JP2023122339A - Power equipment monitoring device, power equipment monitoring method, and power equipment monitoring program - Google Patents

Abstract

To eliminate noise even when the environment changes.SOLUTION: According to an embodiment, the power equipment monitoring device includes a sensor and a noise removal unit. The sensor is attached to power equipment and measures a signal related to partial discharge. The noise removal unit generates a denoising signal by inputting the measured signal into a denoising model. A denoising model is a model learned so that a signal including noise is used as an input and a denoising signal obtained by removing noise is used as an output. The noise removal unit generates the denoising signal using a first denoising model and a second denoising model. The first denoising model is a denoising model learned using a learning data set generated using a signal measured by the sensor. The second denoising model is a denoising model learned using a learning data set generated using signals measured from other power equipment.SELECTED DRAWING: Figure 1

Description

Embodiments of the present invention relate to a power device monitoring apparatus, a power device monitoring method, and a power device monitoring program.

In electric power equipment, the insulation performance of the insulator on the surface of the electric power equipment or inside the electric power equipment deteriorates due to secular change. When the insulation performance deteriorates, partial discharge occurs at the deteriorated portion. If the insulation performance further deteriorates, dielectric breakdown occurs. Therefore, it is required to determine the presence or absence of partial discharge in maintenance work of electric power equipment.

The partial discharge waveform can be obtained by measuring in advance in a laboratory or the like. On the other hand, since noise exists in the environment in which power equipment is installed, it is required to remove the noise from the measured signal in order to determine partial discharge. The noise may change as the environment in which the power equipment is installed changes. In this case, there is a possibility that the noise cannot be properly removed after the environmental change, resulting in an erroneous determination of partial discharge.

Japanese Patent Application Laid-Open No. 2020-012726

A problem to be solved by the present invention is to provide a power device monitoring apparatus, a power device monitoring method, and a power device monitoring program capable of removing noise even when the environment changes.

A power equipment monitoring apparatus according to an embodiment has a sensor and a noise elimination unit. The sensor is attached to the power equipment and measures a signal related to partial discharge. The noise removing unit generates a denoising signal by inputting the measured signal to the denoising model. A denoising model is a model trained to receive a signal containing noise as an input and output a denoising signal from which the noise has been removed. The noise removal unit generates the denoising signal using the first denoising model and the second denoising model. The first denoising model is a denoising model trained using a training data set generated using signals measured by the sensors. The second denoising model is a denoising model learned using a learning data set generated using signals measured from other power equipment.

Schematic diagram showing the configuration of a detection system according to the first embodiment. 1 is a schematic block diagram showing the configuration of a power equipment monitoring device according to a first embodiment; FIG. 4 is a flowchart showing the operation of the power equipment monitoring apparatus according to the first embodiment; FIG. 5 is a diagram showing a recording example of determination results according to the first embodiment; The schematic block diagram which shows the structure of the electric power equipment monitoring apparatus which concerns on 2nd Embodiment. 9 is a flowchart showing the operation of the power equipment monitoring apparatus according to the second embodiment; FIG. 11 is a diagram showing a recording example of determination results according to the second embodiment; The schematic block diagram showing the structure of the electric power equipment monitoring apparatus which concerns on 3rd Embodiment. 9 is a flow chart showing the operation of the power equipment monitoring apparatus according to the third embodiment; FIG. 11 is a diagram showing a recording example of determination results according to the third embodiment; 10 is a flow chart showing the operation of the power equipment monitoring apparatus according to the fourth embodiment; The schematic block diagram which shows the structure of the electric power equipment monitoring apparatus which concerns on 5th Embodiment. 14 is a flowchart showing model re-learning processing by the power equipment monitoring apparatus according to the fifth embodiment; The schematic block diagram showing the structure of the electric power equipment monitoring apparatus which concerns on 6th Embodiment. An example of a result display screen by the power equipment monitoring device of the sixth embodiment. The figure which shows an example of the noise source information which concerns on 7th Embodiment. An example of a result display screen by the power equipment monitoring device of the seventh embodiment.

Hereinafter, a power device monitoring apparatus, a power device monitoring method, and a power device monitoring program according to embodiments will be described with reference to the drawings.

