Switch equipment fault early warning method based on multi-source data fusion

Abstract

The invention relates to the technical field of fault prediction, in particular to a switch equipment fault early warning method based on multi-source data fusion, which comprises the following steps of: acquiring a channel transition time sequence to generate a time interval, marking a behavior label to generate a time sequence, constructing a sliding window to form a combined alarm, analyzing a cross matrix to extract an abnormal time period, and performing early warning. And identifying the leading channel and sending out early warning. According to the method, alignment processing is carried out through the time intervals of multi-channel state transition responses, dynamic linkage recognition among multiple signals is achieved, time sequence correlation among abnormal behaviors is enhanced by means of behavior label combination and a sliding statistical strategy, and the sensitivity and accuracy of abnormal recognition are enhanced through aggregation expression of composite alarm events. In combination with cross-channel behavior label matrix analysis, the continuity of abnormal labels is clear and key channels are identified, so that the accuracy and timeliness of early warning of the switch equipment are improved, the risk of false alarm and missing alarm is reduced, and the actual demand of multi-source sensing data fusion analysis of the switch equipment is met.

Description

Technical Field

[0001] The present invention relates to the technical field of fault prediction, and particularly to a fault warning method for switchgear based on multi-source data fusion. Background Art

[0002] The technical field of fault prediction involves the identification and anticipation of potential faults through the analysis of equipment operation status data, including fault modeling based on historical data, dynamic comparison of real-time monitoring data, identification of abnormal behavior patterns, and assessment of equipment degradation trends. It usually combines sensor monitoring, data mining, and pattern recognition methods to analyze the operation status of key facilities such as industrial equipment, power systems, and transportation equipment to provide a basis for early warning. Among them, the traditional fault warning method for switchgear refers to the method of identifying abnormal states and estimating fault risks based on a single data source such as partial discharge signals or infrared temperature data. A single signal type is extracted through a specific data acquisition channel, and statistical rules are used to set warning thresholds to judge whether the equipment status is abnormal.

[0003] During the existing switchgear fault prediction process, it mainly relies on a single data source to obtain local signal feature information, lacking a time series alignment and behavior collaborative judgment mechanism between multiple types of channel information. It is difficult to effectively associate abnormal behaviors in the case of multiple signal overlaps or hidden fault symptoms. The warning thresholds set for a single channel are easily affected by external environmental fluctuations, resulting in an increase in misjudgments. The failure to form a continuous behavior label sequence makes the identification of abnormal states deviate from the actual operation situation. A typical scenario is that when temperature fluctuations are accompanied by voltage fluctuations, synchronous analysis cannot be carried out, resulting in isolated processing of fault symptoms and delaying equipment maintenance response. Summary of the Invention

[0004] The purpose of the present invention is to solve the deficiencies existing in the prior art, and a fault warning method for switchgear based on multi-source data fusion is proposed.

[0005] To achieve the above purpose, the present invention adopts the following technical scheme: A fault warning method for switchgear based on multi-source data fusion, including the following steps: S1: Obtain the first time stamps corresponding to the state transition points in the voltage channel, temperature channel, and vibration channel of the switchgear, sort them in chronological order to obtain a set of response time intervals; S2: Based on the set of response time intervals, horizontally align the sampled data, detect whether there are situations of voltage stagnation, temperature passivation, and vibration sudden drop labels, and record the corresponding time stamps to obtain a group of behavior label time series; S3: Based on the behavior label time series group, construct a sliding window, and count the number of non-repeating behavior labels that appear in each sliding window. When the number of labels is greater than or equal to 2, combine the corresponding sliding window number, label type, and corresponding channel number, and uniformly number them as the composite alarm event ID to obtain a combined trigger window set; S4: According to the combined trigger window set, perform cross-matrix analysis on the marked corresponding channel numbers and corresponding behavior labels, and extract all time periods during which abnormal labels continuously appear on the time axis to obtain a multi-channel concurrent abnormal period; S5: According to the channel numbers involved in the multi-channel concurrent abnormal period, calculate the single-channel abnormal contribution frequency, mark the abnormal dominant channel, and send a switch failure warning message.

