Power distribution system fault prediction and cause estimation system

Abstract

To provide a power distribution system accident sign cause estimation system capable of estimating a sign cause leading to a supply disruption accident of a power distribution line with higher accuracy than before.SOLUTION: Protective power distribution devices 4 and 5 including switches arranged in a power distribution system 1 are individually provided with slave stations 8. The slave stations 8 are connected to each other via a communication line 9, and also connected to a higher-level server 11. The slave stations 8 each include: an accident sign cause estimation unit 83 that estimates a sign cause of an accident in the power distribution system 1 based on waveform data measured by a measuring unit 81 that performs signal processing on current and voltage waveform data detected by the protective power distribution devices 4 and 5; and a data transmitting / receiving unit 84 that transmits, to the server 11, an estimation result of the sign cause of the accident of the power distribution system 1 obtained by the accident sign cause estimation unit 83. The accident sign cause is autonomously estimated by mutual communication of the slave stations 8.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present application relates to a system for estimating the cause of a fault in a power distribution system. [Background technology]

[0002] In recent years, there have been many natural disasters, such as typhoons, earthquakes, and snow damage, both in Japan and overseas. As a result, many accidents have occurred in power distribution systems due to these environmental factors, making it necessary to improve the efficiency of on-site work to eliminate the causes of accidents.

[0003] Conventionally, when an accident such as a ground fault or short circuit occurs, the accident site is visually inspected, the cause of the accident is identified, and workers carry out work to eliminate the cause of the accident. However, when a large-scale natural disaster occurs, it is necessary to respond at multiple locations, making it difficult to restore operations quickly. Therefore, a system has been established to identify signs that could lead to accidents such as supply disruptions on distribution lines in advance, which makes it possible to repair facilities before an accident occurs, and optimize labor costs by making capital investments required for efficient countermeasures, or by carrying out repairs, inspections, and restoration work.

[0004] As a mechanism for detecting signs of potential accidents in a power distribution system at a sub-station, efforts have been made to set up a power distribution system monitoring device and collect information between sub-stations via communication, or to realize a system that uses neurocomputing to create its own judgment criteria (see, for example, Patent Document 1 below). [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2020-174493 A Summary of the Invention [Problem to be solved by the invention]

[0006] In conventional accident prediction and cause estimation systems, each automatic switch installed in a distribution system is equipped with a detector that detects the current, voltage, etc. flowing in the distribution line, and communication equipment. Meanwhile, a slave station that forms a communication network for each automatic switch is individually provided, and each slave station is equipped with a measurement data processing function. Waveform data of current, voltage, etc. detected by the detector installed in the automatic switch is transmitted to the slave station via the communication equipment. Each slave station then uses its measurement data processing function to perform signal processing on the acquired waveform data, before transmitting it to a higher-level server located at a representative base.

[0007] In this case, the slave stations generally collect waveform data required to estimate the signs that will lead to the occurrence of an accident, for example, at the time of periodic collection, at the time of detecting an accident sign, or at the time of tripping a breaker when an accident is detected, etc. Then, at the time of collecting waveform data that will be a sign of an accident, at the time of periodic collection, or at the time of restoration of power after tripping an FCB (Feeder Circuit Breaker), etc., each slave station works with a higher-level server to estimate the signs that will lead to the occurrence of an accident based on the waveform data that the server has stored in advance.

[0008] In the case of conventional technology, the waveform data sent from each slave station to a host server is often measured at a sampling period exceeding several thousand points per second, and it is common to take into consideration the load on the network when sending data and to limit the data size, transfer speed, etc. However, the number of slave stations individually installed in each automatic switch in a distribution system can exceed tens of thousands, and in situations where there are frequent distribution line accidents due to large-scale natural disasters such as typhoons and earthquakes, this can put a significant strain on the network load.

[0009] Regarding the transmission of waveform data between the upper server and each slave station, it is possible to alleviate the load on the network to a certain extent by making the network optical. However, under the above-mentioned circumstances, a large amount of waveform data flows through the network, which puts a strain on the network load and may cause problems such as delays or transmission abnormalities. This can lead to problems in the operation of distribution systems that use the same network, such as the inability to monitor the status of slave stations or failure to control them.

