Partial discharge diagnostic device, partial discharge diagnostic method, and partial discharge diagnostic system

Abstract

The invention provides a partial discharge diagnosis system, a partial discharge diagnosis device, and a partial discharge diagnosis method, which can restrain reduction of determination precision of partial discharge even if false partial discharge occurs. A partial discharge diagnostic device for a device according to an embodiment of the present invention is a partial discharge diagnostic device for performing a cause determination of a partial discharge signal of an insulator, and is provided with a processing unit and a learning model generation unit. The processing unit performs generation processing on data of a predetermined form in which a pseudo partial discharge signal in the electrical signal that varies according to the phase is suppressed. The learning model generation unit generates, on the basis of the data, a learning model for determining at least one of the cause of the partial discharge and the presence or absence of the partial discharge.

Description

Technical Field

[0001] Embodiments of the present invention relate to a partial discharge diagnostic device, a partial discharge diagnostic method, and a partial discharge diagnostic system. Background Art

[0002] Plant equipment includes power plants such as generators, motors, inverters, switchgear, and cables. These devices have insulation on their conductors, and the insulation performance of these insulation materials degrades over time. For example, repeated thermal expansion and contraction due to temperature fluctuations in the generator coils of a generator can cause insulation failure. This deterioration can lead to failures in power plants due to insulation breakdown.

[0003] It is known that partial discharge occurs from power equipment when insulation deteriorates. Therefore, the insulation condition is determined based on the occurrence of partial discharge.

[0004] As such a technique, refer to Japanese Patent Application Laid-Open No. 2022-161713 (Patent Document 1) and Japanese Patent Application Laid-Open No. 2023-180397 (Patent Document 2).

[0005] However, pseudo partial discharge, which is similar to partial discharge, may occur, and in this case, the accuracy of determining the insulation state may be reduced. Summary of the Invention

[0006] An object of the present invention is to provide a partial discharge diagnostic system, a partial discharge diagnostic device, and a partial discharge diagnostic method that can suppress a decrease in the accuracy of partial discharge determination even when a false partial discharge occurs.

[0007] A partial discharge diagnostic device for equipment according to an embodiment of the present invention is configured to determine the cause of a partial discharge signal in an insulator. The device comprises a processing unit and a learning model generator. The processing unit generates data in a predetermined format after suppressing false partial discharge signals in an electrical signal due to phase fluctuations. The learning model generator generates a learning model based on the data to determine at least one of the cause of the partial discharge and the presence or absence of the partial discharge.

[0008] Effects of the Invention

[0009] According to the present invention, it is possible to improve the accuracy of determining the cause of partial discharge in electric power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1This is a diagram illustrating different forms of the φ-qn characteristic diagram depending on the cause of partial discharge.

[0011] Figure 2 : is a diagram showing an example of a partial discharge signal and a pseudo-discharge signal.

[0012] Figure 3 This is a diagram showing an example in which partial discharge signals generated in the U-phase and the W-phase overlap with the V-phase.

[0013] Figure 4 This is a block diagram showing a configuration example of a partial discharge diagnostic system.

[0014] Figure 5 This is a block diagram showing an example of the configuration of a data processing unit.

[0015] Figure 6 This is a diagram showing an example of a method for generating a φ-q characteristic map by the calculation unit.

[0016] Figure 7 This is a diagram showing an example of a method for generating a φ-qn characteristic map by the calculation unit.

[0017] Figure 8 This is a diagram showing an example of performing Fast Fourier Transformation on the time interval of time-series data of an electrical signal.

[0018] Figure 9 This is a diagram showing an example of spectrogram generation by the calculation unit.

[0019] Figure 10 This is a diagram showing an example of normalization of time-series data of an electrical signal.

[0020] Figure 11 This is a diagram showing an example in which a logarithmic transformation is performed on a two-dimensional matrix of a φ-qn characteristic diagram.

[0021] Figure 12 This is a diagram for explaining an increase processing method using averaging.

[0022] Figure 13 This is a block diagram showing a configuration example of a feature extraction unit.

[0023] Figure 14 This is a diagram showing an example of a frequency spectrum generated by performing fast Fourier transform on an analog electrical signal.

[0024] Figure 15 This is a diagram showing an example of setting a frequency range based on the value of the spectrum of an analog electrical signal.

[0025] Figure 16 This figure shows an example of selecting a frequency range based on the size of the spectrum and the number of occurrences.

[0026] Figure 17 This figure shows an example of selecting a frequency range using a spectrum graph.

[0027] Figure 18 This is a diagram illustrating a method of selecting a phase range based on the φ-qn characteristic diagram.

[0028] Figure 19 It is a diagram showing an example of generating a charge amount histogram (histogram) with respect to the charge amount.

[0029] Figure 20 This is a diagram illustrating a method of selecting a phase range based on the peak position of a data point in a φ-q characteristic diagram.

[0030] Figure 21 The arrows in the φ-q characteristic diagram indicate pseudo partial discharge.

[0031] Figure 22 This is a diagram showing an example of integrating data points on the φ-q characteristic diagram.

[0032] Figure 23 This is a diagram illustrating a method of selecting a phase range based on a spectrum diagram.

[0033] Figure 24 This is a flowchart showing a processing example of the partial discharge diagnostic system.

[0034] Figure 25 It is a diagram showing an example of imaging processing performed by the image generating unit.

[0035] Figure 26 This is a diagram showing an example of masking processing performed by a filtering unit.

[0036] Figure 27 This is a diagram schematically showing an example of the arrangement of a hydro-generator coil and a sensor.

[0037] Figure 28 This is a diagram showing an example of base point adjustment processing of an applied voltage by a filter unit.

[0038] Figure 29 It is a diagram showing an example of masking processing performed by the filter unit.

[0039] Description of Reference Signs

[0040] 1: Partial discharge diagnostic system, 10: Measuring instrument, 20: Display device, 40: Partial discharge diagnostic processing device, 100: Data acquisition unit, 102: Electric signal generation unit, 104: Feature extraction unit, 105: Processing unit, 106: Data processing unit, 108: Image generation unit, 110: Learning model generation unit, 112: Partial discharge determination unit, 200: Calculation unit, 202: Scale adjustment unit, 204: Data expansion unit, 206: Threshold processing unit, 208: Filtering unit, 302: Frequency range selection unit, 304: Phase range selection unit. DETAILED DESCRIPTION

[0041] Hereinafter, the partial discharge diagnostic device, partial discharge diagnostic method and partial discharge diagnostic system according to the embodiment of the present invention will be described in detail with reference to the accompanying drawings. In addition, the embodiment shown below is an example of an embodiment of the present invention, and the present invention is not limited to these embodiments. In addition, in the drawings referred to in the present embodiment, the same reference numerals or similar reference numerals are marked for parts having the same parts or the same functions, and their repeated descriptions are sometimes omitted. In addition, the dimensional ratios of the drawings are sometimes different from the actual ratios for the sake of convenience of explanation, and sometimes a part of the structure is omitted from the drawings. Hereinafter, first, using Figures 1 to 3 Partial discharge and pseudo partial discharge will be described.

