JP2021027676A - Information processing device, information processing method, and program - Google Patents

Abstract

To provide an information processing device, an information processing method and a program capable of appropriately selecting learning data suitable for learning of a model which specifies an accident occurring in a power distribution system.SOLUTION: According to an embodiment, an information processing device includes an accident cause record information acquisition unit, a measurement result information acquisition unit, a feature amount calculation unit, an accident cause estimation unit, a determination unit and a leaning data selection unit. The accident cause record information acquisition unit acquires accident cause record information recording an accident cause. The measurement result information acquisition unit acquires measurement result information. The feature amount calculation unit calculates a feature amount on the basis of the measurement result information at an accident. The accident cause estimation unit estimates an accident cause on the basis of the feature amount. The determination unit determines whether the estimated first accident cause coincides with a second accident cause indicated by the acquired accident cause record information. The learning data selection unit selects the measurement result information at the accident of object equipment as learning data for creating a learned model in the case of determining that the accident causes are coincident.SELECTED DRAWING: Figure 1

Description

The present invention relates to an information processing device, an information processing method, and a program.

Conventionally, as a method of identifying the cause of an accident occurring in a power distribution system, a technique of identifying the cause of an accident using a learned model in which the measurement result of the power distribution system at the time of an accident is learned as learning data is known (for example). , Patent Document 1). By the way, the learning data used for learning the machine learning model is required to be data in which the measurement result of the distribution system at the time of the accident and the cause of the accident are appropriately associated with each other. However, with the conventional technique, it may be difficult to select appropriate data for learning the learning model.

Japanese Unexamined Patent Publication No. 11-142466

The problem to be solved by the present invention is to provide an information processing device, an information processing method, and a program capable of appropriately selecting learning data suitable for learning a model for identifying an accident occurring in a distribution system. is there.

The information processing device has an accident cause record information acquisition unit, a measurement result information acquisition unit, a feature amount calculation unit, an accident cause estimation unit, a determination unit, and a learning data selection unit. The accident cause record information acquisition unit acquires accident cause record information in which the accident cause of the target equipment is recorded. The measurement result information acquisition unit acquires measurement result information indicating the result of measuring the electric power in the target equipment. The feature amount calculation unit calculates the feature amount based on the measurement result information at the time of the accident of the target equipment among the measurement result information acquired by the measurement result information acquisition unit. The accident cause estimation unit estimates the accident cause based on the feature amount calculated by the feature amount calculation unit. The determination unit determines whether or not the first accident cause estimated by the accident cause estimation unit and the second accident cause indicated by the accident cause record information acquired by the accident cause record information acquisition unit match. judge. When the learning data selection unit determines that the first accident cause and the second accident cause match by the determination unit, the target of the measurement result information acquired by the measurement result information acquisition unit. The measurement result information at the time of an equipment accident is selected as learning data for creating a learned model that outputs information indicating the cause of the accident when the information related to the measurement result information is input.

It is a figure which shows an example of the structure of the information processing apparatus 10 of an embodiment. It is a figure which shows an example of the measurement result information MR. It is a figure which shows an example of the contents of the accident cause record information AC. It is a figure which shows an example of the contents of the accident data 152. It is a figure which shows an example of the output contents of the trained model 158. It is a figure which shows an example of the contents of the data set of a feature quantity. It is a figure which shows an example of a simple synthetic wave. It is the figure which expressed the 2nd order harmonic vector by the complex plane. It is a flowchart which shows the series flow of processing of a pre-processing function part. It is a flowchart which shows the series flow of the process of specifying the learning data of a learning function part. It is a flowchart which shows the series flow of the process which concerns on learning of a learning function part.

Hereinafter, the information processing apparatus, the information processing method, and the program of the embodiment will be described with reference to the drawings.

