD13, RD11 and RD14 are presented in.

[0044] compares the measurement results of the healthy case with cases RD1, RD3 and RD6. Also,compares the simulation results of similar signals. It can be seen that the change in the location of the radial deformation leads to a change in the scattering parameter tests, and therefore it can be said that this method can detect the radial deformation.shows that the order of deformation can also change the results. The effect of changing the radial deformation angle at the same height is shown in. It can be seen that the radial deformation angle also affects the results.

[0045] All figures 4-8 are indicative of the value of parameter S in three frequency intervals with a more negative value, so we have used these changes in these points to perform fault localization. In the magnified part, it can be understood that as the error occurrence distance from the antenna (reference point) increases, the value of parameter S increases in the frequency range of 12.6 GHz and its opposite is observed in the frequency range of 9.8 GHz, i.e. where the value of parameter S decreases by increasing the error occurrence interval. In the zoomed area in, it can be seen that, by changing the angle of incidence of the error from the antenna (zero degrees), the size of the parameter S does not change significantly, but the displacement of the nulls can be observed in the same frequency range.

[0046] In this patent, the pattern recognition technique is used to detect and estimate the location of the error. In order to detect the presence of errors, the LOF outlier data removal algorithm has been used in the data preprocessing stage from the three characteristics of mean absolute amplitude (MAMD), correlation (CSM), and Euclidean distance (MED). An outlier is an observation that is far away from other data. Therefore, due to the dispersion of the data, a threshold limit of 1.15 has been used for this purpose, then appropriate data mining techniques such as collective bagging algorithm and regression tree have been used to classify the type of error.

[0047] The main idea of collective algorithms is to apply several learning methods and combine their prediction results as a group of classifiers and increase the overall learning accuracy. In this design, the regression tree algorithm is used, whose flowchart is shown in. This flowchart consists of three main parts.

[0048] The first part: forming a data bank by applying indicators

[0049] Second part: data training based on the regression tree method

[0050] The third part: calculating the error function between the measured value and the value recorded in the data bank and detecting the error and its location based on the following classification:

[0051] Each feature separately classifies the behavior of the transformer in four states including healthy, radial fault in the upper, middle and lower part of the low voltage winding. To detect the location of the first type of error, we divided the low voltage bobbin into three parts, the distance 0-70 mm from the beginning of the bobbin is the upper part, 210-70 mm is the middle part, and 500-210 mm is the lower part of the bobbin. The accuracy of the classifiers in detecting the location of the error is shown in [Table 3]. According to [Table 3], the most accurate is the regression technique with a correlation index of 97.83%, which is used in this project. The Accuracy of this method can be calculated using [Formula. 12].Advantage Effects of the Invention

[0052] • Diagnosing and locating the radial deformation error on the low voltage windingin the transformer

[0053] • The possibility of installation with the minimal changes in the structure of the transformer (4 square centimeters).

[0054] • The possibility of taking tests completely online and without risk for users

[0055] • Preventing the occurrence of short circuit due to the radial fault of the transformer winding

[0056] • Independent testing method regrardless of the transformer being on load or not

[0057] Shows a general view of the claimed process.

[0058] Declares the placement of different compnants of the transformer.

[0059] Shows a view of the studied transformer.

[0060] Shows a line graph of the real and simulated healthy state comparison.

[0061] Shows a line graph of the measured results.

[0062] Displays a line graph of the measured and compared results.

[0063] Captures another line graph of the compared results.

[0064] Shows a similar line graph of the compared results.

[0065] Displays a flowchart of the process for the proposed system.

[0066] Shows a view of the designed process for the claimed system which includes different sections including:

[0067] A. Scattered parameter network analyzer (500MHZ-20GHZ)

[0068] B. Calculating and processing unit

[0069] C. Raw data

[0070] D. Signal processing

[0071] E. Data analysis

[0072] F. Display of the processed data

[0073] X. Signal path

[0074] Y. Data path

[0075] Displays the placement of different componants on the transformer cylinder including the location of the transmitter antenna at the entrance of the channel in the transformer 4.5 / 10 kV 50 MVA and the following:

[0076] 1. The core

[0077] 2. Low voltage winding

[0078] 3. High voltage winding

[0079] 4. Transmitter antenna

[0080] 5. Antenna entrance box

[0081] Shows a view of the one-phase model of the studied and simulated transformer.

[0082] Shows a line graph wherein the scattering parameter (S11) is used in order to run a comparison of the healthy state in the practical and simulated state. It can be seen that the simulation can predict the real change process.

[0083] Displays a line graph of the measurement and simulation results of cases RD1, RD3 and RD6.It compares the measurement results of the healthy case with cases RD1, RD3 and RD6.

[0084] Also shows a line graph that compares the measured results of cases RD1, RD3 and RD6. It can be seen that the change in the location of the radial deformation leads to a change in the scattering parameter tests, and therefore it can be said that this method can detect the radial deformation.

