Accident point location device, learning device, accident point location system, accident point location method, and program

Abstract

An accident point location device (6-1) according to the present disclosure comprises: a data acquisition unit (61) that acquires measurement data, which is at least one of a current and a voltage measured at an electric-supply station when a system accident occurred in a power system; and a location result inference unit (62) that acquires an inference result for an accident point by inputting, into a trained model generated by machine learning for inferring the accident point from a feature amount including at least part of the measurement data which is measured at the electric-supply station and / or processed data which is calculated from at least part of said measurement data, a feature amount including at least part of the measurement data which was acquired by the data acquisition unit (61) when the system accident occurred and / or processed data which is calculated from at least part of said measurement data and that corrects the inference result using an estimation result for the cause of the system accident and correction information indicating the probability for the accident point determined in accordance with the cause of the accident.

Description

Fault point locating device, learning device, fault point locating system, fault point locating method and program

[0001] The present disclosure relates to a fault point locating device, a learning device, a fault point locating system, a fault point locating method, and a program for locating a fault point in a power system.

[0002] When an abnormal condition such as a short circuit or a ground fault (hereinafter referred to as a system fault or an accident) occurs in a power system such as a power transmission system or a power distribution system, it is desirable to locate the location of the accident, that is, the point where the accident occurred, in order to carry out recovery work from the accident.

[0003] Patent Document 1 discloses a fault section locating method in which a current sensor is provided for each of a plurality of steel towers to measure the current flowing in an overhead ground wire, and the fault section of a power transmission line is located by a neural network technique using current information acquired by the current sensor at the time of a fault.

[0004] Japanese Patent Application Laid-Open No. 2003-114249

[0005] When a fault occurs in a power system, it is thought that the impact on the current will differ depending on the cause of the fault. The fault section location method described in Patent Document 1 does not take the cause of the fault into consideration, so there is a possibility that the location accuracy will be low.

[0006] The present disclosure has been made in view of the above, and aims to provide a fault point locating device that can improve the accuracy of locating a fault point in a power system.

[0007] In order to solve the above-mentioned problems and achieve the objectives, the fault point location device of the present disclosure comprises: a data acquisition unit that acquires measurement data, which is at least one of current and voltage measured at a power station when a system fault occurs in an electric power system; and a location result inference unit that obtains an inference result of the fault point by inputting features including at least one of at least a portion of the measurement data acquired by the data acquisition unit when a system fault occurs and processed data calculated from at least a portion of the measurement data into a trained model generated by machine learning for inferring the fault point from features including at least a portion of the measurement data measured at the power station and processed data calculated from at least a portion of the measurement data, and corrects the inference result using an estimated result of the cause of the system fault and correction information indicating the probability of the fault point determined depending on the cause of the fault.

[0008] The fault point locating device according to the present disclosure has the effect of improving the accuracy of locating a fault point in a power system.

[0009] FIG. 1 is a diagram showing an example of the configuration of an accident point locating system according to the first embodiment. FIG. 2 is a diagram showing an example of the configuration of an accident point locating device according to the first embodiment. FIG. 3 is a diagram showing an example of the configuration of an accident point locating server according to the first embodiment. FIG. 4 is a diagram showing an example of the configuration of an accident point locating system according to the first embodiment. FIG. 5 is a diagram showing an example of the configuration of an accident point locating system according to the first embodiment.

[0010] An accident point locating device, a learning device, an accident point locating system, an accident point locating method, and a program according to embodiments will be described in detail below with reference to the accompanying drawings.

[0011] First Embodiment. FIG. 1 is a diagram illustrating an example of the configuration of a fault point locating system according to a first embodiment. The fault point locating system 10 of this embodiment locates a fault point in a power system. FIG. 1 also illustrates a simplified view of the power system to be subjected to fault point location, i.e., the power system to be subjected to fault point location. In the example shown in FIG. 1, the power system to be subjected to fault point location includes substations 2-1 to 2-3, such as a power plant and a substation. In FIG. 1, substations 2-1 and 2-2 are connected by an electric wire 1, such as a transmission line or a distribution line. Furthermore, the electric wire 1 branching off between substations 2-1 and 2-2 is connected to substation 2-3. Hereinafter, when referring to substations 2-1 to 2-3 without distinguishing them individually, they will be referred to as substations 2. Substations 2-1 to 2-3 are examples of one or more substations 2, and the number of substations 2 in the power system is not limited to the example shown in FIG. 1. Although Fig. 1 shows an example in which two electric power stations 2 are connected by two electric wires 1, the number of electric wires 1 connecting the electric power stations 2 is not limited to two. Furthermore, Fig. 1 is only an example, and the configuration of the electric power system that is the target of fault point location is not limited to the example shown in Fig. 1. Hereinafter, the electric power system that is the target of fault point location will also be simply referred to as the electric power system.

[0012] The electric power station 2-1 is provided with current transformers (CT) 3-1 and 3-2, which are an example of a current sensor that measures the current of the electric wire 1, and a voltage transformer (VT) 4-1, which is an example of a voltage sensor that measures the bus voltage of the electric power station 2-1. Similarly, the electric power station 2-2 is provided with CTs 3-3 and 3-4, and a VT 4-2. The electric power station 2-3 is provided with CTs 3-5 and 3-6, and a VT 4-3. Hereinafter, when CTs 3-1 to 3-6 are referred to without distinction, they will be referred to as CT3, and when VTs 4-1 to 4-3 are referred to without distinction, they will be referred to as VT4.

[0013] The fault point locating system 10 includes fault point locators 6-1 to 6-3 installed in the electric power stations 2-1 to 2-3, respectively, and a fault point locating server 8. Hereinafter, when the fault point locators 6-1 to 6-3 are referred to without being individually distinguished, they will be referred to as a fault point locator 6. The fault point locators 6-1 to 6-3 may have the same configuration, or some of the fault point locators 6-1 to 6-3 may not have the function of performing the fault point locating process described below. Furthermore, the fault point locators 6 may be installed in all electric power stations 2 in the power system that is the target of fault point locating, or only in some electric power stations 2. The fault point locators 6 can communicate with the fault point locating server 8 and the central monitoring system 7 via a communication network 9. The communication line between the accident point locating device 6 and the accident point locating server 8, and the communication line between the accident point locating device 6 and the central monitoring system 7 may be a wired line or a wireless line, or may be a mixture of wired and wireless lines.

[0014] The fault point locating server 8 is an example of a learning device that generates a trained model for locating the fault point. A location processing execution device, which is one of the fault point locating devices 6-1 to 6-3, locates the fault point using the trained model generated by the fault point locating server 8, corrects the fault point location result using correction information indicating the probability of the fault point determined according to the cause of the accident, and notifies the central monitoring system 7 of the corrected location result as the fault point location result. This embodiment thereby improves the accuracy of locating the fault point in the power system compared to when the cause of the accident is not considered. In the example shown in FIG. 1 , the fault point locating device 6-1 is the location processing execution device, but the location processing execution device may also be the fault point locating device 6-2 or 6-3.

[0015] The central monitoring system 7 is an EMS (Energy Management System) or a SCADA (Supervisory Control And Data Acquisition) system or a DAS (Distribution Automation System) that is installed in a central load control center, a main system load control center, a local load control center, or the like, and monitors and controls the power system.

[0016] In the example shown in FIG. 1 , the electric power station 2-1 is further provided with an accident cause estimation device 5. The accident cause estimation device 5 is capable of communicating with a central monitoring system 7 via a communication network 9. The communication line between the accident cause estimation device 5 and the central monitoring system 7 may be a wired line or a wireless line, or a combination of wired and wireless lines. The accident cause estimation device 5 is also capable of communicating with an accident point locator 6. The communication line between the accident cause estimation device 5 and the accident point locator 6 may be a wired line or a wireless line, or a combination of wired and wireless lines.

[0017] When an accident occurs, the accident cause estimation device 5 estimates the cause of the accident and transmits the estimation result to the central monitoring system 7 and the accident point locator 6. For example, the accident cause estimation device 5 acquires current and voltage from CTs 3-1, 3-2, VT 4-1, etc., performs clustering according to the cause of the accident based on the content rate of frequency components in waveform data of the current or voltage in the power system when past accidents occurred, calculates the average value of each cluster and the probability value of each cause of the accident as cluster data, and when an accident occurs, estimates the cause of the accident using the content rate of frequency components in the waveform data acquired at the time of the accident and the cluster data. The method of estimating the cause of the accident in the accident cause estimation device 5 is not limited to this example and may be any method.

[0018] 1 illustrates an example in which the accident cause estimation device 5 is provided, but as will be described later, the cause of an accident may be estimated manually when an accident occurs, in which case there is no need to provide the accident cause estimation device 5. Furthermore, in the example shown in FIG. 1, the accident cause estimation device 5 is provided in the electric power station 2-1, but the accident cause estimation device 5 may also be provided outside the electric power station 2-1.

[0019] The accident cause estimation device 5 may be integrated with the accident point locating device 6, or may be integrated with the accident point locating server 8. At least one of the accident cause estimation device 5 and the central monitoring system 7 may be included in the accident point locating system 10.

[0020] As described above, in this embodiment, the location processing execution device of the accident point locator 6 locates the accident point using the trained model generated by the accident point locator server 8 and corrects the location result of the accident point using correction information indicating the probability of the accident point determined according to the cause of the accident. Examples of the correction information include a probability distribution of arc resistance for each accident cause, and the probability of an accident occurring for each type of surrounding condition for each accident cause. For example, since the arc resistance value is thought to differ depending on the cause of the accident, the accuracy of locating the accident point can be improved by considering the probability distribution of arc resistance according to the cause of the accident. Furthermore, it is thought that there are accident causes that are more likely to occur depending on the type of surrounding condition of the power line 1. For example, tree contact is more likely to occur in forests than on flat land. Therefore, the accuracy of locating the accident point can be improved by performing correction taking into account the type of surrounding condition according to the "presumed cause of the accident." Examples of the type of surrounding condition include, but are not limited to, at least one of forest, flat land, mountainous area, coastal area, urban area, and farmland.