[0006] As a further solution of the present invention, the response time interval set includes the channel response sequence, timestamp pairing structure, and time interval arrangement sequence. The behavior label time series group includes a voltage stagnation label, a temperature passivation label, and a vibration sudden drop label. The combined trigger window set includes a sliding window number, a composite alarm event ID, and a label type matching result. The multi-channel concurrent abnormal period includes the channel number corresponding matrix analysis result, the abnormal label continuous section, and the abnormal label cross-mapping relationship. The switch failure warning message includes the behavior label frequency distribution record, the channel number and abnormality correlation record, and the switch failure warning real-time record.

[0007] As a further solution of the present invention, the specific obtaining steps of the response time interval set are as follows: S111: Obtain the original signal sequences monitored in the switch device voltage channel, temperature channel, and vibration channel, identify and extract the first timestamp value corresponding to the state transition point, and record them as the transition start time points of each channel respectively, to form a corresponding mapping set composed of channel numbers and timestamp values, and obtain the channel first jump time set; S112: According to the channel first jump time set, sort all channel numbers in ascending order according to the corresponding first jump timestamp value, record the original serial number corresponding to the channel number after ascending order as the current sorting serial number, and construct a paired value pair set between the sorting serial number and the channel first jump time to obtain a time series mapping data group; S113: According to the time series mapping data group, calculate the response time difference between adjacent channels based on the timestamp values between adjacent sorting serial numbers, and sort and organize the time intervals between all channel sorting serial numbers into a unified set to obtain the response time interval set.

[0008] As a further solution of the present invention, the specific obtaining steps of the behavior label time series group are as follows: S211: Based on the set of response time intervals, perform unified horizontal alignment processing on the sampling time axes of the voltage channel, temperature channel, and vibration channel. Use the minimum time step as the reference step size to construct a unified sampling time axis, extract the sampling amplitude records at the corresponding time points in each channel, fill the unsampled areas using the previous value holding method to complete data alignment. After all the sampling data is aligned, establish a horizontal channel mapping to obtain a three-channel aligned data matrix; S212: According to the three-channel aligned data matrix, detect whether the amplitude change range within two adjacent periodic sampling segments in the voltage channel is less than the voltage stability threshold. Calculate two groups of temperature differences for every three consecutive sampling points in the temperature channel and compare whether their maximum difference is less than or equal to the reference value. Extract the current sampling value in the vibration channel and determine whether it is lower than 30% of the mean value. Mark the corresponding tags for the channel sampling points that meet the above judgment conditions respectively to generate a tag determination sequence; S213: According to the tag determination sequence, map it to the tags of three categories: voltage stagnation, temperature passivation, and vibration sudden drop respectively, integrate it into a structured record unit in chronological order, obtain the channel numbers and event timestamps corresponding to each type of behavior, and get a behavior tag time series group.

[0009] As a further solution of the present invention, the specific steps for obtaining the combined trigger window set are as follows: S311: Obtain the behavior tags, channel numbers, and timestamps corresponding to each record in the behavior tag time series group. Construct a sliding window on the overall time axis according to the time order, perform time attribution screening on the tag records included in each sliding window, extract all the tag records within the time window range and construct a tag summary table, perform unique counting after removing duplicate behavior tags, and generate a window tag counting result; S312: According to the window tag counting result, screen all the sliding windows with the number of tags greater than or equal to 2, extract the tag types and corresponding channel numbers included in the window, and combine them with the window number to form a unique identification structure for all the windows that meet the conditions. Integrate the information based on the behavior tags and channel numbers to obtain a composite tag identification set; S313: Based on the composite tag identification set, assign numbers to each composite record according to the alarm management specification, combine the time number of the sliding window to construct a unique composite alarm event identification, and file and integrate the number, the start and end times of the window, the involved channels, and the tag types, and construct a unified sequence in chronological order to generate a combined trigger window set.

