JP7402956B2 - Diagnostic method, edge device, partial discharge online diagnostic system and program - Google Patents

Description

The present invention relates to a diagnostic method, an edge device, a server device, and a program, and particularly relates to a partial discharge online diagnostic system, in which a sensor evaluates waveform data obtained by measuring a cable in an online state to determine whether partial discharge is present. related to diagnostic methods for determining whether

Patent Document 1 describes that the degree of progress of partial discharge is determined using a neural network obtained by machine learning.

JP2021-38927A

High-voltage cable accidents occur regularly, but conventional cable diagnosis results in power outages in many cases, resulting in the suspension of business activities. Furthermore, partial discharge has the characteristic of repeatedly occurring and extinguishing. Therefore, short-term, spot-on diagnosis cannot reliably detect partial discharges, and there is a possibility of misjudgment.

Although the technique described in Patent Document 1 uses machine learning, it is difficult to ensure reliability of diagnosis only by machine learning, considering the frequency of occurrence of cable abnormalities.

Therefore, it is an object of the present invention to provide a diagnostic method etc. suitable for realizing reliable cable diagnosis using measurement data obtained by measuring a cable while it is online. .

A first aspect of the present invention is a diagnosis method in which a sensor evaluates waveform data obtained by measuring a cable in an online state in a partial discharge online diagnosis system to determine whether or not a partial discharge exists. , the partial discharge online diagnosis system includes an edge device, the edge device includes a first determination processing unit that determines whether or not there is a suspicion of partial discharge with respect to the waveform data, and the edge device includes a first determination processing unit that determines whether or not there is a suspicion of partial discharge with respect to the waveform data, The obtained waveform acquisition time is divided into one reference segment time and a plurality of auxiliary segment times, and the first determination processing unit calculates the statistical value of the waveform data in each of the auxiliary segment times and the reference segment time. The method includes a first determination step of determining whether there is a suspicion of partial discharge by a first statistical analysis process of comparing statistical values of waveform data included in time.

A second aspect of the present invention is the diagnosis method according to the first aspect, wherein the edge device includes a measurement data processing unit that uses part of the measurement data obtained by measurement by the sensor as the waveform data. , the measurement data processing unit calculates a threshold value from the measurement data measured in the waveform sampling period, and the measurement data measured after that includes a value larger than the threshold value, and the maximum value is in the reference classification. The method includes a waveform data generation step of generating the waveform data so as to exist in time.

A third aspect of the present invention is the diagnostic method according to the first or second aspect, wherein the statistical value indicates a variation in waveform data in the reference segment time and the auxiliary segment time, and the first determination In the step, the first determination processing unit determines that the statistical value of the waveform data in each of the auxiliary segment times is greater than the product of the corresponding first statistical threshold and the statistical value of the waveform data included in the reference segment time. If the variation is small, it is determined that there is a suspicion of partial discharge, and the statistical value of the waveform data in the auxiliary segment time is set to the corresponding first statistical threshold and the statistics of the waveform data included in the reference segment time. If there is a variation that does not show that the variation is smaller than the value multiplied by the value, it is determined that there is no suspicion of partial discharge.

A fourth aspect of the present invention is the diagnostic method according to any one of the first to third aspects, wherein in the first determination step, the first determination processing section determines that the When it is determined that there is a suspicion of a discharge, it is further determined whether or not there is a suspicion of a partial discharge by determination based on the pulse width, and the pulse width is determined based on the peak point farthest from the reference voltage in the waveform data. The first determination processing unit determines that partial discharge is suspected if the pulse width is smaller than a predetermined value as a determination based on the pulse width. However, if the pulse width is larger than a predetermined value, it is determined that there is no suspicion of partial discharge.

A fifth aspect of the present invention is the diagnosis method according to any one of the first to fourth aspects, wherein the partial discharge online diagnosis system further includes a server device, and the edge device further includes an upload processing unit. The server device includes a second determination processing unit, and the upload processing unit transmits the waveform data to the server device when the first determination processing unit determines that there is a suspicion of partial discharge. and a second determination step in which the second determination processing unit determines whether the waveform data is partial discharge or noise through determination processing using one or more artificial intelligence models.

A sixth aspect of the present invention is the diagnostic method according to the fifth aspect, in which in the second determination step, the second determination processing section performs determination processing using one or more artificial intelligence models, and further comprises: It is determined whether it is a partial discharge or noise by a determination process using a second statistical analysis process using statistical values in each of a plurality of sections obtained by dividing the waveform acquisition time.

A seventh aspect of the present invention is the diagnostic method according to the fifth or sixth aspect, in which the server device includes a clustering processing section, and the clustering processing section is configured to perform the diagnostic method according to the fifth or sixth aspect, wherein the clustering processing section The method includes a step of classifying the waveform data into patterns and performing a clustering process, and an improvement step of the second determination processing unit improving the determination process by the artificial intelligence model using the results of the clustering process.

