Electric power information acquisition equipment fault diagnosis method based on deep learning
Through deep learning technology, data item comparison, time delay clustering and time stamp correction of the information packets of the power information acquisition equipment are carried out, and combined with the acquisition protocol, the problem of fault diagnosis in the existing technology is solved, achieving higher accuracy and comprehensive fault identification improve the fault handling efficiency of power information collection equipment.
Patent Information
- Application Number
- CN202411054139.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-12-03
AI Technical Summary
The existing technology has one-sided nature in the fault diagnosis of power information collection equipment, and cannot fully guarantee the diagnostic accuracy of equipment faults, resulting in low fault handling efficiency and insufficient management level and operation and maintenance level.
Adopting a deep learning-based fault diagnosis method, through data item comparison, transmission delay clustering, bandwidth identification and time stamp correction of information packets, combined with power acquisition protocol, it can identify data missing and time delay abnormalities, and quickly identify communication failures. Or collect a fault.
It realizes higher accuracy and comprehensiveness in fault diagnosis of power information collection equipment, quickly identify abnormal nodes and fault types, and improves fault diagnosis efficiency and equipment management level.
Smart Images

Figure CN119071179A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of equipment fault diagnosis, and specifically to a method for fault diagnosis of electric power information acquisition equipment based on deep learning. Background Art
[0002] Electric power information collection equipment is an integral part of a complete network and system for measuring, collecting, storing, analyzing and applying user electricity consumption information. It mainly includes system master station, transmission channels, collection equipment and smart meters.
[0003] The operating status monitoring of domestic power information collection equipment is not interoperable due to different manufacturers and equipment types, and the operating status monitoring of power information collection equipment based on artificial intelligence technology is still in the initial research stage. Equipment fault analysis and processing rely too much on manual means and repetitive mechanical work, resulting in a large waste of human resources. The current fault handling efficiency can no longer meet the demand. At the same time, due to the lack of system information on the status of terminal equipment in the entire region and the ability to predict and analyze faults, the system management level and operation and maintenance level are reduced.
[0004] When diagnosing faults of electric power information collection equipment, it is generally based on the specific collection characteristics of the corresponding electric power collection equipment to assess whether there are any collection missing items, so as to assess whether there are any collection abnormalities in this collection equipment. However, the original assessment method is not comprehensive, because the collection missing items may be packet loss in network transmission, or it may be a collection failure of the collection equipment. Therefore, the original assessment method is relatively one-sided and cannot guarantee the comprehensiveness of its equipment fault diagnosis. Summary of the invention
[0005] In view of the shortcomings of the prior art, the present invention provides a fault diagnosis method for power information acquisition equipment based on deep learning, which solves the problem that the original assessment method is relatively one-sided and cannot guarantee the comprehensiveness of its equipment fault diagnosis.
[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A fault diagnosis method for power information acquisition equipment based on deep learning, comprising the following steps:
[0007] S1), transmitting the information message generated by the power collection device in the current collection stage to the terminal device, checking the data items of the information message received by the terminal device, and identifying whether there is data missing in the information message. If so, executing step S4, if not, executing step S2, the specific sub-steps are:
[0008] Based on the information message received by the terminal device, the different data items in the information message are compared with the standard message template to identify whether each group of different data items has a comparison item. If so, it means that there is no data missing in the information message and step S2 is executed. If there is a group of data items that do not have a comparison item, it means that there is data missing in the information message and step S4 is directly executed, wherein the standard message template is a preset standard template, which is prepared in advance by the operator, and a number of comparison items are marked in the standard template, and the comparison items correspond one-to-one to the collection items of the power collection equipment, and the data collected by the collection items correspond one-to-one to the data items in the information message;