(First embodiment)
FIG. 1 is a schematic diagram showing the configuration of a detection system 1 according to the first embodiment.
A detection system 1 detects partial discharge of power equipment M installed at a plurality of sites S such as substations. Examples of the power equipment M include circuit breakers, disconnectors, current transformers, transformers, gas-insulated switches, generators, motors, reactors, and the like. The power equipment M supplies a high voltage and a large current from the outside through a power cable. The electric power equipment M has a function of interrupting power supply in the event of an abnormality. A ground electrode is connected to the lower portion of the housing that accommodates the power equipment M. As shown in FIG.

The detection system 1 includes power equipment monitoring devices 10 installed at a plurality of sites S. FIG. A plurality of power equipment monitoring apparatuses 10 are connected to each other via a wide area network N such as the Internet.

The power equipment monitoring apparatus 10 includes a sensor 11 and an arithmetic unit 12 installed for each power equipment M. As shown in FIG. The power equipment monitoring apparatus 10 is provided near the power equipment M. As shown in FIG. In the power equipment monitoring apparatus 10, the sensor 11 measures waves generated by the operation of the power equipment M, such as electromagnetic waves, sound waves, transient ground currents, and wall currents derived from electromagnetic waves. Examples of the sensor 11 include a CT (Current Transformer) sensor, a TEV (Transient Earth Voltage) sensor, an AE (Acoustic Emission) sensor, and the like. As the sensor 11, one type of sensor may be used, or a combination of multiple types may be used.

Arithmetic device 12 detects partial discharge generated from power equipment M based on the measurement data of sensor 11 . The measurement data represents the wave waveform. The arithmetic device 12 performs preprocessing of the measurement data, noise removal, and determination of the presence or absence of partial discharge according to predetermined parameters. Parameters include, for example, amplification gain of measurement data, detection threshold, matching threshold, phase resolution, bandpass frequency band, and the like. For example, the arithmetic unit 12 amplifies the measurement data according to the amplification gain and samples the measurement data according to the phase resolution and the bandpass frequency band. The arithmetic device 12 removes noise from the preprocessed measurement data and determines the presence or absence of partial discharge based on the detection threshold and the matching threshold.

FIG. 2 is a schematic block diagram showing the configuration of the power equipment monitoring apparatus 10 according to the first embodiment.
The computing device 12 is an information processing device such as a PC, a smart phone, or a tablet computer. The computing device 12 determines whether or not partial discharge occurs based on the electrical signal output from the sensor 11 . Arithmetic device 12 includes processor 121 , main memory 122 , storage 123 , and interface 124 .

Examples of the processor 121 include a CPU (Central Processing Unit), a GPU (Graphic Processing Unit), a microprocessor, and the like. Note that in other embodiments, the arithmetic unit 12 may include a custom LSI (Large Scale Integrated Circuit) such as a PLD (Programmable Logic Device) in addition to or instead of the above configuration. Examples of PLD include PAL (Programmable Array Logic), GAL (Generic Array Logic), CPLD (Complex Programmable Logic Device), and FPGA (Field Programmable Gate Array). In this case, part or all of the later-described functions implemented by the processor 121 may be implemented by the integrated circuit. Such an integrated circuit is also included as an example of a processor.

Examples of the storage 123 include magnetic disks, magneto-optical disks, optical disks, semiconductor memories, and the like. The storage 123 may be internal media directly connected to the bus of the computing device 12, or may be external media connected to the computing device 12 via the interface 124 or communication line. A denoising model for each of the plurality of sites S is recorded in the storage 123 .

A denoising model is a trained model whose parameters have been learned so that a signal containing noise is input and a signal from which noise is removed (hereinafter referred to as a denoising signal) is output. A denoising model is composed of, for example, a neural network model or a machine learning model such as a decision tree model. The denoising model is learned using a learning data set generated based on the signal acquired from the sensor 11 during a predetermined learning period immediately after the start of operation of the electric power device M. It is assumed that the learning period is sufficiently shorter than the period during which partial discharge due to deterioration of the power equipment M begins to occur. That is, the signal acquired from the sensor 11 during the learning period has a low probability of including a signal related to partial discharge.
For the learning data set, for example, a set in which a signal obtained from the sensor 11 plus a known partial discharge signal is used as an input sample and the partial discharge signal is used as an output sample, or a set in which the signal obtained from the sensor 11 is used as an output sample. A set of input samples and a signal of zero intensity as output samples is included. By using such a training data set, the parameters of the denoising model are learned so that when a signal containing noise is input, a denoising signal from which noise is removed is output. Each sample of the learning data set may be, for example, data representing a waveform of an electrical signal, or data representing an electrical signal in a ΦQN pattern.
The storage 123 stores denoising models related to other sites S in addition to the denoising models related to the site S where the power equipment M to be detected by the power equipment monitoring apparatus 10 is installed. The denoising model related to the other site S is received from the power equipment monitoring device 10 of the other site S. FIG.