An eighth aspect of the present invention is an edge device including a first determination processing unit that evaluates waveform data obtained by measuring a cable with a sensor in an online state and determines whether there is a suspicion of partial discharge. and a first determination processing unit that determines whether or not there is a suspicion of partial discharge with respect to the waveform data, and the waveform acquisition time at which the waveform data is acquired is divided into one reference segment time and a plurality of auxiliary segments. The first determination processing unit performs a first statistical analysis that compares the statistical value of the waveform data in each of the auxiliary divided times with the statistical value of the waveform data included in the reference divided time. Through the process, it is determined whether there is a suspicion of partial discharge.

A ninth aspect of the present invention is the edge device according to the eighth aspect, including a measurement data processing section that uses a part of measurement data obtained by measurement by the sensor as the waveform data, and processes the measurement data. The unit calculates a threshold value from the measurement data measured in the waveform sampling period, and calculates the threshold value so that the measurement data measured after that includes a value larger than the threshold value and the maximum value exists in the reference segment time. The waveform data is generated.

A tenth aspect of the present invention is the edge device according to the eighth or ninth aspect, wherein the statistical value indicates a variation in waveform data in a reference segment time and an auxiliary segment time, and the first determination If the processing unit indicates that the statistical value of the waveform data in each of the auxiliary segment times has a smaller dispersion than the product of the corresponding first statistical threshold and the statistical value of the waveform data included in the reference segment time. For example, if it is determined that there is a suspicion of partial discharge, the statistical value of the waveform data in the auxiliary segment time is multiplied by the corresponding first statistical threshold and the statistical value of the waveform data included in the reference segment time. If there is something that does not indicate that the variation is small, it is determined that there is no suspicion of partial discharge.

An eleventh aspect of the present invention is a server device that determines whether a partial discharge exists by evaluating waveform data obtained by measuring a cable in an online state with a sensor, the edge device A second determination processing unit is provided that determines whether the waveform data is a partial discharge or noise through determination processing using one or more artificial intelligence models when it is determined that there is a suspicion of partial discharge.

A twelfth aspect of the present invention is the server device according to the eleventh aspect, in which the second determination processing unit performs a determination process using one or more artificial intelligence models, and also performs a determination process based on the acquired waveform data. It is determined whether it is a partial discharge or noise by a determination process using a second statistical analysis process using statistical values in each of a plurality of intervals obtained by dividing the waveform acquisition time.

A thirteenth aspect of the present invention is a program for causing a computer to function as an edge device according to any one of the eighth to tenth aspects, or a server device according to the eleventh or twelfth aspect. Note that the present invention may be understood as a computer-readable recording medium that records the program of the thirteenth aspect.

According to each aspect of the present invention, whether or not partial discharge is suspected is determined by statistical analysis processing that compares statistical values in multiple sections for measurement data obtained by measuring a cable in an online state. By doing so, highly reliable cable diagnosis can be realized. Therefore, for example, constant remote diagnosis can be realized. Furthermore, more practical diagnosis can be achieved by using, for example, determination processing using one or more artificial intelligence models, clustering determination processing in which similar waveforms are statistically aggregated, or the like.

1 is a block diagram showing an example of the configuration of a partial discharge online diagnosis system according to an embodiment of the present invention. 2 is a flow diagram showing an example of processing in the measurement data processing section 25 of FIG. 1. FIG. FIG. 3 is a first diagram for specifically explaining the threshold calculation process of step STA3 in FIG. 2; FIG. 3 is a second diagram for specifically explaining the threshold calculation process of step STA3 in FIG. 2; 3 is a diagram for specifically explaining the waveform data generation process in step STA5 in FIG. 2. FIG. FIG. 2 is a flow diagram showing an example of a process in which the first determination processing unit 27 in FIG. 1 evaluates partial discharge. (a) An example of 200 points of waveform data stored in the waveform data storage unit 29, and (b) an example of a determination process based on dispersion. FIG. 7 is a diagram for explaining a specific example of determination based on pulse width. FIG. 2 is a flow diagram showing an example of a process in which the second determination processing unit 45 of FIG. 1 evaluates partial discharge. 2 is a flow diagram illustrating an example of clustering processing by the clustering processing unit 51 of FIG. 1. FIG.

Embodiments of the present invention will be described below with reference to the drawings. Note that the embodiments of the present invention are not limited to the following examples.

FIG. 1 is a block diagram showing an example of the configuration of a partial discharge online diagnosis system according to an embodiment of the present invention.

The partial discharge online diagnosis system 1 includes a cable 3, a sensor 5, an edge device 7, a server device 9, and a management device 11.

The edge device 7 includes an edge communication section 13, an edge processing section 15, an edge storage section 17, and an edge input/output processing section 19.

The edge communication unit 13 is a device that performs processing for communication with other devices. Communication may be wired or wireless. The edge communication unit 13 includes a measurement data acquisition unit 21 and an upload processing unit 23.

