JP7590717B2 - Connection abnormality detection device and distribution board connection abnormality detection system - Google Patents

Description

The present invention relates to a connection abnormality detection device and a distribution board connection abnormality detection system that detects connection abnormalities that occur in conductor connections that form electrical paths within a distribution board, etc.

In a distribution board, the conductor connections that make up the electrical circuit, including the primary terminal of the main breaker that connects the lead-in line from the commercial power source to the main breaker, and its secondary terminal, are made using screws, and connections are made by tightening the screws. Therefore, an abnormality in the electrical circuit due to a loose screw, etc., can cause heat generation, and if left unattended, there is a risk of fire, so there is a device to detect this.
For example, in Patent Document 1, an overheat detector is placed adjacent to a branch breaker to detect the potential difference between the drop line and the main bar to which the branch breaker is connected, and if the potential difference exceeds a certain value, it determines that a connection abnormality has occurred.

JP 2012-257343 A

The connection abnormality detection method of Patent Document 1 mentioned above was effective for monitoring connection abnormalities within a distribution board because it made it possible to collectively monitor multiple terminal parts by simply installing one overheat detector.
However, in the configuration for determining the occurrence of a connection abnormality from the above-mentioned potential difference information, it is difficult to set a threshold value, and there are cases where a connection abnormality cannot be detected even if it occurs, or where a malfunction occurs and the electrical circuit is interrupted.

In view of these problems, the present invention aims to provide a connection abnormality detection device and a distribution board connection abnormality detection system that can collectively monitor the occurrence of connection abnormalities in electrical circuits with high accuracy.

In order to solve the above problems, the invention of claim 1 is a connection abnormality detection device that, when a connection abnormality occurs at a connection part of an electric circuit, detects the accompanying discharge and determines the occurrence of a connection abnormality, and has a noise measurement unit that is placed on the electric circuit and detects electromagnetic noise generated in the electric circuit and outputs its frequency spectrum information, and an abnormality judgment unit that detects electromagnetic noise associated with the discharge from the measured frequency spectrum information and determines the occurrence of a connection abnormality, characterized in that the abnormality judgment unit divides the frequency range into a low frequency range and a high frequency range with a specific frequency as the boundary, and determines that a connection abnormality has occurred if the measured value of the electromagnetic noise in both frequency ranges exceeds a predetermined threshold value .
According to this configuration, if electromagnetic noise of two different frequencies each exceed a predetermined magnitude, it is determined that a connection abnormality has occurred. Therefore, it is possible to determine the specific phenomenon associated with the occurrence of a connection abnormality and detect the connection abnormality with high accuracy.

In addition, a specific frequency is used to divide the frequency range into two, and if electromagnetic noise of a predetermined magnitude occurs in either of these frequency ranges, it is determined that a connection abnormality has occurred, making it possible to determine with even greater accuracy the specific phenomena associated with the occurrence of a connection abnormality.

The invention of claim 2 is characterized in that in the configuration of claim 1 , the specific frequency that divides the frequency band into two is 1 MHz.
According to this configuration, the occurrence of a connection abnormality is determined from the characteristics of electromagnetic noise that occurs in both the low frequency region and the high frequency region with the boundary of 1 MHz, thereby making it possible to accurately detect noise specific to a connection abnormality.

The invention of claim 3 is characterized in that in the configuration of claim 1 or 2 , the judgment frequency of the low frequency region is 0.2 MHz to 0.9 MHz, and the judgment frequency of the high frequency region is 11 MHz to 13 MHz.
According to this configuration, connection abnormalities specific to loose screws can be accurately detected, and erroneous detection due to external noise such as high-frequency noise specific to the load can be eliminated.

The invention of claim 4 is characterized in that, in the configuration described in any one of claims 1 to 3 , the abnormality determination unit learns and sets frequency spectra for determining a connection abnormality in both the low frequency region and the high frequency region using a convolutional neural network.
According to this configuration, the frequency to be judged is set based on learning data using a convolutional neural network, making it possible to make an accurate judgment even if the frequency spectrum characteristics of the electromagnetic noise associated with a connection abnormality differ depending on the type of load.