[0042] [Forms of partial discharge in insulation]

[0043] use Figure 1 The following describes the behavior of partial discharge in insulators. A φ-q characteristic diagram shows the relationship between the phase of an applied voltage and the value of an electrical signal. An electrical signal corresponds to the phase of an applied voltage to, for example, an electrical device. For example, an electrical signal is at least one of charge, current, and voltage that varies with phase. Therefore, in this embodiment, the term "electrical signal" includes at least one of a charge signal, a current signal, and a voltage signal that varies with phase.

[0044] In this embodiment, a charge amount signal is used as the electrical signal for explanation. However, this is not limiting, and either a current signal or a voltage signal can also be used. The φ-qn characteristic diagram is a characteristic diagram that accumulates the φ-q characteristic over a certain period of time and shows the correlation between the partial discharge charge amount, the number of occurrences, and the applied voltage phase.

[0045] Figure 1 This is a φ-q characteristic diagram for each cause of partial discharge. Figure 1The φ-q characteristic diagram shows the relationship between the phase φ of the applied voltage L10 and the charge amount q. The vertical axis is the charge amount, and the horizontal axis is the phase (time). During one cycle of the applied voltage L10, there are multiple insulation degradation causes that can cause partial discharge signals. Therefore, the shape of the φ-q characteristic diagram varies depending on the insulation degradation cause. For example, Figure 1 (a) produces a partial discharge signal between 180° and 270°, Figure 1 (b) A partial discharge signal is generated between 0° and 90° and between 90° and 270°. As will be described later, the partial discharge signal is measured as, for example, a charge amount larger than the average charge amount.

[0046] Figure 1 (c) generates a partial discharge signal on the negative (minus) side of 0° to 90° and the positive side of 90° to 270°. Each of these forms can be associated with the cause of the occurrence. In this embodiment, when performing machine learning for determining the insulation state of the power equipment, learning data associated with the cause of the occurrence can be used. In addition, Figure 1 These are examples of the form of the φ-q characteristic diagram according to the cause of partial discharge, and the pattern of the cause of partial discharge is not limited to these.

[0047] [False partial discharge signal]

[0048] In this embodiment, an electrical signal not caused by the insulation of the detection target is sometimes referred to as a pseudo partial discharge signal. A pseudo partial discharge may be caused, for example, by a discharge signal generated outside the measurement target being transmitted to and measured by the sensor. Figure 2 This is a φ-q characteristic diagram showing an example of a pseudo partial discharge signal. The vertical axis is the charge amount, and the horizontal axis is the phase (time). For example, Figure 2 (a) is an example of a form in which a partial discharge signal is generated in the range of 0° to 90° and 90° to 270°. Figure 2 (b) shows an example of a form in which a pseudo partial discharge signal is generated near 90° and near 270°.

[0049] For example, in the φ-q characteristic diagram, Figure 1 as well as Figure 2 As shown in (a), many partial discharge signals are generated near at least one of the ranges of 0° to 90° and 180° to 270° of the applied voltage phase. In contrast, regarding the pseudo partial discharge signal, as shown in Figure 2 As shown in (b), according to the experimental results of the present applicant, it is known that the pseudo partial discharge signal has a statistical tendency to be out of phase with the generation of the partial discharge signal.

[0050] [False partial discharge signal in three-phase AC]

[0051] Furthermore, according to experimental results conducted by the present applicant, it has been found that signals may overlap between the U-phase, V-phase, and W-phase in a generator and a motor. Figure 3 This diagram shows an example in which partial discharge signals generated in the U-phase and the W-phase overlap with the V-phase. The vertical axis represents the charge amount, and the horizontal axis represents the phase (time). Figure 3 (a) is the φ-q characteristic diagram of the U phase, Figure 3 (b) is the V-phase φ-q characteristic diagram, Figure 3 (c) is a φ-q characteristic diagram of phase W. This example shows an example in which the partial discharge signals of the U and W phases are superimposed on the partial discharge signal S10 of the V phase as pseudo partial discharge signals PU and PW.

[0052] [Composition of the partial discharge diagnosis system]

[0053] Next, the system configuration of partial discharge diagnostic system 1 will be described. Figure 4 FIG. 1 is a block diagram showing an example of the configuration of the partial discharge diagnostic system 1. Figure 4 As shown, partial discharge diagnostic system 1 is a system capable of reducing the influence of false partial discharge signals and determining at least one of the presence or absence of partial discharge and the cause of partial discharge. Partial discharge diagnostic system 1 includes a measuring instrument 10, a display device 20, an operating device 30, and a partial discharge diagnostic processing device 40.

[0054] Measuring instrument 10 supplies time-series data of electrical signals measured by sensors installed in power equipment to partial discharge diagnostic processing device 40. For example, the sensors may be current-based high-frequency current sensors or electromagnetic wave-based electromagnetic antennas. As described above, the electrical signals include at least one of charge signals, current signals, and voltage signals.

[0055] The display device 20 is, for example, a monitor and displays image data supplied from the partial discharge diagnosis processing device 40 .

[0056] The operating device 30 is composed of an input device such as a keyboard, a mouse, etc. The operating device 30 inputs a signal corresponding to an operator's operation to the partial discharge diagnosis processing device 40 .

[0057] like Figure 4As shown, partial discharge diagnostic processing device 40 is a device capable of suppressing the influence of false partial discharges and performing partial discharge diagnosis. Partial discharge diagnostic processing device 40 includes a data acquisition unit 100, an electrical signal generation unit 102, a feature extraction unit 104, a processing unit 105, a learning model generation unit 110, a partial discharge determination unit 112, and a storage unit 116. Processing unit 105 includes a data processing unit 106 and an image generation unit 108.

[0058] The partial discharge diagnostic processing device 40 includes a CPU (Central Processing Unit), such as a computer. By executing a program stored in a storage unit 116, the partial discharge diagnostic processing device 40 can configure the data acquisition unit 100, the electrical signal generation unit 102, the feature extraction unit 104, the processing unit 105, the learning model generation unit 110, and the partial discharge determination unit 112. Furthermore, the data acquisition unit 100, the electrical signal generation unit 102, the feature extraction unit 104, the data processing unit 106, the image generation unit 108, the learning model generation unit 110, and the partial discharge determination unit 112 can also be configured using electronic circuits.

[0059] Data acquisition unit 100 acquires time-series data of electrical signals measured by sensors installed in power equipment from measuring instrument 10. Measuring instrument 10 includes multiple sensors. Data acquisition unit 100 can acquire time-series data of multiple different electrical signals from measuring instrument 10. Furthermore, measuring instrument 10 is not limited to a single instrument; multiple measuring instruments 10 can be used.

[0060] The electric signal generation unit 102 acquires or generates an electric signal (analog electric signal) for learning used in machine learning. For example, the electric signal generation unit 102 generates an analog electric signal having partial discharge signal data. Figure 1 The electrical signal generating unit 102 can also generate an analog electrical signal that does not have a partial discharge signal, as shown in the φ-q characteristic diagram of FIG.

