Low-voltage insulator on-line monitoring method and system based on edge computing

Abstract

The present application relates to the technical field of low-voltage insulator online monitoring, in particular to a low-voltage insulator online monitoring method and system based on edge computing, which comprises the following steps: identifying pulses and pulse interval boundaries based on sampled current, extracting insulation abnormal sections in combination with pulse interval density distribution, and determining abnormal trigger segments according to the overlapping relationship between peak value period and amplitude change; fusing two types of time sequence information to generate insulation state change trajectories, and determining the corresponding insulation deterioration type by matching with sample library trajectories. Through hierarchical identification of time sequence current sampling values, in combination with pulse interval distribution and peak value change characteristics, the present application can effectively distinguish between incidental disturbance and persistent insulation abnormality; and through cross-discrimination of abnormal sections and trigger segments to construct state change trajectories, the present application can further realize accurate identification of insulation deterioration type, thereby improving early warning capability, abnormal positioning accuracy and repair pertinence.

Description

Technical Field

[0001] This invention relates to the field of online monitoring technology for low-voltage insulators, and in particular to an online monitoring method and system for low-voltage insulators based on edge computing. Background Technology

[0002] The field of online monitoring technology for low-voltage insulators encompasses methods for deploying insulation status sensing devices for distribution lines, methods for collecting leakage current from insulator surfaces, coupled detection techniques for temperature, humidity, and pollution levels, edge-side data acquisition and local computation processing workflows, and remote status transmission mechanisms. Its core content revolves around continuously acquiring and determining the operating status of pole-type or pin-type insulators in low-voltage distribution networks. This involves the construction of current sampling circuits, measurement of insulator surface conductivity changes, synchronous acquisition of environmental parameters, data caching and hierarchical uploading at field nodes, and multi-node collaborative deployment methods, thereby forming a complete monitoring system from the sensing end to the transmission end and then to the analysis end.

[0003] Among them, the low-voltage insulator online monitoring method and system based on edge computing refers to an implementation method in which acquisition nodes are deployed close to the insulators, and the acquired data is processed locally by a local computing unit to complete the status determination. Its technical aspects include: sampling leakage current using a sampling resistor and obtaining instantaneous current values ​​through analog-to-digital conversion; calculating the mean and extracting the peak value of the sampling sequence using a fixed time window; segmenting and marking the environment based on temperature and humidity sensor data; classifying and determining the data according to preset threshold intervals; periodically packaging the determination results and raw data; uploading them to the main station via low-power wireless communication; and simultaneously setting up a circular storage area within the edge nodes to retain historical sampling records.

[0004] Existing processing methods tend to focus on average, peak, and threshold classifications, which can reflect the level of values ​​but are difficult to show the evolution process of anomalies from their inception to expansion. When faced with pulsed or intermittent leakage currents, short-term dense anomalies are easily diluted by the average, and periodic changes and clustering characteristics are also difficult to reveal, resulting in inconspicuous early signs. Although environmental and current data are collected simultaneously, the results mostly remain at the level output, and the ability to compare across time periods and judge the deterioration trend is weak, which can easily lead to false alarms and missed alarms, and affect the scheduling of maintenance priorities. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide a method and system for online monitoring of low-voltage insulators based on edge computing. The technical solution is as follows:

[0006] A method for online monitoring of low-voltage insulators based on edge computing includes the following steps:

[0007] S1: Obtain the current sampling value and timestamp of the low-voltage insulator, identify the pulse time point that reaches the set leakage current pulse amplitude within the continuous rising interval, construct the pulse interval sequence, determine the start and end positions of the pulse interval sequence, and generate the boundary of the insulator pulse interval sequence.

[0008] S2: Based on the boundary of the insulator pulse interval sequence, obtain the pulse interval sequence and its start and end positions, determine the pulse interval sequence selection interval, divide multiple time segments, extract continuous clustered time segments, form an abnormal time segment time range sequence, and generate a set of insulation abnormal segments.

[0009] S3: Based on the boundary of the insulator pulse interval sequence, correlate the current sampling value and the timestamp, extract local high points to form a peak sequence, construct a periodic expression sequence and an amplitude change expression sequence, extract the overlapping time range of the two types of sequences, form a trigger time segment time range sequence, and generate an insulation anomaly trigger segment;

[0010] S4: Based on the time range sequence of abnormal time segments, call the time range sequence of trigger time segments, compare them one by one and assign corresponding state identifiers to form a state sequence expression, check the change direction of adjacent state identifiers to form a direction expression sequence and convert it into a discrete coding sequence to generate the insulation state change trajectory.

[0011] S5: Extract the discrete coding sequence from the insulation state change trajectory, compare it with the coding sequence of each trajectory sample in the trajectory sample library, screen candidate trajectory samples with consistent arrangement structure and change trend, read the corresponding insulation degradation type, establish the correspondence with the insulation state change trajectory, and output the insulation degradation type result.

[0012] As a further aspect of the present invention, the step of obtaining S1 is as follows:

[0013] S101: Obtain the current transformer output current sampling value and corresponding sampling timestamp of the low-voltage insulator. In the same distribution branch, the current sampling values ​​of each low-voltage insulator are divided into a fixed sampling window in chronological order. The adjacent current sampling values ​​are compared point by point. The continuously rising sampling points are merged into continuously rising intervals. The start and end positions of each continuously rising interval are recorded to obtain the rising interval position sequence.

[0014] S102: Based on the current sampling value and timestamp in the rising interval position sequence, the sampling time corresponding to the sampling point where the current sampling value in each consecutive rising interval reaches the set leakage current pulse amplitude is recorded as the pulse corresponding time point. The time interval between each pulse corresponding time point is arranged in chronological order to form a pulse interval sequence.

[0015] S103: Based on the pulse interval sequence, check the arrangement position of each pulse interval within the sampling window, identify the time positions corresponding to the first and last pulse intervals in the pulse interval sequence, and mark them as the start position and end position of the pulse interval sequence, respectively, to generate the boundary of the insulator pulse interval sequence.

[0016] As a further aspect of the present invention, the step of obtaining S2 is as follows:

[0017] S201: Based on the boundary of the insulator pulse interval sequence, check the correspondence between each pulse interval and the start and end positions according to the pulse interval sequence order, mark each pulse interval falling between the two boundaries, and remove the contents outside the boundaries to form the pulse interval sequence selection range.

[0018] S202: Select an interval based on the pulse interval sequence, read the current sampling period of the corresponding sampling window, divide the sampling window time axis into multiple time segments in an integer multiple order of the sampling period, assign segments according to the time position of each pulse interval, and form a time segment distribution sequence.

