A method for establishing a motor stator bar partial discharge fingerprint library based on an ultrahigh frequency method

Abstract

A kind of motor stator bar partial discharge fingerprint library establishment method based on ultra-high frequency method, comprising the following steps: 1) build stator bar typical defect partial discharge detection platform;2) alternating current power supply to the segmented stator bar to be detected pressurization;Make stator bar to be detected occur partial discharge;Ultra-high frequency antenna receives the electromagnetic wave signal radiated outward by the partial discharge of stator winding to be detected in real time, and sends electromagnetic wave signal to high-speed oscilloscope;3) high-speed oscilloscope draws and displays the waveform of the partial discharge of stator winding to be detected and corresponding power frequency periodic waveform;4) host computer carries out discharge pulse extraction to the discharge pulse group collected;5) calculate the partial discharge characteristic value database under different insulation defects and different discharge conditions;6) construct the partial discharge diagnosis fingerprint library with partial discharge characteristic value as the basis for judging, the present application has strong anti-interference ability and strong sensitivity, can effectively receive the electromagnetic wave signal radiated outward by partial discharge.

Description

Technical Field

[0001] This invention relates to the statistical analysis of partial discharge of typical defects in large hydro-generators, specifically a method for establishing a fingerprint database of partial discharge of motor stator bars based on the ultra-high frequency method. Background Technology

[0002] The stator winding is a critical component of large generators. For large generators, the stator winding withstands very high voltages and currents. With the continuous development of the economy and society, the installed capacity of motors is constantly increasing, and the rated voltage of the stator windings is also constantly rising, which places higher demands on the reliability of motor operation.

[0003] However, with long-term operation of the generator, factors such as aging of insulation materials and air gaps caused by mechanical vibration can lead to partial discharge at the ends and slots of the stator windings. Prolonged partial discharge deteriorates the insulation, further exacerbating the hazards of partial discharge and significantly impacting the reliability of the motor's safe operation.

[0004] Currently, the main methods for detecting partial discharge in motor windings according to standards are visual inspection and ultraviolet imaging. Visual inspection has a significant subjective element; ultraviolet imagers are affected by various factors such as environmental conditions, ultraviolet gain, and monitoring distance, making calibration difficult. Alternatively, a pulse current method using a detection resistor and coupling capacitor can be used to detect partial discharge, but this is not suitable for online live detection and is also easily affected by other interference on the ground wire. Summary of the Invention

[0005] The purpose of this invention is to provide a method for establishing a partial discharge fingerprint database for motor stator bars based on the ultra-high frequency method, comprising the following steps:

[0006] 1) Build a partial discharge detection platform for typical defects in stator bars.

[0007] The stator bar typical defect partial discharge detection platform is used to detect the electromagnetic wave signals radiated by the stator bar under test.

[0008] The stator bar typical defect partial discharge detection platform includes the stator bar to be tested, an AC power supply, an ultra-high frequency antenna, a high-speed oscilloscope, and a host computer.

[0009] The stator bars to be tested include defect-free stator bars and defective stator bars.

[0010] 2) The AC power supply applies voltage to the stator windings under test in segments. During the voltage application process, partial discharge occurs in the stator windings under test. The ultra-high frequency antenna receives the electromagnetic wave signals radiated by the partial discharge of the stator windings under test in real time and sends the collected electromagnetic wave signals to a high-speed oscilloscope.

[0011] 3) After receiving the electromagnetic wave signal, the high-speed oscilloscope plots and displays the waveform of the partial discharge of the stator winding to be tested and the corresponding power frequency period waveform.

[0012] 4) The high-speed oscilloscope sends the discharge data of the UHF antenna channel and the power frequency cycle channel to the host computer, and the host computer extracts the discharge pulses from the collected discharge pulse groups.

[0013] 5) Replace the stator bar to be tested and repeat steps 2) to 4) to obtain discharge pulses under different insulation defects and discharge conditions;

[0014] 6) Based on the discharge pulses extracted by the host computer under different insulation defects and discharge conditions, a database of partial discharge characteristic values ​​under different insulation defects and discharge conditions is calculated.

[0015] 7) Based on the partial discharge characteristic value database of different insulation defects and different discharge conditions, construct a fingerprint database for partial discharge diagnosis based on partial discharge characteristic values.

[0016] Furthermore, the ultra-high frequency antenna is placed on the side where typical defects in the stator winding are detected, and is connected to the input of an oscilloscope via an RF amplifier and a coaxial cable.

[0017] The host computer is connected to the output of the oscilloscope via a coaxial cable.

[0018] Furthermore, the output impedance of the ultra-high frequency antenna is consistent with the wave impedance of the coaxial cable.

[0019] Furthermore, the AC power supply applies voltage to the stator winding under test in segments, including the following steps:

[0020] 1) Apply voltage U0 to the stator winding to be tested, where U0 is the rated voltage corresponding to the bar. Observe whether there is obvious partial discharge at the insulation defect.

[0021] 2) If there is no obvious partial discharge phenomenon after each cycle T, the voltage of the bar is gradually increased until the bar begins to show obvious partial discharge phenomenon. Then the voltage is stopped and the voltage applied at this time is used as the starting voltage U1 for data recording.

[0022] 3) Apply voltage U based on the starting voltage U1 in step 2) and start recording the discharge pulse signal data at this time.

[0023] 4) Set ΔU as the segmented voltage value. Each time the applied voltage U increases by ΔU, record the discharge pulse signal data at this time.

[0024] 5) Set the maximum applied voltage to U maxCompare the voltage U and the maximum applied voltage U. max , if U>U max Then the pressurization ends if U ≤ U max Then return to step 4).

[0025] Further steps to observe whether there is obvious partial discharge at the insulation defect include recording the curves of ambient sound, voltage and ultraviolet photon count.

[0026] When the difference between the discharge voltage and the rated voltage of the conductor is less than the preset voltage threshold, if the ambient sound level is greater than the preset decibel threshold, there will be obvious partial discharge at the insulation defect.