[0009] S2), based on the data capacity of the information message, identify the transmission delay related to the information message data with the same data capacity from the past information message data, perform numerical clustering on the identified groups of transmission delays to confirm the standard interval, and confirm whether the transmission delay of the information message meets the standard based on the standard interval, the specific sub-steps are:
[0010] S21, based on the data capacity R of this information message i , where i represents different information messages, based on this data capacity R i Identify other information messages with the same data capacity from past information message data, and calibrate the transmission delay caused by other information messages as SY k , k represents different information messages;
[0011] S22, based on the determined transmission delays SY of the plurality of groups k , generate a time delay sequence according to the order of values from small to large, and fit this time delay sequence to a set of metric lines, the initial point of which is the same as the first group of SY in the time delay sequence k Correspondingly, based on the second group SY k With the first group SY k The delay difference is calibrated on the measurement line for the second group of SY k The numerical point at which the delay sequence is located is then based on the different SY k With the first group SY k The delay difference is determined one by one on the measurement line. k The numerical point where ;
[0012] S23. Based on the plurality of numerical points determined on the measurement line, the measurement length DLq between adjacent numerical points is confirmed one by one, where q represents the numerical segment between different adjacent numerical points, and a set of virtual segments with a length of DLqmax is generated:
[0013] Move the virtual segment from the initial point to the end point of the measurement line, record the number of numerical points Gp contained in the virtual segment of each different moving process, where p represents a different moving process, and use DLqmax÷Gp=Fp to confirm the mean value Fp associated with each different moving process;
[0014] S24, increasing the length of the virtual segment by a set of measurement units, and executing step S23 to confirm the mean values Fp of several components associated with the virtual segment;
[0015] S25, repeating step S24 until the length of the virtual segment increases to the same length as the measuring line, and confirming the mean values Fp of several components associated with virtual ends of different lengths one by one;
[0016] S26, from the several component mean values Fp confirmed in step S23-step S25, select the maximum value Fpmax, and record the virtual segment corresponding to the maximum value Fpmax and the numerical points contained in the virtual segment, and determine the SY corresponding to the numerical point k , from the determined groups SY k Select the minimum and maximum values to generate the standard interval;
[0017] S27, select the information message transmission delay CS, confirm whether the CS belongs to the standard interval, if CS∈standard interval, no processing is performed; if In the standard interval, execute step S3 to perform correlation analysis;
[0018] S3) confirm the bandwidth used by the information message during transmission and the standard bandwidth of the corresponding transmission channel, compare the bandwidth used with the standard bandwidth to determine the recognition result, and then generate and display the delay signal based on the recognition result. The specific sub-steps are:
[0019] S31, performing average processing on the bandwidth used by several groups of the information message during the transmission process to determine the average bandwidth Dk;
[0020] S32, comparing the confirmed average bandwidth DK with the standard bandwidth BZ, where BZ is a preset value, if DK<BZ-Y1, where Y1 is a floating value, then generating a transmission congestion delay signal for display; if DK≥BZ-Y1, then generating a device communication failure signal for display;
[0021] S4), according to the timestamp of the data item in the information message, determine its timestamp sequence, correct the internal events of the timestamp sequence, determine the standard time difference sequence, and then identify whether the power collection device has a collection fault based on the collection protocol of the power collection device; the specific sub-steps are:
[0022] S41, according to the data items determined in sequence in the information message, the timestamps of the corresponding data items are sorted in the same way as the previous and next data items to generate a timestamp sequence, and then a standard time difference sequence is generated based on the time difference values of adjacent timestamps in the timestamp sequence;
[0023] S42, confirming the time intervals between corresponding collection items according to the set collection protocol, and generating a time interval sequence;
[0024] S43. According to the location of the data missing item in the information message, identify the time intervals before and after the data missing item from the time interval sequence, then sum the time intervals to determine the total time value, and select the time difference value corresponding to the data missing item from the standard time difference sequence. If the total time value = the time difference value, then directly generate a communication fault signal for display; if the total time value ≠ the time difference value, then directly generate a corresponding collection item node collection fault signal for display.