[0019] S203: For the time segment distribution sequence, read the number of pulse intervals in each time segment and check them against the preset pulse interval density threshold. Mark the time segments with the number exceeding the threshold and adjacent to each other as continuous clustering time segments, and record their start and end times in sequence to obtain the clustering time range sequence.

[0020] S204: Based on the aggregation time range sequence, read the start and end times of each continuous aggregation time segment, check the time interval between adjacent segments, splice the beginning and end of segments with an interval of zero or less than the preset segment interval threshold, keep segments with an interval greater than the threshold independent, and output them in chronological order to generate a set of insulation abnormal segments.

[0021] As a further aspect of the present invention, the step of obtaining S3 is as follows:

[0022] S301: Based on the boundary of the insulator pulse interval sequence, associate the current sampling value and the timestamp within the same sampling window, compare the amplitude relationship of the current sampling value corresponding to adjacent sampling points, mark the position where the current sampling value corresponding to the adjacent sampling points is lower than the current sampling value of the current sampling point as a local high point, organize the timestamp and current sampling value corresponding to each local high point, and obtain the local high point sequence.

[0023] S302: Based on the local high point sequence, read the timestamps and current sampling values ​​corresponding to adjacent local high points, check the direction of change of time interval between adjacent local high points and the state of change of current sampling values ​​along the time sequence, connect the time intervals of adjacent local high points that show a continuous decrease in time interval between adjacent local high points to form a periodic contraction segment, and connect the time intervals of adjacent local high points that show a continuous increase in current sampling values ​​between adjacent local high points to form an amplitude expansion segment, and arrange the periodic contraction segment and amplitude expansion segment according to the time axis to obtain overlapping candidate segments;

[0024] S303: Call the overlapping candidate segment, extract the overlapping part of the periodic contraction segment and the amplitude expansion segment on the time axis as the overlapping time range, record the start time and end time corresponding to each overlapping time range, and arrange them in the order of the start time to form a trigger time segment time range sequence, and generate an insulation abnormality trigger segment.

[0025] As a further aspect of the present invention, the step of obtaining S4 is as follows:

[0026] S401: Based on the time range sequence of abnormal time segments in the set of insulation abnormal segments, call the start time and end time of each abnormal time segment in the unified sampling window, retrieve the time range of each trigger time segment in the time range sequence of trigger time segments in the insulation abnormal trigger segment, and arrange them according to the same sampling window to obtain the window comparison sequence.

[0027] S402: According to the window comparison sequence, check the time range of abnormal time segments and the time range of trigger time segments in each sampling window, assign the first type of status identifier to the sampling window that appears simultaneously with two types of time ranges, assign the second type of status identifier to the sampling window that appears only with one type of time range, assign the basic status identifier to the sampling window that does not appear with either type of time range, and sort them according to time sequence to obtain the status sequence expression.

[0028] S403: Call the state sequence expression, read the positions of adjacent state identifiers, check the change direction between adjacent state identifiers in turn, and concatenate the change directions in sequence to generate a direction expression sequence;

[0029] S404: Establish a pre-established coding correspondence between the state change direction and the discrete marker. For the direction expression sequence, map each change direction to a corresponding discrete marker according to the coding correspondence, form a discrete coding sequence along the time axis, and organize the discrete coding sequence with the sampling window order to generate the insulation state change trajectory.

[0030] As a further aspect of the present invention, the step of obtaining S5 is as follows:

[0031] S501: Extract the discrete coding sequence from the insulation state change trajectory, read the arrangement position of each discrete marker, retrieve the coding sequence of each trajectory sample in the trajectory sample library, and compare the arrangement relationship of each discrete marker in the coding sequence item by item according to the coding segment with the same length as the discrete coding sequence. Retain the trajectory samples in the trajectory sample library that correspond to the discrete coding sequence in terms of arrangement structure, and sort them in order according to the position in the sample library to obtain the structure-corresponding sample sequence.

[0032] S502: Based on the sample sequence corresponding to the structure, read the corresponding encoding sequence of each trajectory sample, check the connection and change relationship between each discrete mark and the adjacent discrete mark along the encoding arrangement direction, and compare the connection and change relationship with the connection and change relationship of the corresponding position of the discrete encoding sequence in the insulation state change trajectory segment by segment, select the trajectory samples with consistent change direction, and arrange them according to the order of the sample library to obtain the candidate trajectory sample set.

[0033] S503: Call the candidate trajectory sample set, read the insulation degradation type corresponding to each trajectory sample, extract the corresponding mark content of each insulation degradation type in the order of the trajectory samples arranged in the sample library, establish a correspondence between each insulation degradation type and the insulation state change trajectory and output it. When there are multiple insulation degradation types in the candidate trajectory sample set, determine the insulation degradation type according to the preset matching frequency and output it to generate the insulation degradation type result.

[0034] As a further aspect of the present invention, the specific operation of arranging the trigger time segment time range sequence according to the chronological order of the start time is as follows:

[0035] The start and end times corresponding to each overlapping time range are extracted one by one. The overlapping time ranges are sorted in ascending order of start time. During the sorting process, the temporal relationship between adjacent overlapping time ranges is checked in turn. When there are overlapping time ranges with the same start time or overlapping time, they are sorted in ascending order of end time. The sorted overlapping time ranges are arranged in the order of time axis to form the trigger time segment time range sequence.

[0036] An online monitoring system for low-voltage insulators based on edge computing, the system comprising:

[0037] Pulse sampling and recognition module: acquires low-voltage insulator current sampling values ​​and timestamps, identifies pulse time points that reach the set leakage current pulse amplitude, constructs pulse interval sequences and determines start and end positions, and generates insulator pulse interval sequence boundaries;

[0038] Pulse interval identification module: Based on the boundary of the insulator pulse interval sequence, determine the pulse interval sequence selection interval, divide the time segment, extract the continuous clustered time segment, form the time range sequence of abnormal time segment, and generate a set of insulation abnormal segments;

[0039] Periodic amplitude analysis module: Based on the boundary of the insulator pulse interval sequence, correlated sampling data, extracted peak sequence, constructed periodic and amplitude change expression sequence, determined the overlapping time range, formed the trigger time segment time range sequence, and generated insulation anomaly trigger segment;

[0040] State change coding module: Based on the time range sequence of abnormal time segments, compare the abnormal time segments with the trigger time segments, assign state identifiers, form a state sequence, and convert it into a discrete coding sequence to generate the insulation state change trajectory;

[0041] Trajectory Sample Comparison Module: Extracts discrete coded sequences from insulation state change trajectories, compares them with the trajectory sample library, filters out matching samples, reads the corresponding insulation degradation type, and outputs the insulation degradation type result.