[0061] The electrical signal generator 102 acquires, via the data acquisition unit 100, a simulated electrical signal obtained by simulating the cause of partial discharge in a test apparatus that replicates the entire power plant or a portion thereof. Furthermore, the electrical signal generator 102 can also generate data by adding pseudo partial discharge signal data to the data containing partial discharge signal data. This data is stored in the storage unit 116 in association with the cause of partial discharge.

[0062] Furthermore, the electrical signal generation unit 102 may also use simulation as a method for generating the analog electrical signal. Alternatively, the electrical signal generation unit 102 may combine the analog electrical signal obtained by the test apparatus with the data generated by the simulation. This increases the amount of data for partial discharge signals from various causes, further improving the accuracy of machine learning determination. Similar to the electrical signal obtained by the data acquisition unit 100, the analog electrical signal generated by the electrical signal generation unit 102 represents the amount of charge, voltage, current, and the like relative to phase (time).

[0063] The feature extraction unit 104 extracts features indicating the generation of a partial discharge signal for each cause of the partial discharge signal from the analog electrical signal generated by the electrical signal generation unit 102. The feature extraction unit 104 extracts features indicating the generation of a partial discharge signal from, for example, the magnitude and frequency of the electrical signal, and generates, as features, information on the range of the partial discharge signal generation, such as the phase range, frequency range, and magnitude range of the partial discharge signal. Specifically, these features refer to information on the range of the partial discharge signal generation, such as the phase range, frequency range, and magnitude range of the partial discharge signal. Furthermore, by using an analog electrical signal, the influence of noise, such as false partial discharge signals, is minimized, enabling efficient extraction of features related to the partial discharge signal for each cause of the signal generation. The feature extraction unit 104 will be described in detail later.

[0064] The data processing unit 106 performs pre-processing on the electrical signals acquired by the data acquisition unit 100 and the analog electrical signals generated by the electrical signal generation unit 102. Furthermore, the data processing unit 106 suppresses false partial discharge signals based on the information related to the partial discharge signal generated by the feature extraction unit 104. For example, the data processing unit 106 can suppress electrical signals in a range where there is a high probability that no partial discharge signal exists. Suppression in this embodiment also includes deletion. The details of the data processing unit 106 will be described later.

[0065] Image generation unit 108 can generate data in a predetermined format using data processed by data processing unit 106 or data processed by data processing unit 106 before processing. For example, image generation unit 108 generates data in a predetermined format as image data having two-dimensionally arranged numerical values. In this manner, processing unit 105, including data processing unit 106 and image generation unit 108, generates data in a predetermined format after suppressing false partial discharge signals in electrical signals that vary depending on the phase.

[0066] More specifically, the image generation unit 108 can generate an image in the form of a grayscale image or a color image. In addition, in the present embodiment, the data used in the machine learning or determination processing is sometimes configured in a two-dimensional matrix, but is not limited to this. Sometimes the learning data and determination data configured in a two-dimensional matrix are referred to as image data. That is, sometimes the learning data and determination data that can be visualized by configuring the data in a two-dimensional matrix are referred to as image data. By displaying this image data on the display device 20, the characteristics of the learning data and determination data can be grasped by visual observation.

[0067] The learning model generation unit 110 generates a learning model (discriminator) using the data processed by the data processing unit 106 as learning data. Furthermore, the learning model generation unit 110 can also generate a learning model using the image data generated by the image generation unit 108 and arranged in a two-dimensional matrix and capable of being visualized. The learning model generation unit 110 associates the cause of partial discharge, the presence or absence of partial discharge, and other information, serving as teaching signals, with the data processed by the data processing unit 106 as learning data. The machine learning performed by the learning model generation unit 110 can utilize neural networks or other discriminant learning algorithms, and the learning algorithm is not limited thereto. In this manner, the learning model generation unit 110 generates a learning model using learning data that associates at least one of the cause of partial discharge and the presence or absence of partial discharge with the data generated by the processing unit 105.

[0068] Partial discharge determination unit 112 uses the learning model generated by learning model generation unit 110, for example, based on the processing results of data processing unit 106 to determine the presence or absence of a partial discharge signal and its cause in the electrical signal to be diagnosed (determination data). Display control unit 114 can display the two-dimensional data generated by image generation unit 108 on display device 20.

[0069] Storage unit 116 is comprised of, for example, an HDD (hard disc drive) or an SSD (solid state drive). Storage unit 116 stores various data and programs used by partial discharge diagnostic apparatus 40. For example, storage unit 116 stores data acquired by data acquisition unit 100, learning data, determination data, and data related to a completed learning model.

[0070] Here, the details of the data processing unit 106 will be described. Figure 5 1 is a block diagram showing an example of the configuration of the data processing unit 106. Figure 5As shown, the data processing unit 106 includes a calculation unit 200, a scale adjustment unit 202, a data expansion unit 204, a threshold value processing unit 206, and a filter unit 208. The calculation unit 200 performs calculations such as statistical processing, fast Fourier transform, wavelet transform, linear transform, logarithmic transform, and other nonlinear transformations, standardization, normalization, regularization, and averaging on the electrical signal acquired by the data acquisition unit 100 and the analog electrical signal generated by the electrical signal generation unit 102. Alternatively, the calculation unit 200 can perform any of these processes individually or in combination.

[0071] In addition, the calculation unit 200 can output, as calculation processing results, for example, a φ-q characteristic diagram, a φ-qn characteristic diagram, a spectrogram, and a scalogram. Figure 6 as well as Figure 7 An example of how the calculation unit 200 generates a φ-qn characteristic map will be described.

[0072] Figure 6 1 is a diagram showing an example of a method for generating a φ-q characteristic map by the calculation unit 200 . Figure 6 (a) is a time-series electric signal acquired by the data acquisition unit 100. The vertical axis represents the charge amount, and the horizontal axis represents the phase (time) from 0 degrees to 360 degrees in one cycle of the applied voltage. Figure 6 (b) is to divide one period into multiple intervals and make Figure 6 The data in (a) becomes a graph of discrete values ​​for each interval. Furthermore, the time series data can be for a single period or multiple periods. That is, the number of times the time series data is overlapped can be set to any number. In this way, the calculation unit 200 can generate a φ-q characteristic graph by plotting the charge amount versus the phase of the applied voltage.

[0073] Figure 7 This is a diagram showing an example of a method for generating a φ-qn characteristic diagram by the calculation unit 200. The vertical axis represents the charge amount, and the horizontal axis represents the phase (time) from 0 degrees to 360 degrees in one cycle of the applied voltage. Figure 6 In the φ-q characteristic diagram, the vertical axis represents the charge amount and the horizontal axis represents the phase, which are divided into arbitrary intervals N10. The number of times the charge occurs in each interval N10 is accumulated and added to form the frequency of occurrences A to F, etc., and then displayed in a two-dimensional histogram. The accumulation of the number of charge occurrences can be performed for a single cycle of the electrical signal or for multiple cycles of the electrical signal.

[0074] Here, use Figure 8 as well as Figure 9 An example of generating a spectrogram by the calculation unit 200 will be described. Figure 8 This is a diagram showing an example of performing Fast Fourier Transformation (FFT) on time series data of an electrical signal for an arbitrary time interval. Figure 8 (a) is a diagram showing an example of converting time-series data of an electrical signal into a frequency spectrum. The vertical axis represents the charge amount, and the horizontal axis represents the phase (time) from 0 to 360 degrees in one cycle of the applied voltage. Figure 8 (b) is a diagram showing a spectrum in one section TSn. The vertical axis represents the spectrum, and the horizontal axis represents the frequency.