The edge processing unit 15 is, for example, a processor, and is a device that performs information processing using a program. The edge processing section 15 includes a measurement data processing section 25 and a first determination processing section 27.

The edge storage unit 17 is, for example, a memory, and is a device that stores data. The edge storage section 17 includes a measurement data storage section 28 , a waveform data storage section 29 , an upload storage section 30 , and a first statistical threshold storage section 31 .

The edge input/output processing unit 19 is a device, such as a keyboard, a mouse, or a touch panel, through which the user of the edge device 7 inputs data and outputs data to the user.

The server device 9 includes a server communication section 33, a server processing section 35, and a server storage section 37.

The server communication unit 33 is a device that performs processing for communication with other devices. Communication may be wired or wireless. The server communication section 33 includes a waveform data acquisition section 39 and a diagnosis management communication section 41 .

The server processing unit 35 is, for example, a processor, and is a device that performs information processing using a program. The server processing section 35 includes a second determination processing section 45 , a batch processing section 49 , a clustering processing section 51 , and a data visualization processing section 53 .

The server storage unit 37 is, for example, a memory, and is a device that stores data. The server storage section 37 includes a waveform data storage section 55 , a second statistical threshold storage section 57 , and a data storage section 58 .

The management device 11 includes a management communication section 61, a management processing section 63, and a management input/output processing section 65.

The management communication unit 61 is a device that performs processing for communication with other devices. Communication may be wired or wireless.

The management processing unit 63 is, for example, a processor, and is a device that performs information processing using a program.

The management input/output processing unit 65 is a device, such as a keyboard, a mouse, or a touch panel, through which a user of the edge device 7 inputs data and outputs data to the user.

Partial discharge (also referred to as "PD") refers to a discharge that occurs partially in an insulator between electrodes when a voltage is applied between them. Discharges that completely bridge the electrodes are not included. For example, if there is a minute void-like defect in the insulation of an electrical device such as a cable, electrolysis will concentrate in that area, causing a weak electrical discharge. This is called partial discharge. This deteriorates the insulator and may even lead to dielectric breakdown.

In the partial discharge online diagnosis system 1, the sensor 5 is a CT (Current Transformer) type sensor that measures the voltage value of the cable 3 in an online state. Here, the online state is an uninterrupted state and a state in which there is no power outage. The sensor 5 may, for example, measure periodically, continuously, or irregularly. The sensor 5 sequentially transmits the measured data to the edge device 7 . Cable 3 is three-phase AC. The sensor 5 acquires voltage values in each layer (R phase, S phase, T phase). Note that the number of cables and/or sensors may be one or more. If there are multiple cables and/or sensors, the measurement data is transmitted in a state where the cable and/or sensor that was measured can be identified.

The measurement data acquisition unit 21 sequentially receives and acquires measurement data from the sensor 5. The measurement data acquisition unit 21 stores the received measurement data in the measurement data storage unit 28.

The measurement data processing section 25 generates waveform data including suspicion of partial discharge (PD) from the received measurement data and stores it in the waveform data storage section 29 . Waveform data generation processing will be described with reference to FIGS. 2 to 5. FIG. 2 is a flow diagram showing an example of processing in the measurement data processing section 25. 3 to 5 are diagrams for explaining each process in FIG. 2.

Furthermore, the measurement data processing unit 25 generates waveform data at regular time intervals even if there is no suspicion of partial discharge and stores it in the upload storage unit 30. For example, an error may occur in determining the suspicion of partial discharge, and waveform data may not be generated appropriately, resulting in the edge device 7 not uploading the data appropriately. By generating waveform data at regular time intervals, it can be utilized for maintenance and inspection of the edge device 7 in the server device 9 and the like.

The measurement data processing unit 25 starts a threshold update cycle timer and a waveform acquisition timer (step STA1). The measurement data processing unit 25 determines whether the waveform acquisition timer has passed the waveform sampling period T2 (step STA2). If the waveform sampling period T2 has not yet elapsed, wait until it has elapsed. When the waveform sampling period T2 has elapsed, the measured data processing unit 25 calculates a threshold value by threshold calculation processing from the measurement data from the time when the timer is started until the waveform sampling period T2 has elapsed (step STA3). The threshold value is calculated corresponding to each phase (R phase, S phase, T phase).

The measurement data processing unit 25 uses the measurement data measured in each phase to determine whether to perform waveform data generation processing that uses a threshold value to generate waveform data that includes suspicion of PD (step STA4). ). If waveform data generation processing is not to be performed, the process advances to step STA6. If the waveform data generation process is to be performed, the measurement data processing unit 25 generates waveform data including suspicion of PD (step STA5), and proceeds to step STA6.

In step STA6, the measurement data processing unit 25 determines whether the threshold update cycle timer has passed the threshold update cycle T1. If the threshold update period T1 has not elapsed, the processes of steps STA4 and STA5 are repeated until the threshold update period T1 has elapsed. When the threshold update period T1 has elapsed, the threshold update period timer and the waveform acquisition timer are reset (step STA7), and the process returns to step STA1.