[0025] The present invention provides a fault diagnosis method for power information collection equipment based on deep learning. Compared with the prior art, it has the following beneficial effects:
[0026] The present invention confirms the data capacity of the information message, and then determines the standard delay interval from different information message data based on different information messages corresponding to the data capacity, and then evaluates whether the delay data of the information message is normal based on the standard delay interval. This evaluation method can fully guarantee the specific accuracy of its evaluation and achieve a better delay evaluation effect;
[0027] When there are missing data items in the information message, the time difference between the timestamps of the corresponding data items is evaluated, and then the time difference is determined based on the collection protocol to determine whether it is abnormal. Based on the specific determination results, it is evaluated whether the power collection equipment has collection abnormalities. This evaluation method can quickly identify the corresponding abnormal nodes and abnormal situations, and evaluate whether the collection equipment has communication failures or collection failures. The fault diagnosis method is more comprehensive and can achieve better equipment fault diagnosis results. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 It is a schematic diagram of the process of the present invention;
[0029] Figure 2 Schematic diagram of the processing of the measurement line of the present invention. DETAILED DESCRIPTION
[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0031] Example 1
[0032] See also Figure 1 The present application provides a method for diagnosing faults of electric power information collection equipment based on deep learning. When the electric power information collection equipment performs fault diagnosis, it is generally based on the specific collection items of the corresponding electric power information. The collection equipment generates relevant messages for the collection items of each stage and transmits them to the corresponding terminal. Based on whether there are missing items or response delays in the messages, fault diagnosis is performed to identify whether the collection equipment has communication failures or collection failures, and relevant displays are made in a timely manner. The method specifically includes the following steps:
[0033] S1), the information message generated by the power collection device in the current collection stage is transmitted to the terminal device, and the data item of the information message received by the terminal device is checked to identify whether there is data missing in the information message. If so, step S4 is executed, if not, step S2 is executed. Specifically, for the generated information message, data item comparison is performed first to identify whether there are missing items in the information message after transmission. If there are missing items, it may be that a certain power collection node of the power information collection device has a fault, or it may be a communication fault. If there are no missing items, delay analysis is performed to identify whether there is a problem with its transmission delay. If the transmission delay does not meet the standard, it may be that the terminal device has network congestion, or it may be a communication fault. One by one analysis is performed to diagnose the faults existing in the power collection device one by one, wherein the specific sub-steps for identifying whether there is data missing are as follows:
[0034] S11. Based on the information message received by the terminal device, the different data items in the information message are compared with the standard message template to identify whether each group of different data items has a comparison item. If so, it means that there is no data missing in the information message and step S2 is executed. If there is a group of data items that do not have a comparison item, it means that there is data missing in the information message and step S4 is directly executed. The standard message template is a preset standard template, which is prepared in advance by the operator, and a number of comparison items are marked in the standard template. The comparison items correspond to the collection items of the power collection device one by one, and the data collected by the collection items correspond to the data items in the information message one by one. Specifically, the data items are different data items collected by the power collection device, which are current data, voltage data, power data and other power data;
[0035] S2), based on the data capacity of the information message, identify the transmission delay related to the information message data with the same data capacity from the past information message data, perform numerical clustering on the identified groups of transmission delays to confirm the standard interval, and confirm whether the transmission delay of the information message meets the standard based on the standard interval, wherein the specific sub-steps of confirming are:
[0036] S21, based on the data capacity R of this information message i , where i represents different information messages, based on this data capacity R i Identify other information messages with the same data capacity from past information message data, and calibrate the transmission delay caused by other information messages as SY k , k represents different information messages;
[0037] S22, based on the determined transmission delays SY of the plurality of groups k , generate a time delay sequence according to the order of values from small to large, and fit this time delay sequence to a set of metric lines, the initial point of which is the same as the first group of SY in the time delay sequence k Correspondingly, based on the second group SY k With the first group SY k The delay difference is calibrated on the measurement line for the second group of SY k The numerical point at which the delay sequence is located is then based on the different SY k With the first group SY k The delay difference is determined one by one on the measurement line. k The numerical point where ;