[0027] If the slope of the curve of voltage versus ultraviolet photon number changes abruptly, there will be obvious partial discharge at the insulation defect.

[0028] Furthermore, the processing performed by the host computer on the acquired discharge pulse groups also includes noise suppression.

[0029] Furthermore, the corresponding PRPD spectrum is constructed using the discharge pulses extracted by the host computer. The PRPD spectrum is used to calculate a partial discharge characteristic value database under different discharge conditions.

[0030] Furthermore, the partial discharge characteristic values ​​include characteristic values ​​in the time domain and the frequency domain.

[0031] Furthermore, the time-domain eigenvalues ​​include: the ratio of positive to negative half-cycle discharge counts N1, the ratio of positive to negative half-cycle discharge amounts N2, the correlation coefficient between positive and negative half-axis C1, the correlation coefficient between positive and negative half-cycles C2, and the kurtosis K. u .

[0032] The characteristic values ​​of the frequency domain include: bandwidth P. width Maximum frequency point P max .

[0033] Furthermore, the ratio of positive to negative half-cycle discharge times N1 is as follows:

[0034]

[0035] In the formula, N + N and N- represent the number of discharges in the positive and negative half-cycles of the power frequency cycle, respectively.

[0036] The ratio of positive to negative half-cycle discharge N2 reflects the comparison of the severity of partial discharge during the negative and positive half-cycles of the power frequency voltage, as shown below:

[0037]

[0038] In the formula, k1 and k2 are the number of phase segments in the positive and negative half-cycles, respectively, and qi + q i - These represent the partial discharge quantities defined within the i-th phase segment of the positive and negative half-cycles of the power frequency voltage, respectively.

[0039] A partial discharge signal The similarity of the three-dimensional spectrum along the positive and negative half-axis of amplitude is represented by the correlation coefficient C1, which is shown below:

[0040]

[0041] In the formula, n 1i n 2i This represents the average number of discharges across all amplitude ranges on both the positive and negative half-axis. k max This represents the absolute value of the maximum amplitude of the spectrum on the positive and negative half-axis.

[0042] A partial discharge signal The similarity of the three-dimensional spectrum in the positive and negative half-cycles of amplitude is represented by the correlation coefficient C2, which is shown below:

[0043]

[0044] In the formula, n' 1i 、n' 2i This represents the average number of discharges in each phase interval of the positive and negative half-cycles. N is the number of divisions in the phase intervals of the positive and negative half-cycles.

[0045] The kurtosis K u This reflects the sharpness of the peaks in the spectrum, as shown below:

[0046]

[0047] In the formula, For the i-th phase interval on the spectrum, In the phase interval The probability of the number of discharges on the surface. for The mean number of discharges on the plane, where σ is the variance.

[0048] The bandwidth P width It is calculated by performing FFT transformation on several typical waveforms that have been collected.

[0049] The maximum frequency point P max It is calculated by performing FFT transformation on several typical waveforms that have been collected.

[0050] The technical effects of this invention are undeniable. The ultra-high frequency method used in this invention has strong anti-interference capabilities and high sensitivity. It can also effectively receive electromagnetic wave signals radiated by partial discharge, which not only avoids interference from various factors, but also greatly improves the reliability of motor safe operation.

[0051] Instruction manual illustrations

[0052] Figure 1 This is a schematic diagram of a partial discharge detection platform for typical defects in stator bars.

[0053] Figure 2 The partial discharge signal PRPD diagram of a defect-free bar at 22kV;

[0054] Figure 3 The partial discharge signal PRPD diagram of the anti-corona lap joint defect (surface defect) at 22kV;

[0055] Figure 4 PRPD image of the partial discharge signal of an internal delamination defect at 22kV;

[0056] Figure 5 The amplitude-frequency distribution diagram is shown for a typical waveform of a defect-free wire bar.

[0057] Figure 6 Amplitude-frequency distribution diagram of typical waveforms to prevent poor overlap defects (surface defects);

[0058] Figure 7 This is an amplitude-frequency distribution diagram of a typical waveform of an internal defect.

[0059] Figure 8 This is an example diagram illustrating the use of the tangent method in ultraviolet imaging to determine discharge phenomena. Detailed Implementation

[0060] The present invention will be further described below with reference to embodiments, but it should not be construed that the scope of the present invention is limited to the following embodiments. Various substitutions and modifications made based on ordinary technical knowledge and common practices in the art without departing from the above-described technical concept of the present invention should be included within the scope of protection of the present invention.

[0061] Example 1:

[0062] See Figures 1 to 8 A method for establishing a partial discharge fingerprint database for motor stator bars based on ultra-high frequency method includes the following steps:

[0063] 1) Build a partial discharge detection platform for typical defects in stator bars.

[0064] The stator bar typical defect partial discharge detection platform is used to detect the electromagnetic wave signals radiated by the stator bar under test.

[0065] The stator bar typical defect partial discharge detection platform includes the stator bar to be tested, an AC power supply, an ultra-high frequency antenna, a high-speed oscilloscope, and a host computer.

[0066] The stator bars to be tested include defect-free stator bars and defective stator bars.

[0067] Typical defects in the stator bars include poor anti-corona overlap defects (surface defects) and internal delamination defects.

[0068] 2) The AC power supply applies voltage to the stator windings under test in segments. During the voltage application process, partial discharge occurs in the stator windings under test. The ultra-high frequency antenna receives the electromagnetic wave signals radiated by the partial discharge of the stator windings under test in real time and sends the collected electromagnetic wave signals to a high-speed oscilloscope.

[0069] 3) After receiving the electromagnetic wave signal, the high-speed oscilloscope plots and displays the waveform of the partial discharge of the stator winding to be tested and the corresponding power frequency period waveform.