[overall structure]
FIG. 1 is a diagram showing an example of the configuration of the information processing apparatus 10 of the embodiment. The information processing device 10 is a device that identifies the cause of an accident that occurs in the distribution system DS by using the learning model while learning the learning model. In the information processing device 10, for example, one or more measuring devices 20, an accident cause recording device 30, and a display device 40 are connected directly or via a network. The network is, for example, a communication network such as a dedicated communication line or the Internet. Hereinafter, a case where the distribution system DS is a three-phase AC distribution system having A phase, B phase, and C phase will be described.

The measuring device 20 acquires the measurement result related to the distribution system DS. The measurement results include, for example, measurement results related to the distribution system DS, such as zero-phase voltage data, zero-phase current data, A-phase voltage data, A-phase current data, B-phase voltage data, B-phase current data, and C-phase. Includes voltage data and C-phase current data. Each data shows, for example, a voltage value or a current value from the acquisition timing of the measurement result in a time series. The measuring device 20 acquires the measurement result at predetermined time intervals. The predetermined time interval is, for example, about several tens to several hundreds [ms] intervals. The measuring device 20 transmits to the information processing device 10 information (hereinafter, measurement result information MR) summarizing the measurement results acquired at a predetermined time (for example, the time for one cycle of the electric power supplied by the distribution system DS). To do.

FIG. 2 is a diagram showing an example of the measurement result information MR. The measurement result information MR includes, for example, a record in which the acquired measurement result, the measurement date and time when the measurement result is measured, and the measurement device identification information that can identify the measurement device 20 are associated with each other for only one cycle. Information. Hereinafter, the measuring device 20 shall transmit the measurement result information MR to the information processing device 10 every time the time for one cycle elapses. That is, the measurement result information MR is the discretized zero phase and the waveform data for one cycle of each phase.

The measurement result information MR may not include the data of each phase as long as the data of at least the zero phase is included.

Returning to FIG. 1, when an accident occurs in the distribution system DS, the accident cause recording device 30 inputs information related to the cause of the accident (hereinafter referred to as accident cause recording information AC) input by a worker or the like who has recovered from the accident. Record. FIG. 3 is a diagram showing an example of the contents of the accident cause record information AC. The accident cause recording device 30 includes, for example, accident identification information capable of identifying the accident, information indicating the cause of the accident, information indicating the basis for identifying the cause of the accident, and accident equipment in which the accident occurred. A record in which the accident equipment identification information that can identify the accident equipment and the information indicating the weather at the time of the accident are associated with each other is input. The accident cause recording device 30 generates or updates the accident cause recording information AC based on the input record. The accident cause recording device 30 transmits the accident cause recording information AC to the information processing device 10 every time a record is input or at a predetermined time interval. The accident cause shown in the accident cause record information AC is an example of the "second accident cause".

Returning to FIG. 1, the display device 40 is a device that displays an image showing the cause of the accident in the distribution system DS identified by the information processing device 10.

[Configuration of information processing device 10]
Hereinafter, the configuration of the information processing device 10 will be described. The information processing device 10 includes, for example, a control unit 100 and a storage unit 150. The control unit 100 includes, for example, a measurement result information acquisition unit 102 and an accident cause record information acquisition unit 104 by executing a program (software) stored in the storage unit 150 by a processor such as a CPU (Central Processing Unit). , Accident identification unit 106, harmonic correction unit 108, feature amount calculation unit 110, accident cause estimation unit 112, learning data selection unit 114, learning processing unit 116, accident cause acquisition unit 118, and accident cause. Each function unit with the output unit 120 is realized. In addition, some or all of these components are hardware (circuits) such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), GPU (Graphics Processing Unit), etc. It may be realized by the part (including circuitry), or it may be realized by the cooperation of software and hardware.