[0042] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0043] In this invention, by performing hierarchical identification around the time-series current sampling value, the corresponding time point of the pulse is extracted from the continuous rising interval and the pulse interval sequence boundary is formed. Based on the pulse interval distribution in each time segment, continuous clustering segments are identified, which can distinguish between occasional disturbances and continuous anomalies. The trigger segment is extracted by combining the time span change and amplitude change of the local peak. After cross-judgment with the abnormal time segment, the state change trajectory is formed. The degradation type is output by comparing the sample sequence, which improves the early identification capability, anomaly location accuracy and maintenance direction. Attached Figure Description

[0044] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0045] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0046] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0047] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0048] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0049] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0050] Example 1

[0051] Please see Figure 1 This embodiment provides a technical solution: an online monitoring method for low-voltage insulators based on edge computing, comprising the following steps:

[0052] S1: Obtain the current transformer output current sampling value and corresponding sampling timestamp of the low-voltage insulator. In the same distribution branch, the current sampling values ​​of each low-voltage insulator are divided into a fixed sampling window in chronological order. The adjacent current sampling values ​​are compared point by point along the time sequence. The position that reaches the set leakage current pulse amplitude in the continuous rising interval is selected as the pulse corresponding time point. The pulse interval sequence is constructed based on the time interval between the corresponding time points of adjacent pulses. The lower boundary position and upper boundary position of the time span are identified in the pulse interval sequence and marked as the start position and end position of the pulse interval sequence, respectively, to generate the insulator pulse interval sequence boundary.

[0053] S2: Based on the boundary of the insulator pulse interval sequence, obtain the pulse interval sequence, the start position of the pulse interval sequence, and the end position of the pulse interval sequence for the corresponding sampling window. Mark the range between the start position and the end position of the pulse interval sequence as the pulse interval sequence selection interval. Divide the time into multiple time segments according to integer multiples of the current sampling period. Map each pulse interval to the corresponding time segment and count the pulse interval distribution within the time segment. Mark the time segments that are adjacent and whose number of pulse intervals is higher than the preset pulse interval density threshold as continuous aggregation time segments. Record the start and end times of each continuous aggregation time segment. Form an abnormal time segment time range sequence in chronological order to generate an insulation abnormal segment set.

[0054] S3: Based on the boundary of the insulator pulse interval sequence, associate the current sampling value and the timestamp within the same sampling window, extract local high points in the current sampling value sequence to form a peak sequence, construct a periodic expression sequence based on the time span between adjacent peaks, and construct an amplitude change expression sequence based on the amplitude change between adjacent peaks. Match the continuously shortened intervals and continuously extended intervals on the time axis, extract the time overlap to form one or more sets of overlapping time ranges, record the start and end times of each set of overlapping time ranges, and arrange them in chronological order to form a trigger time segment time range sequence, generating an insulation anomaly trigger segment;

[0055] S4: Based on the time range sequence of abnormal time segments in the set of abnormal insulation segments, call the time range sequence of trigger time segments in the set of abnormal insulation trigger segments. Within a unified sampling window, compare the time range sequence of abnormal time segments with the time range sequence of trigger time segments one by one according to the time axis position. Assign a first type of status identifier to the sampling window that contains both the time range of abnormal time segments and the time range of trigger time segments. Assign a second type of status identifier to the sampling window that contains only the time range of abnormal time segments or only the time range of trigger time segments. Assign a basic status identifier to the sampling window that does not contain either the time range of abnormal time segments or the time range of trigger time segments. Arrange the status identifiers in chronological order to form a status sequence expression. Sequentially concatenate the change directions between adjacent status identifiers to form a direction expression sequence. Convert the direction expression sequence into a discrete coding sequence to generate the insulation state change trajectory.

[0056] S5: Extract the discrete coding sequence from the insulation state change trajectory, compare the discrete coding sequence with the coding sequence of each trajectory sample in the trajectory sample library, find the coding sequence of the trajectory sample that is consistent with the discrete coding sequence in terms of arrangement structure and change trend, and classify the trajectory samples in the trajectory sample library that meet the consistency relationship into the candidate trajectory sample set. Read the insulation degradation type corresponding to each trajectory sample in the candidate trajectory sample set, establish the correspondence between the insulation degradation type contained in the candidate trajectory sample set and the insulation state change trajectory, and output the insulation degradation type result.

[0057] Example 2

[0058] This embodiment expands and supplements each step in Embodiment 1.

[0059] Specifically, the steps to obtain S1 are as follows:

[0060] S101: Obtain the current transformer output current sampling value and corresponding sampling timestamp of the low-voltage insulator. In the same distribution branch, the current sampling values ​​of each low-voltage insulator are divided into a fixed sampling window in chronological order. The adjacent current sampling values ​​are checked point by point. The continuously rising sampling points are merged into continuously rising intervals. The start and end positions of each continuously rising interval are recorded to obtain the rising interval position sequence.

[0061] Obtain the current transformer output current sample value and corresponding sampling timestamp of the low-voltage insulator, and call the first sample in the sampling time sequence. timestamp With the corresponding current amplitude Set the sampling frequency to And the fixed sampling window duration is ,Will If consecutive sampling points are grouped into the same vector group, adjacent sampling points in that vector group are extracted. and Perform subtraction to obtain the difference. ,like Then it is determined that the current sampling point is in an upward trend, for example, in time and time The difference is when If the condition of being greater than zero is met, then consecutive sampling points that meet this condition will be included. to Extract and compare to find the local minimum points of the sequence. and its timestamp Using these as initial parameters, the local maximum points of the sequence are obtained through comparison. and its timestamp As the end parameter, and Store coordinate pairs In the middle, when The current coordinate pair is terminated and the next set of interpolation calculations begins, covering the entire sampling window. The sampling points are iterated through repeatedly. If there are three obvious fluctuations within the window, three sets of coordinate pairs are generated. , and The start and end time values ​​of each coordinate pair are extracted and arranged in chronological order to obtain the sequence of rising interval positions.

[0062] S102: Based on the current sampling value and timestamp in the rising interval position sequence, the sampling time corresponding to the sampling point where the current sampling value in each consecutive rising interval reaches the set leakage current pulse amplitude is recorded as the pulse corresponding time point. The time interval between each pulse corresponding time point is arranged in chronological order to form a pulse interval sequence.