[0010] As a further solution of the present invention, the specific steps for obtaining the multi-channel concurrent abnormal time period are as follows: S411: Extract all the sliding window numbers sorted in chronological order according to the sliding window numbers, corresponding channel numbers, and behavior labels recorded in the combined trigger window set. Calculate whether the difference between adjacent numbers in the window number sequence is 1. Continuously merge the window numbers with a difference equal to 1 to construct a set of time - continuous window segment sequences, and obtain a set of continuous window labels; S412: Based on the set of continuous window labels, construct a cross - matrix structure with the channel number as the row index and the behavior label as the column index. In each cross - cell, count the frequency of the combination of the channel and the label in the continuous window segment. At the same time, extract the co - occurrence times of each group of channel - behavior pairs, calculate the behavior coupling degree of the channel and the label in the time - series window, and determine whether it exceeds the behavior anomaly determination threshold. Screen all channel - behavior combinations that meet the coupling strength determination standard to generate a channel - label coupling degree matrix; S413: According to the channel - label coupling degree matrix, extract the corresponding time index and integrate the continuously distributed segments in the time set to determine the continuous segment sequence of the abnormal label appearing on the time axis. Integrate the start and end times, corresponding channels, and label combinations recorded in each segment in chronological order to obtain the multi - channel concurrent abnormal time period.

[0011] As a further solution of the present invention, the specific steps for obtaining the switch fault warning information are as follows: S511: According to the channel numbers involved in the multi - channel concurrent abnormal time period, count the number of behavior labels corresponding to each channel number within the covered time segment, extract the label frequency information of the channel in the concurrent abnormal segment, and calculate the sum of the label records in each channel respectively to obtain the channel - label frequency data; S512: According to the channel - label frequency data, combine the total time length of each channel in the concurrent abnormal segment, the number of historical normal behavior label types, and the number of abnormal co - occurrence labels, calculate the abnormal contribution frequency of the channel to the current concurrent abnormal segment, and statistically obtain the abnormal contribution quantity; S513: According to the abnormal contribution quantity, screen all channel numbers with an abnormal contribution quantity greater than or equal to 2, construct a mapping table with the corresponding switch identification port, mark all qualified channels as the dominant state numbers, load the dominant state numbers and synchronously output them to the switch state identification port, and send out the switch fault warning information.

[0012] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, by aligning the time intervals of multi-channel state transitions, dynamic linkage recognition between multiple signals is achieved. The temporal correlation between abnormal behaviors is strengthened by means of behavior label combinations and sliding statistical strategies. The sensitivity and accuracy of abnormal recognition are enhanced through the aggregated expression of composite alarm events. By combining cross-channel behavior label matrix analysis, the persistence of abnormal labels is clarified and key channels are identified, improving the accuracy and timeliness of switchgear early warning, reducing the risk of false alarms and missed alarms, supporting the real-time recognition of concurrent and progressive complex faults, while enhancing the comprehensiveness of the recognition logic and taking into account the clarity of the decision-making basis for alarm results, adapting to the actual needs of multi-source perception data fusion analysis of switchgear, and enhancing the traceability and judgment value of fault evolution trends. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is the main flowchart of the present invention; Figure 2 is the flowchart for obtaining the set of response time intervals of the present invention; Figure 3 is the flowchart for obtaining the time series group of behavior labels of the present invention; Figure 4 is the flowchart for obtaining the set of combined trigger windows of the present invention; Figure 5 is the flowchart for obtaining the multi-channel concurrent abnormal time period of the present invention; Figure 6 is the flowchart for obtaining switch fault early warning information of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0015] In the description of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0016] Please refer to Figure 1 , a switchgear fault early warning method based on multi-source data fusion, includes the following steps: S1: Obtain the first timestamps corresponding to the state transition points (matching the "Power Quality Monitoring Standard") in the voltage channel, temperature channel, and vibration channel of the switchgear. Arrange the channel numbers in chronological order and record the sorting sequence numbers. Combine the sorting sequence numbers with the first timestamps of the corresponding channels to form time series tuples (using the TimestampWithQuality structure in the IEC61850 standard) to obtain the response time interval set; S2: Based on the response time interval set, horizontally align the sampled data of the three types of channels. Detect whether the amplitude change of the voltage channel within two adjacent cycles is less than the voltage stability threshold. Detect whether the numerical difference of the temperature channel among three consecutive sampling points is less than or equal to the reference value (using the temperature gradient anomaly in IEEE C57.91-2011: temperature difference between adjacent sampling points ≤ 0.5 °C / min). Detect whether there is vibration energy attenuation in the vibration channel (according to the ISO10816-8 standard: ≤ 30% of the average value). If the corresponding conditions are met, mark them as voltage stagnation, temperature passivation, and vibration sudden drop labels respectively and record the corresponding timestamps to obtain the behavior label time series group; S3: According to the behavior label time series group, construct a sliding window (set the window length to 1 second and the step size to 0.5 second). Count the number of non-repeated behavior labels that appear in each sliding window. When the number of labels is greater than or equal to 2, combine the corresponding sliding window numbers, label types, and corresponding channel numbers, and uniformly number them as composite alarm event IDs (following the IEC62682 alarm management specification) to obtain the combined trigger window set; S4: According to the continuously appearing window numbers in the combined trigger window set, perform cross-matrix analysis on the marked corresponding channel numbers and corresponding behavior labels (using the fault correlation matrix method in "Machine Condition Monitoring and Diagnosis"), and extract all time periods with continuously appearing abnormal labels on the time axis to obtain the multi-channel concurrent abnormal time period; S5: According to the channel numbers involved in the multi-channel concurrent abnormal time period, combined with the occurrence frequencies of each channel behavior label within the interval, calculate the single-channel abnormal contribution frequency. Mark the channels with a contribution quantity greater than or equal to 2 as abnormal dominant channels, output them to the switch state recognition port, and send out switch fault warning information.