The storage unit 150 may be realized by a storage device (a storage device including a non-transient storage medium) such as an HDD (Hard Disk Drive) or a flash memory, and is a removable storage medium such as a DVD or a CD-ROM. It may be realized by (non-transient storage medium), or it may be a storage medium mounted on a drive device. Further, a part or all of the storage unit 150 may be an external device such as NAS or an external storage server that can be accessed by the information processing device 10. In addition to the program, the storage unit 150 stores information such as accident data 152, selection learning data 154, machine learning model 156, and learned model 158. The machine learning model 156 is a model in which parameters are updated by the learning processing unit 116 using the learning data selected by the learning data selection unit 114, which will be described later. The trained model 158 is a machine learning model 156 at the time when the parameter update is completed by the learning processing unit 116. Details of the accident data 152 and the selection learning data 154 will be described later.

The measurement result information acquisition unit 102 acquires the measurement result information MR from the measurement device 20. The accident cause record information acquisition unit 104 acquires the accident cause record information AC from the accident cause recording device 30.

Here, among the functional units included in the control unit 100, the accident identification unit 106 and the harmonic correction unit 108 estimate the cause of the accident and the machine learning model 156 for various data included in the measurement result information MR. This is a preprocessing function unit that performs preprocessing when inputting to and inputting to the trained model 158. Further, among the functional units included in the control unit 100, the feature amount calculation unit 110, the accident cause estimation unit 112, the learning data selection unit 114, and the learning processing unit 116 are preprocessed by the preprocessing function unit. This is a learning function unit that selects the measurement result information MR used for learning the machine learning model 156 from the measurement result information MRs and learns the machine learning model 156 by the selected measurement result information MR. Further, among the functional units included in the control unit 100, the accident cause acquisition unit 118 and the accident cause output unit 120 are based on various data preprocessed by the preprocessing function unit and the trained model 158. It is an operation function unit that outputs the cause of the accident that occurred in the distribution system DS.

[About the pre-processing function]
First, the details of the preprocessing function unit will be described. The accident identification unit 106 identifies a time zone in which an accident occurs in the distribution system DS (hereinafter, a time zone during an accident) based on the measurement result information MR acquired by the measurement result information acquisition unit 102. The accident identification unit 106 calculates the effective value of the zero-phase current based on the zero-phase current data included in the measurement result information MR. The accident identification unit 106 specifies a time zone in which the calculated effective value of the zero-phase voltage exceeds the zero-phase current threshold value as a time zone during an accident. The zero-phase current threshold is a value that can classify the zero-phase current generated when an accident occurs in the distribution system DS and the zero-phase current generated when an accident does not occur.

Next, when the measurement result information MR includes the data of each phase, the accident identification unit 106 identifies the accident phase in the specified time zone during the accident. The accident identification unit 106 calculates the current effective value of each phase based on the current data of each phase indicated by the measurement result information MR. The accident identification unit 106 is based on the calculated current effective value of each phase, and is the average value of the current effective values of each phase related to the time zone before the accident (for example, one minute immediately before the accident occurs in the distribution system DS). , Calculate the average current effective value of each phase during the time of the accident. In the accident identification unit 106, the value obtained by dividing the average value of the current effective values in the time zone during the accident by the average value of the current effective values in the time zone before the accident (hereinafter referred to as the current index value) is larger than the accident phase current threshold value. Identify the phase. The accident phase current threshold value is a value that can classify a value that can be taken by the current index value in the accident phase at the time of occurrence of an accident in the distribution system DS and a value that can be taken by the current index value in a phase other than the accident phase.

The accident identification unit 106 calculates the effective voltage value of each phase based on the voltage data of each phase indicated by the measurement result information MR. Based on the calculated voltage effective value of each phase, the accident identification unit 106 averages the average voltage value of each phase in the time zone before the accident and the voltage effective value of each phase in the time zone during the accident. Calculate with the value. In the accident identification unit 106, the value obtained by dividing the average value of the effective voltage values during the time period during the accident by the average value of the effective voltage values during the time period before the accident (hereinafter referred to as the voltage index value) is smaller than the fault phase voltage threshold value. Identify the phase. The fault phase voltage threshold value is a value that can classify a value that can be taken by the voltage index value in the fault phase at the time of occurrence of an accident in the distribution system DS and a value that can be taken by the voltage index value in a phase other than the fault phase.