[0075] like Figure 8 As shown in (a), the operation unit 200 divides the phase range into n time partitions TS1 to TSn. n can be set to any number. The operation unit 200 performs a fast Fourier transform on the time series data of the electrical signal according to each time partition TS1 to TSn corresponding to the phase. In other words, Figure 8 The frequency spectrum corresponding to the charge amount in (b) represents the component value for each frequency obtained by fast Fourier transform of time series data of the electric signal in one interval.

[0076] Figure 9 1 is a diagram showing an example of spectrogram generation by the calculation unit 200 . Figure 9 (a) is a schematic representation of Figure 8 (b) is a diagram of a spectrum diagram of each time partition TS1 to TSn corresponding to the phase. Figure 9 (b) is to Figure 9 The phase of (a) is divided into multiple partitions, and the spectrum values ​​of each partition S10 are expressed as a two-dimensional histogram. The horizontal axis is phase, and the vertical axis is frequency. The values ​​a to f within each partition are the spectrum values ​​of each partition S10. In this way, the spectrum diagram is a two-dimensional matrix representation of the spectrum corresponding to the charge amount by performing a fast Fourier transform on the time series data of the electrical signal for arbitrary time partitions, with frequency on the vertical axis and phase relative to the applied voltage on the horizontal axis.

[0077] To generate the quantity map, the calculation unit 200 applies a wavelet transform to the time-series data of the electrical signal, instead of a fast Fourier transform. Specifically, the frequency of occurrence of frequency components based on the wavelet transform results is represented as a two-dimensional histogram. Thus, the quantity map is a two-dimensional matrix representing the frequency components corresponding to the charge amount, with frequency on the vertical axis and phase relative to the applied voltage on the horizontal axis, by applying a wavelet transform to the time-series data of the electrical signal at arbitrary time intervals.

[0078] The scale adjustment unit 202 can perform scale conversion on the time-series data of the electrical signal as a result of the calculation processing. Figure 10 This diagram shows an example of normalization of time-series data of an electrical signal. The horizontal axis represents time corresponding to the phase, and the vertical axis represents the charge amount. Figure 10 (a) and (c) are the data before normalization. Figure 10 (b) and (d) are normalized data.

[0079] like Figure 10 As shown, the scale adjustment unit 202 transforms the value of the time series data of the electrical signal in such a way that the reference value measured in an arbitrary time range becomes a prescribed value. For example, the prescribed value is set to 1. By such a standardization process based on the scale adjustment unit 202, even if the signal strength of the electrical signal of the actual machine obtained by the data acquisition unit is different from the scale (range of signal strength) of the analog electrical signal generated in the electrical signal generation unit 102, it is possible to achieve unification. In this way, it is possible to suppress the influence of the situation where the maximum and minimum values ​​are not determined and the situation where there are statistically sudden deviation values. Thus, even if the analog electrical signal generated in the electrical signal generation unit 102 is different from the signal strength of the electrical signal obtained in the actual machine, it is possible to prevent the judgment accuracy of the learning model from being reduced.

[0080] In addition, any value can be used as the reference value for standardization. For example, by setting the maximum value in one cycle of applied voltage or the maximum value in multiple cycles as the reference value, the scaling becomes more stable. Figure 10 In the example, the charge amount of the time series data is normalized, but the present invention is not limited to this. The scale adjustment unit 202 can also perform normalization on other data. In addition, the scale adjustment unit 202 can also normalize the data by using the minimum and maximum values ​​within a predetermined range instead of normalization.

[0081] The scale adjustment unit 202 can also perform scaling transformations on the two-dimensional matrix data of the φ-qn characteristic diagram, the spectrum diagram, and the quantity diagram. The scaling transformations performed by the scale adjustment unit 202 include linear transformation, logarithmic transformation, sigmoid transformation, binarization, standardization, normalization, and regularization.

[0082] Figure 11 This figure shows an example in which a logarithmic transformation is performed as a scaling transformation on a two-dimensional matrix of a φ-qn characteristic diagram. Figure 11 (a) is before transformation, Figure 11(b) is after transformation. The horizontal axis is the phase and the vertical axis is the charge amount. In this way, if the scale transformation is performed using logarithmic transformation, the electrical signal of the partial discharge based on the charge with a low frequency of occurrence compared to the normal discharge waveform can be emphasized. In addition, the scale adjustment unit 202 Figure 11 In the logarithmic transformation of , 10 is used as the base of the logarithm, but the present invention is not limited to this. For example, the scale adjustment unit 202 may use another base such as Napier's constant.

[0083] The data expansion unit 204 performs data augmentation. For example, the acquired data may have varying amounts of data depending on the cause of partial discharge. In such cases, the learning model generation unit 110 tends to prioritize factors with larger amounts of data when constructing the learning model. This can lead to reduced accuracy in determining less frequent phenomena.

[0084] Therefore, the data expansion unit 204 performs data augmentation processing on this data to suppress statistical bias in the learning data. This allows for statistically unbiased learning data to be generated when constructing the learning model, improving the accuracy of cause determination. Similarly, by performing augmentation processing on the analog electrical signals, the data expansion unit 204 can reduce the time and effort required to acquire data through testing and simulation. In this way, the data expansion unit 204 can even out the amount of learning data for each cause.

[0085] Specifically, the data expansion unit 204 may perform a data augmentation process, such as adding random noise or scaling the data to at least one of the electrical signal (analog electrical signal) generated by the electrical signal generation unit and the electrical signal obtained from a simulator or the like via the data acquisition unit. Furthermore, the data expansion unit 204 may also perform data augmentation processes that combine these data. For example, the data expansion unit 204 may perform an augmentation process using averaging (or addition) on at least one of the generated electrical signal and the obtained electrical signal.

[0086] use Figure 12 , the addition processing method using averaging performed by the data expansion unit 204 is described in more detail. Figure 12 This figure explains the increase processing method using averaging. Figure 12As shown, data expansion unit 204 obtains 50 data points from an electrical signal (which may also include an analog electrical signal) related to a single partial discharge cause and extracts eight data points. Data expansion unit 204 then averages the charge amounts at each time point in the eight extracted data points, generating an electrical signal with a charge amount different from the original data. This increases the number of data combinations for which the average values ​​are obtained (50×8).

[0087] In addition, Figure 12 The example of selecting 8 data combinations from 50 data is shown, but this is not limited to this. For example, the number of data used as the basis and the number of data to be extracted can be set arbitrarily. For example, n can be set so that the number of n combinations (mCn) extracted from m data is maximized, or the number of n can be set considering the computational cost. In addition, the addition process is sometimes referred to as the inflated process.

[0088] Furthermore, data can be further randomly extracted from the increased mCn data. This makes it possible to even out the amount of data for each partial discharge cause. Thus, when used as learning data, the time required for learning can be adjusted while maintaining a uniform amount of data.

