JP7739828B2 - Accident cause estimation device, and control method and program for accident cause estimation device - Google Patents

Description

The present invention relates to an accident cause estimation device, a control method for an accident cause estimation device, and a program.

When an accident occurs in a power distribution system, the power company will use its distribution line automation system to identify and isolate the affected section and restore power to healthy sections. At the same time, the power company will locate the location of the accident while estimating the cause of the accident based on various circumstances at the time of the accident and past experience, and will then begin restoration work.

Technology for estimating the cause of such power distribution system accidents is disclosed, for example, in Patent Document 1.

Japanese Patent Application Laid-Open No. 2017-131006

However, estimating the cause of an accident is influenced by the past experience of the person in charge, and for example, it is difficult for an inexperienced person to accurately estimate the cause of an accident. In addition, different people may have different perspectives when analyzing the situation, and the estimated cause of an accident may differ depending on the person in charge.

The present invention was made in light of this background, and aims to provide an accident cause estimation device, a control method for an accident cause estimation device, and a program that can accurately estimate the cause of an accident in a power distribution system.

One of the present inventions for achieving the above-mentioned object is an accident cause estimation device that outputs a presumed cause of an accident that has occurred in a distribution system that is divided into a plurality of sections by switches, and that includes: a learning model memory unit that stores a learning model created based on past accident information including the type of accident that has occurred in the distribution system, the date and time the accident occurred, and the section in which the accident occurred, and the cause of the accident, so that when new accident information is input, the device outputs one or more presumed causes of the new accident and the probability that each of the presumed causes is the true cause; an accident information acquisition unit that acquires the new accident information; an presumed cause acquisition unit that acquires one or more presumed causes of the accident output from the learning model and the probability that each of the presumed causes is the true cause by inputting the new accident information into the learning model; and an presumed cause output unit that outputs either one or more presumed causes of the accident and the probability that each of the presumed causes is the true cause, or a presumed cause whose probability is equal to or greater than a predetermined value, depending on the selection result of whether to output the probability that each of the presumed causes is the true cause, or to output only presumed causes whose probability is equal to or greater than a predetermined value .

Furthermore, the problems disclosed in this application and the solutions thereto will be made clear in the detailed description and drawings.

This invention makes it possible to accurately estimate the cause of an accident in a power distribution system.

FIG. 1 is a diagram illustrating an accident cause estimation system. FIG. 2 is a hardware configuration diagram of the accident cause estimation device. FIG. 2 is a diagram illustrating a storage device of the accident cause estimation device. FIG. 10 is a diagram showing an accident information management table. FIG. 10 is a diagram showing an example of an input screen for inputting the cause of an accident. FIG. 1 is a diagram illustrating a learning model. FIG. 2 is a functional block diagram of the accident cause estimation device. 3 is a flowchart showing a processing flow of the accident cause estimation device. 3 is a flowchart showing a processing flow of the accident cause estimation device. FIG. 10 is a diagram showing an example of a display screen for the estimated cause of an accident.

At least the following points will become clear from the description in this specification and the accompanying drawings. Below, the present invention will be described in accordance with one embodiment with reference to the accompanying drawings.

==Overall Structure==
FIG. 1 shows the overall configuration of an accident cause estimation system 1000 according to an embodiment of the present invention.

The accident cause estimation system 1000 is configured by an accident cause estimation device 100 and a power distribution line automation system 200 that are communicatively connected via a network 500 such as the Internet, a LAN (Local Area Network), or a telephone network. The accident cause estimation device 100 is also communicatively connected to a weather information providing device 300 via the network 500.

The distribution line automation system 200 is a computer system that monitors and controls equipment such as switches and transformers installed in the distribution system, collects measurement data such as voltage and current, and detects and recovers from faults.

The power distribution system is managed by dividing it into multiple sections using switches. The distribution line automation system 200 identifies the type of fault (such as a short circuit, ground fault, or open circuit) and the section in which the fault occurred based on the open/close status of each switch and measurement data such as voltage and current obtained from various sensors installed in the power distribution system. The distribution line automation system 200 then controls the switches to isolate the section in which the fault occurred from the other sections and cause a power outage, while ensuring that power is supplied from the substation to healthy sections where no fault has occurred.

When the distribution line automation system 200 detects such an accident in the power distribution system, it transmits the date and time of the accident, the section in which the accident occurred, and the type of accident (short circuit, ground fault, open circuit, etc.) to the accident cause estimation device 100.