[0070] 4) The high-speed oscilloscope sends the discharge data of the UHF antenna channel and the power frequency cycle channel to the host computer, and the host computer extracts the discharge pulses from the collected discharge pulse groups.

[0071] 5) Replace the stator bar to be tested and repeat steps 2) to 4) to obtain discharge pulses under different insulation defects and discharge conditions;

[0072] 6) Based on the discharge pulses extracted by the host computer under different insulation defects and discharge conditions, a database of partial discharge characteristic values ​​under different insulation defects and discharge conditions is calculated.

[0073] 7) Based on a database of partial discharge characteristic values ​​under different insulation defects and discharge conditions, a fingerprint database for partial discharge diagnosis is constructed, using partial discharge characteristic values ​​as the basis for judgment. This fingerprint database can determine the different insulation defect types corresponding to the discharge and assess the severity of the partial discharge.

[0074] The ultra-high frequency antenna is positioned on the side where typical defects in the stator winding are detected. It is connected to the input of an oscilloscope via an RF amplifier and a coaxial cable.

[0075] The host computer is connected to the output of the oscilloscope via a coaxial cable.

[0076] The output impedance of the ultra-high frequency antenna is consistent with the wave impedance of the coaxial cable.

[0077] The AC power supply applies voltage to the stator winding under test in segments, including the following steps:

[0078] 1) Apply voltage U0 to the stator winding to be tested, where U0 is the rated voltage corresponding to the bar. Observe whether there is obvious partial discharge at the insulation defect.

[0079] 2) If there is no obvious partial discharge phenomenon after each cycle T, the voltage of the bar is gradually increased until the bar begins to show obvious partial discharge phenomenon. Then the voltage is stopped and the voltage applied at this time is used as the starting voltage U1 for data recording.

[0080] 3) Apply voltage U based on the starting voltage U1 in step 2) and start recording the discharge pulse signal data at this time.

[0081] 4) Set ΔU as the segmented voltage value. Each time the applied voltage U increases by ΔU, record the discharge pulse signal data at this time.

[0082] 5) Set the maximum applied voltage to U max Compare the voltage U and the maximum applied voltage U. max , if U>U max Then the pressurization ends if U ≤ U max Then return to step 4).

[0083] The steps for observing whether there is obvious partial discharge at the insulation defect include: recording the curves of ambient sound, voltage and ultraviolet photon count.

[0084] When the difference between the discharge voltage and the rated voltage of the conductor is less than a preset voltage threshold, if the ambient sound level is greater than a preset decibel threshold, there will be obvious partial discharge at the insulation defect. The voltage threshold and decibel threshold are set in advance according to the actual situation.

[0085] The standard for judging the sound is that, compared with the quiet environment when it is not discharged, the discharge sound is significantly increased. This is relatively subjective, and a noticeable sound can usually be heard when it is close to the rated voltage of the rod.

[0086] If the slope of the curve of voltage versus ultraviolet photon number changes abruptly, there will be obvious partial discharge at the insulation defect.

[0087] The criterion for judging the effect of the ultraviolet imager is to plot a curve of applied voltage versus ultraviolet photon count. When the slope of the curve changes abruptly, it indicates a violent discharge phenomenon (generally, an ultraviolet photon count >300 is considered a relatively significant discharge). In this case, the intercept point between the tangent to the curve and the voltage on the horizontal axis is taken as the starting voltage for the corona initiation, as shown in the example. Figure 8 (Note: This method is mainly applicable to surface discharge. For internal discharge, since the discharge occurs inside the insulation, it is impossible to collect the number of ultraviolet photons. In this case, the sound judgment standard is more suitable.)

[0088] The processing performed by the host computer on the acquired discharge pulse groups also includes noise suppression.

[0089] The corresponding PRPD spectrum is constructed from the discharge pulses extracted by the host computer. The PRPD spectrum is used to calculate a database of partial discharge characteristic values ​​under different discharge conditions.

[0090] The partial discharge characteristic values ​​include characteristic values ​​in the time domain and the frequency domain.

[0091] The time-domain eigenvalues ​​include: the ratio of positive to negative half-cycle discharge counts N1, the ratio of positive to negative half-cycle discharge amounts N2, the correlation coefficient between positive and negative half-axis C1, the correlation coefficient between positive and negative half-cycles C2, and the kurtosis K. u .

[0092] The characteristic values ​​of the frequency domain include: bandwidth P. width Maximum frequency point P max .

[0093] The ratio of positive to negative half-cycle discharge times, N1, is shown below:

[0094]

[0095] In the formula, N + and N - These represent the number of discharges during the positive and negative half-cycles of the power frequency cycle, respectively.

[0096] The larger N1 is, the more times the negative half-axis is discharged.

[0097] The ratio of positive to negative half-cycle discharge N2 reflects the comparison of the severity of partial discharge during the negative and positive half-cycles of the power frequency voltage, as shown below:

[0098]

[0099] In the formula, k1 and k2 are the number of phase segments in the positive and negative half-cycles, respectively, and q i + q i - These represent the partial discharge quantities defined within the i-th phase segment of the positive and negative half-cycles of the power frequency voltage, respectively. The larger N2 is, the more intense the partial discharge is within the negative half-cycle of the power frequency cycle.

[0100] A partial discharge signal The similarity of the three-dimensional spectra along the positive and negative half-axis of amplitude is represented by the correlation coefficient C1. In the three-dimensional spectrum, n represents the number of discharges, and q represents the partial discharge quantity. Indicates phase.

[0101] The correlation coefficient C1 between the positive and negative half-axis is shown below:

[0102]

[0103] In the formula, n 1i n 2i This represents the average number of discharges across all amplitude ranges on both the positive and negative half-axis. k max This represents the absolute value of the maximum amplitude of the spectrum on the positive and negative half-axis.

[0104] A partial discharge signal The similarity of the three-dimensional spectrum in the positive and negative half-cycles of amplitude is represented by the correlation coefficient C2, which is shown below:

[0105]

[0106] In the formula, n' 1i 、n' 2i This represents the average number of discharges in each phase interval of the positive and negative half-cycles. N is the number of divisions in the phase intervals of the positive and negative half-cycles.