The distribution board connection abnormality detection system of the invention of claim 5 is characterized in that a connection abnormality detection device described in any of claims 1 to 4 is installed in a feeder line to a distribution board, and a first filter circuit for blocking electromagnetic noise generated upstream of the feeder line is arranged upstream of the feeder line on which the connection abnormality detection device is installed.
According to this configuration, electromagnetic noise generated upstream of the drop line is not detected, and only electromagnetic noise generated downstream, i.e., in the distribution board or on the electrical circuit downstream of that, is detected. This makes it possible to detect connection abnormalities in the distribution board or further downstream, and limits the detection range of connection abnormalities.

The invention of claim 6 is characterized in that, in the configuration described in claim 5 , a second filter circuit is arranged for each branch electric circuit branched off by the distribution board, which blocks electromagnetic noise generated in a load connected to the branch electric circuit from flowing into the distribution board.
According to this configuration, a second filter circuit that cuts out electromagnetic noise is disposed in the branch electric circuit, so that the connection abnormality detection device will not malfunction due to electromagnetic noise generated at the load connected to the end of the branch electric circuit or at the connection part of the branch electric circuit, such as an outlet.

According to the present invention, if electromagnetic noise of two different frequencies exceeds a predetermined level, it is determined that a connection abnormality has occurred. This makes it possible to determine the specific phenomenon that accompanies the occurrence of a connection abnormality, and to detect the connection abnormality with high accuracy.

1 is a configuration diagram showing an example of a connection abnormality detection device according to the present invention; 2 is a configuration diagram of a distribution board connection abnormality detection system in which the connection abnormality detection device of FIG. 1 is attached to a distribution board. 1 is a flowchart showing a flow of learning based on a convolutional neural network. 13 is a flowchart for detecting a loose terminal. 1 is a flowchart of learning loose terminals based on a convolutional neural network. 1 shows the frequency spectrum of the measured electromagnetic noise, where (a) shows a state in which the electrical circuit is normal, and (b) shows a state in which an abnormality has occurred in the electrical circuit due to a loose screw. 5A and 5B show frequency spectra of measured electromagnetic noise, in which (a) shows a state in which the second filter circuit is present, and (b) shows a state in which the second filter circuit is not present.

The following describes in detail an embodiment of the present invention with reference to the drawings. FIG. 1 is a block diagram showing an example of a connection abnormality detection device according to the present invention. The connection abnormality detection device 1 is composed of a measuring device (noise measurement unit) 2 that measures noise and a detection unit 3 as an abnormality determination unit that detects connection abnormalities, to which a cutoff mechanism unit 4 that cuts off the electrical circuit upon receiving the detection result from the detection unit 3, an alarm unit 5 that issues an alarm, a contact output unit 6 that issues an alarm to the outside, etc. are connected.

The measuring device 2 is installed on an electric circuit to detect electromagnetic noise occurring in the electric circuit and output its frequency spectrum information.
The detection unit 3 includes a classifier (CNN classifier) 31 based on a convolutional neural network (hereinafter referred to as "CNN"), a memory 32 for storing measurement data, a learning unit 33 for performing deep learning by CNN, and a determination unit 34 for determining a connection abnormality based on the classification result of the CNN classifier 31. The detection unit 3 is composed of one MCU (Micro Controller Unit).

Frequency spectrum information of the detected electromagnetic noise is output from the measuring device 2 to the CNN classifier 31 and the memory 32. The measuring device 2 detects electromagnetic noise generated on an electric circuit by, for example, a current transformer, and outputs the frequency spectrum data.
The CNN identifier 31 extracts frequency spectrum information of electromagnetic noise caused by a connection abnormality from the frequency spectrum of the electromagnetic noise measured by the measuring instrument 2 based on information from the learning unit 33, and outputs it to the judgment unit 34 in a specified programming language.
The learning unit 33 learns and accumulates information for selecting, from the frequency spectrum of electromagnetic noise, a specific frequency spectrum that occurs in association with a connection abnormality such as a loose screw of a conductive part.