[0021] FIG. 2 is a diagram showing an example of the configuration of an accident point locator 6-1 according to this embodiment. As shown in FIG. 2, the accident point locator 6-1 includes a data acquisition unit 61, a location result inference unit 62, a storage unit 63, a communication unit 64, and a display data generation unit 65. FIG. 2 shows an example of the configuration of a location processing execution device, in which the accident point locator 6-1 is the location processing execution device. The location processing execution device is not limited to this, and may be the accident point locator 6-2 or the accident point locator 6-3, as described above. The location processing execution device performs a data acquisition process for acquiring measurement data and an accident point location process. Note that the accident point locator 6 other than the location processing execution device does not necessarily need to perform the accident point location process. It may be a device with the same configuration as the location processing execution device, or it may be a data acquisition device without the function of performing the accident point location process. This data acquisition device has a configuration similar to that of the accident point locator 6-1 shown in FIG. 2, with the location result inference unit 62 and the display data generation unit 65 removed. In addition, the fault point locators 6 other than the location processing execution device may have the same configuration as the location processing execution device, and the fault point locator 6 that serves as the location processing execution device may be switchable. In the following, it is assumed that all the fault point locators 6 provided in each electric power station 2 have the same configuration.

[0022] The data acquisition unit 61 acquires measurement data, which is at least one of the current and voltage measured at the substation 2 when a grid fault occurs in the power system. For example, the data acquisition unit 61 acquires at least the current and voltage measured at the substation 2-1 when the fault occurs, but may also acquire current and voltage measured at times other than when the fault occurs. Specifically, the data acquisition unit 61 acquires "voltage (actual)" from the VT 4-1 installed at the substation 2-1, acquires "current (actual)" from the CTs 3-1 and 3-2 installed at the substation 2-1, and stores the "voltage (actual)" and "current (actual)" as measurement data in the storage unit 63. The "voltage (actual)" is a measurement value obtained by the VT 4-1, and the "current (actual)" is a measurement value obtained by the CTs 3-1 and 3-2. Furthermore, the data acquisition unit 61 receives, from the other fault point locators 6, the "voltage (actual)" and "current (actual)" measured by the CT 3 and VT 4 at the electric power station 2 where the fault point locator 6 is installed, from the other fault point locators 6-2 and 6-3. The data acquisition unit 61 stores the received "voltage (actual)" and "current (actual)" as measurement data in the memory unit 63 for each corresponding electric power station 2. In the fault point locators 6-2 and 6-3, the communication unit 64 transmits the measurement data ("voltage (actual)" and "current (actual)") stored in the memory unit 63 to the fault point locator 6-1. The communication unit 64 may also receive the "voltage (actual)" and "current (actual)" from the other fault point locators 6 and store them in the memory unit 63. In this case, the communication unit 64 and the data acquisition unit 61 can be considered as a data acquisition unit in a broad sense.

[0023] The communication unit 64 communicates with external devices. For example, the communication unit 64 receives information such as grid-related information (described later) and a trained model for locating an accident point from the accident point locating server 8, and stores the received information in the storage unit 63. The communication unit 64 also transmits measurement data stored (saved) in the storage unit 63 to the accident point locating server 8. The communication unit 64 also transmits an accident point location result (described later) output from the location result inference unit 62 to the central monitoring system 7. The communication unit 64 also receives an accident cause estimation result from the accident cause estimation device 5, and stores the received accident cause estimation result in the storage unit 63.

[0024] The storage unit 63 stores system-related information, measurement data, accident cause estimation results, learned models, and calculation data. The system-related information is information indicating the configuration, state, characteristics, etc. of the power system (hereinafter also referred to as the system), and includes, for example, "system topology" indicating the system configuration, "impedance information" indicating the system impedance, "system surrounding situation information," and "statistical models." Note that the storage unit 63 in the fault point locator devices 6-2 and 6-3 does not need to store the system-related information, learned models, accident cause estimation results, and calculation data.

[0025] The "impedance information" is information indicating the impedance per unit length expressed in Ω / km, for example. In this case, the "system topology" includes distance information indicating the distance from the substation 2 to each point in the system. Alternatively, the "impedance information" may be information indicating the impedance itself of each point in the system. The "system surrounding condition information" includes the type of surrounding condition for each section (each position in the power system) described below, and a correction coefficient for each type of surrounding condition for each cause of the accident. The correction coefficient for each type of surrounding condition for each cause of the accident is information indicating the probability of an accident occurring for each type of surrounding condition for each cause of the accident, and is an example of correction information. The "statistical model" is a probability distribution of arc resistance for each cause of the accident, and is an example of correction information. Details of the "statistical model" will be described later.

[0026] As described above, the measurement data includes the "voltage (actual)" and "current (actual)" acquired within the electric power station 2-1 by the data acquisition unit 61, and the "voltage (actual)" and "current (actual)" measured at the other electric power station 2 and received by the data acquisition unit 61 from the other fault point locating device 6. The measurement data is stored in the storage unit 63 in association with the electric power station 2 where the measurement was performed. As described above, the trained model is received from the fault point locating server 8 by the communication unit 64. As described above, the accident cause estimation result is received from the accident cause estimation device 5 by the communication unit 64. Note that the accident cause estimation result may be manually input. In this case, an input receiving unit (not shown) of the fault point locating device 6-1 may receive the input, or another device such as the fault point locating server 8 may receive the input, and the communication unit 64 may receive the accident cause estimation result from the other device. The calculated data is data calculated by the accident point locating device 6-1, and includes, for example, an "accident point (calculated value)" which is an inference result of the accident point calculated by the location result inference unit 62.

[0027] The location result inference unit 62 acquires an inference result of the fault point by inputting feature quantities including the current and voltage acquired by the data acquisition unit 61 when a grid fault occurs into the trained model, and corrects the inference result using the estimated result of the fault cause of the grid fault and correction information indicating the probability of the fault point determined according to the fault cause. Specifically, the location result inference unit 62 outputs the "fault point (calculated value)" obtained by locating the fault point to the communication unit 64 as the fault point location result. Specifically, the location result inference unit 62 calculates "processed data" using the "voltage (actual)" and "current (actual)" at the time of the fault. Details of the "processed data" will be described later, but the "processed data" is calculated by calculation using at least one of the current and voltage measured at the power station 2. The location result inference unit 62 calculates the "processed data" in the same manner as the "processed data" calculated when the fault point location server 8 generates learning data. The location result inference unit 62 inputs the "voltage (actual)," "current (actual)," and "processed data" at the time of the accident as feature quantities into the trained model stored in the storage unit 63, thereby obtaining an inference result of the fault point as an output from the trained model. Note that the feature quantities are not limited to these and may include, for example, at least one of the "voltage (actual)," "current (actual)," and "processed data" at the time of the accident. In other words, the feature quantities may include at least one of at least a portion of measurement data, which is at least one of the voltage and current measured at the substation 2, and processed data calculated from at least a portion of the measurement data. For example, the feature quantities may be "current (actual)" and "processed data" obtained by processing the "voltage (actual)." The location result inference unit 62 then updates the "fault point (calculated value)," which is the inference result of the fault point obtained as an output from the trained model, by correcting the "fault point (calculated value)" using at least one of the "statistical model" and "surrounding situation information of the system." The time of accident occurrence is not limited to the time when the accident occurred, but may be, for example, a certain period including the estimated time of accident occurrence, which is the time when it is determined that an abnormality caused by the accident has occurred, but is not limited to this.For example, if the accident point cannot be identified during a patrol after the accident, and relocation (e.g., changing settings, recalculation) is performed, the location may be performed after a certain period of time, such as one week after the accident. Such cases are also included in the occurrence of the accident point. Furthermore, the certain period (data acquisition period) at the time of the accident when used as a feature may be different from the acquisition period of the measurement data at the time of the accident used to calculate the arc resistance value described below. The acquisition period of the data used as a feature is, for example, the period from before to after the accident, and may be several cycles to several tens of cycles, or may be a period of one minute or more, and is not particularly limited. The acquisition period of the measurement data used to calculate the arc resistance is, for example, a period during the accident (the period from immediately after the accident to the time the accident is cleared by opening and closing a circuit breaker), but is not limited to this as long as the arc resistance value can be calculated.

[0028] For example, the location result inference unit 62 corrects the "accident point (calculated value)" using the "statistical model", updates the "accident point (calculated value)" in the storage unit 63 with the corrected value, and further corrects the updated "accident point (calculated value)" using the "surrounding condition information of the system", and updates the "accident point (calculated value)" in the storage unit 63 with the corrected value. Here, an example will be described in which both correction using the "statistical model" and correction using the "surrounding condition information of the system" are performed, but this is not limiting, and at least one of correction using the "statistical model" and correction using the "surrounding condition information of the system" may be performed.