[0089] Alternatively, a method for adding data directly obtained from the diagnostic power equipment can be used to add to the learning electrical signals. In particular, data can be generated by combining electrical signals obtained when the diagnostic power equipment is stopped. Random noise generated by the power equipment during shutdown is reduced, thus suppressing statistical bias in the learning data while generating noise-reduced learning data. This data addition process can be performed not only after conversion to time-series data but also after conversion to two-dimensional matrix data such as φ-qn characteristic diagrams, spectrograms, and quantity diagrams.

[0090] The threshold processing unit 206 sets the threshold. Furthermore, the threshold processing unit 206 can perform preliminary determination processing using the threshold. For example, the threshold processing unit 206 can determine the presence of a partial discharge signal based on a threshold for the maximum discharged charge amount. More specifically, the threshold processing unit 206 determines that a partial discharge signal is present if the number of times per second that the discharged charge amount exceeds the threshold exceeds a predetermined value. For example, if the threshold processing unit 206 determines that a partial discharge signal is present, it can also perform partial discharge cause classification using a learning model (discriminator) based on machine learning. In this way, the partial discharge determination unit 112 can also use the threshold processing unit 206 and the learning model to perform partial discharge determination, cause determination, and other functions.

[0091] Filtering unit 208 uses the information related to the partial discharge signal extracted by feature extraction unit 104 to generate characteristic ranges from the data used in the learning or determination process, as learning data or normal determination data. For example, filtering unit 208 can perform filtering processes such as bandpass filtering, windowing, and masking on the electrical signal. Furthermore, filtering unit 208 can also perform alignment processes to align reference points of the phase of the electrical signal.

[0092] Here, the feature extraction unit 104 will be described in detail. Figure 13 1 is a block diagram showing an example of the configuration of the feature extraction unit 104. Figure 13 As shown, the feature quantity extraction unit 104 includes a frequency range selection unit 302 and a phase range selection unit 304 .

[0093] The frequency range selection unit 302 extracts characteristics related to the frequency of the partial discharge signal based on, for example, the electrical signal (which may include an analog electrical signal) generated or acquired by the electrical signal generation unit 102. The frequency range selection unit 302 uses the electrical signal for each generation factor to extract characteristics related to the frequency of the partial discharge signal for each generation factor.

[0094] Figure 14 This diagram shows an example of a frequency spectrum generated by performing a fast Fourier transform on an electrical signal. The horizontal axis represents frequency, and the vertical axis represents spectrum. Figure 14 (a) is the simulated electrical signal with partial discharge, Figure 14 (b) is the electrical signal without partial discharge.

[0095] When the general Figure 14 When (a) and (b) are compared, there is a range in which the value of the spectrum of the analog electrical signal with partial discharge becomes larger than that of the analog electrical signal without partial discharge. The frequency range selection unit 302 generates, for example, a frequency range that exceeds a specified threshold as a feature. In addition, the frequency range selection unit 302 automatically sets the frequency range based on, for example, the size and frequency of the spectrum. Alternatively, the frequency range selection unit 302 may generate, as a feature, a frequency range determined by a person through the operating device 30 through confirmation. In addition, the frequency range may be one region or multiple regions. Moreover, the high-speed Fourier transform may be performed on an electrical signal of one cycle or on an electrical signal of multiple cycles. Alternatively, the high-speed Fourier transform may be performed on a region obtained by dividing the electrical signal into regions of arbitrary lengths.

[0096] Figure 15 This diagram shows an example of setting a frequency range based on the value of the spectrum of an electrical signal having multiple partial discharges. The horizontal axis represents frequency, and the vertical axis represents spectrum. Figure 15(a) to (c) are simulated electrical signals with partial discharge. Figure 15 As shown, the frequency range selection unit 302 can also select a characteristic frequency range for the partial discharge signal from multiple frequency spectra. In this case, the frequency range selection unit 302 selects the largest frequency range among the frequency ranges determined by each frequency spectra as the frequency range of the partial discharge signal. This can suppress the influence of single-shot noise during the process of acquiring electrical signals through testing.

[0097] In this case, the frequency range selection unit 302 may select representative data from the plurality of partial discharge-free electrical signals as the partial discharge-free electrical signal to be compared. Alternatively, the frequency range selection unit 302 may use data obtained by averaging the frequency spectra of the plurality of partial discharge-free electrical signals as the partial discharge-free electrical signal to be compared.

[0098] Figure 16 This figure shows an example of how the frequency range selection unit 302 selects a frequency range using a two-dimensional histogram. The horizontal axis is the frequency, and the vertical axis is the spectrum. The numbers in each small divided area are the frequency of occurrence of the spectrum that meets the conditions in the divided area. That is, Figure 16 Is for Figure 15 A two-dimensional histogram is generated by accumulating the frequency and frequency of each data item and the number of occurrences of the frequency spectrum. The frequency range selection unit 302 uses this two-dimensional histogram to select a frequency range based on the size of the frequency spectrum and the number of occurrences. For example, the frequency range selection unit 302 selects the frequency of the divided area where the frequency spectrum exceeds the specified value and the occurrence frequency also exceeds the specified value as frequency range S20. On the other hand, the frequency range N20, although its occurrence frequency exceeds the specified value, does not exceed the specified value in the frequency spectrum, so the frequency range selection unit 302 does not select it.

[0099] The frequency range selection unit 302 may extract features related to the frequency of the partial discharge signal from the spectrum graph and the quantity graph, or may combine these. For example, when combining these, the frequency range selection unit 302 selects the largest frequency range among the frequency ranges determined in each two-dimensional data set as the frequency range of the partial discharge signal.

[0100] Figure 17 1 is a diagram showing an example in which the frequency range selection unit 302 selects a frequency range using a spectrum diagram. Figure 17 (a) is a graph showing an electric signal when partial discharge occurs, wherein the horizontal axis represents time and the vertical axis represents the electric signal. Figure 17 (b) is a spectrum diagram of the electrical signal with partial discharge. The horizontal axis represents time and the vertical axis represents frequency. Figure 17 In the figure, according to the relationship between the display concentrations, the lower the concentration, the stronger the spectrum intensity.

[0101] like Figure 17 As shown in (b), there is a frequency range where the spectral intensity of the electrical signal with partial discharge is greater than that of the electrical signal without partial discharge. The frequency range selection unit 302 can extract the frequency range exceeding a predetermined value as a characteristic value of the partial discharge signal. The characteristic value can also be determined based on multiple spectrograms. In this case, the frequency range selection unit 302 selects the largest frequency range among the frequency ranges determined based on each spectrogram as the frequency range of the partial discharge signal. Frequency-related characteristic values ​​are also extracted from the spectrogram.

[0102] In this case, the frequency range selection unit 302 may select representative data from multiple spectrum graphs without partial discharge as the spectrum graph without partial discharge for comparison. Alternatively, the frequency range selection unit 302 may use data obtained by averaging multiple spectrum graphs without partial discharge as the spectrum graph without partial discharge for comparison. Furthermore, while the above description is based on an electrical signal (which may include an analog electrical signal) generated or obtained by the electrical signal generation unit 102, this is not limiting and any electrical signal may be used for processing. Specifically, the data used by the frequency range selection unit 302 may include an analog electrical signal or an electrical signal obtained from the power equipment being diagnosed. Furthermore, the operator may pre-set the selected frequency range to any desired range.