[0017] The response time interval set includes the channel response sequence, timestamp pairing structure, and time interval arrangement sequence. The behavior label time series group includes voltage stagnation labels, temperature passivation labels, and vibration sudden drop labels. The combined trigger window set includes sliding window numbers, composite alarm event IDs, and label type matching results. The multi-channel concurrent abnormal time period includes the channel number corresponding matrix analysis results, abnormal label continuous sections, and abnormal label cross-mapping relationships. The switch fault warning information includes behavior label frequency distribution records, channel number and abnormality correlation records, and switch fault warning real-time records.

[0018] Please refer to Figure 2 , the specific steps of S1 are as follows: S111: Obtain the original signal sequences monitored in the voltage channel, temperature channel, and vibration channel of the switching device, identify and extract the first timestamp value corresponding to the state transition point, and record them as the transition start time points of each channel respectively, forming a corresponding mapping set composed of channel numbers and timestamp values, and obtaining the channel first-hop time set; Based on the original signal sequences monitored in the voltage channel, temperature channel, and vibration channel of the switching device, respectively call the continuous sampling data sequences collected by the voltage sensor, thermocouple, and vibration accelerometer. Record data every millisecond according to the fixed sampling frequency, and perform state recognition processing on each group of data. Determine the position of the state transition point by comparing whether the mutation amplitude between the data at two consecutive time points exceeds the transition recognition threshold set for this type of channel. For example, if the sampling value at the sampling moment 14500010 ms in the voltage channel is 112.4 V, and the value jumps to 93.2 V at 14500015 ms, then since the amplitude mutation exceeds the set voltage transition threshold of 15 V, this point is determined as the state transition point and the first timestamp of the transition is recorded as 14500015 ms. Similarly, in the temperature channel, the sampling point is 38.9 °C at 14500030 ms and mutates to 41.2 °C at 14500035 ms, exceeding the set temperature transition threshold of 2 °C, it is judged as the transition point and the transition timestamp is recorded. In the vibration channel, taking 0.02 g as the vibration amplitude transition judgment benchmark, if the sampling value is 0.021 g at 14500080 ms and changes to 0.012 g at the next sampling point, then record the transition timestamp as 14500095 ms. All data must satisfy that the jump amplitude exceeds the threshold range of this channel to determine the validity of the state transition point. The state transition threshold needs to be set according to the device type standard. Among them, the voltage channel threshold is set to 15 V, the temperature channel is 2 °C, and the vibration channel is 0.01 g. The transition point detection process should ensure that only the first timestamp that meets the mutation amplitude condition is recognized in the single-channel detection process. Finally, calculate point by point in the sampling data sequence and according to the threshold setting standard of each channel, call the timestamp values corresponding to the first transition points in all channels, and summarize them in combination with the channel numbers to form a corresponding set of channel numbers and timestamps, and obtain the channel first-hop time set.