[0107] The kurtosis K u This reflects the sharpness of the peaks in the spectrum, as shown below:

[0108]

[0109] In the formula, For the i-th phase interval on the spectrum, In the phase interval The probability of the number of discharges on the surface. for The mean number of discharges on the plane, where σ is the variance.

[0110] The power frequency channel contains only a single sinusoidal waveform, representing the phase position corresponding to each pulse occurrence. The ultra-high frequency channel records the discharge pulse signal, i.e., a series of pulse clusters, representing the pulse oscillation and amplitude information. Therefore, the data processed by formulas 1)-5) comes from the phase position information and amplitude position information of these two channels, yielding time and frequency domain characteristic values.

[0111] Different defects have different frequency distribution characteristics in their discharge waveforms, and their bandwidth varies.

[0112] The bandwidth P width It is calculated by performing FFT transformation on several typical waveforms that have been collected.

[0113] The maximum frequency point P max It is calculated by performing FFT transformation on several typical waveforms that have been collected.

[0114] Example 2:

[0115] See Figures 1 to 8 A method for establishing a partial discharge fingerprint database for motor stator bars based on ultra-high frequency method includes the following steps:

[0116] 1) Build a partial discharge detection platform for typical defects in stator bars.

[0117] The stator bar typical defect partial discharge detection platform is used to detect the electromagnetic wave signals radiated by the stator bar under test.

[0118] The stator bar typical defect partial discharge detection platform includes the stator bar to be tested, an AC power supply, an ultra-high frequency antenna, a high-speed oscilloscope, and a host computer.

[0119] The stator bars to be tested include defect-free stator bars and defective stator bars.

[0120] Typical defects in the stator bars include poor anti-corona overlap defects (surface defects) and internal delamination defects.

[0121] 2) The AC power supply applies voltage to the stator windings under test in segments. During the voltage application process, partial discharge occurs in the stator windings under test. The ultra-high frequency antenna receives the electromagnetic wave signals radiated by the partial discharge of the stator windings under test in real time and sends the collected electromagnetic wave signals to a high-speed oscilloscope.

[0122] 3) After receiving the electromagnetic wave signal, the high-speed oscilloscope plots and displays the waveform of the partial discharge of the stator winding to be tested and the corresponding power frequency period waveform.

[0123] 4) The high-speed oscilloscope sends the discharge data of the UHF antenna channel and the power frequency cycle channel to the host computer, and the host computer extracts the discharge pulses from the collected discharge pulse groups.

[0124] 5) Replace the stator bar to be tested and repeat steps 2) to 4) to obtain discharge pulses under different insulation defects and discharge conditions;

[0125] 6) Based on the discharge pulses extracted by the host computer under different insulation defects and discharge conditions, a database of partial discharge characteristic values ​​under different insulation defects and discharge conditions is calculated.

[0126] 7) Based on a database of partial discharge characteristic values ​​under different insulation defects and discharge conditions, a fingerprint database for partial discharge diagnosis is constructed, using partial discharge characteristic values ​​as the basis for judgment. This fingerprint database can determine the different insulation defect types corresponding to the discharge and assess the severity of the partial discharge.

[0127] Example 3:

[0128] A method for establishing a partial discharge fingerprint database for motor stator windings based on ultra-high frequency (UHF) method is described in Example 2. The main steps involve placing the UHF antenna on the side of the stator winding where typical defects are detected, and connecting it to the input of an oscilloscope via an RF amplifier and a coaxial cable.

[0129] The host computer is connected to the output of the oscilloscope via a coaxial cable.

[0130] Example 4:

[0131] A method for establishing a partial discharge fingerprint database for motor stator bars based on ultra-high frequency method is described in Example 3. The main steps are as follows: the output impedance of the ultra-high frequency antenna is consistent with the wave impedance of the coaxial cable.

[0132] Example 5:

[0133] A method for establishing a partial discharge fingerprint database for motor stator windings based on ultra-high frequency method, the main steps of which are described in Example 2, wherein the AC power supply applies voltage to the stator windings to be tested in segments, including the following steps:

[0134] 1) Apply voltage U0 to the stator winding to be tested, where U0 is the rated voltage corresponding to the bar. Observe whether there is obvious partial discharge at the insulation defect.

[0135] 2) If there is no obvious partial discharge phenomenon after each cycle T, the voltage of the bar is gradually increased until the bar begins to show obvious partial discharge phenomenon. Then the voltage is stopped and the voltage applied at this time is used as the starting voltage U1 for data recording.

[0136] 3) Apply voltage U based on the starting voltage U1 in step 2) and start recording the discharge pulse signal data at this time.

[0137] 4) Set ΔU as the segmented voltage value. Each time the applied voltage U increases by ΔU, record the discharge pulse signal data at this time.

[0138] 5) Set the maximum applied voltage to U max Compare the voltage U and the maximum applied voltage U. max , if U>U max Then the pressurization ends if U ≤ U max Then return to step 4).

[0139] Example 6:

[0140] A method for establishing a partial discharge fingerprint database for motor stator bars based on ultra-high frequency method is described in Example 5. The main steps include observing whether there is obvious partial discharge phenomenon at the insulation defect. The steps include recording the curves of ambient sound, voltage and ultraviolet photon count.

[0141] When the difference between the discharge voltage and the rated voltage of the conductor is less than a preset voltage threshold, if the ambient sound level is greater than a preset decibel threshold, there will be obvious partial discharge at the insulation defect. The voltage threshold and decibel threshold are set in advance according to the actual situation.

[0142] The standard for judging the sound is that, compared with the quiet environment when it is not discharged, the discharge sound is significantly increased. This is relatively subjective, and a noticeable sound can usually be heard when it is close to the rated voltage of the rod.