The process in FIG. 2 is executed on a phase-by-phase basis. In a real environment, for example, three phases can execute processing in an independent and distributed manner. Further, it is considered desirable that the waveform sampling period T2 is set to one period with respect to the commercial frequency of the feeder to be sensed. For example, in a 60Hz environment, 167ms (1.66 million samplings) is set.

The threshold calculation process in step STA3 will be specifically explained with reference to FIGS. 3 and 4.

FIG. 3 shows an example of measurement data acquired in the waveform sampling period T2. Let the values of these measurement data be V[i] (i=0,...,n). The measured data processing unit 25 calculates the average value V _ave of the measured data acquired in the waveform sampling period T2 according to equation (eq1). In FIG. 3, line _L1 indicates the average value V _ave .

FIG. 4 shows average values maxV _ave and minV _ave of the maximum and minimum values of the acquired waveforms in the measurement data of FIG. 3 . The measurement data processing unit 25 acquires the maximum value and minimum value of every 10 data from the first data in the waveform sampling period T2. Assume that m+1 data groups of every 10 data are included from 0 to n. In FIG. 4, maxV[0] to maxV[m] are the maximum values in each data group. In FIG. 4, minV[0] to minV[m] are the minimum values in each data group. Then, the measurement data processing unit 25 calculates and obtains the average value maxV _ave of the maximum values and the average value minV _ave of the minimum values according to equations (eq2) and (eq3). In FIG. 4, line L ₂ indicates the average value maxV _ave of the maximum values, and line L ₃ indicates the average value minV _ave of the minimum values.

The measurement data processing unit 25 calculates a threshold value V _ave from the average value V ave of the acquired waveform, the maximum average value maxV _ave , and the minimum average value minV _ave . Specifically, the measurement data processing unit 25 calculates the threshold value thresholdV using equation (eq4) if the average value V _ave is 0 or more, and using equation (eq5) if the average value V _ave is less than 0. Here, a is a constant and is set to 0.00016 here. Noise is a noise magnification, and here it is set to 2.0.

The waveform data generation process in step STA5 will be specifically explained with reference to FIG.

The measurement data processing unit 25 processes 1024 pieces of measurement data from the point where the value exceeds the threshold thresholdV for the measurement data obtained after calculating the threshold value thresholdV after the waveform sampling period T2 has elapsed since the timer was started. get. FIG. 5 shows an example of measurement data acquired after the waveform sampling period T2 has elapsed since the timer was started and the threshold value thresholdV is calculated. In FIG. 5, line _L4 indicates the threshold value thresholdV. In FIG. 5, V[1023] is extracted from V[0]. The measurement data processing unit 25 identifies the point V[k] where the maximum value is stored from the extracted 1024 points. Then, the measurement data processing unit 25 acquires waveform data V[k-99], ..., V[k+100] of 100 points before and after the maximum value peakV[k] (200 points in total). However, if the array position [k] is 99 or less, it is set as 200 points from V[0] to V[199]. If the array position [k] is 923 or more, it is set as 200 points from V[824] to V[1023]. The measurement data processing section 25 stores the acquired waveform data in the waveform data storage section 29.

The evaluation of partial discharge by the first determination processing section 27 and the second determination processing section 45 will be specifically described with reference to FIGS. 6, 7, 8, and 9. For each of the 200 points of waveform data stored in the waveform data storage section 29, the first determination processing section 27 determines whether there is a suspicion of partial discharge by statistical analysis processing. The second determination processing unit 45 determines whether there is a partial discharge by comprehensively evaluating one or more artificial intelligence models (hereinafter referred to as "AI models") and statistical analysis processing. A final judgment will be made as to whether or not to do so.

FIG. 6 is a flow diagram illustrating an example of a process in which the first determination processing unit 27 evaluates partial discharge. The first determination processing section 27 waits until 200 points of waveform data are stored in the waveform data storage section 29 (step STB1). When 200 points of waveform data are stored in the waveform data storage unit 29, the first determination processing unit 27 divides the stored waveform data into a plurality of sections and calculates statistical values in each section as a first determination process. is calculated (step STB2). The first determination processing unit 27 determines whether there is a suspicion of partial discharge by statistical analysis processing (step STB3). If the first determination processing unit 27 determines that there is a suspicion of partial discharge, it stores the waveform data in the upload storage unit 30 (step STB4) and returns to step STB1. If the first determination processing unit 27 does not determine that there is a suspicion of partial discharge, the process returns to step STB1.

In step STB2, the first determination processing unit 27 divides the 200 points of data into a plurality of sections and calculates statistical values in each section. The statistical value indicates how the waveform data is scattered in each section, and includes, for example, variance and standard deviation. One of the plurality of sections is set as a reference section time, and the other sections are set as auxiliary section times. A first statistical threshold is set corresponding to each of the auxiliary segment times. The first statistical threshold storage unit 31 stores a first statistical threshold. The first determination processing unit 27 determines that partial discharge is suspected if the statistical value in each auxiliary segment time is smaller than the value obtained by multiplying the statistical value in the reference segment time by the corresponding first statistical threshold.