[0029] The display data generating unit 65 generates display data for displaying the location result of the fault point located by the location result inference unit 62 on a display screen. Specifically, the display data generating unit 65 generates display data for a display screen that displays the "fault point (calculated value)" stored as calculation data in the storage unit 63, and outputs the generated display data to the display device 66. The display screen that displays the "fault point (calculated value)" will be described later. The display device 66 displays the display screen based on the received display data. While FIG. 2 illustrates an example in which the display device 66 is directly connected to the fault point locating device 6-1, this is not limiting. Alternatively, the display data generating unit 65 may output the display data to the communication unit 64, and the communication unit 64 may transmit the display data to the display device 66. The display device 66 may be provided within the power station 2-1 or outside the power station 2-1. The display device 66 may be provided within the fault point locating device 6-1, within the fault point locating server 8, or as part of the central monitoring system 7. 2 shows an example in which the display data generating unit 65 is provided in the accident point locating device 6-1, but the display data generating unit 65 may be provided in the central monitoring system 7, or the display data generating unit 65 may be provided in the accident point locating server 8. When the display data generating unit 65 is provided in the accident point locating server 8, the "accident point (calculated value)" is transmitted from the accident point locating device 6-1 to the accident point locating server 8.

[0030] 3 is a diagram showing an example of the configuration of the accident point locating server 8 according to this embodiment. As shown in FIG. 3, the accident point locating server 8 includes a communication unit 81, a storage unit 82, an input receiving unit 83, a learning data generating unit 84, an arc resistance calculating unit 85, a statistical model generating unit 86, and a learning model generating unit 87.

[0031] The input accepting unit 83 accepts input from an operator or the like. For example, the input accepting unit 83 accepts input of system-related information such as "system topology," "impedance information," and "system surrounding situation information," and stores the input system-related information in the storage unit 82. Note that, although an example in which the input accepting unit 83 accepts input of system-related information has been described here, the present invention is not limited to this. For example, the input accepting unit 83 may be linked to a database of an external system present on the communication network 9, such as an EMS, SCADA, DAS, or digital twin system, so that the system-related information is automatically input. Furthermore, when an accident occurs, the input accepting unit 83 accepts input of accident-related information such as "accident point (actual)" and "accident cause (actual)" related to the accident, and stores the input accident-related information in the storage unit 82. While the example described above shows the input receiving unit 83 receiving input of accident-related information, the present invention is not limited to this. For example, the system may be linked to an external system on the communication network 9, such as an EMS, SCADA, DAS, digital twin system, an accident management system for managing accident information, or a database for a distribution facility maintenance system, so that accident-related information can be automatically input. The "accident point (actual)" and "accident cause (actual)" are the accident point and accident cause that were actually confirmed during a patrol after the accident occurred. The accident cause may include, for example, at least one of tree contact, bird and animal contact, metal contact, nesting material, lightning, snow, ice, wind and rain, etc. However, the classification of the accident cause is not limited to these and may be determined in any way.

[0032] The communication unit 81 communicates with an external device. For example, the communication unit 81 transmits, from among the system-related information stored in the storage unit 82, the "system topology," "impedance information," and "system surrounding situation information" to the fault point locator 6-1. The communication unit 81 also transmits, from among the calculated data stored in the storage unit 82, the "statistical model," to the fault point locator 6-1. The communication unit 81 also transmits, to the fault point locator 6-1, the trained model generated by the trained model generation unit 87 and stored in the storage unit 82. The communication unit 81 also receives measurement data (the "voltage (actual)" and the "current (actual)" at the time of the accident) from the fault point locator 6-1 and stores the received measurement data in the storage unit 82. The measurement data received by the communication unit 81 may be measurement data measured at the electrical power station 2-1 corresponding to the fault point locating device 6-1, or measurement data measured at the electrical power stations 2-1 to 2-3 corresponding to the fault point locating devices 6-1 to 6-3, i.e., measurement data measured at multiple electrical power stations 2.

[0033] The storage unit 82 stores system-related information, measurement data, accident-related information, learned models, and calculation data. The system-related information stored in the storage unit 82 includes the "system topology," "impedance information," and "system surrounding situation information" received by the input receiving unit 83, as well as an "instantaneous value analysis model" (described later) generated by the learning data generating unit 84. The measurement data includes the "voltage (actual)" and "current (actual)" at the time of the accident, which are received from the fault point locating device 6-1 by the communication unit 81. The accident-related information includes the "fault point (actual)" and "cause of accident (actual)" received by the input receiving unit 83, as well as the "arc resistance (calculated value)" (described later) calculated by the arc resistance calculation unit 85. The "voltage (actual)" and "current (actual)" are stored in association with each accident. For example, information about the same accident may be associated by adding the time of the accident or identification information of the accident to the information, or the measurement data and accident-related information may be stored as a set for each accident, or the association may be achieved by other methods. The trained model is generated by the trained model generation unit 87 as described below. The calculated data includes "current (calculated value)" and "voltage (calculated value)" calculated by a system simulation as described below, and a "statistical model" as described below generated by the statistical model generation unit 86.

[0034] The arc resistance calculation unit 85 calculates, for each of a plurality of system accidents that have already occurred in the power system, an arc resistance value corresponding to the system accident, using actual values ​​of measurement data ("current (actual)" and "voltage (actual)"), which are current and voltage measured at the sub-station 2-1 at the time of the accident, an actual value of the accident point of the identified system accident ("accident point (actual)"), and system-related information. Specifically, the arc resistance calculation unit 85 calculates an arc resistance value, which is the value of the arc resistance, using the "current (actual)," "voltage (actual)," and "accident point (actual)" for each of a plurality of past accidents that are stored in the storage unit 82, as well as the "system topology" and "impedance information," and stores the calculated arc resistance value as an "arc resistance (calculated value)" in the storage unit 82 as accident-related information.

[0035] For example, in the case of a short-circuit fault, the arc resistance calculation unit 85 calculates the impedance for each fault using the “current (actual)” and “voltage (actual)” at the time of the fault. If the calculated impedance is a first impedance, the first impedance is the original impedance up to the fault point plus the arc resistance (arc resistance due to the fault). The arc resistance calculation unit 85 calculates a second impedance using the “fault point (actual),” “system topology,” and “impedance information.” The second impedance is the original impedance, i.e., the impedance from the substation 2-1 to the fault point, and the arc resistance calculation unit 85 calculates the arc resistance value by subtracting the second impedance from the first impedance. Alternatively, the arc resistance calculation unit 85 may calculate the current and voltage at the substation 2-1 by sequentially changing the arc resistance value at the “fault point (actual)” using the “system topology,” “impedance information,” and “fault point (actual),” and then find the arc resistance value at which the calculated current and voltage are closest to the “current (actual)” and “voltage (actual).” In the case of a ground fault, the arc resistance calculation unit 85 calculates the arc resistance value by performing a similar calculation. Note that the method for calculating the arc resistance value can be a general method, and is not limited to the above example, and any method may be used. In addition, the ground resistance at the neutral point of the transformer may be taken into consideration.

[0036] In addition, here we will explain an example in which the arc resistance calculation unit 85 calculates the arc resistance (calculates the arc resistance value), but the arc resistance of each accident may be calculated by an external device or manually and input into the accident point location server 8, and the ``arc resistance (calculated value)'' corresponding to each accident may be stored in the memory unit 82 as accident-related information.

[0037] The statistical model generation unit 86 generates a "statistical model" using arc resistance values ​​and actual values ​​of accident causes corresponding to the arc resistance values ​​("accident causes (actual)") for multiple accidents. The statistical model generation unit 86 may generate a "statistical model" using "accident causes (estimated)" or may generate a "statistical model" using a mixture of "accident causes (actual)" and "accident causes (estimated)." Specifically, the statistical model generation unit 86 generates a "statistical model" indicating a probability distribution of arc resistance for each accident cause using the "accident causes (actual)" and "arc resistance (calculated value)" stored in the storage unit 82, and stores the generated "statistical model" in the storage unit 82 as calculated data. Details of the processing by the statistical model generation unit 86 will be described later.

[0038] The arc resistance calculation unit 85 and the statistical model generation unit 86 may calculate the arc resistance value and the statistical model using measurement data from one electric power station, i.e., the electric power station 2-1, or may calculate the arc resistance value and the statistical model using measurement data from multiple electric power stations 2. When calculating the arc resistance value using measurement data from multiple electric power stations 2, for example, the arc resistance calculation unit 85 calculates the arc resistance value using measurement data from the other electric power stations 2 in a manner similar to the case where the arc resistance calculation unit 85 calculates the arc resistance value using measurement data from the electric power station 2-1, which is the electric power station 2 at the power source end. Then, when generating the statistical model, the statistical model generation unit 86 may calculate statistical values ​​(such as average values, median values, and mode values) of the arc resistance values ​​corresponding to the multiple electric power stations 2 for each accident and generate the statistical model using the calculated statistical values. Alternatively, the statistical model generation unit 86 may generate the statistical model using all the calculated arc resistance values ​​(by treating the arc resistance values ​​calculated using measurement data from multiple electric power stations 2 that actually correspond to the same accident as virtually different accidents). Alternatively, the arc resistance calculation unit 85 may use a least squares method to find a solution by forming a plurality of equations used to calculate the arc resistance value into simultaneous equations. The method of calculating the arc resistance value and the statistical model using the measurement data of a plurality of electric power stations 2 is not limited to the above-described example, and any method may be used.

[0039] The learning data generation unit 84 sets multiple conditions including the accident point and accident aspect of a system accident in the power system, and calculates the calculated values ​​of the current and voltage measured at each of the power stations 2-1 to 2-3 ("current (calculated value)" and "voltage (calculated value)") by performing a system simulation under the multiple conditions, and generates learning data including the calculated values ​​and the set accident point.