A plurality of sensors 11 , a communication device 125 , an input device 126 and a display device 127 are connected to the interface 124 . The communication device 125 communicates with an external communication device via the wide area network N. FIG. The input device 126 receives input of operations from the user. Examples of the input device 126 include a touch panel, mouse, keyboard, and the like. The display device 127 displays the calculation result of the calculation device 12 to the user. Examples of the display device 127 include a CRT (Cathode Ray Tube) display, a liquid crystal display, an organic EL (Electro Luminescence) display, and the like.

When the processor 121 executes the power equipment monitoring program stored in the storage 123 , the processor 121 functions as a signal acquisition section 131 , a noise removal section 132 and a partial discharge determination section 133 . The signal acquisition unit 131 acquires signals measured by the sensor 11 . The noise removing unit 132 obtains a denoising signal by inputting the signal acquired by the signal acquiring unit 131 into the denoising model stored in the storage 123 . The partial discharge determination unit 133 analyzes the denoising signal and determines the presence or absence of partial discharge. For example, the partial discharge determination unit 133 determines that there is a partial discharge when the denoising signal includes a signal component having an intensity equal to or greater than a predetermined threshold. The partial discharge determination unit 133 may record the partial discharge determination result in the storage 123 .

FIG. 3 is a flow chart showing the operation of the power equipment monitoring apparatus 10 according to the first embodiment. The power equipment monitoring apparatus 10 executes the following partial discharge determination process at predetermined intervals (for example, one hour).

The signal acquisition unit 131 acquires a signal measured by the sensor 11 (step S1). Next, the power equipment monitoring apparatus 10 selects one of the plurality of denoising models stored in the storage 123 (step S2), and executes the following steps S3 and S4 for the selected denoising model.

The noise removing unit 132 obtains a denoising signal by inputting the signal acquired in step S1 to the denoising model selected in step S2 (step S3). The partial discharge determination unit 133 analyzes the denoise signal and determines the presence or absence of partial discharge (step S4).

After determining the presence or absence of partial discharge using each denoising model, the partial discharge determination unit 133 records the determination result in the storage 123 (step S5). FIG. 4 is a diagram showing a recording example of determination results according to the first embodiment. As shown in FIG. 4, the storage 123 stores the determination results of the presence or absence of partial discharge using the denoising models of a plurality of sites S for the signals measured at the same day and at the same time at one site S (sensor 11). Recorded.

As described above, the electric power equipment monitoring apparatus 10 according to the first embodiment can be expected to improve the accuracy of determining the presence or absence of partial discharge by removing noise using the denoising models related to a plurality of sites S. For example, when only the denoising model generated by noise data of a certain site S is used, there is a possibility that correct determination cannot be made if the environmental noise of the site S changes. At this time, if the cause of the environmental noise change is due to the additional installation of noise source equipment (for example, inverters, relays, etc.), if equivalent equipment has already been installed at another site S, the equipment are considered to be similar to each other, there is a possibility that the noise can be removed correctly by using the denoising model of the other site S.

Although the solution of a trained model generated by machine learning is not always correct, it is expected that the judgment accuracy will be improved by making a judgment (ensemble) using a plurality of models.
It is conceivable to construct a single denoising model using noise signals from a plurality of sites S, but this poses the following problems.
- In order to learn a denoising model from noise signals of a plurality of sites S, it is necessary to transmit the noise signals through the wide area network N, which imposes a communication load.
・It is necessary to process a large amount of noise signals with a large data size, and the amount of calculation is large.
• Every time a site S is added, the common denoising model for all sites S needs to be updated.
- Depending on the content of the contract with the site S, it may be necessary to rebuild the denoising model from which the noise signal of the site S with which the contract has been terminated is removed.
Therefore, by sharing denoising models for a plurality of sites S like the power equipment monitoring apparatus 10 according to the first embodiment, highly accurate denoising signals can be obtained with a small load.