When the phase specified based on the current index value and the phase specified based on the voltage index value match, the accident identification unit 106 identifies the phase as the accident phase. Further, the accident identification unit 106 stores the data of the phase specified as the accident phase in the measurement result information MR and the data of the accident phase in the time zone during the accident as the accident data 152 in the storage unit 150. The accident identification unit 106 is an example of a “determination unit”.

FIG. 4 is a diagram showing an example of the contents of the accident data 152. The accident data 152 is, for example, information in which the accident identification information, the measurement date and time, the measuring device identification information, and the accident phase data (in the figure, the A phase data) are associated with each other. Accidents are managed by, for example, an accident database (not shown), and the accident database stores information in which accident identification information and accident date and time are associated with each other. For example, the accident identification unit 106 acquires the accident identification information of the accident corresponding to the specified time zone during the accident from the accident database, and mutually obtains the acquired accident identification information and the measurement result information MR of the time zone during the accident. Accident data 152 is generated (updated) in association with each other. The accident identification information may be input by the user of the information processing device 10.

Returning to FIG. 1, the harmonic correction unit 108 corrects the accident phase data included in the accident data 152 with respect to the accident phase specified by the accident identification unit 106 with the accident phase data in the time zone before the accident. Specifically, the harmonic correction unit 108 Fourier transforms the data of the accident phase in the time zone before the accident among the measurement results included in the measurement result information MR and decomposes them into each harmonic component. The harmonic correction unit 108 removes the waveform obtained by performing the inverse Fourier transform on each of the decomposed harmonic components from the accident phase data included in the accident data 152. As a result, the harmonic correction unit 108 removes the harmonics existing before the accident from the data of the accident phase during the time zone during the accident.

The harmonic correction unit 108 may not be able to identify the accident phase by the accident identification unit 106 (that is, the phase specified based on the current index value and the phase specified based on the voltage index value do not match), or When the measurement result information MR does not include the data of each phase, the zero-phase data of the time zone during the accident is corrected by the zero-phase data of the time zone before the accident. Since the correction process is the same as the process for the accident phase data, the description thereof will be omitted. As a result, the harmonic correction unit 108 can remove the influence of the harmonics generated on the distribution system DS from the accident phase data or the zero phase data, and can select more suitable learning data from these data. .. In addition, the correction related to the accident phase data or the zero phase data should normalize the harmonic component of the data in the time zone before the accident and subtract it from the normalization of the harmonic component of the data in the time zone after the accident. May be done by.

[About the learning function department]
Next, the details of the learning function unit will be described. The feature amount calculation unit 110 calculates the feature amount based on the data of the accident phase in the time zone during the accident corrected by the harmonic correction unit 108. FIG. 6 is a diagram showing an example of the contents of the feature amount data set. The feature amount calculated by the feature amount calculation unit 110 is, for example, as shown in FIG. 6, showing the data of the accident phase by the harmonic components from the second order to the 25th order. Specifically, the feature quantity is indicated by a record in which the magnitude of the harmonic, the in-phase component with respect to the fundamental wave, and the orthogonal component with respect to the fundamental wave are associated with each other for each order of the harmonic. The feature amount data set stores a feature amount record calculated by the feature amount calculation unit 110. Here, an example of an actual waveform including harmonics will be described with reference to FIG. 7.