[0010] For these reasons, in order to realize sufficient network bandwidth or speed between the upper server and the child stations in the event of frequent power line accidents, a large amount of capital investment and advanced network technology for system construction is required. In addition, the upper server needs to predict the signs of accidents based on the huge amount of waveform data collected from the child stations, so server resources to process this huge amount of waveform data and complex server application technology are required.

[0011] In addition, since the child stations also acquire weak waveform data that is not strong enough to cause a power outage in the power distribution system as a warning sign, not all of the waveform data sent to the server is necessarily waveform data that has been caused by an actual accident on the power distribution line.

[0012] Furthermore, the signs that lead to an accident do not occur only in the child station in question, but are detected in conjunction with surrounding child stations as well, so it is not possible to deduce where and what type of accident is occurring or its cause until the waveform data is analyzed by the upper server.

[0013] In other words, since it is difficult to accurately judge the signs of an accident using only waveform data detected by a single slave station, by comparing information from multiple slave stations, it is possible to estimate the approximate location of the accident signs, and by taking into account past accident cases at that location, the cause of the accident can be inferred, and ultimately the actual situation cannot be understood without visually checking the target slave station and adjacent sites (power distribution lines, adjacent slave stations, etc.) on-site. As such, the conventional method requires the use of multiple systems and manual intervention, which is time-consuming and labor-intensive, and there is room for improvement.

[0014] The present application discloses technology for solving the problems described above, and aims to provide a power distribution system accident precursor cause estimation system that can autonomously determine signs that may lead to an accident with high accuracy at the child stations without sending waveform data to a higher-level server, and that can estimate the cause of the signs that may lead to an accident by sharing information between the child stations and combining waveform data from multiple child stations. [Means for solving the problem]

[0015] The power distribution system fault prediction cause estimation system disclosed in the present application comprises: A protective distribution device having a plurality of switches and arranged in a distribution system, Each switch is It is equipped with a detector that detects the current and voltage flowing through the distribution line, and the child stations that make up the communication network are installed separately. 、 The above each The slave stations are connected to each other via communication lines and to a higher-level server, The above each The slave station includes an operation unit that operates the protective power distribution device in response to a detection result by the detector, and a measurement unit that processes signal waveform data of the current and voltage detected by the detector. The system comprises a learning device and an inference device having an artificial intelligence function, an accident precursor cause estimating unit that estimates a precursor cause of an accident in the power distribution system based on the waveform data measured by the measuring unit; and a data transmitting / receiving unit that transmits an estimation result of a precursor cause of an accident in the power distribution system obtained by the accident precursor cause estimating unit to the server. and Equipped with Each of the slave stations detects a change in the waveform data and a change in a cumulative number of times that a sign of an accident is detected, and propagates the change in the waveform data and the change in the cumulative number of times that a sign of an accident is detected to a slave station of a higher-level switch among the plurality of switches; The above each The child station is The fault location in the power distribution system is identified based on the fluctuation of the waveform data and the fluctuation of the cumulative number of times the fault sign is detected, and the fault sign cause estimation unit autonomously analyzes the waveform data. Estimate the causes of accident precursors. Effect of the Invention

[0016] According to the power distribution system accident precursor cause estimating system of the present application, when estimating the cause of an accident precursor, communication is performed only between related child stations without transmitting waveform data to a higher-level server each time, so that the cause of an accident precursor in a power distribution line can be estimated without putting a strain on the network load. [Brief description of the drawings]