(Second embodiment)
FIG. 5 is a schematic block diagram showing the configuration of the power equipment monitoring apparatus 10 according to the second embodiment. The power equipment monitoring apparatus 10 according to the second embodiment further includes a comprehensive determination unit 134 as a function of the processor 121 in addition to the configuration of the first embodiment. The comprehensive determination unit 134 makes a comprehensive determination of the presence or absence of partial discharge based on the partial discharge determination results for each of the plurality of denoising models recorded in the storage 123 . For example, the comprehensive determination unit 134 determines that there is partial discharge when it is determined that there is partial discharge in all of the plurality of determination results, and determines that there is no partial discharge when it is determined that there is no partial discharge even in one of the determination results. judge. This is because when the input signal is only noise, if there is even one denoising model that has learned the noise, the signal is removed from the data after noise removal using that denoising model, and the partial discharge judgment unit This is because the result of 133 can be expected to be determined as no partial discharge.

FIG. 6 is a flow chart showing the operation of the power equipment monitoring apparatus 10 according to the second embodiment. The power equipment monitoring apparatus 10 executes the following partial discharge determination process at predetermined intervals (for example, one hour). Note that the processing from step S1 to step S4 is the same as in the first embodiment. When determination results are obtained for all denoising models, the comprehensive determination unit 134 performs comprehensive determination based on the recorded determination results (step S11), and records the results in the storage 123 (step S12). FIG. 7 is a diagram showing a recording example of determination results according to the second embodiment. As shown in FIG. 7, the storage 123 stores the determination result of the presence or absence of partial discharge using the denoising model of a plurality of sites S for signals measured at the same day and at the same time at one site S (sensor 11). Also, the comprehensive determination result of the partial discharge at the site S at the same time on the same day is recorded.

According to the second embodiment, the power equipment monitoring apparatus 10 can attach a comprehensive determination result to the determination result of the presence or absence of partial discharge for each denoising signal. Thereby, the user can easily recognize whether or not the site S has a partial discharge.

(Third Embodiment)
FIG. 8 is a schematic block diagram showing the configuration of the power equipment monitoring apparatus 10 according to the third embodiment. The power equipment monitoring apparatus 10 according to the third embodiment further includes an unlearned determination unit 135 as a function of the processor 121 in addition to the configuration of the second embodiment. The unlearned determination unit 135 determines whether or not the signal to be determined is unlearned for the denoising model of the site S. Specifically, the unlearned determining unit 135 determines that there is partial discharge in the denoising signal generated from the denoising model related to the site S, and that there is no partial discharge in the comprehensive determination. If so, it is determined that the signal has not been learned.

FIG. 9 is a flow chart showing the operation of the power equipment monitoring apparatus 10 according to the third embodiment. The power equipment monitoring apparatus 10 executes the following partial discharge determination process at predetermined intervals (for example, one hour). Note that the processing from step S1 to step S11 is the same as in the second embodiment. When the comprehensive determination unit 134 performs comprehensive determination in step S11, the unlearned determination unit 135 determines that the partial discharge determination result for the denoising signal generated from the denoising model related to the site S matches the comprehensive determination result. It is determined whether or not to do so (step S21). If the determination result of the denoising signal for the site S and the overall determination result match (step S21: YES), it is determined that the noise included in the measured signal has been learned (step S22). On the other hand, if the determination result for the denoising signal related to the site S does not match the overall determination result (step S21: NO), it is determined that the noise included in the measured signal has not been learned (step S23). . Then, the unlearned determination unit 135 records the determination result in the storage 123 (step S24).

FIG. 10 is a diagram showing a recording example of determination results according to the third embodiment. As shown in FIG. 10, the storage 123 stores the determination result of the presence or absence of partial discharge using the denoising model of a plurality of sites S for signals measured at the same day and time at one site S (sensor 11), A comprehensive determination result and whether or not the signal has been learned are recorded.