The accident cause estimation device 100 is a computer that estimates and outputs the causes of accidents that occur in a power distribution system, and stores a history of accident information that records the date and time of accidents, the section where the accident occurred, and the type of accident that has been previously obtained from the power distribution line automation system 200. The accident cause estimation device 100 then analyzes past accident information and the causes of those accidents to generate a learning model (AI model) that outputs the estimated cause of a new accident when new accident information is input. By using this learning model, the accident cause estimation device 100 outputs the estimated cause of a new accident when new accident information is input.

This method makes it possible to accurately estimate the cause of a power distribution system accident. For example, it can reduce the variation in the estimated cause of an accident that arises due to differences in the experience and perspective of situation analysis of the personnel responding to the power distribution system accident. This also makes it possible to more accurately identify the actual location of the accident and plan the personnel and equipment required for restoration work.

The weather information providing device 300 is a computer that provides weather information such as weather, temperature, humidity, air pressure, and wind speed.

When a new accident occurs in the power distribution system, the accident cause estimation device 100 acquires predetermined types of weather information, such as the weather and temperature at that time, from the weather information providing device 300, and stores this information as accident information in association with information acquired from the power distribution line automation system 200, such as the time of the new accident, the section of the accident, and the type of accident.

By including weather information in the accident information, the accuracy of the learning model is improved, enabling the accident cause estimation device 100 to output a more accurate estimated cause of the accident.

More details will be provided below.

==Accident cause estimation device==
Next, the accident cause estimation device 100 will be described.

An example of the hardware configuration of the accident cause estimation device 100 is shown in Figure 2. The accident cause estimation device 100 is composed of a CPU (Central Processing Unit) 110, memory 120, communication device 130, storage device 140, input device 150, output device 160, and recording medium reader 170.

The CPU 110 is responsible for overall control of the accident cause estimation device 100, and realizes the various functions of the accident cause estimation device 100 by reading into memory 120 and executing or processing the accident cause estimation device control program 700, which is composed of code for performing various operations related to this embodiment stored in the storage device 140, and various data.

For example, the CPU 110 executes or processes the accident cause estimation device control program 700 and various data, and by cooperating with hardware devices such as the memory 120, communication device 130, and storage device 140, functions such as the learning model storage unit 101, accident information acquisition unit 102, estimated cause acquisition unit 103, and estimated cause output unit 104, which will be described later, are realized.

The accident cause estimation device control program 700 is a general term for programs that realize the functions of the accident cause estimation device 100, and includes, for example, application programs, an OS (Operating System), and various libraries that run on the accident cause estimation device 100.

Memory 120 can be composed of, for example, a semiconductor memory device.

The storage device 140 is a device that provides physical storage space for storing various programs, data, tables, etc., such as a hard disk drive, SSD (Solid State Drive), or flash memory. In this embodiment, as shown in FIG. 3, the storage device 140 stores the accident cause estimation device control program 700, as well as various data such as an accident information management table 600, learning data 610, learning model 620, and cause estimation results 630.

An example of the accident information management table 600 is shown in Figure 4.

When an accident occurs in the power distribution system and accident information including the date and time of the accident, the section where the accident occurred, and the type of accident (e.g., short circuit, ground fault, or disconnection) is transmitted from the power distribution line automation system 200, the accident cause estimation device 100 stores this accident information in the accident information management table 600.

The accident cause estimation device 100 also obtains a specified type of weather information for the date and time the accident occurred from the weather information providing device 300 and stores it as accident information in the accident information management table 600. For example, the types of weather information may include information that may have some effect on an accident in a power distribution system, such as weather (sunny, cloudy, rain, snow, thunder, hail, etc.), temperature, humidity, wind speed, wind direction, and air pressure.

In the accident information management table 600 shown in Figure 4, the information listed in the "Date and Time," "Type," and "Section" columns is information obtained from the power distribution line automation system 200, and the information listed in the "Weather Information" column is information obtained from a weather information provider.

The information listed in the "Inspection Results" column relates to the inspection results of the distribution system equipment installed in the section where the accident occurred, and the information listed in the "Period of Use" column indicates the period of use of the distribution system equipment installed in the section where the accident occurred.

Power distribution systems are equipped with a variety of equipment, including pole transformers, insulators, cross arms, and switches, and electric power companies and their partner companies regularly inspect and replace these pieces of equipment for maintenance. The electric power company then records and manages information about this equipment, such as inspection results and the dates of installation and replacement, in a computer (not shown). The accident cause estimation device 100 obtains this information from the computer at the time the accident occurred and records it as accident information in the "Inspection Results" and "Duration of Use" columns of the accident information management table 600.