[0040] For example, the learning data generation unit 84 uses the "system topology" and "impedance information" stored in the storage unit 82 to generate an "instantaneous value analysis model," which is a model of the power system for performing instantaneous value analysis, which is an example of a system simulation, and stores the generated model in the storage unit 82 as system-related information. The learning data generation unit 84 performs instantaneous value analysis using the "instantaneous value analysis model" stored in the storage unit 82 under various conditions, i.e., under a plurality of conditions in which at least one of the accident point, accident aspect, and accident cause differs, thereby calculating the "current (calculated value)" and "voltage (calculated value)" observed at the substation 2-1 when an accident occurs. Note that the "current (calculated value)" and "voltage (calculated value)" at the time of the accident are not limited to the "current (calculated value)" and "voltage (calculated value)" at the time of the accident, but are also the "current (calculated value)" and "voltage (calculated value)" for a certain period including the time of the accident. For the instantaneous value analysis, for example, EMTP (Electro-Magnetic Transient Program), XTAP (eXpandable Transient Analysis Program), PSCAD (registered trademark), and the like can be used, but are not limited to these. Here, an example of performing instantaneous value analysis as a system simulation will be described, but the system simulation may be a power flow calculation, a detailed stability calculation, or the like, and is not limited to instantaneous value analysis.

[0041] Furthermore, the learning data generation unit 84 calculates "processed data" using the calculated "current (calculated value)" and "voltage (calculated value)," and sets the "current (calculated value)," "voltage (calculated value)," and "processed data" together with the fault point set as a condition to form "learned data," and stores the "learned data" as calculated data in the storage unit 82. Note that the "processed data" may be calculated for the "current (calculated value)" and "voltage (calculated value)" corresponding to the electric power station 2-1, or may be calculated for the "current (calculated value)" and "voltage (calculated value)" corresponding to the electric power stations 2-1 to 2-3.

[0042] The "processed data" is data calculated using at least one of the "current (calculated value)" and the "voltage (calculated value)," and is, for example, at least one of the following: the total distortion rate, the third harmonic content rate, the crest factor, the maximum value (for example, the maximum effective value from the time the accident occurs until the time the accident is cleared (the circuit breaker is opened)), the amount of change over a certain period of time, the results of frequency component analysis, the impedance value (the impedance value calculated from the "current (calculated value)" and the "voltage (calculated value)"), the rate of change of current before and after the accident, and the rate of change of voltage before and after the accident, but is not limited to these.

[0043] The total distortion factor is the ratio of the distortion component of the entire zero-phase current waveform to the fundamental component, and is defined by the following formula (1): where THD is the total distortion factor (unit: %), and I １ is the fundamental component of the zero-phase current (unit: A), and I ｎ is the n-th harmonic component of the zero-phase current (unit: A), and N is the maximum harmonic order to be considered. Note that although the definition of the total distortion factor of current is given here, the total distortion factor of voltage can also be defined in a similar manner.

[0044]

[0045] The third harmonic content is the ratio of the third harmonic component contained in the zero-phase current to the fundamental component, and is defined by the following equation (2): ３ＨＣ is the third harmonic content (unit: %), and I １ is the fundamental component of the zero-phase current (unit: A), and I ３is the third harmonic component of the zero-phase current (unit: A). Note that although the definition of the third harmonic content of the current is shown here, the third harmonic content of the voltage can also be defined in the same way.

[0046]

[0047] The crest factor is the ratio of the peak value to the effective value of the zero-phase current, and is defined by the following equation (3): where CFI is the crest factor (unit: dimensionless), and I ｍａｘ is the peak value of the zero-phase current (unit: A), and I ｒｍｓ is the effective value of the zero-phase current (unit: A). ｒｍｓ is the k-th instantaneous value data of "current (calculated value)" ｋ Then, it can be calculated by the following formula (4): where M is the number of instantaneous value data (number of sampling points). Note that although the definition of the current crest factor has been shown here, the voltage crest factor can also be defined in the same way.

[0048]

[0049] The learning model generation unit 87 uses the learning data to generate a trained model for inferring the fault point from feature quantities including the current and voltage measured at the electric power stations 2-1 to 2-3 through machine learning. For example, the learning model generation unit 87 generates a trained model by performing supervised machine learning using multiple sets of "training data" in which "current (calculated value)," "voltage (calculated value)," and "processed data" are feature quantities and "fault point (set)" is the answer (correct answer), and stores the generated trained model in the storage unit 82. Note that the feature quantities are not limited to these and may include, for example, at least one of "current (calculated value)," "voltage (calculated value)," and "processed data." Note that the "fault point (set)" is a value set as a condition for instantaneous value analysis. Note that, for example, machine learning (classification), i.e., machine learning using a classification method (supervised machine learning using a classification method), can be used as the machine learning. However, this is not limited thereto, and machine learning (regression), i.e., machine learning using a regression method, may also be used. Specifically, for example, a random forest, gradient boosting, a neural network, a support vector machine, or the like may be used as a machine learning algorithm, or other algorithms may be used.

[0050] As described above, the trained model generated by the trained model generation unit 87 is transmitted to the accident point locator 6-1. When an accident occurs, the accident point locator 6-1 uses the trained model to locate the accident point.

[0051] FIG. 4 is a diagram illustrating an example of a location result when a machine learning classification method is used to locate an accident point according to this embodiment. FIG. 5 is a diagram illustrating an example of a location result when a machine learning regression method is used to locate an accident point according to this embodiment. In FIGS. 4 and 5 , an occurrence point 200 indicates the location where an accident occurred. As shown in FIG. 4 , when machine learning classification is used, the accident point locator 6-1 can calculate the probability of each section into which the electric wire 1 is divided (the probability of an accident point existing in each section). In other words, a trained model generated using machine learning classification is a trained model that can output information indicating the probability of an accident point existing in each section in the system as an inference result of the accident point. Note that the sections may be defined in any way, such as dividing the electric wire 1 so that the distance therebetween is equal, or may be determined taking into account equipment such as circuit breakers in the power system, or may be divided in some other way.

[0052] On the other hand, when using machine learning based on a regression method, although it is not possible to calculate probability, it is possible to pinpoint the location of the accident point, as shown in Figure 5. The machine learning method to use can be determined by considering the advantages and disadvantages of each method, but below, an example using machine learning based on a classification method will be mainly described as an example that emphasizes the advantage of being able to assign probability. When machine learning based on a classification method is used, the probability of each section is calculated, as shown in Figure 4, so that, for example, when working on-site to respond to an accident, it is possible to determine which direction to search for the accident point from the section with the highest probability, which is the most efficient direction.

[0053] FIG. 6 is a diagram illustrating an example of applying machine learning based on a classification method to the accident point location system 10 of this embodiment. In FIG. 6, dashed arrows indicate the flow of data in the learning process (learning phase), and solid arrows indicate the flow of data in the inference process (inference phase). In this embodiment, the learning process is performed by the accident point location server 8, and the inference process is performed by the accident point locator 6-1. In the learning process, "learning data" is generated using the analysis results of a system simulation, such as instantaneous value analysis, as described above. Note that the method of generating "learning data" is not limited to this. If there is sufficient information about past accidents, actual values ​​for each accident corresponding to multiple past accidents may be used, or both actual values ​​and analysis results from the system simulation may be used. In the inference process, information about the accident to be located, i.e., information obtained at the time of the accident, is input into a trained model to infer the accident point and obtain an accident point location result. Since FIG. 6 illustrates an example of applying machine learning based on a classification method, the probability of an accident point in each section is obtained as the accident point location result. 6 shows an example in which a random forest is used as the machine learning algorithm, but as described above, the machine learning algorithm is not limited to this. Also, as described above, in this embodiment, correction is performed on the fault point location result obtained as the output of the trained model.

[0054] Here, an example will be described in which the accident point locating server 8 performs the learning process and the accident point locating device 6-1 performs the inference process. However, the present invention is not limited to this. For example, the accident point locating device 6-1 may perform both the learning process and the inference process. That is, the accident point locating device 6-1 may have the same components as the accident point locating server 8 and may also function as a learning device. Furthermore, the accident point locating server 8 may have a location result inference unit 62 and may perform both the learning process and the inference process. In this case, the communication unit 81 functions as a data acquisition unit, and receives "current (actual)" and "voltage (actual)" from the power stations 2-1 to 2-3 when an accident occurs. For example, a communication device (not shown) installed in each of power stations 2-1 to 2-3 may acquire the "current (actual)" and "voltage (actual)" from CT3 and VT4 and transmit the "current (actual)" and "voltage (actual)" to the fault point locating server 8, or CT3 and VT4 may each have a communication function and transmit the "current (actual)" and "voltage (actual)" to the fault point locating server 8, or the "current (actual)" and "voltage (actual)" may be transmitted to the fault point locating server 8 by a method other than those mentioned above.

[0055] Next, the operation of this embodiment will be described. FIG. 7 is a sequence diagram showing an example of the fault point locating process in the fault point locating system 10 of this embodiment. As shown in FIG. 7, the fault point locating server 8 accepts input of grid-related information (step S1) and stores the grid-related information (step S2). Specifically, the input receiving unit 83 accepts input of "grid topology," "impedance information," and "grid surrounding situation information," and stores the input grid-related information in the storage unit 82. As described above, the grid-related information may be automatically input in cooperation with an external system present on the communication network 9, such as a database of an EMS, SCADA, DAS, or digital twin system.

[0056] The fault point locating server 8 transmits the grid-related information to the fault point locating device 6-1 (step S3). Specifically, the communication unit 81 transmits the "grid topology," "impedance information," and "grid surrounding situation information" stored in the storage unit 82 to the fault point locating device 6-1. The fault point locating device 6-1 stores the grid-related information (step S4). Specifically, the communication unit 64 receives the "grid topology," "impedance information," and "grid surrounding situation information" from the fault point locating server 8, and stores the received "grid topology," "impedance information," and "grid surrounding situation information" in the storage unit 63 as grid-related information. The above steps S1 to S4 are pre-processing for setting grid conditions. Note that these processes are performed again if the grid conditions change.