[0143] If the slope of the curve of voltage versus ultraviolet photon number changes abruptly, there will be obvious partial discharge at the insulation defect.

[0144] The criterion for judging the effect of the ultraviolet imager is to plot a curve of applied voltage versus ultraviolet photon count. When the slope of the curve changes abruptly, it indicates a violent discharge phenomenon (generally, an ultraviolet photon count >300 is considered a relatively significant discharge). In this case, the intercept point between the tangent to the curve and the voltage on the horizontal axis is taken as the starting voltage for the corona initiation, as shown in the example. Figure 8 (Note: This method is mainly applicable to surface discharge. For internal discharge, since the discharge occurs inside the insulation, it is impossible to collect the number of ultraviolet photons. In this case, the sound judgment standard is more suitable.)

[0145] Example 7:

[0146] A method for establishing a partial discharge fingerprint database for motor stator bars based on ultra-high frequency method is described in Example 2. The main steps of the method are as follows: the processing of the collected discharge pulse groups by the host computer also includes noise suppression.

[0147] Example 8:

[0148] A method for establishing a partial discharge fingerprint database for motor stator bars based on ultra-high frequency (UHF) method is described in Example 2. The main steps involve constructing a corresponding partial discharge pulse (PRPD) spectrum from the discharge pulses extracted by a host computer. The PRPD spectrum is used to calculate a database of partial discharge characteristic values ​​under different discharge conditions.

[0149] Example 9:

[0150] A method for establishing a partial discharge fingerprint database for motor stator bars based on ultra-high frequency method is described in Example 2. The main steps are as follows: the partial discharge feature values ​​include time domain and frequency domain feature values.

[0151] Example 10:

[0152] A method for establishing a partial discharge fingerprint database for motor stator bars based on ultra-high frequency method is described in Example 2. The main steps include: the ratio of positive to negative half-cycle discharge counts N1, the ratio of positive to negative half-cycle discharge amounts N2, the correlation coefficient between positive and negative half-axis C1, the correlation coefficient between positive and negative half-cycles C2, and the kurtosis K.u .

[0153] The characteristic values ​​of the frequency domain include: bandwidth P. width Maximum frequency point P max .

[0154] Example 11:

[0155] A method for establishing a partial discharge fingerprint database for motor stator bars based on ultra-high frequency method, the main steps of which are shown in Example 10, wherein the ratio of positive and negative half-cycle discharge times N1 is as follows:

[0156]

[0157] In the formula, N + and N - These represent the number of discharges during the positive and negative half-cycles of the power frequency cycle, respectively.

[0158] The larger N1 is, the more times the negative half-axis is discharged.

[0159] The ratio of positive to negative half-cycle discharge N2 reflects the comparison of the severity of partial discharge during the negative and positive half-cycles of the power frequency voltage, as shown below:

[0160]

[0161] In the formula, k1 and k2 are the number of phase segments in the positive and negative half-cycles, respectively, and q i + q i - These represent the partial discharge quantities defined within the i-th phase segment of the positive and negative half-cycles of the power frequency voltage, respectively. The larger N2 is, the more intense the partial discharge is within the negative half-cycle of the power frequency cycle.

[0162] A partial discharge signal The similarity of the three-dimensional spectra along the positive and negative half-axis of amplitude is represented by the correlation coefficient C1. In the three-dimensional spectrum, n represents the number of discharges, and q represents the partial discharge quantity. Indicates phase.

[0163] The correlation coefficient C1 between the positive and negative half-axis is shown below:

[0164]

[0165] In the formula, n 1i n 2i This represents the average number of discharges across all amplitude ranges on both the positive and negative half-axis. k max This represents the absolute value of the maximum amplitude of the spectrum on the positive and negative half-axis.

[0166] A partial discharge signal The similarity of the three-dimensional spectrum in the positive and negative half-cycles of amplitude is represented by the correlation coefficient C2, which is shown below:

[0167]

[0168] In the formula, n' 1i 、n' 2i This represents the average number of discharges in each phase interval of the positive and negative half-cycles. N is the number of divisions in the phase intervals of the positive and negative half-cycles.

[0169] The kurtosis K u This reflects the sharpness of the peaks in the spectrum, as shown below:

[0170]

[0171] In the formula, For the i-th phase interval on the spectrum, In the phase interval The probability of the number of discharges on the surface. for The mean number of discharges on the plane, where σ is the variance.

[0172] The power frequency channel contains only a single sinusoidal waveform, representing the phase position corresponding to each pulse occurrence. The ultra-high frequency channel records the discharge pulse signal, i.e., a series of pulse clusters, representing the pulse oscillation and amplitude information. Therefore, the data processed by formulas 1)-5) comes from the phase position information and amplitude position information of these two channels, yielding time and frequency domain characteristic values.

[0173] Different defects have different frequency distribution characteristics in their discharge waveforms, and their bandwidth varies.

[0174] The bandwidth P width It is calculated by performing FFT transformation on several typical waveforms that have been collected.

[0175] The maximum frequency point P max It is calculated by performing FFT transformation on several typical waveforms that have been collected.

[0176] Example 12:

[0177] See Figures 1 to 8 A method for establishing a partial discharge fingerprint database for motor stator bars based on ultra-high frequency method, characterized by the following main steps:

[0178] 1) Build a partial discharge detection platform for typical defects in stator windings, including the stator winding to be tested, AC power supply, UHF antenna, high-speed oscilloscope, and host computer.

[0179] The stator winding is a stator bar.

[0180] The AC power supply powers the stator bar under test, causing it to exhibit noticeable partial discharge. The UHF antenna is placed next to a typical defect on the stator bar under test. The UHF antenna is connected to the input of the oscilloscope via an RF amplifier and a coaxial cable. The host computer is connected to the input of the oscilloscope via a coaxial cable.

[0181] 2) Apply AC power to the stator bars under test in segments.