[0103] The phase range selection unit 304 extracts features related to the phase of the partial discharge signal based on the electrical signal (which may include an analog electrical signal) generated or acquired by the electrical signal generation unit 102. The phase range selection unit 304 uses the electrical signal for each generation factor to extract features related to the phase range of the partial discharge signal for each generation factor.

[0104] Figure 18 This diagram illustrates a method for selecting a phase range from a φ-qn characteristic diagram. This diagram shows a φ-q characteristic diagram in which the number of times charge appears is accumulated over a full period for a plurality of electrical signals. Figure 18 (a), (b), and (c) are φ-q characteristic diagrams for different analog electrical signals. The horizontal axis is phase (time), and the vertical axis is charge.

[0105] like Figure 18As shown in (a), (b), and (c) of FIG, phase range selection unit 304 selects a phase range for generating a partial discharge signal based on the φ-q characteristic diagram for each partial discharge cause, and selects the largest phase range among the phase ranges for each partial discharge cause as the phase range for the partial discharge signal. Phase range selection unit 304 can automatically set the phase range based on the magnitude and number of occurrences of the partial discharge signal. Alternatively, phase range selection unit 304 can determine the phase range based on an operator input signal to operating device 30.

[0106] In addition, Figure 18 In the embodiment, the phase range selection unit 304 sets the phase ranges S30 and S32 to two regions. This is because it is known that partial discharge signals are generated near at least one of the ranges of 0° to 90° and 180° to 270° of the applied voltage phase. However, any number of regions may be set.

[0107] use Figure 19 The method of automatically setting the phase range of the partial discharge signal by the phase range selection unit 304 will be described in more detail. Figure 19 FIG. 1 is a diagram showing an example of generating a charge amount histogram with respect to the charge amount in the φ-q characteristic diagram. Figure 19 (a) is a φ-q characteristic diagram of an electric signal with a partial discharge signal. The horizontal axis represents the phase, and the vertical axis represents the charge amount. Figure 19 (b) is Figure 19 (a) shows a charge amount histogram of the φ-q characteristic diagram. The horizontal axis represents the frequency of charge generation, and the vertical axis represents the charge amount.

[0108] like Figure 19 As shown in (b), the phase range selection unit 304 sets the data outside the predetermined range as a partial discharge signal, centering on the charge amount with the maximum frequency in the charge amount histogram, and extracts the phase range where the partial discharge signal occurs as feature amounts S40 and S42.

[0109] Figure 20 This is a diagram illustrating a method of selecting a phase range based on the peak position of a data point in a φ-q characteristic diagram. Figure 20 (a), (b), and (c) are φ-q characteristic diagrams for different electrical signals. The horizontal axis is phase (time) and the vertical axis is charge. Figure 20 In (a), (b), and (c), the peak values ​​of the analog electrical signals are indicated by arrows.

[0110] like Figure 20As shown in (a), (b), and (c) of FIG. 1 , the phase range selection unit 304 can also select the phase range of the partial discharge signal based on the peak position of the analog electric signal in the φ-q characteristic diagram of each partial discharge cause. The peak position can be selected as the data point where the charge amount reaches a maximum (minimum), and can also be selected as described later. Figure 22 The peak position is selected based on the result of comprehensive data processing. In addition, the peak position can also be implemented on the signal data of one cycle. In this case, the phase range is set to a range with statistical deviation.

[0111] Figure 21 The arrows represent the φ-q characteristic diagram of pseudo partial discharge. The horizontal axis is the phase (time) and the vertical axis is the charge amount. As mentioned above, the phase range selection unit 304 uses Figure 19 When the phase ranges S34 and S36 of the partial discharge signal are set, for example, even if Figure 21 In this way, even when a pseudo partial discharge occurs over a wide range relative to the phase, it can be statistically distinguished from a partial discharge signal.

[0112] Figure 22 This is a diagram showing an example of comprehensive processing of data points on the φ-q characteristic diagram. The horizontal axis is phase (time) and the vertical axis is charge amount. The phase range selection unit 304 can also determine the peak position based on the following Figure 22 The peak position is selected based on the result of comprehensive data processing.

[0113] Furthermore, the phase range selection unit 304 can select a phase range based on the centroid position of the data points of the partial discharge signal as well as the peak position as a method of selecting a feature amount related to the phase of the partial discharge signal.

[0114] Furthermore, the phase range selection unit 304 may extract features related to the phase of the partial discharge signal from the spectrum graph and the quantity graph, or may combine these. For example, when combining these, the phase range selection unit 304 selects the largest range among the phase ranges determined in each two-dimensional data set as the phase range of the partial discharge signal.

[0115] Figure 23 1 is a diagram showing an example in which the phase range selection unit 304 selects a phase range using a spectrum diagram. Figure 23 (a) is a graph showing a simulated electrical signal with partial discharge, where the horizontal axis represents time and the vertical axis represents the electrical signal. Figure 23 (b) is a spectrum diagram of the simulated electrical signal with partial discharge. The horizontal axis represents time and the vertical axis represents frequency. Figure 23 In the figure, according to the relationship between the display concentrations, the thinner the concentration, the stronger the spectrum intensity.

[0116] like Figure 23 As shown in (b), there is a phase range where the spectral intensity of the electrical signal with partial discharge is greater than that of the electrical signal without partial discharge. Phase range selection unit 304 can extract phase ranges S38 and S40 where the spectral intensity exceeds a predetermined value as the characteristic quantity of the partial discharge signal. It is also possible to determine the characteristic quantity based on multiple spectrograms. In this case, frequency range selection unit 302 selects the largest range among the frequency ranges determined based on each spectrogram as the frequency range of the partial discharge signal. Frequency-related characteristic quantities are similarly extracted from the spectrogram.

[0117] In this case, the frequency range selection unit 302 may select representative data from multiple spectrograms without partial discharge as the spectrogram without partial discharge for comparison. Alternatively, the frequency range selection unit 302 may use data obtained by averaging multiple spectrograms without partial discharge as the spectrogram without partial discharge for comparison. Furthermore, while the data used by the phase range selection unit 304 is described based on the electrical signal (which may include an analog electrical signal) generated or obtained by the electrical signal generation unit 102, this is not limiting and any electrical signal may be used for processing. Specifically, the data used by the frequency range selection unit 302 may include an analog electrical signal or an electrical signal obtained from the power equipment being diagnosed. Furthermore, the operator may pre-set the phase range to any desired range. In this manner, the feature extraction unit 104 can select the frequency range and phase range in which the characteristics of the partial discharge signal appear.

[0118] [Processing example of partial discharge diagnosis system]

[0119] Reference Figures 25 to 29 ,use Figure 24 An example of processing performed by the partial discharge diagnostic system 1 will be described. Figure 24 This is a flowchart illustrating an example of processing performed by the partial discharge diagnostic system 1. Here, the diagnosis is based on determining the cause of a partial discharge signal generated in a turbine generator coil. The electrical signal used for diagnosis is charge data measured by a sensor. Furthermore, the electrical signals used for diagnosis include intermittent data frequently acquired from power equipment and periodically measured intermittent data. While charge data is used as an example of an electrical signal, this is not limiting; signals based on phase fluctuations can also be used. For example, signals other than charge data, such as electrical signals or voltage signals, are also acceptable.