FIG. 7 is a diagram showing an example of a simple synthetic wave. The composite wave in the upper figure is composed of the fundamental wave and the second harmonic shown in the lower figure. The magnitude of the fundamental wave is A, the magnitude of the second harmonic is R, and the second harmonic is the fundamental wave. It is assumed that the phase is advanced by θ. The fundamental wave can be expressed as Asin (ωt) when the angular velocity is ω and the time is t, and the second harmonic can be expressed as Rsin (2ωt + θ). Here, when this second harmonic is expressed by the vector symbol method with respect to the fundamental wave, it can be expressed as Rcosθ + jRsinθ. FIG. 8 is a diagram in which the second-order harmonic vector is represented by a complex plane. As shown in FIG. 8, since the real part of the second harmonic is defined as the in-phase component and the imaginary part of the second harmonic is defined as the orthogonal component, the in-phase component of the second harmonic is Rcosθ and the orthogonal component is Rsinθ. Become.

Returning to FIG. 1, the accident cause estimation unit 112 estimates the cause of the accident that occurred in the accident phase based on the feature amount calculated by the feature amount calculation unit 110. The accident cause estimation unit 112 may estimate the accident cause based on the feature amount calculated by the feature amount calculation unit 110 and the ruling related to the predetermined feature amount, and the user of the information processing device 10 may estimate the cause of the accident. The cause of the accident may be acquired by inputting the cause of the accident estimated by referring to the feature amount calculated by the feature amount calculation unit 110 into the information processing apparatus 10. The accident cause estimated by the accident cause estimation unit 112 is an example of the "first accident cause".

The learning data selection unit 114 determines whether or not the accident causes match based on the accident cause record information AC acquired by the accident cause record information acquisition unit 104 and the estimation result of the accident cause estimation unit 112. For example, the learning data selection unit 114 searches the accident cause record information AC using the accident identification information included in the accident data 152 as a search key, and the accident cause to which the same accident identification information is associated with the accident cause record information AC. To identify. When the identified accident cause and the accident cause estimated by the accident cause estimation unit 112 match, the learning data selection unit 114 corrects the accident by the harmonic correction unit 108 corrected by the harmonic correction unit 108. The data of the accident phase in the middle time zone is stored (updated) in the storage unit 150 as the selection learning data 154 as the learning data of the machine learning model 156. In this case, the estimation result of the accident cause estimation unit 112 is used as teacher data. Further, the learning data selection unit 114 determines that the accident data 152 is not used as the learning data of the machine learning model 156 when the identified accident cause and the accident cause estimated by the accident cause estimation unit 112 do not match. ..

The learning processing unit 116 performs a process of learning the machine learning model 156 based on the selection learning data 154 selected by the learning data selection unit 114. For example, when the learning progress of the machine learning model 156 reaches a predetermined progress based on the learning curve of the machine learning model 156, the learning processing unit 116 sets the machine learning model 156 at that time as the learned model 158. ..

[About the operation function department]
Next, the details of the operation function unit will be described. The accident cause acquisition unit 118 inputs the accident phase data in the time zone during the accident corrected by the harmonic correction unit 108 into the trained model 158, and acquires an output indicating the accident cause. FIG. 5 is a diagram showing an example of the output contents of the trained model 158. The output of the trained model 158 is, for example, information in which one or more accident causes, the occurrence probability for each accident cause, and the accident identification information associated with the input accident phase data are associated with each other. .. The accident cause output unit 120 outputs and displays the information indicating the output of the learned model 158 acquired by the accident cause acquisition unit 118 on the display device 40. As a result, the information processing device 10 can present the cause of the accident to the user.

The accident cause output unit 120 may output the accident cause having the highest probability of occurrence among the outputs of the learned model 158 to the display device 40, and the learned model 158 outputs the cause of the accident having the highest probability of occurrence. It may be learned to output.

Further, in the above description, the accident cause acquisition unit 118 has described the case where the accident phase data in the time zone during the accident corrected by the harmonic correction unit 108 is input to the trained model 158, but the present invention is not limited to this. .. The accident cause acquisition unit 118 may input the zero-phase data included in the measurement result information MR acquired by the measurement result information acquisition unit 102 or the data of each phase into the trained model 158.