[0057] The fault location server 8 acquires a "statistical model" for each cause of the accident (step S5). For example, the fault location server 8 calculates an arc resistance value using measurement data ("current (actual)" and "voltage (actual)") corresponding to past accidents, the "fault point (actual)," and the grid-related information. The fault location server 8 then acquires a "statistical model" that indicates a probability distribution of arc resistance for each cause of the accident using the calculated arc resistance value and the "cause of the accident (actual)." It is assumed that the measurement data and accident-related information from past accidents are stored in the storage unit 82 by processing similar to steps S16 and S17 described below. Here, an example is described in which a "statistical model" indicating a probability distribution of arc resistance is generated as correction information. However, as described above, "system surrounding situation information," i.e., information indicating the probability of an accident point depending on the surrounding situation for each cause of the accident, may also be used as correction information. Details of step S5 will be described later.

[0058] The fault point locating server 8 generates "learned data" through a system simulation (step S6). For example, the fault point locating server 8 performs a system simulation under various conditions to calculate the "current (calculated value)" and "voltage (calculated value)" observed at each of the substations 2-1 to 2-3 when an accident occurs, and generates "learned data" using the "current (calculated value)" and "voltage (calculated value)." The fault point locating server 8 generates a trained model using the generated "learned data" (step S7). The fault point locating server 8 transmits the trained model and the "statistical model" to the fault point locating device 6-1 (step S8). The fault point locating device 6-1 stores the received trained model and "statistical model" (step S9). In detail, the communication unit 64 receives the trained model and the "statistical model," stores the trained model in the storage unit 63, and stores the "statistical model" in the storage unit 63 as system-related information. The timing of transmission of the trained model and the "statistical model" from the accident point location server 8 is not limited to the example shown in FIG. 7 , and the trained model and the "statistical model" may be transmitted at different times. For example, the trained model may be transmitted when the trained model is updated, and the "statistical model" may be transmitted when the statistical model is updated. Details of steps S6 to S9 will be described later.

[0059] The above steps S6 to S9 are part of the learning process, which may be performed periodically, for example, once a month, when the "system topology" or the like changes, or at other times. Here, an example is shown in which the "statistical model" is generated together with the learning process. However, since the "statistical model" is generated for correction purposes (described later), it may be generated at a different time from the learning process. The "statistical model" may also be generated by the fault point locator 6-1. In this case, the fault point locator 6-1 includes an arc resistance calculator 85 and a statistical model generator 86, which generates the "statistical model" and stores the generated "statistical model" in the storage unit 63 as system-related information.

[0060] When a grid fault occurs, an inference process is performed. Specifically, the fault point locator 6-1 acquires measurement data at the time of the fault (step S11) and inputs feature quantities including the acquired measurement data into the trained model to obtain an inference result of the fault point (step S12). The fault point locator 6-1 may also acquire an estimation result of the cause of the accident in step S11, or may acquire the estimation result of the cause of the accident separately from step S11. The estimation result of the cause of the accident only needs to be acquired before the correction process. The measurement data acquired in step S11 includes measurement data at the electric power station 2-1 and measurement data at the electric power stations 2-2 and 2-3 received from the fault point locators 6-2 and 6-3. The fault point locator 6-1 performs a correction process on the inference result of the fault point (step S13) and transmits the corrected inference result of the fault point to the central monitoring system 7 as the fault point locator result (step S14). As described above, the correction process includes, for example, correction using a "statistical model" and processing using "information on the surrounding conditions of the grid." The above steps S11 to S14 are processing in the inference process. Details of steps S11 to S14 will be described later.

[0061] The accident point locating device 6-1 also transmits measurement data (measurement data at the time of the accident) to the accident point locating server 8 (step S15). The accident point locating server 8 accepts input of accident-related information (step S16) and stores the measurement data and the accident-related information in association with each other (step S17). As described above, the accident-related information may be automatically input by linking with external systems present on the communication network 9, such as EMS, SCADA, DAS, digital twin systems, accident management systems for managing accident information, and databases for distribution facility maintenance systems. Steps S15 to S17 are processes for storing measurement data and accident-related information from past accidents in the storage unit 82, which the accident point locating server 8 uses to generate a "statistical model." In FIG. 7, steps S15 to S17 are performed after an accident occurs, but steps S15 to S17 do not have to be performed for each accident. Since the learning process does not need to be performed frequently, the "current (actual)" and "voltage (actual)" related to past accidents do not need to be acquired frequently. For example, the "current (actual)" and "voltage (actual)" for a certain period of time may be transmitted in bulk from the accident point locating device 6-1 to the accident point locating server 8 and stored in the memory unit 82.

[0062] The accident point locating device 6-1 also performs a display process (step S18). Specifically, for example, the display data generating unit 65 generates display data showing a display screen for displaying the accident point locating result, i.e., the "accident point (calculated value)," and outputs the generated display data to the display device 66. Note that step S18 does not necessarily have to be performed. As described above, the display data may be generated by the accident point locating server 8.

[0063] Next, the details of the "statistical model" acquisition process, i.e., step S5 shown in FIG. 7, will be described. FIG. 8 is a flowchart showing an example of the "statistical model" acquisition process according to this embodiment. The fault point location server 8 calculates and stores an "arc resistance (calculated value)" using the "measurement data (actual results)," "fault point (actual results)," "system topology," and "impedance information" at the time of the accident (step S21). Specifically, the arc resistance calculation unit 85 calculates an arc resistance value using the "current (actual results)," "voltage (actual results)," and "fault point (actual results)" for each of multiple past accidents, as well as the "system topology" and "impedance information," which are stored in the storage unit 82, and stores the calculated arc resistance value as the "arc resistance (calculated value)" in the storage unit 82 as accident-related information.

[0064] Next, the accident point location server 8 generates and stores a "statistical model" from the "accident cause (actual)" and the "arc resistance (calculated value)" (step S22). Specifically, the statistical model generation unit 86 generates a "statistical model" indicating the probability distribution of arc resistance for each accident cause using the "accident cause (actual)" and the "arc resistance (calculated value)" stored in the storage unit 82, and stores the generated model as calculated data in the storage unit 82. As described above, the statistical model generation unit 86 may generate a "statistical model" using the "accident cause (estimated)" or may generate a "statistical model" using a mixture of the "accident cause (actual)" and the "accident cause (estimated)." Specific examples of the "statistical model" include a histogram, or a probability density function, probability distribution function, cumulative distribution function, or the like obtained from the histogram using a parametric or non-parametric method. An example of a parametric method is the least squares method, and an example of a non-parametric method is the kernel density estimation method, but the method used to generate the "statistical model" may be any method and is not limited to these.

[0065] Next, the details of the "learning data" generation process and the learned model generation process, i.e., steps S6 to S9 shown in FIG. 7, will be described. FIG. 9 is a flowchart showing an example of the "learning data" generation process and the learned model generation process of this embodiment. The fault point location server 8 generates and stores an "instantaneous value analysis model" of the power system using the "system topology" and "impedance information" (step S31). In detail, the learning data generation unit 84 uses the "system topology" and "impedance information" stored in the storage unit 82 to generate an "instantaneous value analysis model" of the power system for performing instantaneous value analysis, which is an example of a system simulation, and stores the generated model in the storage unit 82 as system-related information. As described above, the system simulation is not limited to instantaneous value analysis.

[0066] The accident point locating server 8 sets various conditions (accident points and accident aspects) (step S32). Specifically, the learning data generating unit 84 sets a plurality of conditions of the accident points and accident aspects to be set as conditions for the instantaneous value analysis so that at least one of the accident points and accident aspects is different for each condition.

[0067] The fault point location server 8 performs instantaneous value analysis using the "instantaneous value analysis model" to calculate the current and voltage under various conditions (step S33). In detail, the learning data generation unit 84 performs instantaneous value analysis using the "instantaneous value analysis model" stored in the storage unit 82 under various conditions to calculate the current and voltage observed at the substations 2-1 to 2-3 when the fault occurs for each condition.

[0068] The fault point location server 8 calculates "processed data" using the "current (calculated value)" and "voltage (calculated value)" (calculated current and voltage), associates the "current (calculated value)," "voltage (calculated value)," and "processed data" with the conditions (fault point) of the instantaneous value analysis, and stores them as "learned data" (step S34). In detail, the learning data generation unit 84 calculates "processed data" using the calculated "current (calculated value)" and "voltage (calculated value)," and sets the "current (calculated value)," "voltage (calculated value)," and "processed data" together with the conditions (fault point) as "learned data," and stores the "learned data" in the storage unit 82 as calculated data.

[0069] Note that, although the above description has been given of an example in which the learning data generation unit 84 generates "learning data" by system simulation, the present invention is not limited to this. If there is sufficient information about past accidents, "processed data" may be calculated using the "current (actual)" and "voltage (actual)" corresponding to each of a plurality of past accidents, and the situations at the time of the plurality of past accidents (accident points (actual)) may be associated with the "current (actual)," "voltage (actual)," and "processed data," and the "learning data" may be stored in the storage unit 82 as "learning data." Furthermore, the learning data generation unit 84 may generate both "learning data" using calculated values ​​obtained by system simulation and "learning data" using actual values, and store a mixture of these in the storage unit 82.