2 shows the configuration of a distribution board connection abnormality detection system in which a connection abnormality detection device 1 is attached to a distribution board 7, and shows a configuration in which the connection abnormality detection device 1 detects an abnormality that occurs at a connection part of an electric circuit 10 in the distribution board 7. The connection abnormality detection device 1 is installed on a lead-in line L of the distribution board 7 that forms part of the electric circuit.
A first filter circuit 11 is provided on the lead-in line L of the commercial power source P, which is upstream of the connection abnormality detection device 1, to block electromagnetic noise generated on the commercial power source P side so that it does not flow into the connection abnormality detection device 1 and cause it to malfunction. Furthermore, each branch electric circuit 10a arranged from the distribution board 7 is provided with a second filter circuit 12 to block high-frequency noise generated in a load 15 connected to the branch electric circuit 10a so that it does not flow into the distribution board 7 via the branch electric circuit 10a and cause the connection abnormality detection device 1 to malfunction.

A main breaker 13 and a number of branch breakers 14 are attached to the distribution board 7, and screws are used for the electrical circuit connections of the primary terminal, secondary terminal, etc. of the main breaker 13, the electrical circuit connections of the branch breakers 14, and also the connections of the copper bars that make up the electrical circuit 10. If any loosening occurs in the connections made by these screws, it is configured to detect this.
The connection abnormality detection device 1 may be installed inside the distribution board 7 as long as it is installed on the upstream side of the main breaker 13 (on the commercial power source P side).

In this distribution board connection abnormality detection system, when the connection abnormality detection device 1 detects a connection abnormality in the electric circuit 10, it issues an alarm and performs measures such as cutting off the electric circuit 10.
Specifically, when the determination unit 34 determines that there is a connection abnormality, an abnormality occurrence signal is output. In response to this abnormality occurrence signal, the main breaker 13 constituting the cutoff mechanism unit 4 performs a cutoff operation. The main breaker 13 is equipped with an external signal input unit, and performs a cutoff operation in response to the abnormality occurrence signal.
The alarm unit 5 includes a buzzer, an LED, etc., and the buzzer sounds and the LED emits light. The contact output unit 6 outputs a contact signal.
The alarm unit 5 and the contact output unit 6 are supplied with power from the primary side of the main breaker 13 .

3 is an explanatory diagram of the learning unit 33, showing a flowchart showing the flow of learning based on CNN. The learning unit 33 learns based on CNN, and the CNN is configured to have a combination of multiple layers such as a convolution layer T2, a pooling layer T3, an activation layer T4, and a fully connected layer T5 between an input layer T1 and an output layer T6.
The convolution layer T2 convolves the input frequency spectrum data. The pooling layer T3 reduces the data by replacing the data in the kernel domain with the maximum value or average value. The activation layer T4 performs a process of nonlinearly transforming the input value using an activation function. By combining multiple of these layers, the input value is output through the fully connected layer T5, which is a process of connecting the input value to all nodes.
By learning using this CNN, it becomes possible to distinguish between the frequencies of conductive noise specific to loads such as electrical appliances and the frequencies of conductive noise associated with connection abnormalities caused by loose terminals (loose screws) among the individual waveforms in the frequency spectrum of detected noise, thereby detecting abnormalities in the connections.

4 shows a flow of detecting a loose terminal. The determination unit 34 receives the identification result from the CNN identifier 31 and determines whether or not there is a connection abnormality.
The frequency spectrum measured by the measuring device 2 is input to the CNN classifier 31 (S1), and is analyzed by comparing it with the learning data accumulated in the learning unit 33 (S2). Through this analysis, frequency spectrum data associated with the connection anomaly is extracted and output to the determination unit 34.