[0017] [Figure 1] 1 is a configuration diagram showing an example of a power distribution system to which a power distribution system accident sign cause estimating system in a first embodiment is applied. [Diagram 2] 2 is a block diagram showing the configuration of a switch and a slave station provided individually for the switch in the power distribution system of FIG. 1. [Diagram 3] FIG. 2 is a configuration diagram showing an example of a communication network configured by each slave station and a higher-level server. [Figure 4] 1 is a characteristic diagram showing a change tendency of waveform data of current, voltage, etc. measured by a measurement unit of a slave station and a cumulative detection count when the power distribution system is normal. FIG. [Diagram 5] 1 is a characteristic diagram showing the change trends of waveform data of current, voltage, etc. measured by a measurement unit of a slave station and the cumulative number of detections when there is a sign of an accident in a power distribution system. [Figure 6] 10 is a characteristic diagram showing a change tendency of waveform data of current, voltage, etc. measured by a measuring unit of a slave station and a cumulative detection count when an abnormality occurs in a power distribution system. FIG. [Figure 7] FIG. 1 is a system diagram showing an example of a case where an accident occurs in a part of a power distribution system. [Figure 8] 8 is an explanatory diagram showing waveform data of current, voltage, etc. measured by a measuring unit of each slave station in the case of FIG. 7, and changes in the cumulative number of detections. FIG. [Figure 9]FIG. 11 is a system diagram showing another example of a case where an accident occurs in a part of a power distribution system. [Figure 10] 10 is an explanatory diagram showing waveform data of current, voltage, etc. measured by a measuring unit of each slave station in the case of FIG. 9, and changes in the cumulative number of detections. [Figure 11] FIG. 2 is a block diagram showing a configuration of a learning device. [Figure 12] FIG. 2 is a block diagram showing a configuration of an inference device. [Figure 13] 11 is a flowchart showing an example of a procedure for generating a learning model that infers a precursor cause that leads to an accident by a learning device. [Figure 14] 11 is a flowchart showing an example of a procedure for generating a learning model that infers a precursor cause that leads to an accident by a learning device. [Figure 15] 11 is a flowchart showing an example of a procedure for generating a learning model that infers a precursor cause that leads to an accident by a learning device. [Figure 16] 13 is a flowchart showing the process performed by the server after the results of inference of the cause of an accident sign obtained at each slave station are transmitted to a higher-level server via a network. [Figure 17] 2 is a block diagram showing an example of hardware of an accident sign cause estimating unit. FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] Embodiment 1 FIG. 1 is a configuration diagram showing an example of a power distribution system to which a power distribution system accident sign cause estimating system according to the first embodiment is applied.

[0019] The power distribution system fault sign cause estimating system according to the first embodiment is a system that estimates the cause of a sign that may lead to the occurrence of a fault in a power distribution system. As shown in FIG. 1, for example, a power distribution system 1 includes a substation 2 at the distribution source, a transformer 3 at a branch point, and protective power distribution equipment (in FIG. 1, a circuit breaker 4, a switchgear (normally closed) 5, and a switchgear (normally open) 6 correspond to the protective power distribution equipment), which are connected via a power distribution line 7.

[0020] In particular, here, a detector (not shown) for detecting the current, voltage, etc. flowing through a distribution line 7 is mounted on a switch (normally closed) 5 and a switch (normally open) 6 constituting the protective distribution device. Further, a slave station 8 for forming a communication network is provided for each of the switches 5 and 6 as shown in Fig. 2. The slave stations 8 are connected to each other via a communication line 9 such as an optical cable, and are also connected to a higher-level server 11 via an IP network 10 as shown in Fig. 3.

[0021] Each slave station 8 is electrically connected to the switches 5 and 6 via a signal line 12. Each slave station 8 includes an operation unit 81 that receives waveform data of current, voltage, etc. detected by a detector (not shown) provided in the switches 5 and 6 via the signal line 12 and opens and closes the switches 5 and 6, a measurement unit 82 that receives the waveform data of current, voltage, etc. detected by the detector and processes the signals, and a human operator that estimates the cause of a fault in the power distribution system 1 based on the waveform data obtained by the measurement unit 82. Engineering The power distribution system 1 includes an accident precursor cause estimating unit 83 having an artificial intelligence (AI) function, and a data transmitting / receiving unit 84 that transmits the result of estimation of the precursor cause of an accident in the power distribution system 1 obtained by the accident precursor cause estimating unit 83 to a higher-level server 11 and acquires information from the server 11. The configuration and contents of the accident precursor cause estimating unit 83 will be described later.