By including this information in the accident information, the accuracy of the learning model 620 is improved, enabling the accident cause estimation device 100 to output a more accurate estimated cause of the accident.

In addition, the information listed in the "Seasonal Factors" column indicates seasonal phenomena that occur in the area where the section where the accident occurred and may be the cause of an accident, as well as the times when such phenomena are likely to occur. Information listed in the "Regional Factors" column indicates geographically specific phenomena that exist in the area where the section where the accident occurred and may be the cause of an accident.

For example, some areas are located along windier coasts than other areas, and there is a tendency for short-circuit accidents in power lines to increase during strong winds in bad weather. In other areas, there are many wild birds that build nests on utility poles to raise their young during certain periods each year (for example, March to July), which can lead to an increase in short-circuit accidents. Furthermore, there are also areas where the number of accidents involving snakes shorting out power lines increases during certain periods (for example, April to October).

This information on potential accident causes specific to a particular region or season is recorded in a computer (not shown), and the accident cause estimation device 100 can obtain this information from the computer and record it as accident information in the "seasonal factors" and "regional factors" columns of the accident information management table 600.

In this way, by including this information in the accident information, the accuracy of the learning model 620 is further improved, and the accident cause estimation device 100 is able to output a more accurate estimated cause of the accident.

The "Cause of Accident" column also lists the true cause of the accident that was discovered during a post-accident investigation. These causes of the accident are recorded in the accident information management table 600, for example, when the accident cause estimation device 100 outputs a screen such as that shown in Figure 5 to the output device 160 and the person in charge inputs the cause of the accident via the input device 150. In the example shown in Figure 5, the person in charge can select the cause of the accident from multiple candidates, but the person in charge can also input the cause of the accident directly.

Returning to Figure 3, the accident cause estimation device 100 generates learning data 610 using past accident information and accident causes recorded in the accident information management table 600.

The training data 610 is data (teaching data) used to train a machine learning model (training model 620) that outputs the estimated cause of a new accident when new accident information is input.

The accident cause estimation device 100 presents accident information from a predetermined period in the past (one month, six months, one year, etc.; hereafter referred to as the "time interval") to the person in charge as explanatory variables (features), and accepts from the person in charge the setting of the accident cause, which is the objective variable (label) for the accident information. The accident cause estimation device 100 then generates learning data 610 by associating the accident information with the accident cause.

Note that the task of matching accident information with accident causes does not need to be performed on a time interval basis; it is preferable to perform the task each time the cause of an accident is identified. In this case, the accident cause estimation device 100 extracts data from the accident information for a time interval that has been matched with the accident cause as learning data 610. The above time interval is set empirically, for example, based on the frequency of accidents. Furthermore, learning models 620 may be generated for multiple time intervals, and these learning models 620 may be used to infer the cause of an accident in a power distribution system. The above time interval is set, for example, to a length that improves the accuracy of the accident cause, based on past cases.

Alternatively, when generating learning data 610 using past accident information and accident causes recorded in the accident information management table 600, the accident cause estimation device 100 may generate learning data 610 using data on accidents that occurred in a specific region. In this manner, a learning model 620 specific to that region can be generated, thereby improving the accuracy of estimating the cause of an accident.

Similarly, when generating learning data 610 using past accident information and accident causes recorded in the accident information management table 600, the accident cause estimation device 100 may generate learning data 610 using data on accidents that occurred during a specific period. In this manner, a learning model 620 specific to that period can be generated, thereby improving the accuracy of estimating the cause of an accident.

Furthermore, when generating learning data 610 using past accident information and accident causes recorded in the accident information management table 600, the accident cause estimation device 100 may generate learning data 610 using data on accidents that occurred in a specific region and time period. In this manner, a learning model 620 specific to that region and time period can be generated, thereby improving the accuracy of estimating the cause of the accident.

In this embodiment, the learning model 620 is a DNN (Deep Neural Network), but it may also be implemented using other types of models, such as gradient boosting (GBDT (Gradient Boosting Decision Tree)).