[0038] S23. Based on the number of numerical points determined on the measurement line, the measurement length DLq between the adjacent numerical points is confirmed one by one, where q represents the numerical segment between different adjacent numerical points, and a set of virtual segments with a length of DLqmax is generated (this DLqmax is the maximum value of the number of measurement lengths DLq):
[0039] Move the virtual segment from the initial point to the end point of the measurement line, record the number of numerical points Gp contained in each virtual segment of each different movement process, where p represents a different movement process, and each movement process is separated by a set of measurement units, which are marked on the measurement line, and use DLqmax÷Gp=Fp to confirm the mean value Fp associated with each different movement process;
[0040] S24, increasing the length of the virtual segment by a set of measurement units, and executing step S23 to confirm the mean values Fp of several components associated with the virtual segment;
[0041] S25, repeating step S24 until the length of the virtual segment increases to the same length as the measuring line (that is, the length of the virtual line = the length of the measuring line), and confirming the mean values Fp of several components associated with virtual ends of different lengths one by one;
[0042] S26, from the several component mean values Fp confirmed in step S23-step S25, select the maximum value Fpmax, and record the virtual segment corresponding to the maximum value Fpmax and the numerical points contained in the virtual segment, and determine the SY corresponding to the numerical point k , from the determined groups SY k Select the minimum and maximum values to generate the standard interval;
[0043] S27, select the information message transmission delay CS, confirm whether this CS belongs to this standard interval, if CS∈standard interval, no processing is performed, if If the error is in the standard range, it means that there is a problem with the transmission delay of the information message, and a fault diagnosis is required to identify whether it is a communication problem or a device problem, and execute step S3 for relevant analysis;
[0044] Specific, combined Figure 2 , the generated virtual segment moves on the metric line, Figure 2 In order to better illustrate, the virtual segment is placed above the measurement line. When actually moving, the virtual segment is located on the measurement line. When the virtual segment moves, it moves from the initial point to the end point, so that the numerical points contained in the virtual segment can be determined, and the length of the virtual segment in each processing process also changes. The length of the virtual segment in the first stage is the smallest, and the length of the virtual segment in the last stage is the largest. The length of the virtual segment changes from small to large, and the unit length of the virtual segment changes one by one, each time changing a measurement length.
[0045] S3) confirming the bandwidth used by the information message during transmission and the standard bandwidth of the corresponding transmission channel, comparing the used bandwidth with the standard bandwidth to determine the recognition result, and then generating and displaying the delay signal based on the recognition result, wherein the specific sub-steps of recognition are:
[0046] S31, performing average processing on the bandwidth used by several groups of the information message during the transmission process, and determining the average bandwidth Dk (the bandwidth used is affected by the congestion of other transmission data, and the bandwidth transmission will change accordingly, which can be understood as the transmission capacity per unit time);
[0047] S32. Compare the confirmed average bandwidth DK with the standard bandwidth BZ, where BZ is a preset value, which is prepared in advance by the operator. If DK<BZ-Y1, where Y1 is a floating value, generally taking a value of 0.5, which is prepared in advance by the operator, a transmission congestion delay signal is generated for display for external personnel to view. If DK≥BZ-Y1, a device communication fault signal is generated for display for external personnel to view. Specifically, if the bandwidth of the message during transmission is the specific bandwidth under normal use, then the corresponding delay caused by the communication fault of the device itself during transmission of such information message, if the bandwidth of the message during transmission is not the relevant bandwidth under normal use, then such information message is affected by the corresponding congestion during transmission, resulting in relevant delays in the information message generated by the collection;
[0048] S4), according to the timestamp of the data item in the information message, determine its timestamp sequence, correct the internal events of the timestamp sequence, determine the standard time difference sequence, and then identify whether the power collection device has a collection fault based on the collection protocol of the power collection device, wherein the specific sub-steps of identification are:
[0049] S41. When the power collection device collects power data, it collects data from different power nodes based on the corresponding collection protocol. Then the collected data items are generated one by one, and the generated data items all have corresponding timestamps, and the collected data items are collected one by one. That is to say, there are three nodes A, B, and C, and the collection protocol is ABC. Then the generated data items are A node data, B node data, and C node data. According to the data items determined in sequence in the information message, the timestamps of the corresponding data items are sorted in the same way as the previous and next data items to generate a timestamp sequence, and then based on the time difference between adjacent timestamps in the timestamp sequence, a standard time difference sequence is generated;