[0063] Based on the current sampling values ​​and timestamps in the rising interval position sequence, retrieve each rising interval. Amplitude of all sampling points inside Set the leakage current pulse amplitude threshold. for This threshold is based on the average background stray current of the insulator under power frequency operation. The setting is multiplied, for example, the average background current is [value missing]. hour;

[0064] ;

[0065] In the first upward zone Internal point-by-point comparison and The size relationship, if detected Sample value at time Satisfy greater than If the conditions are met, then record. The time point corresponding to the pulse in this interval Similarly, compare within the second interval to find The condition is met at all times and recorded as Compare within the third interval The condition is met at all times and recorded as Extract adjacent pulse time points and perform subtraction operation:

[0066] ;

[0067] calculate ;

[0068] ;

[0069] The calculated value and The pulse interval sequence is formed by storing the pulses in the order they are generated.

[0070] S103: Based on the pulse interval sequence, check the arrangement position of each pulse interval within the sampling window, identify the time position corresponding to the first pulse interval and the last pulse interval in the pulse interval sequence, and mark them as the start position and end position of the pulse interval sequence respectively, and generate the boundary of the insulator pulse interval sequence.

[0071] Based on the pulse interval sequence, extract the first element from the sequence. With the last element tracing Associated original pulse time points as well as Associated original pulse time points ,Will Time value Mark the relative position with respect to the starting zero point of the sampling window, and Time value Mark the relative position with respect to the starting zero point of the sampling window, obtain the start position and end position of the pulse interval sequence, and truncate on the time axis. The segment is used as the effective analysis domain to generate the boundary of the insulator pulse interval sequence.

[0072] Furthermore, the steps to obtain S2 are as follows:

[0073] S201: Based on the boundary of the insulator pulse interval sequence, check the positional relationship between each pulse interval and the start and end positions of the pulse interval sequence in sequence. Mark the pulse intervals located between the start and end positions of the pulse interval sequence one by one to obtain the pulse interval sequence selection interval.

[0074] Based on the boundary of the insulator pulse interval sequence, retrieve the starting position of the pulse interval sequence determined in S103. End position of pulse interval sequence Extract each element from the original pulse interval sequence ,Will The corresponding first pulse time The offset is obtained by subtracting from the starting boundary. ,Will The corresponding last pulse time The offset is obtained by subtracting from the termination boundary. Determine the offset is at The pulse intervals within the closed interval all belong to the valid interval. Traversing each pulse in the sequence... Elements, such as In and Between them, the time span falls entirely within Within the range, for those that satisfy the position constraint Assign validity labels Similarly, for Where Perform a logical AND operation on the range, and assign a label if the logical result is true. All those with labels The pulse intervals are logically extracted according to the original index number to obtain the pulse interval sequence selection range.

[0075] S202: Select an interval based on the pulse interval sequence, read the current sampling period within the corresponding sampling window, divide the time axis range corresponding to the sampling window into multiple time segments in order according to integer multiples of the current sampling period, call the time position corresponding to each pulse interval in the pulse interval sequence, and assign each pulse interval to the corresponding time segment according to its time position to generate a time segment distribution sequence.

[0076] Based on the pulse interval sequence, select an interval and extract the sampling frequency. The reciprocal of the current is used as the current sampling period. Set the time segment length coefficient The span of a single time segment is obtained through multiplication:

[0077] ;

[0078] Total duration The sampling window is set by Divide into equal parts Sub-intervals, such as intervals for And the interval for Retrieve the timestamp coordinates of each element in the pulse interval sequence ,Will and The boundary values ​​are compared to determine the size. Greater than and less than Thus Into Similarly, Into ,Will Into Count the number of pulses falling within each sub-interval. ,like If there are multiple pulses in memory, their count values ​​are accumulated, and the count values ​​of each sub-interval are arranged in index order to form a one-dimensional vector, generating a time segment distribution sequence.

[0079] S203: For the time segment distribution sequence, read the pulse interval distribution in each time segment, check the number of pulse intervals in adjacent time segments one by one and compare them with the preset pulse interval density threshold. Mark the time segments with the number of pulse intervals exceeding the pulse interval density threshold and adjacent to each other as continuous aggregation time segments, and record the start time and end time of each continuous aggregation time segment in sequence to obtain the aggregation time range sequence.

[0080] For the time-segmented distribution sequence, extract each segment Pulse count value within Set a preset pulse interval density threshold. for The specific setting of this pulse interval density threshold is based on the average number of pulses per millisecond under historical normal conditions. times, that is For example, in The historical average pulse is One rule:

[0081] ;

[0082] Take here Increase the rigor of the judgment, of and Perform size comparison and determine If true, read adjacent segments simultaneously. pulse number ,determination If true, and Mark as aggregated state, extract left boundary and right boundary Connect the segments and record the start time of each segment. and the end time ,like of determination If the clustering is not included, this comparison process continues until all clusters have been traversed. The time range sequence of the aggregation is obtained by dividing the data into segments.

[0083] S204: Based on the aggregation time range sequence, call the start and end times of each continuous aggregation time segment, check the time interval between adjacent continuous aggregation time segments, and splice the beginning and end of adjacent continuous aggregation time segments with a time interval of zero or less than the preset segment interval threshold to form a merged time range. Keep the continuous aggregation time segments with a time interval greater than the preset segment interval threshold as independent time ranges, and output each merged time range and independent time range in chronological order to form an abnormal time segment time range sequence and generate an insulation abnormal segment set.

[0084] Based on the aggregation time range sequence, retrieve the first aggregation range. With the second group of clusters Preset segment interval threshold for The interval threshold for this segment is determined by the sampling period. of Multiplier redundancy setting ;

[0085] Extract the end time of the first group The start time of the second group The interval duration is obtained by subtraction. ,Will and Compare and determine If the splicing conditions are met, perform the start-end splicing operation to merge the two ranges into one. If a third cluster exists The difference between its start time and the end time of the previous merged group is Determine if greater than Then keep it as an independent scope. Arrange all merged and independent intervals in ascending order of their time axis coordinates to form a set. This forms a time range sequence of abnormal time segments, generating a set of insulation abnormal segments.

[0086] Furthermore, the steps to obtain S3 are as follows:

[0087] S301: Based on the boundary of the insulator pulse interval sequence, associate the current sampling value and the timestamp within the same sampling window, compare the amplitude relationship of the current sampling value corresponding to the adjacent sampling points, mark the position where the current sampling value corresponding to the adjacent sampling points is lower than the current sampling value of the current sampling point as a local high point, and sort out the timestamp and current sampling value corresponding to each local high point to obtain the local high point sequence.