[0022] Next, features of the fault sign detection in the power distribution system fault sign detection system according to the first embodiment will be described.

[0023] First, using Figures 4, 5, and 6, we will explain the changes in waveform data of current, voltage, etc. measured by the measurement unit 82 of each slave station 8, as well as the cumulative value of the number of detections indicating signs or abnormalities (hereinafter referred to as the cumulative number of detections) in each of the following cases: when the distribution system 1 is normal, when there are signs of an accident occurring in part of the distribution line 7, and when an abnormality such as a short circuit accident has occurred in part of the distribution line 7.

[0024] Fig. 4 shows a case where the power distribution system 1 is normal, and the current waveform measured by the slave station 8 (upper part of Fig. 4) is within a certain threshold range that has been set in advance. In addition, the cumulative number of detections (lower part of Fig. 4) does not increase over time. In Fig. 4, the number of detections indicating an abnormality is zero. Figure 5 shows a case where there are signs of an accident occurring in part of the power distribution line 7, and the current waveform measured by the slave station 8 (upper part of Figure 5) frequently exceeds a certain preset threshold range. In addition, the cumulative number of detections (lower part of Figure 5) also gradually increases over time. Fig. 6 shows a case where an abnormality such as a short circuit occurs in part of the distribution line 7, and the current waveform measured at the slave station 8 exceeds a preset threshold range for determining an accident. Fig. 6 shows the prediction criteria and accident criteria, and the accident criteria are exceeded when an abnormality occurs.

[0025] FIG. 7 is a system diagram showing an example of a case where an accident occurs in a part of a power distribution system. FIG. 8 is an explanatory diagram showing the current waveform measured by the measuring unit 82 of each substation 8 and the change in the cumulative number of detections when a short-circuit accident occurs at the location of the distribution line 7 indicated by the symbol P1 in a power distribution system such as that shown in FIG. 7 is adopted.

[0026] As can be seen from Fig. 8, the current waveform and the cumulative detection count fluctuations measured at the slave station 8e of the switch 5e due to a short circuit occurring in the distribution line 7 connected to a lower switch (switch 5e in this case) far from the substation 2 are also propagated to the current waveform and the cumulative detection count measured at the slave stations 8b and 8a of the upper switches (switches 5b and 5a in this case) closer to the substation 2. Note that the slave stations 8c, 8g, 8h, and 8i corresponding to the switches indicated by the symbols 5c, 5g, 5h, and 5i are not affected by the current waveform and the cumulative detection count fluctuations due to the accident. Therefore, the location of the distribution line 7 where the accident occurred can be identified by communicating with each other between the slave stations 8 to check whether there is an influence on the current waveform and the cumulative detection count fluctuations.

[0027] FIG. 9 is a system diagram showing an example of a case where an accident occurs in a part of a power distribution system. FIG. 10 is an explanatory diagram showing the current waveform measured by the measurement unit 82 of each substation 8 and the change in the cumulative number of detections when a short-circuit accident occurs at the location of the distribution line 7 indicated by the symbol P2 in a power distribution system such as that shown in FIG. 9 is adopted.

[0028] As can be seen from FIG. 10, the fluctuations in the current waveform and cumulative detection count measured at slave station 8c of switch 5c due to a short circuit occurring in distribution line 7 connected to a lower switch (switch 5c in this case) distant from substation 2 are also propagated to the current waveform and cumulative detection count measured at slave stations 8b and 8a of higher-level switches (switches 5b and 5a in this case) closer to substation 2. Note that slave stations 8d and 8f corresponding to switches indicated by symbols 5d and 5f are not affected by the current waveform and cumulative detection count fluctuations due to the accident. Therefore, even in this case, the location of the distribution line 7 where the accident occurred can be identified by communicating between each slave station 8 to check whether there is any influence on the current waveform and cumulative detection count fluctuations. but can.