Thus, according to the third embodiment, the storage 123 accumulates the determination result as to whether or not the noise generated at the site S is unlearned. The user can recognize the increasing trend of unlearned noise in the judgment result accumulated in the storage 123 as to whether or not it is unlearned. If the unlearned noise is increasing, the user can determine that the denoising model of the site S is no longer able to remove the actual noise. As a result, the user can expect to maintain and recover the noise removal accuracy of the denoising model by recollecting the noise of the site S and relearning the denoising model.

(Fourth embodiment)
The power equipment monitoring apparatus 10 according to the third embodiment records in the storage 123 whether the noise is unlearned or not. On the other hand, the power equipment monitoring apparatus 10 according to the fourth embodiment records in the storage 123 a signal determined to contain unlearned noise. The configuration of the power equipment monitoring apparatus 10 according to the fourth embodiment is the same as that of the third embodiment.

FIG. 11 is a flow chart showing the operation of the power equipment monitoring apparatus 10 according to the fourth embodiment. The power equipment monitoring apparatus 10 executes the following partial discharge determination process at predetermined intervals (for example, one hour). Note that the processing from step S1 to step S23 is the same as in the third embodiment. When the unlearning determining unit 135 determines that the learning is unlearned in step S23, it records the signal acquired in step S1 in the storage 123 (step S31). Then, in step S<b>24 , the unlearned determination unit 135 records the determination result in the storage 123 .

Thus, according to the fourth embodiment, the power equipment monitoring apparatus 10 records unlearned noise in the denoising model of the site S in the storage 123 . As a result, when re-learning the denoising model, the power equipment monitoring apparatus 10 can efficiently re-learn the denoising model by using the unlearned noise recorded in the storage 123 .

(Fifth embodiment)
FIG. 12 is a schematic block diagram showing the configuration of the power equipment monitoring apparatus 10 according to the fifth embodiment. The power equipment monitoring apparatus 10 according to the fifth embodiment further includes a learning unit 136 as a function of the processor 121 in addition to the configuration according to the fourth embodiment. The learning unit 136 re-learns the denoising model using the unlearned noise recorded in the storage 123 .

A power device monitoring apparatus 10 according to the fifth embodiment executes the partial discharge determination process shown in FIG. 11 at predetermined intervals (for example, one hour), as in the fourth embodiment. As a result, unlearned noise is accumulated in the storage 123 .
FIG. 13 is a flowchart showing model relearning processing by the power equipment monitoring apparatus 10 according to the fifth embodiment. The electric power equipment monitoring apparatus 10 according to the fifth embodiment performs model relearning processing in parallel with the partial discharge determination processing shown in FIG. 11 .

The learning unit 136 determines whether or not a denoising model update condition is satisfied (step S41). The denoising model update condition may be, for example, that the number of samples of unlearned noise recorded in the storage 123 has reached a certain amount, or that a certain period of time or more has elapsed since the previous denoising model re-learning. It may be that If the update condition is not satisfied (step S41: NO), the learning unit 136 does not re-learn the denoising model.

On the other hand, if the update condition is satisfied (step S41: YES), the learning unit 136 generates a learning data set using the unlearned noise recorded in the storage 123 (step S42). The learning unit 136 uses, as a learning data set, for example, a set of unlearned noise plus a known partial discharge signal as an input sample, a set of the unlearned noise as an output sample, or an unlearned noise as an input sample, Generate a set of zero intensity signals as output samples. It should be noted that the signal recorded in the storage 123 as unlearned noise has a high probability of containing only noise because it has been determined in the comprehensive determination that there is no partial discharge. Also, the unlearned noise used in the learning data set may be accumulated after the previous learning of the denoising model, or may be the most recent predetermined number of noises.

The learning unit 136 duplicates the current denoising model of the site S stored in the storage 123, uses the parameters of the duplicated denoising model as initial values, and uses the generated learning data set to perform denoising. The parameters of the noise model are learned (step S43). During this learning process, the power equipment monitoring apparatus 10 can execute the partial discharge determination process shown in step S13 using the original denoising model. When parameter learning is completed, the learning unit 136 updates the current denoising model stored in the storage 123 to the re-learned denoising model (step S44).