[0088] Based on the boundary of the insulator pulse interval sequence, the sampling window is called with the... and The original sequence is constructed, and the sliding window span is set to... Extract the amplitude of the current sampling point from each sampling point. its left adjacent point and adjacent points on the right Perform numerical comparison and calculate the difference:

[0089] ;

[0090] ;

[0091] Judgment when and When established at the same time, It is determined to be a local maximum point, for example, in time and , The result obtained through subtraction is and If all values ​​are greater than zero, then record the coordinates. Set noise filtering threshold The noise filtering threshold is set with reference to the standard deviation of the sensor's static noise. times For the extracted Perform subtraction operation If the result is greater than zero, keep the point; otherwise, discard the interfering point. Repeat this process. to All sampled data within the boundary are sequentially stored in an array, with the coordinates satisfying the two-sided decreasing logic and above the noise level. The timestamps and current sampling values ​​corresponding to each local high point are collected to obtain the local high point sequence.

[0092] S302: Based on the local high point sequence, read the timestamps and current sampling values ​​corresponding to adjacent local high points, check the direction of change of time interval between adjacent local high points and the state of change of current sampling values ​​along the time sequence, connect the time intervals of adjacent local high points that show a continuous decrease in time interval between adjacent local high points to form a periodic contraction segment, and connect the time intervals of adjacent local high points that show a continuous increase in current sampling values ​​between adjacent local high points to form an amplitude expansion segment. Arrange the periodic contraction segment and the amplitude expansion segment according to the time axis to obtain overlapping candidate segments;

[0093] Extract the time coordinates of adjacent peaks based on the local high point sequence. With amplitude To obtain the periodic sequence, perform subtraction operations on the time axis:

[0094] ;

[0095] Compare adjacent periods and The size, if it satisfies The condition is that the time interval shows a continuously decreasing trend, for example... and , The corresponding time interval Marked as a periodic contraction segment Simultaneously, a subtraction operation is performed on the amplitude to obtain the change sequence:

[0096] ;

[0097] If satisfied The condition is that the amplitude shows a continuous increasing trend, for example... and , The corresponding time interval Marked as amplitude extension segment Set the coefficient of determination for the rate of change. The rate of change determination coefficient is set with reference to the empirical constant of the insulator surface discharge development rate. If the sequence change amplitude Then the trend is confirmed to be valid, and all identified trends will be... and Align the two segments sequentially according to their starting coordinates on the time axis, extract their distribution positions within the same sampling period, and obtain overlapping candidate segments.

[0098] S303: Call the overlapping candidate segment, extract the overlapping part of the periodic contraction segment and the amplitude expansion segment on the time axis as the overlapping time range, record the start time and end time corresponding to each overlapping time range, and arrange them in the order of the start time to form the trigger time segment time range sequence, and generate the insulation abnormality trigger segment.

[0099] The specific operation to form the trigger time segment time range sequence is as follows: extract the start time and end time corresponding to each overlapping time range one by one, sort the overlapping time ranges in ascending order of start time, check the time sequence between adjacent overlapping time ranges in turn during the sorting process, and when there are overlapping time ranges with the same start time or overlapping time, perform secondary sorting in ascending order of end time, and arrange the sorted overlapping time ranges in the order of time axis to form the trigger time segment time range sequence.

[0100] Retrieve overlapping candidate segments and extract the start time of the periodic contraction segment. With end time and the start time of the amplitude expansion segment With end time Perform logical operations on the four parameters and take the following:

[0101] ;

[0102] and ;

[0103] Judgment if Then output the overlap time range, for example, the contraction segment is... And the extended segment is The calculated overlapping segment is Store all generated overlapping segments into a set. ,extract The start times of each element are sorted in ascending order, and a threshold is set to determine if the times are the same. If the absolute value of the start time difference between the two overlapping segments is less than Then retrieve the end times of both. and Numerical comparison is performed, and the segment with the earlier end time is placed at the front. Adjacent elements in the set are traversed to perform logical overlap verification. If the end time of the previous segment is greater than the start time of the next segment, it is determined that there is time overlap and the current sequence position is maintained. The sorted time segments are logically concatenated in turn to form a time range sequence of trigger time segments, and an insulation anomaly trigger segment is generated.

[0104] Furthermore, the steps to obtain S4 are as follows:

[0105] S401: Based on the time range sequence of abnormal time segments in the set of abnormal insulation segments, call the start and end times of each abnormal time segment in the unified sampling window, retrieve the time range of each trigger time segment in the time range sequence of trigger time segments in the insulation abnormal trigger segment, and arrange them according to the same sampling window to obtain the window comparison sequence.

[0106] Extract the start time of the abnormal time segment sequence within the first sampling window from the abnormal time segment sequence of the insulation abnormal segment set. The time is 10.25ms and the end time. The value is 14.85ms, and the start time of the second abnormal time segment within this window is also extracted. The time is 32.10ms and the end time. The value is 36.50ms. The time range sequence of the trigger time segment within the insulation fault trigger segment is called to obtain the start time of the trigger time segment within the same window. The time is 11.05ms and the end time. The time parameters belonging to the same 100ms sampling window are mapped to a unified time axis coordinate system, and the abnormal segment set is identified through timestamp alignment logic. With trigger fragment set By placing them at the same data structure location and comparing the physical distribution of time intervals of different dimensions on the time axis, the sampling window sequence is analyzed. All time vectors are clustered and grouped, and floating-point data containing time start and end information is stored in the window index table to generate a window reference sequence containing multiple sets of time interval mapping relationships.

[0107] S402: According to the window comparison sequence, check the time range of abnormal time segments and the time range of trigger time segments in each sampling window. Assign the first type of status label to sampling windows that have both types of time ranges, assign the second type of status label to sampling windows that have only one type of time range, assign the basic status label to sampling windows that have no sampling windows that have neither type of time range, and sort them in chronological order to obtain the status sequence expression.

[0108] Read the number as Extract the set of abnormal time periods contained in the sampling window data. With the set of trigger time segments , Determine set With sets Whether they are all non-empty, because If there is an overlap on the timeline and both sets contain elements, set the window identifier value to 1 and read the number. The sampling window data, and its set of abnormal time periods. Triggering a set of time segments for an empty set If the value is not empty, it is determined that only one time range appears. The window identifier value is set to 2, and the number is read. If both sets are empty, the sampling window is set to 0, and the state assignment function is used to assign the value. The above logic is applied to 50 consecutive sampling windows, mapping the value 1 to the first type of state identifier, the value 2 to the second type of state identifier, and the value 0 to the basic state identifier. The identifier characters are then chained together according to the time sequence of the sampling windows to obtain the state sequence representation.