[0182] 3) During the pressurization process, the UHF antenna receives the electromagnetic wave signal radiated by the stator bar under test when partial discharge occurs in real time, and sends it to the oscilloscope. After receiving the electromagnetic wave signal collected by the UHF antenna, the oscilloscope plots and displays the discharge pulse group of partial discharge and the corresponding power frequency period waveform. Then the high-speed oscilloscope sends the discharge data of the UHF antenna channel and the power frequency period channel to the host computer.

[0183] 4) The host computer performs noise suppression and discharge pulse extraction on the collected discharge pulse dataset. Discharge pulse signals under different insulation defects and voltages are extracted, and the phase and amplitude of the discharge pulses are statistically analyzed. Based on this, a database of partial discharge characteristic values ​​under various discharge conditions is calculated.

[0184] 5) Based on a database of partial discharge characteristic values ​​under different insulation defects and voltages, a fingerprint database for partial discharge diagnosis is finally constructed, which is based on partial discharge characteristic values. This database can determine the different types of insulation defects corresponding to the discharge and assess the severity of the partial discharge.

[0185] The main steps for applying AC power to the stator bars under test in segments are as follows:

[0186] 1) First, apply a voltage U0 to the stator winding to be tested, where U0 is the rated voltage corresponding to that bar. Then observe whether there is obvious partial discharge at the insulation defect, using sound and ultraviolet imaging as the judgment criteria.

[0187] Specific judgment criteria:

[0188] The standard for judging the sound is that, compared with the quiet environment when it is not discharged, the discharge sound is significantly increased. This is relatively subjective, and a noticeable sound can usually be heard when it is close to the rated voltage of the rod.

[0189] The criterion for judging the effect of ultraviolet (UV) imagers is to plot a curve of applied voltage versus UV photon count. When the slope of the curve changes abruptly, indicating a significant increase in the number of UV photons, it signifies a more intense discharge phenomenon, and the signal at this point can be acquired (generally, a UV photon count >300 is considered a significant discharge). The intercept point between the tangent to the curve and the horizontal axis voltage is taken as the initiation voltage for the halo effect. See example... Figure 8 (Note: This method is mainly applicable to surface discharge. For internal discharge, since the discharge occurs inside the insulation, it is impossible to collect the number of ultraviolet photons. In this case, the sound judgment standard is more suitable.)

[0190] 2) If there is no obvious partial discharge phenomenon after each cycle T, gradually increase the voltage of the bar until the bar begins to show obvious partial discharge phenomenon, then stop increasing the voltage and use it as the starting voltage for data recording.

[0191] 3) Based on the starting voltage in 2), start recording the partial discharge signal at this time, including the signal of the UHF antenna channel and the power frequency sinusoidal voltage signal at this time, so as to facilitate the recording of phase information.

[0192] 4) Gradually increase the applied voltage U = U + ΔU, where ΔU is the segmented voltage value. Record the partial discharge signal at this point.

[0193] 5) Determine if U>Umax is true. If true, end the pressurization process; otherwise, return to step 4.

[0194] The output impedance of an ultra-high frequency antenna should be consistent with the wave impedance of the coaxial cable.

[0195] Based on the collected partial discharge signals, the corresponding PRPD spectrum is constructed. A database of partial discharge characteristic values ​​under various conditions is then calculated using this data.

[0196] The partial discharge features to be extracted include those in the time domain and those in the frequency domain.

[0197] The time-domain features include: the ratio of positive to negative half-cycle discharge counts N1, the ratio of positive to negative half-cycle discharge quantities N2, the correlation coefficient between positive and negative half-axis C1, the correlation coefficient between positive and negative half-cycles C2, and the kurtosis Ku. The frequency-domain features are mainly the spectral distribution of typical waveforms, including: bandwidth Pwidth and maximum frequency point Pmax.

[0198] ① The ratio of positive and negative half-cycle discharge times to N1

[0199] Given the amplitude values ​​at each point of the PRPD spectrum, we have:

[0200]

[0201] In the formula, N+ and N- represent the number of discharges in the positive and negative half-cycles of the power frequency cycle, respectively.

[0202] The larger N1 is, the more times the negative half-axis is discharged.

[0203] ② The ratio of positive to negative half-cycle discharge N2

[0204]

[0205] In the formula, k1 and k2 are the number of phase segments in the positive and negative half-cycles, respectively, and q i + q i - These represent the partial discharge quantities defined within the i-th phase segment of the positive and negative half-cycles of the power frequency voltage, respectively.

[0206] N2 reflects the comparison between the severity of partial discharge during the negative half-cycle and the positive half-cycle of the power frequency voltage. That is, the larger N2 is, the more severe the partial discharge is during the negative half-cycle of the power frequency cycle.

[0207] ③ Correlation coefficient C1 between positive and negative half axes

[0208] A partial discharge signal The similarity of the three-dimensional spectrum along the positive and negative half-axis of amplitude is represented by the correlation coefficient C1. The amplitude range of the positive and negative half-axis is divided into kma parts, where kma is the absolute value of the maximum amplitude of the spectrum along the positive and negative half-axis.

[0209]

[0210] In the formula n 1i n 2i This represents the average number of discharges across all amplitude ranges of the positive and negative half-axis.

[0211] ④ Correlation coefficient C2 (positive and negative half-cycles)

[0212] A partial discharge signal The similarity of the three-dimensional spectrum in the positive and negative half-cycles of the amplitude is represented by the correlation coefficient C1, and the phase intervals of the positive and negative half-cycles are divided into N parts.

[0213]

[0214] In the formula n 1i n 2i This represents the average number of discharges in each phase interval of the positive and negative half-cycles.

[0215] ⑤ Ku

[0216] Peak sharpness reflects the sharpness of the peaks in a spectrum. The amplitude at each point in the spectrum is then:

[0217]

[0218] In the formula, For the i-th phase interval on the spectrum, In order to be in The probability of the number of discharges on the surface. for The mean number of discharges on the plane, where σ is the variance.