As described above, according to the power equipment monitoring apparatus 10 according to the fifth embodiment, the denoise model can be adapted to changes in noise in the vicinity of the power equipment M by automatically re-learning the denoise model. can be kept up to date.

(Sixth embodiment)
FIG. 14 is a schematic block diagram showing the configuration of the power equipment monitoring apparatus 10 according to the sixth embodiment. The power equipment monitoring apparatus 10 according to the sixth embodiment further includes a display control unit 137 in addition to the configuration of the power equipment monitoring apparatus 10 according to the fifth embodiment. The display control unit 137 associates at least one of the waveform (time waveform or frequency waveform) of the denoising signal generated by the noise removing unit 132 and the determination result by the partial discharge determining unit 133 with the corresponding denoising model. Displayed on the display device 127 .

FIG. 15 is an example of a result display screen by the power equipment monitoring device 10 of the sixth embodiment. The result display screen shown in FIG. 15 includes the denoising signal for each denoising model, the partial discharge determination result, and the comprehensive determination result. In the example shown in FIG. 15, it can be seen that the denoising model trained using the past noise of the site 001 cannot remove the noise from the signal detected at the site 001. FIG. On the other hand, it can be seen that the denoising model trained using the past noise of the site 002 can remove the noise of the signal of the site 001 .

As described above, according to the sixth embodiment, the power device monitoring apparatus 10 displays the noise removal result and the partial discharge determination result for each applied denoising model, thereby letting the user know that the noise removal is going well. It can allow you to identify a denoising model and consider why.

(Seventh embodiment)
A power equipment monitoring apparatus 10 according to the seventh embodiment has the same configuration as that of the sixth embodiment. On the other hand, in the storage 123 of the seventh embodiment, noise source information, which is information about the noise source of each site S, is recorded. FIG. 16 is a diagram showing an example of noise source information according to the seventh embodiment. The noise source information is associated with the site S and stores the type of noise source (for example, compressor, inverter, relay, etc.) provided at the site S.

FIG. 17 is an example of a result display screen by the power equipment monitoring device 10 of the seventh embodiment.
The display control unit 137 according to the seventh embodiment associates, in addition to the denoising signal and the partial discharge determination result, the noise source information related to the site S at which the noise used for learning the denoising model is measured with the denoising model. is displayed on the display device 127.

As described above, the power equipment monitoring apparatus 10 according to the seventh embodiment displays the noise source information for each applied denoising model, thereby allowing the user to consider the noise source of the noise contained in the signal to be diagnosed. can be made

According to at least one embodiment described above, the power equipment monitoring apparatus 10 has the sensor 11 and the noise removal section 132 . The sensor 11 is attached to the power equipment M and measures a signal related to partial discharge. The noise removing unit 132 generates a denoising signal by inputting the measured signal to the denoising model. A denoising model is a model trained to receive a signal containing noise as an input and output a denoising signal from which the noise has been removed. The noise removal unit 132 generates a denoising signal using the first denoising model and the second denoising model. The first denoising model is a denoising model trained using a learning data set generated using signals measured by the sensor 11 . The second denoising model was trained using a training data set generated using signals measured from other power equipment M. It is a denoising model. As a result, the power device monitoring apparatus 10 can remove noise from the measured signal even if the environment of the power device M changes.

While several embodiments of the invention have been described, these embodiments have been presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. These embodiments and their modifications are included in the scope and spirit of the invention, as well as the scope of the invention described in the claims and equivalents thereof.

For example, in other embodiments, the power equipment monitoring device 10 is provided for each site S, but the present invention is not limited to this. For example, in another embodiment, part or all of the functions of the power equipment monitoring apparatus 10 may be realized by an information processing apparatus connected to the wide area network N. FIG. For example, the information processing apparatus may centrally manage the denoising models of a plurality of sites S, and each power equipment monitoring apparatus 10 may download and use the denoising models of other sites S from the information processing apparatus. In this case, when the denoising model is updated, the power equipment monitoring device 10 uploads the updated denoising model to the information processing device.