[0070] The fault point location server 8 performs machine learning using "learning data" in which the "current (calculated value)," "voltage (calculated value)," and "processed data" are used as feature quantities and the corresponding "fault point (setting)" is used as an answer, to generate and store a trained model (step S35). Note that the feature quantities are not limited to these and may include, for example, at least one of the "voltage (calculated value)," "current (calculated value)," and "processed data" at the time of the accident. That is, the feature quantities may be only the "voltage (calculated value)," only the "current (calculated value)," only the "processed data," or two of the "voltage (calculated value)," "current (calculated value)," and "processed data." In detail, the training model generation unit 87 performs supervised machine learning using multiple sets of "learning data" in which the "current (calculated value)," "voltage (calculated value)," and "processed data" stored in the storage unit 82 are used as feature quantities and the "fault point (setting)" is used as an answer (correct answer), thereby generating a trained model, and stores the generated trained model in the storage unit 82.

[0071] The accident point locating server 8 transmits the trained model to the accident point locating device 6-1 (step S36). Specifically, the communication unit 81 transmits the trained model stored in the storage unit 82 to the accident point locating device 6-1. Upon receiving the trained model, the accident point locating device 6-1 stores the trained model in the storage unit 63.

[0072] Next, the processing of the inference process, i.e., steps S11 to S14 shown in FIG. 7, will be described in detail. FIG. 10 is a flowchart showing an example of the processing of the inference process in this embodiment. The fault point locator 6-1 acquires and stores "current (actual)" and "voltage (actual)" from the CTs 3 and VTs 4 of the substations 2-1 to 2-3 (step S41). Specifically, the data acquisition unit 61 acquires the measured values ​​from the CTs 3-1, 3-2, and VTs 4-1 at the time of the accident from the CTs 3-1, 3-2, and VTs 4-1 and stores them as measurement data in the storage unit 63. The data acquisition unit 61 also receives "current (actual)" and "voltage (actual)" from the fault point locators 6-2 and 6-3, respectively, and stores them as measurement data in the storage unit 63.

[0073] The accident point locator 6-1 acquires and stores the "accident cause (estimated)" from the accident cause estimation device 5 (step S42). In particular, the communication unit 64 receives the accident cause estimation result ("accident cause (estimated)") from the accident cause estimation device 5 and stores the received accident cause estimation result in the storage unit 63. As described above, the accident cause estimation result is not limited to being received from the accident cause estimation device 5, but may be input to the accident point locator 6-1, or may be input to a device other than the accident point locator 6-1 and transmitted from that device to the accident point locator 6-1.

[0074] The fault point locating device 6-1 generates "processed data," inputs the "current (actual)," "voltage (actual)," and "processed data" into the trained model, and stores the "fault point (calculated value)" output from the trained model (step S43). In detail, the location result inference unit 62 generates the "processed data" using the "voltage (actual)" and "current (actual)" at the time of the accident that are stored as measurement data in the storage unit 63. Then, the location result inference unit 62 inputs the "current (actual)," "voltage (actual)," and "processed data" as feature quantities into the trained model stored in the storage unit 63, and stores the "fault point (calculated value)," which is the output of the trained model, as calculated data in the storage unit 63.

[0075] The fault point locator 6-1 calculates and stores the probability distribution of distance using the "statistical model" (step S44). Specifically, the location result inference unit 62 uses the "statistical model" and the "accident cause (estimate)" to determine the probability distribution of arc resistance corresponding to the "accident cause (estimate)," and converts the probability distribution of arc resistance into a probability distribution of distance using the "system topology" and "impedance information," and stores the converted probability distribution of distance in the storage unit 63 as calculated data.

[0076] The accident point locating device 6-1 corrects the "accident point (calculated value)" using the "accident point (calculated value)" and the probability distribution of the distance, thereby updating and saving the "accident point (calculated value)" (step S45). In detail, the location result inference unit 62 shifts the "accident point (calculated value)" stored in the storage unit 63 using the probability distribution of the distance stored in the storage unit 63. For example, if the probability distribution of the distance is expressed as a probability for each of m discrete distances, the location result inference unit 62 multiplies the "accident point (calculated value)" by the probability corresponding to the distance of each point, performs processing to shift the "accident point (calculated value)" by the distance of each point, and adds up the processing results for m points, thereby shifting the "accident point (calculated value)" using the probability distribution of the distance.

[0077] Next, the accident point locating device 6-1 corrects the "accident point (calculated value)" by using the "accident cause (estimated)" and the "system surrounding situation information", thereby updating and storing the "accident point (calculated value)" (step S46). In detail, the location result inference unit 62 corrects the "accident point (calculated value)" by using the "accident cause (estimated)" and the "system surrounding situation information" stored in the storage unit 63, and updates the "accident point (calculated value)" stored in the storage unit 63 to the corrected "accident point (calculated value)", thereby storing the corrected "accident point (calculated value)" in the storage unit 63.

[0078] The "system surrounding condition information" includes, for example, the type of surrounding condition for each section and a correction coefficient for each type of surrounding condition for each cause of the accident. FIG. 11 is a diagram showing an overview of the correction process using the "system surrounding condition information" of this embodiment. In the example shown in FIG. 11, of sections #01 to #14, the surrounding conditions of sections #01 to #03 and #09 to #14 correspond to flat land, and the surrounding conditions of sections #04 to #08 correspond to forest. In the example shown in FIG. 11, the "accident cause (estimated)" is assumed to be tree contact, and the correction coefficient corresponding to tree contact is 1 for flat land and 3 for forest. FIG. 11 shows an example in which machine learning based on classification is used as machine learning, and the "accident point (calculated value)" is the probability of an accident point being present in each section.

[0079] As shown in the left diagram of Figure 11, the "fault point (calculated value)" obtained by the processing of step S43 is assumed to be 2.5% in section #04, 7.5% in section #05, 35% in section #06, 5% in section #07, 2.5% in section #10, 7.5% in section #11, 35% in section #12, and 5% in section #13. Note that if location is performed using only impedance information for a power system (transmission line, distribution line) with branches, the branched regions will be equivalent in terms of electrical calculations, and so the location result of the fault point may result in two regions as shown here.

[0080] The location result inference unit 62 multiplies the probability of each section by a correction coefficient corresponding to the type of surrounding conditions of each section, which differs for each accident cause. FIG. 11 shows an example in which the cause of the accident is estimated to be tree contact. For example, because section #04 is a forest, the location result inference unit 62 multiplies 2.5, the value obtained as the "accident point (calculated value)," by a correction coefficient of 3 to obtain a calculation result of 7.5. Furthermore, because section #10 is flat land, the location result inference unit 62 multiplies 2.5, the value obtained as the "accident point (calculated value)," by a correction coefficient of 1 to obtain a calculation result of 2.5. In this way, the location result inference unit 62 performs similar calculations for each section to obtain calculation results, and then normalizes the total of the calculation results (200 in the example shown in FIG. 11 ) to 100. By performing normalization, the total probability of the existence of an accident point at all points can be set to 100%. Note that this normalization is not essential and is performed to make the location results easier to view. Therefore, it does not have to be performed. Instead, the value obtained by multiplying the probability of each section by a correction coefficient may be used as the corrected "accident point (calculated value)." Note that the number of sections, types of surrounding conditions, and values ​​of correction coefficients shown in FIG. 11 are merely examples, and the number of sections, types of surrounding conditions, and values ​​of correction coefficients are not limited to the examples shown in FIG. 11. For example, the value of the correction coefficient (e.g., in the case of contact with trees, the correction coefficient for forest is set to 3 and the correction coefficient for flat land is set to 1) may be determined for each system, or may be determined by weighting each section. Conversely, the same correction coefficient may be used for each system and for each accident type. In this way, by performing correction according to the type of surrounding conditions, the accuracy of accident point location can be improved compared to when correction is not performed.

[0081] In the above example, the "system surrounding situation information" is input to the accident point locating server 8, but this is not limiting, and the accident point locating server 8 may calculate the "system surrounding situation information" using the "accident point (record)" and "accident cause (record)" related to past accidents and the type of surrounding situation of each section. Also, in the above example, an example is shown in which the "system surrounding situation information" is calculated using machine learning, but this is not limiting, and the information may be set by a person based on experience, etc.

[0082] Returning to the explanation of Fig. 10, after step S46, the fault point locator 6-1 transmits the "fault point (calculated value)" to the central monitoring system 7 as the fault point locating result (step S47). In detail, the communication unit 64 transmits the "fault point (calculated value)" stored in the storage unit 63 to the central monitoring system 7 as the fault point locating result.

[0083] Furthermore, when using machine learning based on a regression method, information that pinpoints the accident point, such as X km (X is a real number) from the substation 2-1, is obtained as the "accident point (calculated value)." In this case, too, the accident point locator 6-1 can perform corrections using a "statistical model" and "information about the surrounding conditions of the power system." For example, when performing corrections using a "statistical model," the location result inference unit 62 converts the "statistical model" into a probability distribution of distances, as in the case of using machine learning based on a classification method, and shifts the "accident point (calculated value)" by the amount of the probability distribution of distances. When performing corrections using "information about the surrounding conditions of the power system," for example, as in the case of using machine learning based on a classification method, the power system may be divided into sections, and the type of surrounding conditions for each section may be included in the "information about the surrounding conditions of the power system." Alternatively, information indicating the type of surrounding conditions for each range of each power line 1 may be included in the "information about the surrounding conditions of the power system," such as indicating that the area between Y1 km and Y2 km (Y1 and Y2 are real numbers) on the A line 1L is a forest. The A-line 1L is an example of identification information of the electric wire 1.

[0084] Next, an example of a display screen corresponding to the display data generated by the display process of step S18 in FIG. 7, i.e., an example of a display screen displayed on the display device 66, will be described. When machine learning classification is used, the display screen may display, for example, the probability of an accident point existing for each section. FIG. 12 is a diagram showing a first example of a display screen according to this embodiment. In the display screen shown in FIG. 12, "accident points (calculated values)" are displayed in a list format indicating the probability of an accident point existing for each section. The display data generation unit 65 may generate display data for displaying a display screen indicating the probability of an accident point existing for each section in a list format, as shown in FIG. 12.