[0219]

[0220] The power frequency channel contains only a single sinusoidal waveform, representing the phase position corresponding to each pulse occurrence. The ultra-high frequency channel records the discharge pulse signal, i.e., a series of pulse clusters, representing the pulse oscillation and amplitude information. Therefore, the data processed by the above formula comes from the phase and amplitude position information of these two channels, yielding time and frequency domain characteristic values.

[0221] ⑥ Bandwidth P

[0222] Different defects exhibit different frequency distribution characteristics in their discharge waveforms, resulting in variations in their bandwidth. By acquiring multiple typical waveforms, performing FFT transformations, and calculating their bandwidth distribution range, this study aims to determine the optimal frequency distribution range.

[0223] ⑦ Maximum frequency point Pmax

[0224] Collect multiple typical waveforms, perform FFT transformation, and calculate the maximum frequency point in their frequency band.

[0225] Example 13:

[0226] See Figures 1 to 4 A method for establishing a partial discharge fingerprint database for motor stator bars based on ultra-high frequency method mainly includes the following steps:

[0227] (1) Build a partial discharge detection platform for typical defects of stator bars, including the stator winding to be tested, AC power supply, UHF antenna, high-speed oscilloscope, and host computer.

[0228] The AC power supply powers the stator bar under test, causing it to exhibit noticeable partial discharge. The UHF antenna is placed next to a typical defect on the stator bar under test. The UHF antenna is connected to the input of the oscilloscope via an RF amplifier and a coaxial cable. The host computer is connected to the input of the oscilloscope via a coaxial cable.

[0229] (2) Apply AC power to the stator bar under test in segments.

[0230] (3) During the pressurization process, the UHF antenna receives the electromagnetic wave signal radiated by the stator bar under test when it undergoes partial discharge in real time, and sends it to the oscilloscope. After receiving the electromagnetic wave signal collected by the UHF antenna, the oscilloscope plots and displays the discharge pulse group of the partial discharge and the corresponding power frequency period waveform. Then the high-speed oscilloscope sends the discharge data of the UHF antenna channel and the power frequency period channel to the host computer.

[0231] See Figure 2 , Figure 3 , Figure 4 The images show signals acquired during partial discharge of the wire rods exhibiting three main defect types. At lower applied voltages, the partial discharge waveform received by the oscilloscope did not show pulse-like signals. Further increasing the applied voltage resulted in a partial discharge waveform that... Figure 2 , Figure 3 , Figure 4 similar.

[0232] Furthermore, pulse signals typically appear first during the negative half-cycle of the power frequency, and then, after the applied voltage is increased, pulse signals will also appear during the positive half-cycle of the power frequency. The number of discharge pulses will also increase significantly, indicating that the partial discharge is becoming more severe.

[0233] (4) The host computer performs noise suppression and discharge pulse extraction on the collected discharge pulse dataset. Discharge pulse signals under different insulation defects and different voltages are extracted, and the phase and amplitude of the discharge pulses are statistically analyzed. Based on this, a database of partial discharge characteristic values ​​under various discharge conditions is calculated.

[0234] according to Figure 2 , Figure 3 , Figure 4 The aforementioned partial discharge characteristic values ​​in various time domains can be calculated. Taking surface defects and internal defects as examples, they exhibit significant differences in characteristic values ​​such as the ratio of positive to negative half-cycle discharge counts (N1), the ratio of positive to negative half-cycle discharge amounts (N2), the correlation coefficient between positive and negative half-axis (C1), the correlation coefficient between positive and negative half-cycles (C2), and the kurtosis (Ku). For instance, the coefficient C1 for internal defects is approximately equal to 1, while the coefficient C1 for surface defects is significantly less than 1. Setting a threshold can serve as a basis for distinguishing between internal and external defects.

[0235] See Figure 5 , Figure 6 , Figure 7 The amplitude-frequency distribution diagrams of typical waveforms for defect-free rods, surface defects, and internal defects are shown respectively, allowing for the calculation of partial discharge characteristic values ​​in the frequency domain. It is feasible to determine whether the discharge is internal to the stator winding or surface discharge at the winding ends based on the characteristic peaks in different frequency bands. As can be seen from the figures, the bandwidth of internal discharge is significantly larger than that of surface discharge, and the characteristic peak values ​​also show significant differences.

[0236] (5) Based on the database of partial discharge characteristic values ​​under different insulation defects and different voltages, a fingerprint database for partial discharge diagnosis based on partial discharge characteristic values ​​is finally constructed, which can determine the different insulation defect types corresponding to the discharge and assess the severity of the partial discharge.