[Operation flow]
FIG. 9 is a flowchart showing a series of processing flows of the preprocessing function unit. FIG. 10 is a flowchart showing a series of processes for specifying the learning data of the learning function unit. First, the measurement result information acquisition unit 102 acquires the measurement result information MR from the measurement device 20 (step S100). The accident identification unit 106 calculates the effective value of the zero-phase current based on the zero-phase current data included in the measurement result information MR acquired by the measurement result information acquisition unit 102 (step S102). Next, the accident identification unit 106 identifies a time zone in which the calculated effective value of the zero-phase voltage exceeds the zero-phase current threshold value as a time zone during an accident (step S104). The accident identification unit 106 determines whether or not the measurement result information MR acquired by the measurement result information acquisition unit 102 includes the data of each phase (step S106).

When the accident identification unit 106 determines that the measurement result information MR includes the data of each phase, the accident identification unit 106 performs a process of identifying the accident phase. Therefore, each of the accident identification units 106 is based on the current data of each phase indicated by the measurement result information MR. The current effective value of the phase is calculated (step S108). Next, the accident identification unit 106 calculates the average value of the current effective value in the time zone during the accident and the average value of the current effective value in the time zone before the accident based on the calculated current effective value of each phase. (Step S110). In the accident identification unit 106, the value obtained by dividing the average value of the current effective values in the time zone during the accident by the average value of the current effective values in the time zone before the accident (that is, the current index value) is larger than the accident phase current threshold value. Identify the phase (step S112).

Next, the accident identification unit 106 calculates the effective voltage value of each phase based on the voltage data of each phase indicated by the measurement result information MR (step S114). Next, the accident identification unit 106 calculates the average value of the voltage effective value in the time zone during the accident and the average value of the voltage effective value in the time zone before the accident based on the calculated voltage effective value of each phase. (Step S116). In the accident identification unit 106, the value obtained by dividing the average value of the effective voltage values during the time period during the accident by the average value of the effective voltage values during the time period before the accident (that is, the voltage index value) is smaller than the accident phase voltage threshold value. Identify the phase (step S118). When the phase specified in step S112 and the phase specified in step S118 match, the accident identification unit 106 identifies the phase as an accident phase and obtains data of the time zone during the accident of the specified accident phase. It is stored as accident data 152 (step S120).

The harmonic correction unit 108 corrects the accident phase data included in the accident data 152 with respect to the accident phase specified by the accident identification unit 106 based on the accident phase data in the time zone before the accident, and exists before the accident. Harmonics are removed from the accident phase data during the time of the accident (step S122). When the accident identification unit 106 determines that the measurement result information MR does not include the data of each phase, the harmonic correction unit 108 uses the zero-phase data of the time zone during the accident as zero in the time zone before the accident. It is corrected by the phase data, and the harmonics existing before the accident are removed from the zero-phase data in the time zone during the accident (step S124).

Next, the feature amount calculation unit 110 calculates the feature amount based on the data of the accident phase in the time zone during the accident corrected by the harmonic correction unit 108 (step S124). The accident cause estimation unit 112 estimates the accident cause based on the feature amount calculated by the feature amount calculation unit 110 (step S126). The accident cause record information acquisition unit 104 acquires the accident cause record information AC from the accident cause recording device 30 (step S128). In the accident cause estimation unit 112, whether or not the accident cause estimated based on the feature amount (that is, the first accident cause) and the accident cause indicated by the accident cause record information AC (that is, the second accident cause) match. (Step S130). When the learning data selection unit 114 determines that the first accident cause and the second accident cause match by the accident cause estimation unit 112, the accident phase data in the time zone during the accident corrected by the harmonic correction unit 108 Is selected as the learning data of the machine learning model 156, and the selected data is stored (updated) in the storage unit 150 as the selected learning data 154 (step S132). When the learning data selection unit 114 determines that the first accident cause and the second accident cause do not match by the accident cause estimation unit 112, the accident phase of the accident phase in the time zone during the accident corrected by the harmonic correction unit 108 It is determined that the data is not used as the training data of the machine learning model 156 (step S134).