[0029] Next, in the event that a warning sign that may lead to an accident occurs or an abnormality such as a short circuit occurs, the characteristics of the waveform data of the current or voltage measured by the slave station 8 due to this will be explained in the following [N1] to [N7].

[0030] [N1] Cable ground fault (ground fault caused by cable deterioration of distribution line): Waveform generation and recovery are repeated [N2] Insulator earth fault (earth fault caused by insulator damage): Changes over time depending on the state of insulator damage [N3] Animal-based ground fault (ground fault caused by carbonization of distribution line cables after contact with animals): A state close to a complete ground fault occurs [N4] Tree-ground fault (ground fault caused by tree contact with power distribution cable): The ground fault level increases over time. [N5] Foreign object contact ground fault (ground fault caused by contact of a foreign object with a power distribution cable): A needle-like wave occurs immediately after contact due to the gap distance. [N6] Gap earth fault (earth fault caused by a faulty lightning arrester, etc.): The needle-shaped wave continues, so there is no change in the waveform over time. [N7] Distribution line failure (failure such as a broken distribution line): Some slave stations do not respond when communicating with each other.

[0031] Next, the personnel provided by each slave station 8 Engineering The configuration of the accident symptom cause estimating unit 83 having an AI function will be described.

[0032] The accident symptom cause estimating unit 83 shown in FIG. As shown in FIG. 11, the learning device 831 includes a learning data acquisition unit 831a that acquires waveform data acquired by the measurement unit 82 of the slave station 8, a trained model generation unit 831b that acquires the waveform data acquired by the learning data acquisition unit 831a as learning data and generates a trained model for inferring the precursor cause of an accident in the power distribution system 1 from the learning data, and a trained model storage unit 831c that stores the trained model generated by the trained model generation unit 831b.

[0033] In this embodiment 1, the trained model generating unit 831b basically generates a trained model based on a training program by unsupervised learning. Here, unsupervised learning refers to a method of learning features present in training data (current and voltage waveform data in this case) that does not include results (labels) by providing the training device with the training data. That is, the trained model generating unit 831b of the training device 831 performs unsupervised learning according to a grouping method using, for example, the K-means method. Here, the K-means method is a non-hierarchical clustering algorithm, and is a method of classifying a given number of clusters into K clusters using the average of the clusters.

[0034] Specifically, the K-means method is processed in the following manner. First, a cluster is randomly assigned to each data x1. Next, the center Vj of each cluster is calculated based on the assigned data. Next, the distance between each data x1 and the center Vj is calculated, and x1 is reassigned to the cluster with the closest center. Then, if the cluster assignment of all x1s remains unchanged in the above process, or if the amount of change falls below a certain threshold value set in advance, it is determined that convergence has occurred. Then, the trained model generation unit 831b generates and outputs a trained model by executing the above-mentioned training.

[0035] As shown in FIG. 12, the inference device 832 includes an inference data acquisition unit 832a that acquires waveform data input into the measurement unit 82, and an inference unit 832b that compares the waveform data acquired by the inference data acquisition unit 832a with a learned model stored in the learned model memory unit 831c of the learning device 831, estimates precursor causes that may lead to an accident in the power distribution system 1, and outputs the estimated result.

[0036] Next, an example of a procedure for generating a learning model for inferring a precursor cause leading to an accident by the learning device 831 of the accident precursor cause estimation unit 83 will be described with reference to the flowcharts shown in Fig. 13 to Fig. 15. Note that the symbol S denotes each processing step.