[0050] S42, confirming the time intervals between corresponding collection items according to the set collection protocol, and generating a time interval sequence;
[0051] S43, according to the location of the missing data item in the information message, identify the time intervals before and after the missing data item from the time interval sequence (if the missing data item exists continuously, then the several groups of time intervals before and after it are uniformly summed, and the same time interval does not participate in the two summing processes, that is, the data items are ABCDE, and the missing items are C and D, wherein the corresponding time differences are J1-J2-J3-J4, wherein C and D are missing, then the time intervals to be summed are J2-J3-J4), then sum the time intervals to determine the total time value, select the time difference value corresponding to the missing data item from the standard time difference sequence, if the total time value = the time difference value, then it means that the time difference value comparison is normal, then such missing data item has packet loss during the data transmission process, then directly generate a communication fault signal for display, if the total time value ≠ the time difference value, then it means that the time comparison is inconsistent, then the corresponding collection item fails to collect relevant data during collection, then directly generate the corresponding collection item node collection fault signal and display it;
[0052] Specifically, the collection protocol is ABCDE, and several data item nodes are collected in sequence. A corresponding time interval is set between each data item in the collection protocol. Then the time interval sequence belongs to J1-J2-J3-J4, where J1 is the time interval between A and B, J2 is the time interval between B and C, J3 is the time interval between C and D, and J4 is the time interval between D and E;
[0053] When the generated data item is ABCE, then data item D is missing, and the generated time difference sequence is X1-X2-X3. Then the time difference X1 between A and B corresponds to J1, so they are equal, then X2 corresponds to J2. If data D is lost during the transmission process, it means that the device has completed the relevant collection, then X3 = J3 + J4, that is, the total time value = the time difference value;
[0054] If X3≠J3+J4, it means that the corresponding acquisition device has an acquisition failure when acquiring data item D, so it stops acquiring and directly executes the acquisition process of data item E. Then the generated time difference has related errors, resulting in unequal time difference values.
[0055] Some of the data in the above formulas are numerically calculated by removing their dimensions. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.
[0056] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.
Claims
1. A fault diagnosis method for power information collection equipment based on deep learning, characterized in that: The following steps are involved: S1), transmitting the information message generated by the power collection device in the current collection stage to the terminal device, checking the data items of the information message received by the terminal device, and identifying whether there is data missing in the information message, if so, executing step S4, if not, executing step S2; S2), based on the data capacity of the information message, identifying the transmission delay related to the information message data with the same data capacity from the past information message data, clustering the identified groups of transmission delays to determine the standard interval, and determining whether the transmission delay of the information message meets the standard based on the standard interval; S3) confirming the bandwidth used by the information message during transmission and the standard bandwidth of the corresponding transmission channel, comparing the used bandwidth with the standard bandwidth to determine the recognition result, and then generating and displaying a delay signal based on the recognition result; S4) Determine the timestamp sequence based on the timestamp of the data item in the information message, correct the internal events of the timestamp sequence, determine the standard time difference sequence, and then identify whether the power collection device has a collection fault based on the collection protocol of the power collection device.
2. The method for fault diagnosis of electric power information collection equipment based on deep learning according to claim 1, characterized in that: In step S1, the specific sub-steps of identifying whether the information message has data missing are: Based on the information message received by the terminal device, the different data items in the information message are compared with the standard message template to identify whether each group of different data items has comparison items. If so, it means that there is no missing data in the information message and step S2 is executed. If there is a group of data items that do not have comparison items, it means that there is missing data in the information message and step S4 is directly executed, wherein the standard message template is a preset standard template prepared in advance by the operator, and a number of comparison items are calibrated in the standard template, and the comparison items correspond one-to-one to the collection items of the power collection equipment, and the data collected by the collection items correspond one-to-one to the data items in the information message.