FIG. 7 shows (a) an example of 200 points of waveform data stored in the waveform data storage unit 29, and (b) an example of a determination process based on dispersion. In this example, 200 points of data are divided into five sections (section 1, section 2, section 3, section 4, and section 5), and the variance in each section is calculated. Section 3 located in the center is set as the standard section time, and the other sections are set as auxiliary section times. The first statistical threshold for section 1 is 0.5, the first statistical threshold for section 2 is 0.5, the first statistical threshold for section 4 is 0.8, and the first statistical threshold for section 5 is 0.7. Note that the first statistical threshold may be changed depending on the environment such as the season or noise. As shown in FIG. 7B, the first determination processing unit 27 determines that the variances of section 1, section 2, section 4, and section 5 are all 0.5, 0.5, 0.5, If the value is smaller than the product of 0.8 and 0.7, it is determined that partial discharge is suspected. On the other hand, the first determination processing unit 27 multiplies the variance of interval 1, interval 2, interval 4, and interval 5 by 0.5, 0.5, 0.8, and 0.7 by the variance of interval 3. If a value larger than the above value exists, it is determined that there is no suspicion of partial discharge.

In this way, the measurement data processing section 25 generates waveform data at multiple points with the maximum value located at the center and divides it into multiple sections, and the first determination processing section 27 determines that the variation in the waveform data in the center section is If the variation in waveform data is larger than that in the interval, it is determined that there is a suspicion of partial discharge. This makes it possible, for example, to eliminate general noise that tends to cause continuous variations in measurement data, and to identify points where data changes rapidly in a short period of time as waveform data with a high possibility of partial discharge. can.

Note that the plurality of sections may have the same length of time or may have different lengths of time. If the times are different, for example, the time including before and after the time of the largest amplitude is the reference segment time, the time before the reference segment time is one sub segment time, and the time after the reference segment time is one or more sub segment times. Set multiple sections as auxiliary section times. The lengths of the plurality of sections may be fixed or variable.

Furthermore, in the analysis based on statistical values, there is a risk that noise may be determined to be suspected of being a partial discharge, or that partial discharge may be determined to be noise, depending on the first statistical threshold value. Therefore, the first determination processing unit 27 may perform determination based on pulse width in addition to analysis based on statistical values. The first determination processing unit 27 makes a determination based on the pulse width when it is determined that there is a suspicion of partial discharge through the analysis using the statistical values, and when it is determined that there is no suspicion of partial discharge through the analysis using the statistical values, the first determination processing section 27 performs the determination based on the pulse width. No judgment will be made based on Regarding the determination based on the pulse width, the first determination processing unit 27 may perform or not perform the determination based on an instruction from a system administrator or the like.

A specific example of determination based on pulse width will be described with reference to FIG. In FIG. 8, line _L5 indicates waveform data of 200 points. Line L ₆ shows the reference voltage (0 mV in FIG. 8). Point P ₁ indicates the point on line L ₅ that is farthest from line L ₆ (hereinafter referred to as "peak point"). The voltage at the peak point _P1 is called the peak voltage. The first determination processing unit 27 calculates the measurement reference value using the following formula. Here, the pulse width adjustment value is a value of 0 or more and less than 1.
(Measurement reference value) = (absolute value of peak voltage) x (pulse width adjustment value)

The range from (-1) x (measurement reference value) to (measurement reference value) is called the measurement reference range. In FIG. 8, line _L7 indicates the measurement reference value. Line _L8 indicates (-1)×(measurement reference value). The measurement reference range is between line _L7 and line _L8 .

In FIG. 8, the first determination processing unit 27 determines the end point of the pulse width with respect to the peak point _P1 on the line _L5 . Specifically, in the waveform data of 200 points, the first determination processing unit 27 traces back before the peak point P ₁ and identifies a point P ₂ at which the measurement reference range is first changed from exceeding to not exceeding. This is defined as the left end point P ₂ of the pulse width. In addition, the first determination processing unit 27 identifies a point P ₃ after the peak point P ₁ in the 200 points of waveform data, where the measurement reference range is first changed from exceeding the range to not exceeding it, and determines this point. Let _P3 be the right end point of the pulse width. The first determination processing unit 27 calculates the difference between the left end point P ₂ of the pulse width and the right end point P ₃ of the pulse width, and determines the pulse width from the time between these points.