[0037] 13 to 15, when a sign that may lead to an accident occurs (step S1), the learning device 831 communicates with each of the slave stations 8 to determine whether there is a response from each other (step S2). If there is a slave station 8 that does not respond even when communicating with each other, it is assumed that there is a distribution line fault [N7], such as the distribution line 7 being cut (step S3). In addition, when each slave station 8 communicates with each other and there is a response from each other, each slave station 8 identifies the location in the distribution line 7 where the accident occurred by determining whether or not there is a fluctuation in the waveform data of the current and voltage measured by the measurement unit 82 and whether or not there is a change in the cumulative number of detections (step S4). In addition, the accident precursor cause estimation unit 83 provided in each slave station 8 inputs the current and voltage waveform data measured at the slave station 8 into the learning data acquisition unit 831a of the learning device 831, and the learned model generation unit 831b extracts features of the waveform data (the aforementioned [N1] to [N6]) from this acquired current and voltage waveform data based on the learning program (steps S5, S6, S7, S8, S9, S10, S11, S12, S13, S14, S15). Next, the trained model generating unit 831b generates the extracted waveform data features ([N1] to [N7]) as a trained model (information) for predictive judgment. The trained model generated by the trained model generating unit 831b in this manner is stored in the trained model storage unit 831c.

[0038] In this way, in response to the occurrence of signs that may lead to an accident, the learning device 831 of the accident sign cause estimation unit 83 repeats the steps of the flowcharts shown in Figures 13 to 15, gradually improving the accuracy of the trained model generated by the trained model generation unit 831b.

[0039] In the state where the trained model generation unit 831b generates the trained model as described above, if a new sign of an accident occurs in the power distribution system 1, the slave stations 8 communicate with each other, analyze the detection data, and estimate the cause. In this case, the check is performed not only for the slave station 8, but also for higher-level systems such as the circuit breaker 4, transformer 3, and substation 2.

[0040] That is, similarly to the procedures of the flowcharts shown in Figs. 13 to 15, the accident sign cause estimating unit 83 communicates with each of the slave stations 8 and judges whether there is a response from each other (step S2). If there is a slave station 8 that does not respond even when communicating with each other, it is assumed that there is a distribution line fault [N7], such as the distribution line 7 being cut (step S3). In addition, when each slave station 8 communicates with another slave station 8 and there is a response from each other, the inference device 832 provided in each slave station 8 acquires waveform data of the current and voltage measured in the slave station 8 using the inference data acquisition unit 832a. The inference unit 832b then compares the waveform data acquired by the inference data acquisition unit 832a with the learned model stored in the learned model storage unit 831c of the learning device 831, infers which cluster of [N1] to [N6] the waveform data belongs to, i.e., infers the cause of the sign leading to the occurrence of an accident, and outputs the inference result. If the waveform data does not belong to any of the clusters of [N1] to [N7], it is presumed that there is no sign of an accident in the power distribution system 1. The inference unit 832b then transmits the inference results [N1] to [N7] as outputs from the data transmission / reception unit 84 to the upper server 11.

[0041] As shown in the flowchart of Fig. 16, the upper server 11 shown in Fig. 3 receives the inference results [N1]-[N7] transmitted from the slave stations 8 via the IP network 10 (step S20), displays them on a display (not shown) owned by the manager who manages the power distribution system 1, and outputs an alarm to an alarm device if there are signs of an accident (step S21). Based on the inferred results of the signs of an accident, an on-site investigation is carried out on the power distribution line 7 near the slave station 8 where signs of an accident are assumed to be occurring, and the actual situation is confirmed (step S22).

[0042] In this embodiment 1, the power distribution system 1 has been described as outputting an inference result using a trained model trained by the trained model generation unit 831b, but it is also possible to obtain a trained model of another power distribution system from outside and use this trained model to obtain an inference result.

[0043] 2 includes a processor 1000 and a storage device 1010, as shown in FIG 17, which is an example of hardware. The storage device 1010 includes a volatile storage device such as a random access memory, and a non-volatile auxiliary storage device such as a flash memory, both of which are not shown. Also, an auxiliary storage device such as a hard disk may be provided instead of a flash memory. The processor 1000 executes a program input from the storage device 1010. In this case, the program is input from the auxiliary storage device to the processor 1000 via a volatile storage device. The processor 1000 may output data such as a calculation result to a volatile storage device of the storage device 1010, or may store the data in the auxiliary storage device via the volatile storage device.