[0109] S403: Call the state sequence expression, read the positions of adjacent state identifiers, check the change direction between adjacent state identifiers in turn, and concatenate the change directions in sequence to generate a direction expression sequence;

[0110] Extracting state sequences Locate adjacent status identifier pairs Read the first identifier 0 and the second identifier 1, and record the change flow as "basic state". "Type 1 state", read the second and third identifiers 1, and record the change flow as "Type 1 state". "Type 1 state", read the third identifier 1 and the fourth identifier 2, and record the change flow as "Type 1 state". The second type of state is obtained by analogy for the jump trajectory of all adjacent points. The logical trends such as "rising", "leveling", and "falling" are transformed into specific flow descriptors. The state jump points generated by the distribution branch insulators within the 24-hour monitoring cycle are fully extracted. The identified jump descriptors are physically connected according to the order of sampling time to generate a direction expression sequence.

[0111] S404: Pre-establish the coding correspondence between the state change direction and the discrete marker. For the direction expression sequence, according to the coding correspondence, map each change direction to the corresponding discrete marker, form a discrete coding sequence along the time axis, and organize the discrete coding sequence with the sampling window order to generate the insulation state change trajectory.

[0112] Set "basic state" The "first type of state" is mapped to the discrete label "C1", and the "first type of state" is defined as follows: The second type of state is mapped to the discrete label "C2", and the second type of state is defined as follows: The "basic state" is mapped to the discrete label "C3". The jump descriptor in the direction expression sequence is called to extract the first jump action "basic state". The first type of state is retrieved by searching the corresponding discrete tag C1 in the encoding mapping table, and the second jump action "first type of state" is extracted. "Second type of state", retrieve and match discrete tag C2, and calculate discrete code index value:

[0113] ;

[0114] in These are state characteristic values. This refers to the positional weighting coefficient.

[0115] If the calculation result is in The interval is determined to be in the early stage of degradation, and is in the middle of the process. The interval is determined to be medium-term, and the value above 50 is determined to be severe. The discrete markers are serialized and recombined according to the time index of the sampling window to form a character matrix composed of "C1-C2-C1-C3", which generates the insulation state change trajectory.

[0116] Furthermore, the steps to obtain S5 are as follows:

[0117] S501: Extract the discrete coding sequence from the insulation state change trajectory, read the arrangement position of each discrete marker, retrieve the coding sequence of each trajectory sample in the trajectory sample library, and compare the arrangement relationship of each discrete marker in the coding sequence according to the coding segment with the same length as the discrete coding sequence. Retain the trajectory samples in the trajectory sample library that correspond to the discrete coding sequence in terms of arrangement structure, and sort them in order according to the position in the sample library to obtain the structure-corresponding sample sequence.

[0118] Extract the discrete coded sequence from the insulation state change trajectory and retrieve the currently identified real-time discrete numerical sequence. Get the total number of elements in the sequence. Set the sample library retrieval pointer to point to the first sample trajectory. Extract the corresponding encoded sequence The medium length is also All subsequence segments, the first tag value in the real-time sequence Subtract the first label of the sample segment from the first label of the sample segment. If the absolute value of the difference is... If the first character is successfully aligned, then the alignment is performed character by character, and so on, comparing each subsequent character. Three numerical digits are used to set the structural consistency judgment coefficient. The structural consistency determination coefficient is set based on the rigid requirement of perfect matching of discrete-state logic. The constant value, if the sample segment sequence If every value in the sample is equal to that in the real-time sequence, the permutation structure is considered consistent. The sample IDs that satisfy this complete equality constraint are stored in a temporary buffer. The sample database is then traversed. A typical degradation trajectory sample group, if the first one Sample No. and No. If the encoded sub-segments of sample number 1 both conform to this feature, then the two samples are retained, and a linear rearrangement is performed according to the original logical address order of the sample library to obtain the sample sequence corresponding to the structure.

[0119] S502: Based on the sample sequence corresponding to the structure, read the corresponding coding sequence of each trajectory sample, check the connection and change relationship between each discrete mark and the adjacent discrete mark along the coding arrangement direction, and compare the connection and change relationship with the connection and change relationship of the corresponding position of the discrete coding sequence in the insulation state change trajectory segment by segment, select the trajectory samples with consistent change direction, and arrange them according to the order of the sample library to obtain the candidate trajectory sample set.

[0120] Based on the sample sequence corresponding to the structure, retrieve the first... Number and The discrete encoding vector corresponding to sample number 1 Extracting the dynamic trends between adjacent markers in a real-time encoded sequence:

[0121] ;

[0122] The calculated trend vector is as follows: Right now For the sample Perform the same subtraction operation and extract the positive and negative signs, then compare. Change components The validity of a sample is determined by whether the corresponding sign bit vector completely overlaps with the real-time vector. The sample is valid only if the sign of the change at each connection point is consistent. For example, sample... The transition is as follows Its symbol sequence is Real-time sequence changes are Its symbol sequence is also If a logical comparison operation is performed on both and the result is true, the sample is retained and added to the next level set. If the result is true, the sample is removed from the set. If the sign of the change in the third position of the sample is positive, it is determined that the direction has deviated and a removal operation is performed. By performing adjacent position difference sign extraction and item-by-item logical AND operation on all structural corresponding samples, trajectory samples with consistent change direction before and after are selected to obtain a candidate trajectory sample set.

[0123] S503: Call the candidate trajectory sample set, read the insulation degradation type corresponding to each trajectory sample, extract the corresponding mark content of each insulation degradation type along the trajectory sample arranged in the order of the sample library, establish a correspondence between each insulation degradation type and the insulation state change trajectory and output it. When there are multiple insulation degradation types in the candidate trajectory sample set, determine the insulation degradation type according to the preset matching frequency and output it to generate the insulation degradation type result.

[0124] Call the candidate trajectory sample set and extract the samples from the set. The insulation degradation label attribute corresponding to the sample identifies the physical fault type associated with the label as "surface contamination partial discharge". If the set also contains the first... If a sample is labeled "thermal aging of insulating medium", then count the frequency of each label appearing in the set. Set matching frequency determination parameters For example, "partial discharge of surface contaminants" occurs. Secondly, "thermal aging" occurs. Next, the numerical comparison was used to determine the result. Then, the most frequent "surface contamination partial discharge" is selected as the final judgment indicator, and a confidence evaluation coefficient is set. ,in Let be the total number of samples in the candidate set. Greater than the preset classification confidence threshold If the validity of the degradation type result is determined, the identified degradation type string is associated with the current discrete coded trajectory through key-value pair mapping, and the insulation degradation type result is output.