[0120] [Learning model generation process]

[0121] First, the electrical signal generator 102 generates electrical signals for use in machine learning. The electrical signal generator 102 simulates the insulation degradation pattern of the generator coil's insulation material, generating electrical signals for each cause of partial discharge (step S100). The patterns that may cause partial discharge are previously acquired from field equipment. For example, the electrical signal generator 102 simulates seven patterns, A through G, as possible causes of partial discharge, generating electrical signals for each cause of partial discharge. Furthermore, one of the seven patterns, A through G, generates an electrical signal indicating no partial discharge.

[0122] Next, the calculation unit 200 of the data processing unit 106 generates a frequency spectrum corresponding to each cycle based on the generated electrical signal (charge amount signal). The frequency range selection unit 302 of the feature value extraction unit 104 then selects a frequency range for each partial discharge cause based on the frequency spectrum corresponding to each cycle (step S102a), and the phase range selection unit 304 selects a phase range (step S102b).

[0123] Next, the data expansion unit 204 of the data processing unit 106 performs an addition process based on the averaged data on the generated charge amount signal (step S104). In this case, the processing is performed so that the generation frequencies of the seven modes that cause partial discharge are uniform. Figure 12 As shown in FIG, the measured time series data may be averaged. In this way, since the electrical signal without partial discharge is included, it is possible to make a comprehensive judgment on the data without partial discharge signal.

[0124] Next, the filter unit 208 of the data processing unit 106 performs bandpass filtering on each of the charge amount signals after the increase processing based on the selected frequency range and phase range (step S106 a ).

[0125] Next, the scale adjustment unit 202 of the data processing unit 106 normalizes the charge amount of the charge amount signal after the bandpass filtering process. In this case, the normalization is performed to the value at which the amplitude reaches the maximum value in one cycle of the applied voltage (step S108).

[0126] Next, the calculation unit 200 of the data processing unit 106 generates a φ-qn characteristic diagram based on the normalized charge amount signal (step S110 ).

[0127] Next, the scale adjustment unit 202 of the data processing unit 106 performs logarithmic transformation as a scaling transformation for frequency on the φ-qn characteristic diagram of the charge amount signal (step S112 ).

[0128] Next, the image generation unit 108 images the φ-qn characteristic map of the charge amount signal into a two-dimensional grayscale image, and generates a φ-qn characteristic map image (step S114 a ). Figure 25 1 is a diagram showing an example of imaging processing performed by the image generating unit 108 . Figure 25 (a) is a diagram showing an example of a φ-qn characteristic diagram of a charge amount signal. Figure 25 (b) is a diagram showing an example of a two-dimensional grayscale image. The display control unit 116 displays the two-dimensional grayscale image of the φ-qn characteristic diagram on the display device 20. This allows the operator to visually observe the characteristics of the φ-qn characteristic diagram.

[0129] Next, the filter unit 208 of the data processing unit 106 masks phase regions outside the characteristic phase range of the partial discharge signal based on the selected phase range for the φ-qn characteristic diagram image of the charge amount signal (step S114b).

[0130] Figure 26 1 and 2 are diagrams showing an example of masking processing performed by the filter unit 208 . Figure 26 (a) is a diagram showing an example of imaging as a two-dimensional grayscale image. Figure 26 (b) is a diagram showing an example of masking a two-dimensional grayscale image. Figure 26 In the embodiment, data is deleted by masking, but the present invention is not limited thereto. For example, the filter unit 208 performs masking in a manner that further reduces the data value of the phase region outside the phase range. In other words, the filter unit 208 suppresses the data value of the phase region outside the phase range.

[0131] Next, the learning model generation unit 110 generates a learning model using the φ-qn characteristic map image of the masked charge quantity signal as learning data (step S116). Here, one of the seven patterns A through G that contribute to partial discharge is associated with the φ-qn characteristic map image as a teaching signal. Thus, the learning model outputs one of the seven patterns A through G that contribute to partial discharge, given a two-dimensional array of input data equivalent to the φ-qn characteristic map image.

[0132] [Judgment process]

[0133] Next, a description will be given of a process of generating a φ-qn characteristic map image and performing determination, similarly to the learning model generation process, for a charge amount signal acquired from a hydro-turbine generator coil to be diagnosed.

[0134] The data acquisition unit 100 acquires a time-series electric signal (charge amount signal) from the measuring instrument 10 disposed in the turbine generator coil to be measured (step S118 ).

[0135] Figure 27 This diagram schematically illustrates an example arrangement of coils and sensors in a hydraulic turbine generator. Sensors 402 are placed around each of the U-phase, V-phase, and W-phase conductors 400. Sensors 402 include partial discharge sensors that measure charge data. By placing these sensors around the lead-out conductors of the hydraulic turbine generator, charge data can be easily acquired even in existing plants.

[0136] like Figure 27 As shown in FIG, sensors are sometimes provided adjacent to the lead conductors of the U phase, V phase, and W phase. Figure 3 As shown in the φ-qn characteristic diagram, heterogeneous phase overlap occurs. Heterogeneous phase overlap means that the partial discharge signal propagating to the lead conductor reaches three sensors, making it possible to detect signals other than the target phase.

[0137] The filter unit 208 of the data processing unit 106 selects the base point of the applied voltage of each phase and changes the phase information of the charge amount signal obtained from the turbine generator coil. At this time, the filter unit 208 changes the phase information so that the phases of the charge amount signals are aligned.

[0138] Figure 28 2 is a diagram showing an example of base point adjustment processing of applied voltage performed by the filter unit 208. The horizontal axis represents time, and the vertical axis represents the amount of charge. Figure 28 (a), (b), and (c) are charge amount signals obtained from sensors surrounding the lead conductors of the U phase, V phase, and W phase, respectively. Figure 28 (d), (e), and (f) are charge amount signals with the selected base point and changed phase information.

[0139] Next, the filter unit 208 of the data processing unit 106 performs bandpass filtering on each charge amount signal after the base point adjustment process based on the selected frequency range and phase range (step S106 b ).

[0140] Next, the scale adjustment unit 202 of the data processing unit 106 performs charge normalization on the charge signal after the bandpass filtering process. In this case, normalization is performed based on the value at which the amplitude in one cycle of the applied voltage reaches the maximum (step S120).

[0141] Next, the calculation unit 200 of the data processing unit 106 generates a φ-qn characteristic diagram based on the normalized charge amount signal (step S122 ).

[0142] Next, the scale adjustment unit 202 of the data processing unit 106 performs logarithmic transformation as a scaling transformation for frequency on the φ-qn characteristic diagram of the charge amount signal (step S1124 ).

[0143] Next, the image generation unit 108 images the φ-qn characteristic map of the charge amount signal after the base point adjustment process into a two-dimensional grayscale image, and generates a φ-qn characteristic map image (step S114 c ).