3. The method for fault diagnosis of electric power information collection equipment based on deep learning according to claim 1 is characterized in that: In step S2, the specific sub-steps of clustering the values of several groups of transmission delays to determine the standard interval are: S21, based on the data capacity R of this information message i , where i represents different information messages, based on this data capacity R i Identify other information messages with the same data capacity from past information message data, and calibrate the transmission delay caused by other information messages as SY k , k represents different information messages; S22, based on the determined transmission delays SY of the plurality of groups k , generate a time delay sequence according to the order of values from small to large, and fit this time delay sequence to a set of metric lines, the initial point of which is the same as the first group of SY in the time delay sequence k Correspondingly, based on the second group SY k With the first group SY k The delay difference is calibrated on the measurement line for the second group of SY k The numerical point at which the delay sequence is located is then based on the different SY k With the first group SY k The delay difference is determined one by one on the measurement line. k The numerical point where ; S23. Based on the plurality of numerical points determined on the measurement line, the measurement length DLq between adjacent numerical points is confirmed one by one, where q represents the numerical segment between different adjacent numerical points, and a set of virtual segments with a length of DLqmax is generated: Move the virtual segment from the initial point to the end point of the measurement line, record the number of numerical points Gp contained in the virtual segment of each different moving process, where p represents a different moving process, and use DLqmax÷Gp=Fp to confirm the mean value Fp associated with each different moving process; S24, increasing the length of the virtual segment by a set of measurement units, and executing step S23 to confirm the mean values Fp of several components associated with the virtual segment; S25, repeating step S24 until the length of the virtual segment increases to the same length as the measuring line, and confirming the mean values Fp of several components associated with virtual ends of different lengths one by one; S26, from the several component mean values Fp confirmed in step S23-step S25, select the maximum value Fpmax, and record the virtual segment corresponding to the maximum value Fpmax and the numerical points contained in the virtual segment, and determine the SY corresponding to the numerical point k , from the determined groups SY k Select the minimum and maximum values to generate the standard interval; S27. Select the information message transmission delay CS and confirm whether the CS belongs to the standard interval. If CS∈standard interval, no processing is performed.
4. The method for fault diagnosis of electric power information collection equipment based on deep learning according to claim 3 is characterized in that: In step S27, if Execute step S3 to perform correlation analysis.
5. The method for fault diagnosis of electric power information collection equipment based on deep learning according to claim 4 is characterized in that: In step S3, the specific sub-steps of identifying and comparing the used bandwidth with the standard bandwidth are as follows: S31, performing average processing on the bandwidth used by several groups of the information message during the transmission process to determine the average bandwidth Dk; S32. Compare the confirmed average bandwidth DK with the standard bandwidth BZ, where BZ is a preset value. If DK<BZ-Y1, where Y1 is a floating value, a transmission congestion delay signal is generated for display.
6. The method for fault diagnosis of electric power information collection equipment based on deep learning according to claim 5 is characterized in that: In step S32, if DK≥BZ-Y1, a device communication fault signal is generated for display.
7. The method for fault diagnosis of electric power information collection equipment based on deep learning according to claim 1, characterized in that: In step S4, the specific sub-steps for identifying whether the power collection device has a collection fault are: S41, according to the data items determined in sequence in the information message, the timestamps of the corresponding data items are sorted in the same way as the previous and next data items to generate a timestamp sequence, and then a standard time difference sequence is generated based on the time difference values of adjacent timestamps in the timestamp sequence; S42, confirming the time intervals between corresponding collection items according to the set collection protocol, and generating a time interval sequence; S43. According to the location of the missing data item in the information message, identify the time intervals before and after the missing data item from the time interval sequence, then sum the time intervals to determine the total time value, select the time difference value corresponding to the missing data item from the standard time difference sequence, if the total time value = the time difference value, directly generate a communication fault signal for display.
8. The method for fault diagnosis of electric power information collection equipment based on deep learning according to claim 7, characterized in that: In the step S43, if the total time value ≠ the time difference value, a corresponding collection item node collection fault signal is directly generated and displayed.