[0125] Example 3

[0126] This embodiment provides an edge computing-based online monitoring system for low-voltage insulators. This system is used in either Embodiment 1 or Embodiment 2, and includes:

[0127] The sampling window processing module acquires the output current sampling value and corresponding sampling timestamp of the low-voltage insulator current transformer. It sorts the insulator sampling data in the distribution branch according to the sampling time order, calls the sampling frequency and timestamp interval to check each item, and classifies the sampling points with consecutive and consistent timestamps into the same fixed sampling window. It outputs the current sampling value in each sampling window in time order and reads the start time and end time of the window to generate the time domain range of the sampling window.

[0128] Pulse interval identification module: Based on the time domain range of the sampling window, it calls the current sampling values ​​within each sampling window, compares the amplitudes of adjacent current sampling values, selects sampling points that rise continuously point by point and extracts the corresponding start and end sampling times, calls the current amplitude and compares it with the set leakage current pulse amplitude, extracts the sampling time corresponding to the position that reaches the amplitude as the pulse time point, reads the time interval of adjacent pulse time points and outputs them in chronological order to generate the pulse interval distribution;

[0129] Abnormal Segment Extraction Module: Based on the pulse interval distribution, read the corresponding time position of each pulse interval, call the current sampling cycle time length to segment the sampling window time axis, allocate each pulse interval to the corresponding time segment according to the position of the pulse time point, read the number of pulse intervals in each time segment, compare the number of pulse intervals in adjacent time segments with the preset pulse interval density value, connect adjacent time segments with densely distributed numbers and read the start and end times of the corresponding time segments, continue to compare the time interval between adjacent connected segments with the preset segment interval value, splice segments with close consecutive time intervals, and output each time range in chronological order to generate the time domain range of the abnormal segment;

[0130] Trigger segment determination module: Based on the time domain range of the abnormal segment, call the current sampling value of the corresponding sampling window, compare the current amplitude of adjacent sampling points before and after to select local high points and extract the corresponding timestamps, read the time interval of adjacent local high points and check the direction of time interval change, and at the same time check the direction of current amplitude change of adjacent local high points. Match the segments with gradually shortened time intervals and segments with gradually increased current amplitudes along the time axis and extract the overlapping time range to generate the trigger segment time window;

[0131] The trajectory matching module calls the corresponding sampling window position according to the time domain range of the abnormal segment and the time window of the trigger segment. It checks the occurrence status of two types of time ranges within the same sampling window and assigns status labels. It reads the change direction of adjacent status labels and concatenates them in chronological order. It replaces the change direction with a preset discrete code and forms the coded content. It calls the coded content in the trajectory sample library to compare and filter the content with the same structure. It reads the corresponding insulation degradation type and counts it according to the number of occurrences to generate the insulation degradation category.

[0132] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for online monitoring of low-voltage insulators based on edge computing, characterized in that, Includes the following steps: S1: Obtain the current sampling value and timestamp of the low-voltage insulator, identify the pulse time point that reaches the set leakage current pulse amplitude within the continuous rising interval, construct the pulse interval sequence, determine the start and end positions of the pulse interval sequence, and generate the boundary of the insulator pulse interval sequence. S2: Based on the boundary of the insulator pulse interval sequence, obtain the pulse interval sequence and its start and end positions, determine the pulse interval sequence selection interval, divide multiple time segments, extract continuous clustered time segments, form an abnormal time segment time range sequence, and generate a set of insulation abnormal segments. S3: Based on the boundary of the insulator pulse interval sequence, correlate the current sampling value and the timestamp, extract local high points to form a peak sequence, construct a periodic expression sequence and an amplitude change expression sequence, extract the overlapping time range of the two types of sequences, form a trigger time segment time range sequence, and generate an insulation anomaly trigger segment; S4: Based on the time range sequence of abnormal time segments, call the time range sequence of trigger time segments, compare them one by one and assign corresponding state identifiers to form a state sequence expression, check the change direction of adjacent state identifiers to form a direction expression sequence and convert it into a discrete coding sequence to generate the insulation state change trajectory. S5: Extract the discrete coding sequence from the insulation state change trajectory, compare it with the coding sequence of each trajectory sample in the trajectory sample library, screen candidate trajectory samples with consistent arrangement structure and change trend, read the corresponding insulation degradation type, establish the correspondence with the insulation state change trajectory, and output the insulation degradation type result.

2. The low-voltage insulator online monitoring method based on edge computing according to claim 1, characterized in that, The steps for obtaining S1 are as follows: S101: Obtain the current transformer output current sampling value and corresponding sampling timestamp of the low-voltage insulator. In the same distribution branch, the current sampling values ​​of each low-voltage insulator are divided into a fixed sampling window in chronological order. The adjacent current sampling values ​​are compared point by point. The continuously rising sampling points are merged into continuously rising intervals. The start and end positions of each continuously rising interval are recorded to obtain the rising interval position sequence. S102: Based on the current sampling value and timestamp in the rising interval position sequence, the sampling time corresponding to the sampling point where the current sampling value in each consecutive rising interval reaches the set leakage current pulse amplitude is recorded as the pulse corresponding time point. The time interval between each pulse corresponding time point is arranged in chronological order to form a pulse interval sequence. S103: Based on the pulse interval sequence, check the arrangement position of each pulse interval within the sampling window, identify the time positions corresponding to the first and last pulse intervals in the pulse interval sequence, and mark them as the start position and end position of the pulse interval sequence, respectively, to generate the boundary of the insulator pulse interval sequence.

3. The low-voltage insulator online monitoring method based on edge computing according to claim 1, characterized in that, The steps for obtaining S2 are as follows: S201: Based on the boundary of the insulator pulse interval sequence, check the correspondence between each pulse interval and the start and end positions according to the pulse interval sequence order, mark each pulse interval falling between the two boundaries, and remove the contents outside the boundaries to form the pulse interval sequence selection range. S202: Select an interval based on the pulse interval sequence, read the current sampling period of the corresponding sampling window, divide the sampling window time axis into multiple time segments in an integer multiple order of the sampling period, assign segments according to the time position of each pulse interval, and form a time segment distribution sequence. S203: For the time segment distribution sequence, read the number of pulse intervals in each time segment and check them against the preset pulse interval density threshold. Mark the time segments with the number exceeding the threshold and adjacent to each other as continuous clustering time segments, and record their start and end times in sequence to obtain the clustering time range sequence. S204: Based on the aggregation time range sequence, read the start and end times of each continuous aggregation time segment, check the time interval between adjacent segments, splice the beginning and end of segments with an interval of zero or less than the preset segment interval threshold, keep segments with an interval greater than the threshold independent, and output them in chronological order to generate a set of insulation abnormal segments.