[0085] FIG. 13 is a diagram showing a second example of a display screen according to the present embodiment. In the display screen shown in FIG. 13, an "accident point (calculated value)" is shown on a system diagram. Specifically, in the example shown in FIG. 13, the probability corresponding to each section is shown at a position corresponding to each section of the "accident point (calculated value)" on the system diagram (for example, the position of a representative point of each section). The representative point may be the center of each section, the start point of each section (the end on the side of the substation 2-1), the end point of each section (the end on the opposite side from the substation 2-1), or a location other than these. The display data generating unit 65 may generate display data for displaying a display screen showing the "accident point (calculated value)" on a system diagram, as shown in FIG. 13.

[0086] FIG. 14 is a diagram illustrating a third example of a display screen according to the present embodiment. In the display screen illustrated in FIG. 14 , similar to the example illustrated in FIG. 13 , the “fault point (calculated value)” is displayed on a system diagram, but a different graphic is displayed at the corresponding point in the power system according to the numerical value corresponding to the “fault point (calculated value).” In the example illustrated in FIG. 14 , the probability is divided into levels, such as 10% or less, over 10% to 20%, over 20% to 30%, and so on, and the larger the graphic, the higher the probability level. That is, the graphic is displayed in a different manner for each level. Note that the definition of the level division and the graphic corresponding to each level are not limited to this example. For example, a graphic of a different color may be displayed for each level. Furthermore, the numerical value corresponding to the “fault point (calculated value)” itself may be displayed so that the larger the graphic, the higher the probability level, or the different colors may be used for each level.

[0087] FIG. 15 is a diagram illustrating a fourth example of a display screen according to the present embodiment. In the display screen illustrated in FIG. 15 , an “accident point (calculated value)” is displayed on a map. Specifically, in the example illustrated in FIG. 15 , the probability corresponding to each section of the “accident point (calculated value)” is displayed at a geographical position corresponding to a representative point of each section on the map. As in the second example, the representative point may be the center of each section, the start point of each section, or the end point of each section, or a location other than these. For example, the display data generating unit 65 may include information (map information) indicating the geographical location of each electric wire 1 in the “system topology,” calculate the geographical location of the representative point in the system diagram using the information indicating the geographical location of each electric wire 1, and generate display data for displaying a display screen illustrating the “accident point (calculated value)” on a map, as illustrated in FIG. 15 . When the “accident point (calculated value)” is displayed on a map, the probability may be divided into levels, and each level may be displayed using a different shape or color, as in the example illustrated in FIG. 14 .

[0088] The display screens described above are merely examples, and the display screens are not limited to the above examples. Furthermore, the specific display positions and contents of each display element are not limited to the examples shown in FIGS. 12 to 15 . Furthermore, the display data generation unit 65 may generate two or more display screens (individual display screens) from the first, second, third, and fourth examples, and generate display data so that the two or more generated individual display screens are displayed simultaneously. Furthermore, the two or more generated individual display screens may be displayed sequentially by switching.

[0089] Next, the hardware configuration of the accident point locator 6-1 and the accident point locator server 8 will be described. In the present embodiment, the accident point locator 6-1 functions as the accident point locator 6-1 when a program (computer program) describing the processing of the accident point locator 6-1 is executed on the computer system. Similarly, in the present embodiment, the accident point locator server 8 functions as the accident point locator server 8 when a program describing the processing of the accident point locator server 8 is executed on the computer system. FIG. 16 is a diagram showing an example of the configuration of a computer system that realizes the accident point locator 6-1 and the accident point locator server 8 of the present embodiment. As shown in FIG. 16, this computer system includes a control unit 101, an input unit 102, a storage unit 103, a display unit 104, a communication unit 105, and an output unit 106, which are connected via a system bus 107.

[0090] In FIG. 16 , the control unit 101 is a processor such as a CPU (Central Processing Unit) that executes a program describing the processing of the accident point locator 6-1 or the accident point locator server 8 of this embodiment. The input unit 102 is composed of, for example, a keyboard, a mouse, etc., and is used by the user of the computer system to input various information. The memory unit 103 includes various types of memory, such as RAM (Random Access Memory) and ROM (Read Only Memory), and a storage device, such as a hard disk, and stores the programs to be executed by the control unit 101, necessary data obtained during processing, etc. The memory unit 103 is also used as a temporary storage area for programs. The display unit 104 is composed of a display, LCD (Liquid Crystal Display Panel), etc., and displays various screens to the user of the computer system. The communication unit 105 is a receiver and transmitter that perform communication processing. The output unit 106 is, for example, a printer. Note that FIG. 16 is merely an example, and the configuration of the computer system is not limited to the example shown in FIG. 16 . For example, the computer system that realizes the fault point locator 6-1 does not need to include the display unit 104 and the output unit 106.

[0091] Here, an example of the operation of the computer system until the program of this embodiment is ready to be executed will be described. In the computer system having the above configuration, for example, the program is installed into an auxiliary storage device that is part of the storage unit 103 from a CD-ROM or DVD-ROM inserted in a CD (Compact Disc)-ROM drive or DVD (Digital Versatile Disc)-ROM drive (not shown). Then, when the program is executed, the program read from the auxiliary storage device of the storage unit 103 is stored in the main storage area of ​​the storage unit 103. In this state, the control unit 101 executes the processing as the accident point locating device 6-1 or the accident point locating server 8 of this embodiment in accordance with the program stored in the storage unit 103.

[0092] In the above explanation, a program describing the processing in each of the accident point locating device 6-1 and the accident point locating server 8 is provided using a CD-ROM or DVD-ROM as a recording medium, but this is not limited to this. Depending on the configuration of the computer system, the capacity of the program to be provided, etc., it is also possible to use a program provided by a transmission medium such as the Internet via the communication unit 105.

[0093] The location result inference unit 62 and the display data generation unit 65 shown in Fig. 2 are realized by the control unit 101 shown in Fig. 16 executing a program stored in the storage unit 103 shown in Fig. 16. The storage unit 103 is also used to realize the location result inference unit 62 and the display data generation unit 65. The storage unit 63 shown in Fig. 2 is part of the storage unit 103 shown in Fig. 16. The data acquisition unit 61 and the communication unit 64 shown in Fig. 2 are realized by the communication unit 105 and the control unit 101 shown in Fig. 16. The accident point locator 6-1 may be realized by a plurality of computer systems. The accident point locators 6-2 and 6-3 are similarly realized by computer systems.

[0094] The learning data generation unit 84, the arc resistance calculation unit 85, the statistical model generation unit 86, and the learning model generation unit 87 shown in FIG. 3 are realized by the control unit 101 shown in FIG. 16 executing a program stored in the storage unit 103 shown in FIG. 16. The storage unit 103 is also used to realize the learning data generation unit 84, the arc resistance calculation unit 85, the statistical model generation unit 86, and the learning model generation unit 87. The storage unit 82 shown in FIG. 3 is part of the storage unit 103 shown in FIG. 16. The communication unit 81 shown in FIG. 3 is realized by the communication unit 105 and the control unit 101 shown in FIG. 16. The input reception unit 83 shown in FIG. 3 is realized by the input unit 102 shown in FIG. 16. The accident point locating server 8 may be realized by multiple computer systems. The accident point locating server 8 may be realized by, for example, a cloud system.

[0095] The program of this embodiment causes a computer system to execute, for example, a data acquisition step of acquiring measurement data, which is at least one of the current and voltage measured at a power station 2 when a system accident occurs in the power system, and a location step of acquiring an inference result of the accident point by inputting features including at least one of at least a portion of the measurement data acquired when a system accident occurs and at least one of the processed data calculated from at least a portion of the measurement data into a trained model generated by machine learning for inferring the accident point from features including at least a portion of the measurement data measured at the power station 2 and at least one of the processed data calculated from at least a portion of the measurement data in the data acquisition step, and correcting the inference result using an estimated result of the cause of the system accident and correction information indicating the probability of the accident point determined according to the cause of the accident.

[0096] As described above, in this embodiment, the fault point locator 6-1 corrects the inference results obtained by inputting feature quantities into a trained model generated using training data generated by performing a system simulation using information indicating the probability of an accident point depending on the cause of the accident. This improves the accuracy of locating the fault point in the power system compared to when the cause of the accident is not taken into account. For example, when the fault point locator 6-1 performs correction using a "statistical model" of arc resistance values, it can locate the fault point reflecting differences in arc resistance values ​​for each cause of the accident. Furthermore, when the fault point locator 6-1 performs correction using "system surrounding situation information," it can locate the fault point reflecting the occurrence probability of each accident factor depending on the surrounding situation. Furthermore, when machine learning classification techniques are used, the distribution of the probability of the existence of fault points can be determined, allowing for efficient search for fault points.

[0097] Furthermore, because the measurement data from the existing CT3 and VT4 installed in the substations 2-1 to 2-3 can be used, there is no need to install sensors on the overhead ground wires, which reduces costs. Furthermore, if a sensor is installed on the overhead ground wires and measurement data is used, in the event of a short-circuit fault, no current flows through the overhead ground wires, and in the case of distribution lines, no overhead ground wires exist, so sensors cannot be installed on the overhead ground wires. In this embodiment, because there is no need to use sensors installed on the overhead ground wires, it can also be applied when a short-circuit fault occurs or when no overhead ground wires exist.