In the above-described embodiment, the partial discharge determination unit 133 determines presence/absence of partial discharge by comparing the intensity of the signal component in the denoising signal with the threshold, but the present invention is not limited to this. For example, the partial discharge determination unit 133 according to another embodiment may determine the presence or absence of partial discharge using a learned machine learning model (partial discharge determination model). The learning data set used for learning the partial discharge determination model has input samples as denoised signals and output samples as bit values indicating the presence or absence of partial discharge. For example, the training data set includes a set of known partial discharge signals as input samples and a bit value of "1" indicating the presence of partial discharge as output samples, or a signal denoised by a denoising model as input samples. , a set whose output samples are bit values indicating the presence or absence of partial discharge in the signal. By using such a learning data set, the parameters of the partial discharge determination model are learned so as to output the probability that partial discharge exists when the denoising signal is input.

DESCRIPTION OF SYMBOLS 1... Detection system 10... Electric power apparatus monitoring apparatus 11... Sensor 12... Arithmetic device 121... Processor 122... Main memory 123... Storage 124... Interface 125... Communication apparatus 126... Input device 127... Display device 131... Signal acquisition part 132... Noise Removal unit 133 Partial discharge determination unit 134 Comprehensive determination unit 135 Unlearned determination unit 136 Learning unit 137 Display control unit M Power device N Wide area network S Site

Claims (12)

a sensor that is attached to the power equipment and measures a signal related to partial discharge;
a noise removing unit that generates a denoising signal by inputting the measured signal to a denoising model trained to receive a signal containing noise as an input and output a denoising signal from which the noise has been removed; ,
The noise removal unit includes a first denoising model, which is a denoising model learned using a learning data set generated using the signal measured by the sensor, and a first denoising model measured from other power equipment A power equipment monitoring apparatus that generates the denoising signal using a second denoising model that is a denoising model trained using a learning data set that is generated using the signal. a storage unit that stores a plurality of denoising models including the first denoising model and the second denoising model;
The power equipment monitoring apparatus according to claim 1, wherein the noise removing unit generates the denoising signal using two or more denoising models among a plurality of denoising models stored in the storage unit. 3. The power device monitoring apparatus according to claim 1, further comprising a partial discharge determination unit that determines whether partial discharge has occurred based on the denoising signal. 4. The power device monitoring apparatus according to claim 3, further comprising: a general judgment unit that judges whether partial discharge occurs or not based on the judgment result of occurrence or non-occurrence of partial discharge for each of a plurality of denoising signals generated from different denoising models. The measured signal is 5. The power equipment monitoring apparatus according to claim 4, further comprising an unlearned noise determination unit that determines that the noise is unknown to the first denoising model. A signal storage unit that stores the signal measured by the sensor,
The power equipment monitoring apparatus according to claim 5, wherein the unlearned noise determination unit records the signal determined to be unknown noise in the signal storage unit. A learning unit for learning the parameters of the first denoising model using the learning data set,
The power equipment according to claim 5 or 6, wherein parameters of the first denoising model are re-learned using the learning data set generated using the signal determined to be unknown noise. surveillance equipment. 8. A display control unit for displaying a display screen in which a plurality of denoising signals generated based on the same signal are associated with the denoising model used to generate the denoising signals. 2. The power equipment monitoring device according to claim 1. A display control unit that displays a display screen that associates the determination result of occurrence of partial discharge for each of a plurality of denoising signals generated based on the same signal with the denoising model used to generate the denoising signal. The power equipment monitoring apparatus according to any one of claims 3 to 7. Equipped with a noise source storage unit that stores noise source information for each power device,
10. The power equipment monitoring apparatus according to claim 8, wherein the display control unit causes the display screen to display the noise source information in association with the corresponding denoising model. obtaining a signal related to partial discharge measured by a sensor attached to the power equipment;
A first denoising learned using a learning data set generated using the signal measured by the sensor to receive a signal containing noise as an input and output a denoised signal from which the noise has been removed. generating a first denoising signal by inputting the obtained signal into a model;
Using a learning data set generated using a signal measured by a sensor attached to another power device, a signal containing noise is used as an input and a denoised signal from which the noise is removed is learned as an output. generating a second denoising signal by inputting the obtained signal into a second denoising model;
A power equipment monitoring method comprising: the computer,
A power equipment monitoring program for functioning as the power equipment monitoring device according to any one of claims 1 to 10.

Images