The judgment unit 34 makes a judgment based on the frequency spectrum data output by the CNN classifier 31 in two frequency bands separated by a boundary of 1 MHz, for example, in both a low frequency region from 0.15 MHz to 1 MHz and a high frequency region from 1 MHz to 30 MHz.
At this time, if the peak values of both frequency bands exceed a predetermined threshold, it is determined that a connection abnormality has occurred (Yes in S3). If it is determined that a connection abnormality has occurred, the above-mentioned alarm is issued, the electric circuit 10 is cut off, etc. are performed (S4). If it is determined that there is no looseness (No in S3), monitoring of the frequency spectrum by the measuring device 2 continues.
Furthermore, when a connection abnormality occurs, particularly when a connection abnormality occurs due to a loose screw, it has been confirmed that electromagnetic noise is characterized by noise occurring around 0.3 MHz, 0.8 MHz, and 12 MHz, and it is possible to make judgments based mainly on these frequencies.

Fig. 6 shows the frequency spectrum of the measured electromagnetic noise, where (a) shows the state where the electric circuit 10 is normal, and (b) shows the state where a connection abnormality occurs due to a loose screw. In Fig. 6(b), Q1 indicates the position of 0.3 MHz, Q2 indicates the position of 0.8 MHz, and Q3 indicates the position of 12 MHz. As shown in Fig. 6, when a connection abnormality occurs due to a loose screw, noise occurs mainly at specific frequencies in multiple locations, so the low frequency region of 0.15 MHz to 1 MHz may be further limited to the range of 0.2 MHz to 0.9 MHz, and the high frequency region of 1 MHz to 30 MHz may be limited to the range of 11 MHz to 13 MHz for judgment.
In FIG. 6, load A is a heater, load B is a dryer, and load C is a hot plate.

In this way, the judgment frequency for the low frequency range can be set to 0.2 MHz to 0.9 MHz, and the judgment frequency for the high frequency range can be set to 11 MHz to 13 MHz, allowing accurate detection of connection abnormalities specific to loose screws and eliminating false detections caused by external noise such as high frequency noise specific to the load.

In this series of steps for detecting a connection anomaly, the learning result of the learning unit 33 is updated as follows: Fig. 5 shows a flow for learning a frequency spectrum for determining whether a terminal is loose based on CNN, and the following description will be given with reference to this flow.
The frequency spectrum measured by the measuring device 2 is analyzed by the CNN classifier 31 (S11) and judged by the judgement unit 34 (S12). If it is judged based on this result that there is a connection abnormality (Yes in S12), the frequency spectrum data is updated with data indicating that there is a connection abnormality (S13). On the other hand, if it is judged that there is no connection abnormality (No in S12), the frequency spectrum data is updated with data indicating that there is no connection abnormality (S14).

In this way, the detected electromagnetic noise spectrum is divided into two at the 1 MHz boundary, and if electromagnetic noise of a predetermined magnitude occurs in either frequency range, it is determined that a connection abnormality has occurred. Therefore, it is possible to make a judgment based on a phenomenon specific to the occurrence of a connection abnormality, and to detect the connection abnormality with high accuracy. Furthermore, by making a judgment based on the characteristics of both the low frequency and high frequency with the 1 MHz boundary, it is possible to accurately detect noise specific to a connection abnormality.
In addition, because the frequency to be judged is set based on learning data using a convolutional neural network, accurate judgment is possible even if the frequency spectrum characteristics of the electromagnetic noise associated with a connection abnormality differ depending on the type of load.
Furthermore, by providing a first filter circuit 11 on the upstream side of the feeder line L to the distribution board 7, which blocks electromagnetic noise generated on the upstream side, the connection abnormality detection device 1 will not malfunction due to electromagnetic noise generated on the upstream side.
Here, Fig. 7(a) shows the frequency spectrum of electromagnetic noise measured with the second filter circuit present, and Fig. 7(b) shows the frequency spectrum of electromagnetic noise measured without the second filter circuit present. As shown in Fig. 7, the second filter circuit 12 is disposed in the branch electric circuit 10a, so that electromagnetic noise generated at the load 15 connected to the end of the branch electric circuit 10a or at the connection part of the branch electric circuit 10a such as an outlet can be reduced, and the connection abnormality detection device 1 does not malfunction. Therefore, the detection range of the connection abnormality can be limited to within the distribution board 7, and accurate judgment can be made.