[0144] Next, the filter unit 208 of the data processing unit 106 masks phase regions outside the characteristic phase range of the partial discharge signal based on the selected phase range for the φ-qn characteristic diagram image of the charge amount signal after the base point adjustment process (step S114 d ).

[0145] Figure 29 1 and 2 are diagrams showing an example of masking processing performed by the filter unit 208 . Figure 29 (a) is a diagram showing an example of imaging the φ-qn characteristic diagram of the charge amount signal after the base point adjustment process as a two-dimensional grayscale image. Figure 29 (b) is a diagram showing an example of masking a two-dimensional grayscale image. Figure 29 In the example, data is deleted by masking, but the present invention is not limited to this. For example, filter unit 208 may perform masking to retain partial discharge signals S44 and S48 within the phase range serving as the characteristic quantity, while further reducing the data values ​​of charge quantity signals S42 and S46 in phase regions outside the phase range. In other words, filter unit 208 suppresses the data values ​​in phase regions outside the phase range.

[0146] Next, the partial discharge determination unit 112 uses the learned model to perform a discrimination process using the φ-qn characteristic map image of the masked charge amount signal as discrimination data (step S116). The partial discharge determination unit 112 outputs one of the seven patterns A through G that indicate the cause of partial discharge as a discrimination result. In this way, the partial discharge determination unit 112 determines the presence or cause of partial discharge based on input data in a two-dimensional array equivalent to the φ-qn characteristic map image.

[0147] According to the embodiments described above, data in a predetermined format is generated after suppressing false partial discharge signals in an electrical signal that varies depending on the phase. Based on this data, a learning model is generated that determines at least one of the cause of partial discharge and the presence or absence of partial discharge. Generating a machine-learned learning model based on data that suppresses false partial discharge signals can improve the accuracy of determining the cause of partial discharge in power equipment.

[0148] While the embodiments of the present invention have been described, these are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other ways, and various omissions, substitutions, changes, and combinations can be made without departing from the gist of the invention. These embodiments and their variations are included within the scope and gist of the invention, and are included in the invention described in the claims and their equivalents.

Claims

1. A partial discharge diagnostic device capable of determining the cause of partial discharge in an insulating material, the partial discharge diagnostic device comprising: a processing unit that generates data in a predetermined format after suppressing a false partial discharge signal in the electrical signal that varies according to the phase; and A learning model generating unit generates a learning model for determining at least one of a cause of the partial discharge and the presence or absence of the partial discharge based on the data.

2. The partial discharge diagnostic device according to claim 1, wherein: The processing unit generates the data in which the electrical signal in a predetermined phase range is suppressed as the false partial discharge signal.

3. The partial discharge diagnostic device according to claim 2, wherein: The processing unit generates the data in which the electrical signal having an absolute value greater than or equal to a predetermined value is suppressed as the false partial discharge signal.

4. The partial discharge diagnostic device according to claim 1, wherein: The processing unit generates the data in which the electrical signal within a predetermined frequency range of the electrical signal is suppressed as the false partial discharge signal.

5. The partial discharge diagnostic device according to claim 1, wherein: The electrical signal is a plurality of signals having phase differences, a timed partial discharge signal generated for one of the plurality of signals corresponding to the pseudo partial discharge signal generated for a signal different from the one of the plurality of signals; The processing unit suppresses, for the different signals, the electrical signal in a phase range based on the timing at which the partial discharge signal is generated with respect to the one signal as the false partial discharge signal.

6. The partial discharge diagnostic device according to claim 1, wherein: The electrical signal is a plurality of signals having phase differences, The processing unit changes the phases of the electrical signals so that the phases are aligned.

7. The partial discharge diagnostic device according to claim 1, wherein: The processing unit generates the data as image data having numerical values ​​arranged two-dimensionally, The partial discharge diagnostic device further includes a display control unit configured to cause a display device to display the image data.

8. The partial discharge diagnostic device according to any one of claims 1 to 7, wherein: The device further includes a data acquisition unit that acquires the electrical signal measured by the power equipment or a sensor installed around the power equipment.

9. The partial discharge diagnostic device according to claim 8, wherein: The electric power equipment is at least one of a generator, a motor, an inverter device, a switch mechanism, and a cable, The electric signal is a signal indicating at least one of an amount of charge, a current, and a voltage corresponding to a phase of a voltage applied to the electric device.

10. The partial discharge diagnostic device according to claim 9, wherein: The device further includes an electric signal generating unit configured to generate at least one of a signal for learning and a signal for acquiring the signal via the data acquiring unit.

11. The partial discharge diagnostic device according to claim 10, wherein: The processing unit generates the data based on at least one of the measured electrical signal and the electrical signal for learning.

12. The partial discharge diagnostic device according to claim 11, wherein: The electric signal generating unit generates the electric signal using at least one of test data obtained by simulating an insulation degradation state or a simulation result.

13. The partial discharge diagnostic device according to claim 12, wherein: The device further includes a learning data expansion unit configured to increase the amount of the data by combining at least one of the measured electrical signal and the learning electrical signal.

14. The partial discharge diagnostic device according to claim 13, wherein: further comprising a feature quantity extraction unit that extracts a feature quantity indicating a range of the partial discharge signal based on the electrical signal generated by the electrical signal generation unit when insulation degradation occurs, The processing unit suppresses the false partial discharge signal in the electrical signal based on the feature amount.

15. The partial discharge diagnostic device according to claim 14, wherein: The processing unit is capable of adjusting the signal strength range of the electrical signal.

16. The partial discharge diagnostic device according to claim 15, wherein: The electrical signal is a signal indicating the amount of charge corresponding to the phase of the voltage applied to the electrical device. The processing unit divides the phase range and the charge amount range into a plurality of sections, and generates the charge amount generation frequency as the data for each area represented by the section of the phase range and the section of the charge amount range.

17. The partial discharge diagnostic device according to claim 16, wherein: The processing unit converts the occurrence frequency into a nonlinear form.

18. The partial discharge diagnostic device according to any one of claims 10 to 17, wherein: The processing unit generates the data based on the electrical signal acquired by the data acquisition unit during determination. The partial discharge diagnostic device further includes a partial discharge determination unit that determines at least one of the presence or absence of the partial discharge and the cause of occurrence based on the data using the learning model.

19. A partial discharge diagnostic method for determining the cause of a partial discharge signal of an insulating material, the partial discharge diagnostic method comprising: a processing step of generating data in a predetermined format after suppressing a false partial discharge signal in the electrical signal that varies according to the phase; and The learning model generating step generates a learning model for determining at least one of the presence or absence of the partial discharge and the cause of the occurrence based on the data.

20. A partial discharge diagnostic system capable of determining the cause of partial discharge in an insulating material, the partial discharge diagnostic system comprising: Measuring instruments that measure electrical signals from electrical equipment or sensors installed around it; and Partial discharge diagnostic device, The partial discharge diagnostic device comprises: a processing unit that generates data in a predetermined format after suppressing a false partial discharge signal in the electrical signal that varies according to the phase; and A learning model generating unit generates a learning model for determining at least one of a cause of the partial discharge and the presence or absence of the partial discharge based on the data.

Images