The first determination processing unit 27 determines that partial discharge is suspected if the pulse width is smaller than a predetermined value. The first determination processing unit 27 determines that there is no suspicion of partial discharge if the pulse width is larger than a predetermined value. For example, the first determination processing unit 27 determines that there is a suspicion of partial discharge when the pulse width satisfies the following formula, and when the pulse width does not satisfy the following formula, determines that there is no suspicion of partial discharge and that it is noise. Here, the pulse width threshold is a value given in advance.
Pulse width <= pulse width threshold x 0.1 x (1 - pulse width adjustment value)

When there are multiple peak points, the determination may be made based on the pulse width for each peak point, or the pulse widths may be compared and the determination may be made based on any one peak point. For example, the determination based on the pulse width may be performed at the peak point of the narrowest pulse width, or the determination may be performed based on the pulse width at the peak point of the widest pulse width.

FIG. 9 is a flow diagram illustrating an example of a process in which the second determination processing unit 45 evaluates partial discharge. The waveform data acquisition unit 39 requests the upload processing unit 23 to upload the waveform data stored in the upload storage unit 30 at regular time intervals. The upload processing section 23 transmits the waveform data stored in the upload storage section 30 to the waveform data acquisition section 39 (step STC1). In the server device 9, the waveform data acquisition section 39 receives the waveform data and stores it in the waveform data storage section 55. The second determination processing unit 45 performs data processing on the received waveform data as necessary, and performs a determination process as to whether or not there is a partial discharge (step STC2). The second determination processing unit 45 stores the final result of partial discharge or noise in the data storage unit 58 (step STC3).

For example, the acquisition time interval by the waveform data acquisition section 39 may be made shorter than the interval at which the measurement data processing section 25 unconditionally generates waveform data, and at least one piece of waveform data may be uploaded. If a state in which there is no waveform data in the upload storage section 30 is allowed, the second determination processing section 45 ends the process without performing the second determination process if there is no waveform data.

The second determination processing unit 45 makes a final determination as to whether or not there is a partial discharge by comprehensively evaluating one or more AI models and statistical analysis processing.

The evaluation using the AI model by the second determination processing unit 45 is performed, for example, as follows. AI models include, for example, GRU (Gated Recurrent Unit), which uses a neural network to calculate the probability of partial discharge using the characteristics of the temporal relationship through a recursive network structure; In this method, multiple decision trees are combined to create a tree structure consisting of multiple conditions, and the results are comprehensively judged to calculate the probability of partial discharge. Furthermore, GAN (Generative adversarial networks), Efficient AnoGAN, etc., focus on the restoreability of learned data and determine whether the data is normal or abnormal depending on the degree of restoration. You may also consider adding something that calculates the difference in distance between the front and rear. Generally, it is difficult to prepare training data that covers a huge number of noise patterns. By using an AI model that focuses on restorability, it is necessary to learn with only one type of data (for example, partial discharge), and it is more resistant to unknown noise patterns than other AI models. The second determination processing unit 45 may perform determination by ensemble learning using a plurality of AI models.

Furthermore, the second determination processing unit 45 makes a determination by taking statistical processing into consideration. This is similar to that performed by the first determination processing section 27. That is, the second determination processing unit 45 divides the 200 points of data into a plurality of sections and calculates statistical values in each section. The statistical value indicates how the waveform data is scattered in each section. One of the plurality of sections is set as a reference section time, and the other sections are set as auxiliary section times. A second statistical threshold is set corresponding to each of the auxiliary segment times. The second statistical threshold storage unit 57 stores the second statistical threshold. The second determination processing unit 45 determines that partial discharge is suspected if the statistical value in each auxiliary segment time is smaller than the value obtained by multiplying the statistical value in the reference segment time by the corresponding second statistical threshold.

Here, the statistical analysis processing is performed by the first judgment processing section 27 and the second judgment processing section 45, but the second judgment processing section 45 performs a strict judgment processing compared to the first judgment processing section 27. conduct. Therefore, for example, the number of sections to be divided may be different, or the first statistical threshold and the second statistical threshold may be at least partially different.

Further, the second determination processing unit 45 may perform determination based on pulse width (see FIG. 8) in addition to determination based on statistical processing. Regarding the determination based on the pulse width, the second determination processing unit 45 may perform or not perform the determination based on an instruction from a system administrator or the like.

Note that the necessary execution conditions (cycle, first statistical threshold, second statistical threshold, etc.) can be set by the system administrator as necessary, and even if they are set commonly for the system, they can be set for each company, system, or device. May be set.

The data visualization processing unit 53 performs data visualization processing on the data accumulated in the data storage unit 58. The diagnosis management communication unit 41 transmits the visualized data to the management device 11.

The management device 11 is an information processing device used by, for example, a power company that supplies power using a cable. When the management device 11 receives the visualized data, the management processing section 63 displays it on the management input/output processing section 65 so that the user can visually confirm it. Furthermore, the management processing unit 63 may further perform analysis processing and display it on the management input/output processing unit 65, and may perform determination using a unique AI model, clustering processing, etc.

The data storage unit 58 stores 200 points of waveform data and data regarding the final determination by the second determination processing unit of the waveform data as to whether it is partial discharge or noise. The clustering processing unit 51 performs clustering of waveform data using these. The data visualization processing unit 53 supports visual confirmation using the results of the clustering process. Further, the second determination processing unit 45 uses the results of the clustering process to improve the accuracy of determination using the AI model.