[0044] As described above, according to the power distribution system fault prediction cause estimating system of the first embodiment, Engineering By providing an accident sign cause estimation unit 83 with an AI function, each slave station 8 autonomously analyzes waveform data, eliminating the need to transmit waveform data of accidents and accident signs on the distribution line 7 to the upper server 11 one by one. This makes it possible to estimate the cause of accident signs without putting a strain on the network load. Furthermore, because the slave station 8 operates autonomously using the accident sign cause estimation unit 83 installed therein, it is possible to analyze waveform data in more real time than with conventional methods.

[0045] Furthermore, because multiple slave stations 8 communicate with each other, learning is possible by combining the waveform data of multiple slave stations 8, so it is possible to estimate the cause of an accident precursor by changing the scope of analysis, such as by transformer, distribution line, or substation. This allows the approximate location and cause of an accident precursor to be narrowed down, making it possible to estimate the cause of the precursor with high accuracy. This also leads to reductions in operation and maintenance costs, such as maintaining supply reliability (reducing power outage duration), capital investment required for countermeasures (repairs), and labor costs (inspection and restoration work).

[0046] Although the present application describes exemplary embodiments, the various features, aspects, and functions described in embodiment 1 are not limited to application to a particular embodiment, but may be applied to the embodiments alone or in various combinations.

[0047] Therefore, countless modifications not illustrated are conceivable within the scope of the technology disclosed in this application, including, for example, modifying, adding, or omitting at least one component. [Explanation of symbols]

[0048] 1. Distribution system, 2. Substation, 3. Transformer, 4. Circuit breaker, 5. Switchgear (normally closed), 6 switch (normally open), 7 distribution line, 8 slave station, 81 operation unit, 82 measurement unit, 83 Accident prediction cause estimation unit, 831 Learning device, 831a Learning data acquisition unit, 831b a trained model generation unit, 831c a trained model storage unit, 832 an inference device, 832a inference data acquisition unit, 832b inference unit, 84 data transmission / reception unit, 9 Communication line, 10 IP network, 11 Server, 12 Signal line.

Claims

1. A protective distribution device having a plurality of switches and arranged in a distribution system, Each switch is provided with a detector for detecting a current and a voltage flowing through the distribution line, and a slave station constituting a communication network is provided separately, The child stations are connected to each other via a communication line and to a higher-level server, Each of the slave stations includes an operation unit that operates the protective power distribution device in response to a detection result by the detector, a measurement unit that performs signal processing on waveform data of current and voltage detected by the detector, an accident precursor cause estimation unit that is made up of a learning device and an inference device having an artificial intelligence function and that estimates a precursor cause of an accident in the power distribution system based on the waveform data measured by the measurement unit, and a data transmission / reception unit that transmits to the server the estimation result of the precursor cause of an accident in the power distribution system obtained by the accident precursor cause estimation unit, Each of the slave stations detects a change in the waveform data and a change in a cumulative number of times that a sign of an accident is detected, and propagates the change in the waveform data and the change in the cumulative number of times that a sign of an accident is detected to a slave station of a higher-level switch among the plurality of switches; The slave stations communicate with each other, identify an accident location in the power distribution system based on a change in the waveform data and a change in a cumulative number of times that an accident sign is detected, and autonomously analyze the waveform data by the accident sign cause estimation unit to estimate a cause of the accident sign. A system for estimating the causes of power distribution system accidents.

2. 2. The power distribution system accident precursor cause estimating system according to claim 1, wherein the learning device comprises: a learning data acquisition unit that acquires the waveform data acquired by the learning data acquisition unit as learning data, and a trained model generation unit that acquires the waveform data acquired by the learning data acquisition unit as learning data and generates a trained model for inferring precursor causes of accidents in the power distribution system from the learning data.

3. The power distribution system accident precursor cause estimating system according to claim 2, wherein the inference device comprises an inference data acquisition unit that acquires waveform data input to the measurement unit, and an inference unit that compares the waveform data acquired by the inference data acquisition unit with the learned model generated by the learned model generation unit to estimate a precursor cause of an accident in the power distribution system and output the estimation result.

Images