Figure 6 shows an example of a learning model 620 (neural network structure). As shown in Figure 6, accident information is input to the input layer 621 of the learning model 620. The intermediate layer 622 includes one or more hidden layers consisting of one or more nodes containing parameters adjusted by learning. The intermediate layer 622 calculates one or more predicted values (probabilities) for the output layer 623 based on the accident information provided to the input layer 621. In the learning model 620 of this embodiment, the output layer 623 includes multiple predicted values (multiple causes of accidents and their probabilities).

Returning to Figure 3, the cause estimation result 630 includes information related to the estimated cause of the accident output from the learning model 620. In this embodiment, the cause estimation result 630 includes one or more estimated causes of the accident and the calculation results of the probability that each estimated cause is the true cause. In this manner, it is possible to provide the person investigating the cause of the accident with information that is useful for securing the equipment and personnel necessary for accident restoration work, planning work processes, etc.

Returning to Figure 2, the storage device 140 can be built into the accident cause estimation device 100 or can be externally attached.

The recording medium reader 170 reads programs and data recorded on a recording medium 800 such as a CD-ROM or DVD, and stores them in the storage device 140.

The communication device 130 exchanges data and programs with other computers, such as the power distribution line automation system 200 and the weather information providing device 300, via a network 500 such as the Internet or a local area network (LAN). For example, the above-mentioned accident cause estimation device control program 700 can be stored in another computer (not shown), and the accident cause estimation device 100 can download and execute the accident cause estimation device control program 700 from this computer.

Alternatively, the communication device 130 may periodically receive accident information and weather information from the power distribution line automation system 200 and the weather information providing device 300. Furthermore, the accident cause estimation device 100 may not have a storage device 140, and may instead realize its functions using various data such as the above programs and tables stored in another computer (not shown) communicatively connected via the network 500.

The input device 150 is a device used by personnel and others to input data into the accident cause estimation device 100, and functions as a user interface. Examples of the input device 150 that can be used include a keyboard, mouse, and microphone.

The output device 160 is a device for outputting information to the outside and functions as a user interface. Examples of the output device 160 that can be used include a display, printer, and speaker.

<Functional configuration>
7 shows a functional block diagram of the accident cause estimation device 100 according to this embodiment. The accident cause estimation device 100 includes a learning model storage unit 101, an accident information acquisition unit 102, an estimated cause acquisition unit 103, and an estimated cause output unit 104. These functions are realized by the hardware shown in FIG. 2 executing or processing the accident cause estimation device control program 700 according to this embodiment and various data.

The learning model storage unit 101 stores a learning model 620 that is created to output the estimated cause of a new accident when new accident information is input, based on past accident information including the type of accident that occurred in the power distribution system, the date and time the accident occurred, and the section where the accident occurred, as well as the cause of the accident.

In this embodiment, the learning model storage unit 101 is embodied as a storage device 140.

The accident cause estimation device 100, for example, inputs learning data 610 into the learning model 620, and then learns the learning model 620 by adjusting the parameters that define the learning model 620 using a method such as backpropagation based on the difference between the results output by the learning model 620 and the input learning data 610.

The accident information acquisition unit 102 acquires accident information about new accidents that have occurred in the power distribution system. The accident information includes the type of accident that occurred in the power distribution system, the date and time the accident occurred, and the section in which the accident occurred, but may also include other information.

For example, the accident information may include information regarding the inspection results of the distribution system equipment installed in the section where the accident occurred. Alternatively, the accident information may include information indicating the period of use of the distribution system equipment installed in the section where the accident occurred.

Furthermore, the accident information may include a predetermined type of weather information for the date and time the accident occurred. The accident information may also include information indicating geographically specific phenomena that exist in the area where the section where the accident occurred and that could be the cause of an accident, and information indicating seasonally specific phenomena that occur in the area where the section where the accident occurred and that could be the cause of an accident, and information indicating the times when these phenomena are likely to occur.

By including this information in the accident information, the accuracy of the learning model 620 is further improved, enabling the accident cause estimation device 100 to output a more accurate estimated cause of the accident.

The estimated cause acquisition unit 103 inputs accident information about a new accident that has occurred in the power distribution system into the learning model 620, and thereby acquires the estimated cause of the accident that is output from the learning model 620. The learning model 620 may output one or more estimated causes of the accident. In this manner, possible causes of the newly occurring accident can be appropriately presented to the person in charge. Furthermore, the estimated cause of each accident may include a calculation result of the probability that each estimated cause is the true cause. In this manner, the person in charge can be presented with the degree to which the accident cause estimated by the accident cause estimation device 100 is likely to occur.