FIG. 10 is a flow diagram illustrating an example of clustering processing by the clustering processing unit 51. The clustering processing unit 51 determines whether to perform the clustering process (step STD1), waits until the clustering process is performed, and then performs the clustering process (step STD2). The clustering process may be performed, for example, according to an instruction from a user of the management device 11, or periodically by batch processing.

The clustering processing section 51 performs pattern classification (grouping) of the waveform data using the plurality of waveform data stored in the data storage section 58 and the final determination result, and performs clustering processing. Furthermore, in the clustering process, there is a possibility that some outliers may occur due to technical characteristics, so by performing an outlier test after the clustering process, it is possible to perform classification results with higher accuracy.

1 Partial discharge online diagnosis system 3 Cable 5 Sensor 7 Edge device 9 Server device 11 Management device 13 Edge communication section 15 Edge processing section 17 Edge storage section 19 Edge input/output processing section 21 Measured data acquisition section 23 Upload processing section 25 Measured data processing Section 27 First judgment processing section 28 Measured data storage section 29 Waveform data storage section 30 Upload storage section 31 First statistical threshold storage section 33 Server communication section 35 Server processing section 37 Server storage section 39 Waveform data acquisition section 41 Diagnosis management communication section 45 Second judgment processing section 49 Batch processing section 51 Clustering processing section 53 Data visualization processing section 55 Waveform data storage section 57 Second statistical threshold storage section 58 Data storage section 61 Management communication section 63 Management processing section 65 Management input/output processing section

Claims (9)