Although the above embodiment has been described with reference to an example of a case where loosening of a screw is detected as an electrical circuit connection portion, loosening of a quick-connect terminal that is widely used in branch breakers 14 can also be detected.
In addition, the frequency range is divided into low and high frequency ranges at the boundary of 1 MHz, and if electromagnetic noise of a certain value or more occurs in either range, it is determined that a connection abnormality has occurred, but the boundary of the frequency range does not have to be limited to 1 MHz, and may be, for example, 2 MHz. A change of this magnitude makes almost no difference in the judgment results.
Furthermore, instead of dividing the frequency domain at specific frequencies, it is possible to determine that a connection abnormality has occurred if electromagnetic noise exceeds a predetermined threshold at two different frequencies within one large frequency band (e.g., 0.1 MHz to 30 MHz), making it possible to make a determination with higher accuracy than before.
In addition, although a convolutional neural network is used as the classifier, this is not limiting, and various machine learning techniques such as decision trees and SVMs (support vector machines) can be applied.
Furthermore, if it is desired to detect connection abnormalities occurring downstream of the distribution board in addition to those occurring within the distribution board, the second filter circuit 12 is not necessary.

1: Connection abnormality detection device, 2: Measuring instrument (noise measurement section), 3: Detection section (abnormality judgment section), 4: Shutdown mechanism section, 5: Alarm section, 6: Contact output section, 10: Electrical circuit, 11: First filter circuit, 12: Second filter circuit, 13: Main breaker, 15: Load, 31: CNN identifier, 32: Memory, 33: Learning section, 34: Judgment section, L: Lead-in line (electrical circuit).

Claims (6)

A connection abnormality detection device that, when a connection abnormality occurs in a connection portion of an electric circuit, detects a discharge that occurs as a result of the connection abnormality and determines the occurrence of the connection abnormality,
a noise measuring unit arranged on the electric circuit to detect electromagnetic noise occurring on the electric circuit and output frequency spectrum information thereof;
an abnormality determination unit that detects electromagnetic noise caused by the discharge from the measured frequency spectrum information and determines whether a connection abnormality has occurred;
The abnormality judgment unit divides the frequency range into two, a low frequency range and a high frequency range, with a specific frequency as the boundary, and judges that a connection abnormality has occurred if the measured value of electromagnetic noise in both frequency ranges exceeds a predetermined threshold value . 2. The connection abnormality detection device according to claim 1 , wherein the specific frequency that divides a frequency band into two is 1 MHz. 3. The connection abnormality detection device according to claim 1, wherein the judgment frequency of the low frequency region is 0.2 MHz to 0.9 MHz, and the judgment frequency of the high frequency region is 11 MHz to 13 MHz. 4. The connection anomaly detection device according to claim 1 , wherein the anomaly determination unit learns and sets a frequency spectrum for determining a connection anomaly in both the low frequency region and the high frequency region using a convolutional neural network. 5. A distribution board connection abnormality detection system comprising: a connection abnormality detection device according to claim 1 , which is installed in a service line to a distribution board; and a first filter circuit which blocks electromagnetic noise generated upstream of the service line is disposed upstream of the service line on which the connection abnormality detection device is installed. The distribution board connection abnormality detection system according to claim 5, characterized in that a second filter circuit is arranged for each branch circuit branched out by the distribution board, the second filter circuit blocking electromagnetic noise generated in a load connected to the branch circuit from flowing into the distribution board .

Images