The probable cause output unit 104 outputs the probable cause of the accident. The accident cause estimation device 100 outputs the probable cause of the accident by displaying a screen such as that shown in FIG. 10 on the output device 160.

The above-described aspects make it possible to accurately estimate the cause of an accident in a power distribution system. In this embodiment, the type of accident is either a broken wire, a short circuit, or a ground fault in the power distribution system, but the type of accident is based on the content included in the accident information obtained from the power distribution line automation system 200. Therefore, if the type of accident obtained from the power distribution line automation system 200 includes other information, the estimated cause output unit 104 may output that information as well. Alternatively, the type of accident may be either a ground fault or a short circuit in the power distribution system.

==Processing flow==
Next, with reference to Figures 8 and 9, we will explain the process in which the accident cause estimation device 100 of this embodiment uses learning data 610 to learn the learning model 620, and the process flow when using this learning model 620 to estimate the cause of an accident that occurred in a distribution system.

Figure 8 is a flowchart explaining the processing performed by the accident cause estimation device 100 when it uses the learning data 610 to train the learning model 620. Note that the timing at which the accident cause estimation device 100 performs learning of the learning model 620 is not necessarily limited. However, the accident cause estimation device 100 may perform the learning process of the learning model 620 when, for example, new accident information is added to the accident information management table 600, when an instruction to perform learning is received from a person in charge via a user interface, or at a predetermined timing that occurs at predetermined intervals (such as once a month or once a year).

First, the accident cause estimation device 100 generates data that associates accident information with the cause of the accident as learning data 610 (S1000). For example, the accident cause estimation device 100 performs the association by presenting the accident information (explanatory variables) to the person in charge and accepting from the person in charge the setting of the cause of the accident (objective variable (label)) for the accident information.

Figure 5 shows an example of a screen on which the accident cause estimation device 100 presents the above accident information (explanatory variables) and accepts the setting of the accident cause (objective variable).

Returning to Figure 8, the accident cause estimation device 100 performs a learning process for the learning model 620 using the learning data 610 (S1010).

The accident cause estimation device 100 may also verify the prediction accuracy of the trained learning model 620. In this case, the accident cause estimation device 100 classifies the learning data 610 into training data and verification data in advance, trains the learning model 620 using the training data, and verifies the learning model 620 using the verification data.

Figure 9 is a flowchart explaining the process performed by the accident cause estimation device 100 when estimating the cause of a new accident that has occurred in a power distribution system.

First, the accident cause estimation device 100 acquires accident information including the date, time, section, and type of accident that has newly occurred in the power distribution system (S2000). As mentioned above, this accident information may also include other information such as weather information, equipment inspection results, usage period, seasonal factors, and regional factors.

Next, the accident cause estimation device 100 inputs the acquired accident information into the learning model 620 and obtains the output of the learning model 620 as the cause of the accident (S2010).

The accident cause estimation device 100 then generates a screen listing the estimated cause of the accident, as shown in FIG. 10, and outputs the generated screen to the output device 160 (S2020).

This method makes it possible to accurately estimate the cause of an accident in a power distribution system.

Note that the accident cause estimation device 100 displays the probability that each candidate estimated cause output from the learning model 620 is the true cause on the display screen for estimated causes of accidents shown in Figure 10, but it may also be configured to display only estimated causes with a probability equal to or greater than a predetermined value (e.g., 30% or greater) on the screen. Alternatively, the accident cause estimation device 100 may be configured to allow the person in charge to select whether to display the probability that each candidate estimated cause is the true cause, or to display only estimated causes with a probability equal to or greater than a predetermined value (e.g., 30% or greater). Alternatively, the accident cause estimation device 100 may accept input of a probability value and display on the screen estimated causes with a probability equal to or greater than that value.

In this manner, possible causes of newly occurring accidents can be properly presented to the person in charge.

The accident cause estimation device 100, and the control method and program for the accident cause estimation device 100 according to this embodiment have been described above. According to this embodiment, it is possible to accurately estimate the cause of an accident in a power distribution system.

The above-described embodiments are intended to facilitate understanding of the present invention and are not intended to limit the scope of the present invention. The present invention may be modified or improved without departing from its spirit, and equivalents thereof are also included within the scope of the present invention.