[0098] Second Embodiment Fig. 17 is a diagram showing an example of the configuration of an accident point locating system according to a second embodiment. The accident point locating system 10a of this embodiment includes an accident point locating server 8 and an accident point locating device 6-1 provided in a power station 2-1. Components having the same functions as those in the first embodiment are given the same reference numerals as those in the first embodiment, and redundant explanations will be omitted. The following mainly describes the differences from the first embodiment.

[0099] The configuration of the fault point locator 6-1 is the same as in the first embodiment. In this embodiment, the fault point locators 6-2 and 6-3 described in the first embodiment are not used. That is, there is only one fault point locator 6. Therefore, the fault point locator 6-1 does not receive measurement data from other fault point locators 6, and only the measurement data measured at the power station 2-1 is stored in the storage unit 63. In the process shown in FIG. 9 , the fault point locator server 8 calculates the current and voltage corresponding to the power station 2-1 by instantaneous value analysis in step S33. Then, in step S34, the fault point locator server 8 calculates "processed data" using the calculated current and voltage ("current (calculated value)" and "voltage (calculated value)") and generates "learning data" using the "current (calculated value)," "voltage (calculated value)," and "processed data" as feature quantities and the "fault point (setting)" as the answer.

[0100] As described in the first embodiment, the inference process may be performed by the fault point locator 6-1 or the fault point locator server 8. The fault point locator 6-1 may also perform the learning process in addition to the inference process. The fault point locator 6-1 may also generate the "system model."

[0101] As described in this embodiment, the current and voltage used as the feature quantities may be current and voltage measured at one electric power station 2. That is, the measurement data used as the feature quantities may be current and voltage measured at each of a plurality of electric power stations 2 as described in the first embodiment, or may be current and voltage measured at one electric power station 2 as described in this embodiment. Note that, as described above, the feature quantities are not limited to the example including both current and voltage, and may include at least one of current, voltage, and processed data. This embodiment can also achieve the same effects as those of the first embodiment. Although the present embodiment may have a lower fault point location accuracy than the first embodiment, the system configuration can be simplified compared to the first embodiment.

[0102] The configurations shown in the above embodiments are merely examples, and may be combined with other known technologies, or different embodiments may be combined with each other. It is also possible to omit or modify parts of the configurations as long as they do not deviate from the gist of the invention.

[0103] 1 Power line, 2, 2-1 to 2-3 Electrical station, 3-1 to 3-6 CT, 4-1 to 4-3 VT, 5 Accident cause estimation device, 6, 6-1 to 6-3 Fault point locating device, 7 Central monitoring system, 8 Fault point locating server, 9 Communication network, 10, 10a Fault point locating system, 61 Data acquisition unit, 62 Location result inference unit, 63, 82 Memory unit, 64, 81 Communication unit, 65 Display data generation unit, 66 Display device, 83 Input acceptance unit, 84 Learning data generation unit, 85 Arc resistance calculation unit, 86 Statistical model generation unit, 87 Learning model generation unit.

Claims

1. An accident point locating device comprising: a data acquisition unit that acquires measurement data, which is at least one of current and voltage measured at a power station when a system accident occurs in an electric power system; and a location result inference unit that acquires an inference result of the accident point by inputting feature values ​​including at least one of at least a portion of the measurement data acquired by the data acquisition unit when the system accident occurs and processed data calculated from at least a portion of the measurement data into a trained model generated by machine learning for inferring an accident point from feature values ​​including at least a portion of the measurement data measured at the power station and processed data calculated from at least a portion of the measurement data, and corrects the inference result using an estimated result of the cause of the system accident and correction information indicating the probability of the accident point determined according to the cause of the accident.

2. The fault point locating device of claim 1, characterized in that the correction information is a statistical model indicating a probability distribution of arc resistance generated for each cause of the accident, and the location result inference unit calculates a probability distribution of arc resistance corresponding to the cause of the accident using the statistical model and the estimated result of the cause of the accident, converts the calculated probability distribution of arc resistance into a probability distribution of distance using information indicating the configuration and impedance in the power system, and corrects the inference result using the probability distribution of distance.

3. The fault point locating device according to claim 2, further comprising: an arc resistance calculation unit that calculates, for each of a plurality of system faults that have already occurred in the power system, an arc resistance value corresponding to the system fault using actual values ​​of measurement data that are current and voltage measured at the sub-station at the time the system fault occurred, actual values ​​of the fault points of the identified system faults, and system-related information including information indicating the configuration and impedance of the power system; and a statistical model generation unit that generates the statistical model using the arc resistance value and actual values ​​of the causes of the faults corresponding to the arc resistance values, for the plurality of system faults.

4. The fault point location device described in any one of claims 1 to 3, characterized in that the correction information is a correction coefficient indicating the probability of a system fault occurring for each type of surrounding conditions for each cause of the fault, and the location result inference unit corrects the inference result using the type of surrounding conditions for each position in the power system and the correction information.

5. An accident point locating device as described in any one of claims 1 to 4, characterized in that it is provided with a display data generating unit that generates display data for displaying the location result of the accident point located by the location result inference unit on a display screen.

6. The fault point locating device described in claim 5, characterized in that the machine learning is a supervised machine learning classification method, the trained model outputs information indicating the probability of the existence of the fault point for each section in the power system as an inference result of the fault point, and the display screen displays the probability of the existence of the fault point for each section.

7. The fault point locating device according to claim 6, characterized in that the probability of the existence of the fault point is displayed on the display screen at a position on the system diagram corresponding to the corresponding section.

8. The accident point locating device according to claim 6, characterized in that the probability of the existence of the accident point is displayed on the display screen at a geographical position on a map corresponding to the corresponding section.

9. A fault location device as claimed in any one of claims 1 to 8, characterized in that there are a plurality of said electric power stations.

10. A learning device comprising: a learning model generation unit that generates, by machine learning, a trained model for inferring an accident point of a system accident in a power system from features including at least one of measurement data, which is at least one of current and current measured at a power station, and processed data calculated from at least one of the measurement data; an arc resistance calculation unit that calculates, for each of a plurality of system accidents that have already occurred in the power system, an arc resistance value corresponding to the system accident using actual values ​​of measurement data, which is the current and voltage measured at the power station at the time of the occurrence of the system accident, actual values ​​of the accident points of the identified system accidents, and system-related information including information indicating the configuration and impedance of the power system; and a statistical model generation unit that generates a statistical model indicating a probability distribution of arc resistance for each cause of the accident using actual values ​​of the arc resistance values ​​and corresponding causes of accidents for the plurality of system accidents, wherein the statistical model is used to correct the inference result of the accident point using the trained model.

11. A learning device as described in claim 10, comprising a learning data generation unit that sets a plurality of conditions including an accident point of a system accident in an electric power system, calculates a calculated value of the measurement data measured at a power station by performing a system simulation under the plurality of conditions, and generates learning data including at least one of the calculated value and processed data calculated from the calculated value, and the set accident point, wherein the learning model generation unit generates the trained model using the learning data generated by the learning data generation unit.

12. An accident point location system comprising: a learning model generation unit that generates, by machine learning, a trained model for inferring an accident point of a system accident in an electric power system from features including at least a portion of measurement data, which is at least one of a current and a current measured at an electric power station, and processed data calculated from at least a portion of the measurement data; a data acquisition unit that acquires measurement data, which is at least one of a current and a voltage measured at the electric power station when a system accident occurs in the power system; and a location result inference unit that acquires an inference result of the accident point by inputting features including at least a portion of the measurement data acquired by the data acquisition unit when the system accident occurs and processed data calculated from at least a portion of the measurement data into the trained model, and corrects the inference result using an estimated result of the cause of the system accident and correction information indicating the probability of the accident point determined according to the cause of the accident.

13. The accident point locating system according to claim 12, further comprising: a display data generating unit that generates display data for displaying the location result of the accident point located by the location result inference unit on a display screen.

14. The fault point location system described in claim 13, characterized in that the machine learning is a supervised machine learning classification method, the trained model outputs information indicating the probability of the existence of the fault point for each section in the power system as an inference result of the fault point, and the display screen displays the probability of the existence of the fault point for each section.

15. The fault point locating system according to claim 14, characterized in that the probability of the existence of the fault point is displayed on the display screen at a position on the system diagram corresponding to the corresponding section.

16. The accident point locating system according to claim 14, characterized in that the display screen displays the probability of the existence of the accident point at a geographical position on a map corresponding to the corresponding section.

17. An accident point locating method for an accident point locating device, comprising: a data acquisition step of acquiring measurement data, which is at least one of a current and a voltage measured at a power station when a system accident occurs in an electric power system; and a locating step of acquiring an inference result of the accident point by inputting feature values ​​including at least one of at least a portion of the measurement data acquired at the time of the system accident and processed data calculated from at least a portion of the measurement data in the data acquisition step into a trained model generated by machine learning for inferring an accident point from feature values ​​including at least a portion of the measurement data measured at the power station and processed data calculated from at least a portion of the measurement data, and correcting the inference result using an estimated result of the cause of the system accident and correction information indicating the probability of the accident point determined according to the cause of the accident.

18. A program for causing a computer system to execute the following steps: a data acquisition step of acquiring measurement data, which is at least one of current and voltage measured at a power station when a system accident occurs in an electric power system; and a location step of acquiring an inference result of the accident point by inputting feature values ​​including at least one of at least a portion of the measurement data acquired at the time of the system accident by the data acquisition step and processed data calculated from at least a portion of the measurement data into a trained model generated by machine learning for inferring an accident point from feature values ​​including at least a portion of the measurement data measured at the power station and processed data calculated from at least a portion of the measurement data, and correcting the inference result using an estimated result of the cause of the system accident and correction information indicating the probability of the accident point determined according to the cause of the accident.

Images