FIG. 11 is a flowchart showing a series of processes related to learning of the learning function unit. First, the accident cause acquisition unit 118 inputs the accident phase data in the time zone during the accident corrected by the harmonic correction unit 108 into the trained model 158 (step S200). The accident cause acquisition unit 118 acquires an output from the trained model 158 (step S202). The accident cause output unit 120 outputs and displays the information indicating the output of the learned model 158 acquired by the accident cause acquisition unit 118 on the display device 40 (step S204).

[Other configurations of information processing device 10]
In the above description, the case where the information processing apparatus 10 includes the preprocessing function unit, the learning function unit, and the operation function unit has been described, but the present invention is not limited to this.
The learning function unit may be provided with a learning device that is separate from the information processing device 10. In this case, the information processing device 10 is provided with an accident identification unit 106 which is a preprocessing function unit, a harmonic correction unit 108, an accident cause acquisition unit 118 which is an operation function unit, and an accident cause output unit 120. Further, the separate learning device is provided with a feature amount calculation unit 110, an accident cause estimation unit 112, a learning data selection unit 114, and a learning processing unit 116, which are learning function units, and is corrected by the harmonic correction unit 108. The accident phase data and the accident cause record information AC in the time zone during the accident are input by the user of the learning device or acquired from another device. As a result, the learning device can quickly identify the cause of the accident that occurred in the distribution system DS.

[About harmonic correction unit 108]
Further, in the distribution system DS, a generator using renewable energy such as PV (Photovoltaic) or a storage battery is not connected to or is connected to the distribution system DS in an environment where harmonics are not generated (for example, PV (Photovoltaic)). However, when it is difficult to affect the measurement result of the measuring device 20), the information processing device 10 does not have to correct the data included in the measurement result information MR by the harmonic correction unit 108. In this case, the information processing device 10 does not have to include the harmonic correction unit 108.

[Summary of Embodiment]
According to at least one embodiment described above, the measurement result information acquisition unit 102, the accident cause record information acquisition unit 104, the feature amount calculation unit 110, the accident cause estimation unit 112, the determination unit, and the learning data selection unit. By having the unit 114, it is possible to specify the learning data suitable for learning the machine learning model 156 among the measurement result information MRs measured by the measuring device 20, and to train the machine learning model 156 with high accuracy. Can be done.

Although embodiments of the present invention have been described, these embodiments are presented as examples 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 changes can be made without departing from the gist of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, as well as in the scope of the invention described in the claims and the equivalent scope thereof.

10 ... Information processing device, 20 ... Measuring device, 30 ... Accident cause recording device, 40 ... Display device, 100 ... Control unit, 102 ... Measurement result information acquisition unit, 104 ... Accident cause record information acquisition unit, 106 ... Accident identification unit , 108 ... Harmonic correction unit, 110 ... Feature amount calculation unit, 112 ... Accident cause estimation unit, 114 ... Learning data selection unit, 116 ... Learning processing unit, 118 ... Accident cause acquisition unit, 120 ... Accident cause output unit, 150 ... storage unit, 152 ... accident data, 154 ... selective learning data, 156 ... machine learning model, 158 ... learned model, AC ... accident cause record information, MR ... measurement result information, DS ... power distribution system

Claims (8)