Claims

1. A method for establishing a partial discharge fingerprint library of a motor stator bar based on an ultra-high frequency method, characterized in that, The method comprises the following steps: 1) setting up a stator bar typical defect partial discharge detection platform; The stator bar typical defect partial discharge detection platform is used for detecting electromagnetic wave signals radiated by the stator bar to be detected; The stator bar typical defect partial discharge detection platform comprises a stator bar to be detected, an alternating current power supply, an ultra-high frequency antenna, a high-speed oscilloscope and a host computer; The stator bar to be detected comprises a defect-free stator bar and a defective stator bar; The defects of the stator bar include poor bonding defects and internal layering defects; 2) The alternating current power supply applies segmented voltage to the stator bar to be detected, and in the process of voltage application, the stator bar to be detected generates partial discharge; the ultra-high frequency antenna receives electromagnetic wave signals radiated by the stator winding to be detected in real time, and sends the collected electromagnetic wave signals to the high-speed oscilloscope; The alternating current power supply applies segmented voltage to the stator winding to be detected, and in the process of voltage application, the stator winding to be detected generates partial discharge; the ultra-high frequency antenna receives electromagnetic wave signals radiated by the stator winding to be detected in real time, and sends the collected electromagnetic wave signals to the high-speed oscilloscope; The alternating current power supply applies segmented voltage to the stator winding to be detected, and in the process of voltage application, the stator winding to be detected generates partial discharge; the ultra-high frequency antenna receives electromagnetic wave signals radiated by the stator winding to be detected in real time, and sends the collected electromagnetic wave signals to the high-speed oscilloscope; 2.1) applying voltage U0 to the stator winding to be detected, wherein U0 is the rated voltage corresponding to the stator bar; and observing whether there is obvious partial discharge phenomenon at the insulation defect; 2.2) every time a period T elapses, if there is no obvious partial discharge phenomenon, gradually increasing the voltage applied to the stator winding to be detected until the stator winding to be detected starts to generate obvious partial discharge phenomenon; stopping voltage application; and taking the voltage applied at this time as the starting voltage U1 of data recording; 2.3) taking the starting voltage U1 in step 2.2) as the basis, applying voltage U, and starting to record the discharge pulse signal data at this time; 2.5) Set the maximum applied voltage to U max ; compare the voltage U to the maximum applied voltage U max , if U > U max , then end the voltage application, if U ≤ U max , then return to step 2.4); 2.4) setting ΔU as the segmented voltage increment, increasing the voltage U by ΔU every time, and recording the discharge pulse signal data at this time; 3) after the high-speed oscilloscope receives the electromagnetic wave signals, drawing and displaying the waveform of the partial discharge of the stator winding to be detected and the corresponding power frequency period waveform; 4) the high-speed oscilloscope sends the discharge data of the ultra-high frequency antenna channel and the power frequency period channel to the host computer, and the host computer extracts the discharge pulse from the collected discharge pulse group; 5) replacing the stator bar to be detected, repeating steps 2) to 4), and obtaining discharge pulses under different insulation defects and different discharge conditions; 6) based on the discharge pulses under different insulation defects and different discharge conditions extracted by the host computer, calculating a database of partial discharge characteristic values under different insulation defects and different discharge conditions; 2. The method for establishing a partial discharge fingerprint library of motor stator bars based on the ultra-high frequency method according to claim 1, characterized in that, 7) based on the database of partial discharge characteristic values under different insulation defects and different discharge conditions, constructing a fingerprint library of partial discharge diagnosis taking the partial discharge characteristic values as the judgment basis. The ultra-high frequency antenna is placed on the side of the typical defect of the stator winding to be detected; and is connected to the input end of the oscilloscope through a radio frequency amplifier and a coaxial cable; 3. The method according to claim 2, wherein, The host computer is connected to the output end of the oscilloscope through a coaxial cable.

4. The method of claim 1, wherein the method is characterized by: The output impedance of the ultra-high frequency antenna is consistent with the wave impedance of the coaxial cable. The step of observing whether there is obvious partial discharge phenomenon at the insulation defect comprises: recording the curves of environmental sound, voltage and ultraviolet photon number; When the difference between the discharge voltage and the rated voltage of the stator bar is less than a preset voltage threshold, if the environmental sound decibel number at this time is greater than a preset decibel threshold, there is obvious partial discharge phenomenon at the insulation defect; If the curve slope of voltage and ultraviolet photon number changes suddenly, there is obvious partial discharge at the insulation defect.

5. The method of claim 1, wherein the method is characterized by: The processing of the discharge pulse group collected by the upper computer further includes noise suppression.

6. The method of claim 1, wherein the method is characterized by: The discharge pulse extracted by the upper computer constructs a corresponding PRPD spectrum diagram; the PRPD spectrum diagram is used for calculating a partial discharge characteristic value database under different discharge conditions.

7. The method of claim 1, wherein the method is characterized by: The partial discharge characteristic values include characteristic values in time domain and frequency domain.

8. The method according to claim 7, wherein, The characteristic values of the time domain include: positive and negative half cycle discharge frequency ratio N1, positive and negative half cycle discharge quantity ratio N2, positive and negative half axis correlation coefficient C1, positive and negative half cycle correlation coefficient C2, and kurtosis K u ; The characteristic value of the frequency domain includes: frequency band range width P width , maximum frequency point P max .

9. The method according to claim 8, wherein, The positive and negative half-cycle discharge frequency ratio N1 is as follows: In the formula, N + and N - are the discharge times of the positive and negative half cycles of the power frequency cycle, respectively. The positive and negative half-cycle discharge quantity ratio N2 reflects the comparison of the severity of partial discharge in the negative half-cycle of power frequency voltage and the positive half-cycle of power frequency voltage, and is as follows: The positive and negative half-cycle discharge quantity ratio N2 reflects the comparison of the severity of partial discharge in the negative half-cycle of power frequency voltage and the positive half-cycle of power frequency voltage, and is as follows: In the formula, k1, k2 are the phase segment numbers of positive and negative half cycles respectively, q i + , q i - respectively represent the defined partial discharge quantity in the i th phase segment of positive and negative half cycle of power frequency voltage. a partial discharge signal The similarity of the three-dimensional spectrum in the positive and negative half-axes of the amplitude is represented by a correlation coefficient C1, which is given by: In the formula, n 1i n 2i k is the average number of discharges across all amplitude ranges of the positive and negative half-axis. max This represents the absolute value of the maximum amplitude of the spectrum along the positive and negative half-axis. a partial discharge signal The similarity of the three-dimensional spectrum in the positive and negative half cycles of the amplitude is expressed by a correlation coefficient C2, which is given by: In the formula, n' 1i 、n' 2i is the average number of discharges in each phase interval of the positive and negative half-cycles; N is the number of divisions in the phase intervals of the positive and negative half-cycles. The kurtosis K u reflects the sharpness of the peaks of the spectrum, as follows: wherein is the i-th phase interval on the spectrum, is the discharge frequency probability on the phase interval , is the average discharge frequency on the plane, and σ is the variance, The frequency band range width P width is calculated by FFT transform of several typical waveforms collected; The maximum frequency point P max is calculated by FFT transform of several typical waveforms collected.

Images