[0085] Shows a line graph that similarly to the previous two figures captures the comparison of the measured results for cases RD8 and RD9.

[0086] Displays a line graph that shows the comparison between cases RD10, RD13, RD11 and RD14. It also captures effect of changing the radial deformation angle at the same height. It can be seen that the radial deformation angle also affects the results. By changing the angle of incidence of the error from the antenna (zero degrees), the size of the parameter S does not change significantly, but the displacement of the nulls can be observed in the same frequency range

[0087] Shows a flowchart for the claimed system's process.

[0088] This flowchart consists of three main parts.

[0089] The first part: forming a data bank by applying indicators

[0090] Second part: data training based on the regression tree method

[0091] The third part: calculating the error function between the measured value and the value recorded in the data bank and detecting the error and its location based on the following classification:

[0092] Each feature separately classifies the behavior of the transformer in four states including healthy, radial fault in the upper, middle and lower part of the low voltage winding. To detect the location of the first type of error, we divided the low voltage bobbin into three parts, the distance 0-70 mm from the beginning of the bobbin is the upper part, 210-70 mm is the middle part, and 500-210 mm is the lower part of the bobbin. The accuracy of the classifiers in detecting the location of the error is shown in [Table 3]. According to [Table 3], the most accurate is the regression technique with a correlation index of 97.83%, which is used in this project. The Accuracy of this method can be calculated using [Formula. 11].Examples

[0093] The theory behind this invention relies on certain equations and formulas which will be listed here:

[0094] The distance between two high and low voltage windings and the oil between them can be modeled as a parallel plate waveguide that includes two parallel conductor plates that are separated from each other by the dielectric medium with structural parameters ɛ and µ. We assume that the planes are infinite in the X direction (2). This is because we assume that the fields do not change in the X direction and the edge effects are negligible.

[0095] The intensity of the electric and magnetic fields inside the dielectric region (air is assumed) with a dielectric constant of 1 without charge, satisfies the following Helmholtz vector equations and is calculated by solving the Laplace equation.

[0096] [Formula. 1]

[0097]

[0098] [Formula. 2]

[0099]

[0100] [Formula. 3] The answer to the above equation must apply to the boundary conditions:

[0101]

[0102] [Formula. 4]

[0103]

[0104] [Formula. 5]

[0105]

[0106] [Formula. 6]

[0107]

[0108] is the diffusion constant and can be determined from the above equation. A is the intensity of wave stimulation.

[0109] [Formula. 7]

[0110]

[0111] The cutoff frequency is the frequency that makes the above equation zero.

[0112] [Formula. 8]

[0113]

[0114] [Formula. 9]

[0115]

[0116] [Formula. 10]

[0117]

[0118] [Formula. 11]

[0119]

[0120] In a healthy state, the value of d, which is the distance between the low voltage winding and the high voltage winding, is equal to 12 mm, and with this value, the fields are calculated and the result is recorded as the fingerprint of the transformer in the database. It can be seen that the smallest radial deformation along the length of low and high voltage windings causes a change in the value of d, and the instantaneous values of the field and scattering parameters undergo changes. The instantaneous value of the field in the space between the low and high voltage windings in the healthy stateis simulated in the HFSS software.

[0121] [Formula. 12]

[0122]

[0123] [Formula. 13] includes the following values:

[0124] • TP: The sample is a member of the positive category and is known as a member of the same class (True Positive).

[0125] • FN: The sample is a positive class member and a negative class member (False Negative) is detected.

[0126] • TN: The sample is a member of the negative class and is known as a member of the same class (True Negative).

[0127] • FP: The sample is a negative class member and a positive class member (false positive) is detected.

[0128] The data from the prototype test based on the claimed method is presented in the following tables:

[0129]

[0130]

[0131]

[0132] Considering the high importance of the transformer in the industry of electrical production, transmission and distribution, continuous monitoring of this basic and expensive equipment is essential. This plan has addressed this national and international need to prevent errors and malfunction in this equipment and to increase the reliability of the network.

[0133] This plan has the ability to be industrialized and used in the power generation as well as the transformer manufacturing factories. The presented plan has the ability to continuously monitor the condition of the low voltage coils of the transformer. The implementation of this plan requires minimal changes in the transformer.

Claims

A monitoring system is designed to detect and locate radial deformation errors in low voltage windings of power transformers using electromagnetic waves.According to claim 1, the designed system includes a network analysis unit, X band antenna, high frequency cables and N and SMA connectors.According to claim 1, the system employs data mining techniques such as collective bagging and regression tree algorithms to detect errors, classify error type and estimate the location of radial deformation in low voltage windings.According to claim 1, the system works via installing two small transmitter antennas at top of the low voltage winding and two receiver antennas at the end of the low voltage winding and connecting them to the transformer body or cover through high frequency cables.According to claim 4, by using small antennas and connecting them to the cover or body, it is possible to perform the fault localization tests completely online.

Images