Accident cause record information acquisition department that acquires accident cause record information in which the accident cause of the target equipment is recorded,
A measurement result information acquisition unit that acquires measurement result information indicating the results of measurement related to electric power in the target equipment.
Among the measurement result information acquired by the measurement result information acquisition unit, the feature amount calculation unit that calculates the feature amount based on the measurement result information at the time of the accident of the target equipment, and the feature amount calculation unit.
An accident cause estimation unit that estimates the cause of an accident based on the feature amount calculated by the feature amount calculation unit, and an accident cause estimation unit.
A determination unit that determines whether or not the first accident cause estimated by the accident cause estimation unit and the second accident cause indicated by the accident cause record information acquired by the accident cause record information acquisition unit match. ,
When the determination unit determines that the first accident cause and the second accident cause match, among the measurement result information acquired by the measurement result information acquisition unit, the measurement result at the time of the accident of the target equipment. A learning data sorting unit that sorts information as learning data for creating a trained model that outputs information indicating the cause of an accident when information related to measurement result information is input.
Information processing device equipped with. The information related to the measurement result information includes a feature amount calculated by the feature amount calculation unit based on the measurement result information.
The information processing device according to claim 1. An accident identification unit that identifies the time zone during an accident using the measurement results of the zero-phase current or zero-phase voltage included in the measurement result information.
The harmonics contained in the zero-phase current or the zero-phase voltage in the time zone specified by the accident-identifying unit are corrected by the zero-phase current before the accident or the harmonics contained in the zero-phase voltage before the accident. Further equipped with a harmonic correction unit
The feature amount calculation unit calculates the feature amount based on the zero-phase current in the time zone corrected by the harmonic correction unit or the zero-phase voltage in the time zone.
The information processing device according to claim 1 or 2. An accident identification unit that identifies the time zone during an accident using the measurement results of the current of each phase or the voltage of each phase included in the measurement result information.
The harmonics contained in the accident phase current or the accident phase voltage in the time zone specified by the accident identification unit are corrected by the accident phase current before the accident or the accident phase voltage before the accident. Further equipped with a harmonic correction unit
The feature amount calculation unit calculates the feature amount based on the current of the accident phase in the time zone corrected by the harmonic correction unit or the voltage of the accident phase.
The information processing device according to any one of claims 1 to 3. The accident identification unit uses at least one of the measurement result of the zero-phase current or the measurement result of the zero-phase voltage included in the measurement result information and the measurement result of each phase current or the measurement result of each phase voltage. , Identify the accident phase,
The information processing device according to claim 4. The determination unit is an accident estimated by the accident cause estimation unit based on at least one of a feature amount based on the zero-phase current or zero-phase voltage and a feature amount based on the current of the accident phase or the voltage of the accident phase. Determining whether the cause and the accident cause record information match.
The information processing device according to claim 3 or 4. The computer
Acquire the accident cause record information in which the accident cause of the target equipment is recorded,
Acquire measurement result information showing the result of measurement related to electric power in the target equipment.
Of the acquired measurement result information, the feature amount is calculated based on the measurement result information at the time of the accident of the target equipment.
Based on the calculated features, the cause of the accident is estimated and
It is determined whether or not the estimated first accident cause and the second accident cause indicated by the acquired accident cause record information match.
When it is determined that the first accident cause and the second accident cause match, among the acquired measurement result information, the measurement result information at the time of the accident of the target equipment is input with the information related to the measurement result information. When it is done, it is selected as training data for creating a trained model that outputs information indicating the cause of the accident.
Information processing method. On the computer
Obtain the accident cause record information in which the accident cause of the target equipment is recorded,
Obtain measurement result information indicating the result of measurement related to electric power in the target equipment.
Among the acquired measurement result information, the feature amount is calculated based on the measurement result information at the time of the accident of the target equipment.
Based on the calculated features, the cause of the accident is estimated.
It is made to judge whether or not the estimated first accident cause and the second accident cause indicated by the acquired accident cause record information match.
When it is determined that the first accident cause and the second accident cause match, among the acquired measurement result information, the measurement result information at the time of the accident of the target equipment is input with the information related to the measurement result information. When it is done, it is selected as training data for creating a trained model that outputs information indicating the cause of the accident.
program.

Images