A diagnostic method in an edge device, the method comprising:
The edge device includes a measurement data processing section and a first determination processing section,
The measurement data processing unit converts a portion of the waveform acquisition time, which is a part of the measurement data obtained by measuring the cable while the sensor is online, into waveform data;
The waveform acquisition time includes one or more auxiliary segment times, a reference segment time following this, and one or more auxiliary segment times following this,
The measurement data processing unit calculates a threshold value from the measurement data measured in the waveform sampling period, and the measurement data measured after that includes a value larger than the threshold value, and the maximum value is within the reference segment time. a waveform data generation step in which a portion of the waveform acquisition time existing in the waveform data is set as the waveform data;
The first determination processing unit determines whether a partial discharge is suspected by performing a first statistical analysis process that compares the statistical value of the waveform data in each of the auxiliary segment times with the statistical value of the waveform data included in the reference segment time. A diagnostic method including a first determination step of determining whether or not there is. A diagnostic method in an edge device, the method comprising:
The edge device includes a measurement data processing section and a first determination processing section,
The measurement data processing unit converts a portion of the waveform acquisition time, which is a part of the measurement data obtained by measuring the cable while the sensor is online, into waveform data;
The waveform acquisition time includes one or more auxiliary segment times, a reference segment time following this, and one or more auxiliary segment times following this,
The first determination processing unit determines whether a partial discharge is suspected by performing a first statistical analysis process that compares the statistical value of the waveform data in each of the auxiliary segment times with the statistical value of the waveform data included in the reference segment time. including a first determination step of determining whether or not there is,
The statistical value is the variance and/or standard deviation of the waveform data at the reference segment time and the auxiliary segment time,
In the first determination step, the first determination processing unit:
If the statistical value of the waveform data in each of the auxiliary segment times is smaller than the product of the corresponding first statistical threshold and the statistical value of the waveform data included in the reference segment time, partial discharge is suspected. It is determined that
If there is a statistical value of the waveform data in each of the auxiliary segment times that is not smaller than the product of the corresponding first statistical threshold and the statistical value of the waveform data included in the reference segment time, then A diagnostic method that determines that there is no suspicion of electrical discharge. A diagnostic method in an edge device, the method comprising:
The edge device includes a measurement data processing section and a first determination processing section,
The measurement data processing unit converts a portion of the waveform acquisition time, which is a part of the measurement data obtained by measuring the cable while the sensor is online, into waveform data;
The waveform acquisition time includes one or more auxiliary segment times, a reference segment time following this, and one or more auxiliary segment times following this,
The first determination processing unit determines whether a partial discharge is suspected by performing a first statistical analysis process that compares the statistical value of the waveform data in each of the auxiliary segment times with the statistical value of the waveform data included in the reference segment time. including a first determination step of determining whether or not there is,
In the first determination step,
The first determination processing unit further determines whether there is a suspicion of partial discharge by determination based on the pulse width when determining that there is a suspicion of partial discharge by the determination by the first statistical analysis process,
The pulse width is the width of the portion exceeding the measurement reference range before and after the peak point farthest from the reference voltage in the waveform data,
The first determination processing unit determines that partial discharge is suspected when the pulse width is less than or equal to a predetermined value, and determines that partial discharge is suspected when the pulse width is larger than the predetermined value. A diagnostic method that determines that there is no suspicion. A diagnosis method in a partial discharge online diagnosis system comprising an edge device and a server device,
The edge device includes a measurement data processing section, a first determination processing section, and an upload processing section,
The server device includes a second determination processing section,
The measurement data processing unit converts a portion of the waveform acquisition time, which is a part of the measurement data obtained by measuring the cable while the sensor is online, into waveform data;
The waveform acquisition time includes one or more auxiliary segment times, a reference segment time following this, and one or more auxiliary segment times following this,
The first determination processing unit determines whether a partial discharge is suspected by performing a first statistical analysis process that compares the statistical value of the waveform data in each of the auxiliary segment times with the statistical value of the waveform data included in the reference segment time. a first determination step of determining whether or not there is a
an uploading step in which the upload processing unit transmits the waveform data to the server device when the first determination processing unit determines that there is a suspicion of partial discharge;
A diagnostic method including a second determination step in which the second determination processing unit determines whether the waveform data is partial discharge or noise through determination processing using one or more artificial intelligence models. comprising a measurement data processing section and a first determination processing section,
The measurement data processing unit converts a part of the waveform acquisition time, which is a part of the measurement data obtained by measuring the cable while the sensor is online, into waveform data;
The waveform acquisition time includes one or more auxiliary segment times, a reference segment time following this, and one or more auxiliary segment times following this,
The measurement data processing unit calculates a threshold value from the measurement data measured in the waveform sampling period, and the measurement data measured after that includes a value larger than the threshold value, and the maximum value is within the reference segment time. Let the part of the waveform acquisition time existing in the waveform data be the waveform data,
The first determination processing unit determines whether a partial discharge is suspected by performing a first statistical analysis process that compares the statistical value of the waveform data in each of the auxiliary segment times with the statistical value of the waveform data included in the reference segment time. An edge device that determines whether or not there is. comprising a measurement data processing section and a first determination processing section,
The measurement data processing unit converts a portion of the waveform acquisition time, which is a part of the measurement data obtained by measuring the cable while the sensor is online, into waveform data;
The waveform acquisition time includes one or more auxiliary segment times, a reference segment time following this, and one or more auxiliary segment times following this,
The first determination processing unit determines whether a partial discharge is suspected by performing a first statistical analysis process that compares the statistical value of the waveform data in each of the auxiliary segment times with the statistical value of the waveform data included in the reference segment time. Determine whether there is
The statistical value is the variance and/or standard deviation of the waveform data at the reference segment time and the auxiliary segment time,
The first determination processing unit includes:
If the statistical value of the waveform data in each of the auxiliary segment times is smaller than the product of the corresponding first statistical threshold and the statistical value of the waveform data included in the reference segment time, partial discharge is suspected. It is determined that
If there is a statistical value of the waveform data in each of the auxiliary segment times that is not smaller than the product of the corresponding first statistical threshold and the statistical value of the waveform data included in the reference segment time, then An edge device that determines that there is no suspicion of electrical discharge. comprising a measurement data processing section and a first determination processing section,
The measurement data processing unit converts a portion of the waveform acquisition time, which is a part of the measurement data obtained by measuring the cable while the sensor is online, into waveform data;
The waveform acquisition time includes one or more auxiliary segment times, a reference segment time following this, and one or more auxiliary segment times following this,
The first determination processing unit determines whether a partial discharge is suspected by performing a first statistical analysis process that compares the statistical value of the waveform data in each of the auxiliary segment times with the statistical value of the waveform data included in the reference segment time. Determine whether there is
The first determination processing unit further determines whether there is a suspicion of partial discharge by determination based on the pulse width when determining that there is a suspicion of partial discharge by the determination by the first statistical analysis process,
The pulse width is the width of the portion exceeding the measurement reference range before and after the peak point farthest from the reference voltage in the waveform data,
The first determination processing unit determines that partial discharge is suspected when the pulse width is less than or equal to a predetermined value, and determines that partial discharge is suspected when the pulse width is larger than the predetermined value. An edge device that determines that there is no suspicion. A partial discharge online diagnosis system comprising an edge device and a server device,
The edge device includes a measurement data processing section, a first determination processing section, and an upload processing section,
The server device includes a second determination processing section,
The measurement data processing unit converts a portion of the waveform acquisition time, which is a part of the measurement data obtained by measuring the cable while the sensor is online, into waveform data;
The waveform acquisition time includes one or more auxiliary segment times, a reference segment time following this, and one or more auxiliary segment times following this,
The first determination processing unit determines whether a partial discharge is suspected by performing a first statistical analysis process that compares the statistical value of the waveform data in each of the auxiliary segment times with the statistical value of the waveform data included in the reference segment time. Determine whether there is
The upload processing unit transmits the waveform data to the server device when the first determination processing unit determines that partial discharge is suspected;
The second determination processing unit is a partial discharge online diagnosis system that determines whether the waveform data is partial discharge or noise through determination processing using one or more artificial intelligence models. A program for causing a computer to function as the edge device according to claim 8.

Images