100 Accident cause estimation device 101 Learning model storage unit 102 Accident information acquisition unit 103 Estimated cause acquisition unit 104 Estimated cause output unit 110 CPU
120 Memory 130 Communication device 140 Storage device 150 Input device 160 Output device 170 Recording medium reading device 200 Power distribution line automation system 300 Weather information providing device 500 Network 600 Accident information management table 610 Learning data 620 Learning model 621 Input layer 622 Intermediate layer 623 Output layer 630 Cause estimation result 700 Accident cause estimation device control program 800 Recording medium 1000 Accident cause estimation system

Claims (9)

An accident cause estimation device that outputs an estimated cause of an accident that has occurred in a distribution system that is divided into a plurality of sections by switches,
a learning model storage unit that stores a learning model created based on past accident information including the type of accident that occurred in the power distribution system, the date and time when the accident occurred, and the section where the accident occurred, and the cause of the accident, so as to output one or more presumed causes of the new accident and the probability that each of the presumed causes is the true cause when new accident information is input;
an accident information acquisition unit that acquires the new accident information;
an estimated cause acquisition unit that inputs the new accident information into the learning model and acquires one or more estimated causes of the accident output from the learning model and a probability that each of the estimated causes is a true cause ;
an estimated cause output unit that outputs one or more estimated causes of the accident and the probability that each of the estimated causes is the true cause, or outputs estimated causes whose probabilities are equal to or greater than a predetermined value, depending on a selection result of whether to output the probability that each of the estimated causes is the true cause or to output only estimated causes whose probabilities are equal to or greater than a predetermined value;
An accident cause estimation device comprising: 2. The accident cause estimation device according to claim 1,
The accident information further includes information regarding the inspection results of the equipment of the power distribution system installed in the section where the accident occurred.
Accident cause estimation device. 3. The accident cause estimation device according to claim 1 or 2,
The accident information further includes information indicating a period of use of the distribution system equipment installed in the section where the accident occurred.
Accident cause estimation device. The accident cause estimation device according to any one of claims 1 to 3,
The accident information further includes a predetermined type of weather information at the date and time when the accident occurred.
Accident cause estimation device. The accident cause estimation device according to any one of claims 1 to 4,
The accident information further includes information indicating a geographically specific phenomenon that may be a cause of the accident and exists in the area to which the section where the accident occurred belongs.
Accident cause estimation device. The accident cause estimation device according to any one of claims 1 to 5,
The accident information further includes information indicating seasonal phenomena that occur in the area to which the section in which the accident occurred and that may be the cause of the accident, and information indicating the time periods when the phenomena are likely to occur.
Accident cause estimation device. The accident cause estimation device according to any one of claims 1 to 6 ,
The type of the accident includes at least one of a break in the power distribution system, a short circuit, and a ground fault. A control method for an accident cause estimation device that outputs an estimated cause of an accident that has occurred in a power distribution system that is divided into multiple sections by switches, comprising:
The accident cause estimation device
storing a learning model created based on past accident information including the type of accident that occurred in the power distribution system, the date and time when the accident occurred, and the section where the accident occurred, and the cause of the accident, so as to output one or more presumed causes of the new accident and the probability that each of the presumed causes is the true cause when new accident information is input;
Acquire the new accident information;
inputting the new accident information into the learning model to obtain one or more presumed causes of the accident output from the learning model and a probability that each of the presumed causes is a true cause ;
A control method for an accident cause estimation device that outputs one or more presumed causes of the accident and the probability that each of the presumed causes is the true cause, or outputs presumed causes whose probability is equal to or greater than a predetermined value, depending on the selection result of whether to output the probability that each of the presumed causes is the true cause, or to output only presumed causes whose probability is equal to or greater than a predetermined value . A program for causing a computer to output a presumed cause of an accident that has occurred in a power distribution system that is divided into multiple sections by switches,
a step of storing a learning model created based on past accident information including the type of accident that occurred in the power distribution system, the date and time when the accident occurred, and the section where the accident occurred, and the cause of the accident, so as to output one or more presumed causes of the new accident and the probability that each of the presumed causes is the true cause when new accident information is input;
a step of acquiring the new accident information;
a step of inputting the new accident information into the learning model to obtain one or more presumed causes of the accident output from the learning model and a probability that each of the presumed causes is the true cause ;
a step of outputting one or more presumed causes of the accident and the probability that each of the presumed causes is the true cause, or outputting only presumed causes whose probability is equal to or greater than a predetermined value, depending on the result of selection of whether to output the probability that each of the presumed causes is the true cause, or to output only presumed causes whose probability is equal to or greater than a predetermined value;
A program for causing the computer to execute the above.