4. The low-voltage insulator online monitoring method based on edge computing according to claim 1, characterized in that, The steps for obtaining S3 are as follows: S301: Based on the boundary of the insulator pulse interval sequence, associate the current sampling value and the timestamp within the same sampling window, compare the amplitude relationship of the current sampling value corresponding to adjacent sampling points, mark the position where the current sampling value corresponding to the adjacent sampling points is lower than the current sampling value of the current sampling point as a local high point, organize the timestamp and current sampling value corresponding to each local high point, and obtain the local high point sequence. S302: Based on the local high point sequence, read the timestamps and current sampling values ​​corresponding to adjacent local high points, check the direction of change of time interval between adjacent local high points and the state of change of current sampling values ​​along the time sequence, connect the time intervals of adjacent local high points that show a continuous decrease in time interval between adjacent local high points to form a periodic contraction segment, and connect the time intervals of adjacent local high points that show a continuous increase in current sampling values ​​between adjacent local high points to form an amplitude expansion segment, and arrange the periodic contraction segment and amplitude expansion segment according to the time axis to obtain overlapping candidate segments; S303: Call the overlapping candidate segment, extract the overlapping part of the periodic contraction segment and the amplitude expansion segment on the time axis as the overlapping time range, record the start time and end time corresponding to each overlapping time range, and arrange them in the order of the start time to form a trigger time segment time range sequence, and generate an insulation abnormality trigger segment.

5. The low-voltage insulator online monitoring method based on edge computing according to claim 1, characterized in that, The steps for obtaining S4 are as follows: S401: Based on the time range sequence of abnormal time segments in the set of insulation abnormal segments, call the start time and end time of each abnormal time segment in the unified sampling window, retrieve the time range of each trigger time segment in the time range sequence of trigger time segments in the insulation abnormal trigger segment, and arrange them according to the same sampling window to obtain the window comparison sequence. S402: According to the window comparison sequence, check the time range of abnormal time segments and the time range of trigger time segments in each sampling window, assign the first type of status identifier to the sampling window that appears simultaneously with two types of time ranges, assign the second type of status identifier to the sampling window that appears only with one type of time range, assign the basic status identifier to the sampling window that does not appear with either type of time range, and sort them according to time sequence to obtain the status sequence expression. S403: Call the state sequence expression, read the positions of adjacent state identifiers, check the change direction between adjacent state identifiers in turn, and concatenate the change directions in sequence to generate a direction expression sequence; S404: Establish a pre-established coding correspondence between the state change direction and the discrete marker. For the direction expression sequence, map each change direction to a corresponding discrete marker according to the coding correspondence, form a discrete coding sequence along the time axis, and organize the discrete coding sequence with the sampling window order to generate the insulation state change trajectory.

6. The low-voltage insulator online monitoring method based on edge computing according to claim 1, characterized in that, The steps for obtaining S5 are as follows: S501: Extract the discrete coding sequence from the insulation state change trajectory, read the arrangement position of each discrete marker, retrieve the coding sequence of each trajectory sample in the trajectory sample library, and compare the arrangement relationship of each discrete marker in the coding sequence item by item according to the coding segment with the same length as the discrete coding sequence. Retain the trajectory samples in the trajectory sample library that correspond to the discrete coding sequence in terms of arrangement structure, and sort them in order according to the position in the sample library to obtain the structure-corresponding sample sequence. S502: Based on the sample sequence corresponding to the structure, read the corresponding encoding sequence of each trajectory sample, check the connection and change relationship between each discrete mark and the adjacent discrete mark along the encoding arrangement direction, and compare the connection and change relationship with the connection and change relationship of the corresponding position of the discrete encoding sequence in the insulation state change trajectory segment by segment, select the trajectory samples with consistent change direction, and arrange them according to the order of the sample library to obtain the candidate trajectory sample set. S503: Call the candidate trajectory sample set, read the insulation degradation type corresponding to each trajectory sample, extract the corresponding mark content of each insulation degradation type in the order of the trajectory samples arranged in the sample library, establish a correspondence between each insulation degradation type and the insulation state change trajectory and output it. When there are multiple insulation degradation types in the candidate trajectory sample set, determine the insulation degradation type according to the preset matching frequency and output it to generate the insulation degradation type result.

7. The low-voltage insulator online monitoring method based on edge computing according to claim 4, characterized in that, The specific operation of arranging the trigger time segment time range sequence according to the chronological order of the start time is as follows: The start and end times corresponding to each overlapping time range are extracted one by one. The overlapping time ranges are sorted in ascending order of start time. During the sorting process, the temporal relationship between adjacent overlapping time ranges is checked in turn. When there are overlapping time ranges with the same start time or overlapping time, they are sorted in ascending order of end time. The sorted overlapping time ranges are arranged in the order of time axis to form the trigger time segment time range sequence.

8. A low-voltage insulator online monitoring system based on edge computing, characterized in that, The system is used in the low-voltage insulator online monitoring method based on edge computing as described in any one of claims 1-7, and the system comprises: Pulse sampling and recognition module: acquires low-voltage insulator current sampling values ​​and timestamps, identifies pulse time points that reach the set leakage current pulse amplitude, constructs pulse interval sequences and determines start and end positions, and generates insulator pulse interval sequence boundaries; Pulse interval identification module: Based on the boundary of the insulator pulse interval sequence, determine the pulse interval sequence selection interval, divide the time segment, extract the continuous clustered time segment, form the time range sequence of abnormal time segment, and generate a set of insulation abnormal segments; Periodic amplitude analysis module: Based on the boundary of the insulator pulse interval sequence, correlated sampling data, extracted peak sequence, constructed periodic and amplitude change expression sequence, determined the overlapping time range, formed the trigger time segment time range sequence, and generated insulation anomaly trigger segment; State change coding module: Based on the time range sequence of abnormal time segments, compare the abnormal time segments with the trigger time segments, assign state identifiers, form a state sequence, and convert it into a discrete coding sequence to generate the insulation state change trajectory; Trajectory Sample Comparison Module: Extracts discrete coded sequences from insulation state change trajectories, compares them with the trajectory sample library, filters out matching samples, reads the corresponding insulation degradation type, and outputs the insulation degradation type result.

Images