[0019] S112: According to the channel first-hop time set, sort all channel numbers in ascending order according to the corresponding first-hop timestamp values, record the original serial numbers corresponding to the channel numbers after the ascending order as the current sorting serial numbers, and construct a paired value pair set between the sorting serial numbers and the channel first-hop times, and obtain the time series mapping data group; According to the channel first jump time set, the collected voltage channel 1 time of 14500015ms, temperature channel 2 of 14500035ms, and vibration channel 3 of 14500095ms are first arranged in ascending order. After sorting, the order is voltage channel 1, temperature channel 2, and vibration channel 3. The original sequence number 1 of the voltage channel is assigned to the sorting sequence number 1, the original sequence number 2 of the temperature channel is assigned to the sorting sequence number 2, and the original sequence number 3 of the vibration channel is assigned to the sorting sequence number 3. The sorting sequence number is paired with its original transition time to form a value pair record, and the timestamp quality identification value is recorded in the IEC61850 standard structure. The quality identification is composed of parameters such as synchronization accuracy and time source consistency. According to the above mapping rules, three groups of record values are formed, namely, sorting sequence number 1-14500015ms, sorting sequence number 2-14500035ms, and sorting sequence number 3-14500095ms. Finally, a pairing set between the sorting sequence number and the first jump time is established to obtain a time series mapping data group.

[0020] S113: According to the time series mapping data group, based on the timestamp values between adjacent sorting numbers, calculate the response time differences between adjacent channels, sort the time intervals between all channel sorting numbers in sequence and classify them into a unified set, and obtain a response time interval set; According to the obtained sorting number and corresponding timestamp combination in the time series mapping data group, the response time difference is calculated for the timestamp difference between the two adjacent sorting numbers 1 and 2, 2 and 3. The channel transition time of sorting number 1 is 14500015ms, and the channel transition time of sorting number 2 is 14500035ms. The time interval is 14500035ms-14500015ms=20ms. The time difference between sorting numbers 2 and 3 is 14500095ms-14500035ms=20ms. 5ms=60ms, the above two time differences 20ms and 60ms are recorded in sequence to form a response time interval set. Each interval value in the set needs to meet the setting standard of the logical time response between devices. For example, each channel in the switch device should complete the sequence response process within 100ms, so the response time difference threshold is set to 100ms. When a channel interval exceeds this value, it is considered to exceed the response window. This value is calibrated according to the device response frequency in the actual system test and obtained by averaging 5 groups of samples. The sampling sample data is shown in Table 1.

[0021] Table 1 Response time interval test data table Sample Number Voltage Channel Timestamp (ms) Temperature Channel Timestamp (ms) Vibration Channel Timestamp (ms) Response Interval 1 (ms) Response Interval 2 (ms) 1 14500010 14500035 14500095 25 60 2 14500012 14500030 14500100 18 70 3 14500008 14500033 14500090 25 57 4 14500014 14500034 14500093 20 59 5 14500011 14500032 14500096 21 64

[0022] As shown in Table 1, the average response time intervals between the voltage, temperature, and vibration channels in the test samples are 22.8 ms and 62 ms respectively. Therefore, the upper threshold of the system response time difference is set to 100 ms. The response time involved in the calculation should be derived from the time stamp difference of continuous valid transition signals, rather than using the difference of repeated transition signals or invalid disturbance signals. Finally, the difference values between the above channels are recorded as the response time interval set.

[0023] Please refer to Figure 3 , and the specific steps of S2 are as follows: S211: Based on the response time interval set, perform unified horizontal alignment processing on the sampling time axes of the voltage channel, temperature channel, and vibration channel. Use the minimum time step as the reference step to construct a unified sampling time axis. Extract the sampling amplitude records at the corresponding time points in each channel, and complete data alignment by using the previous value retention method for the unsampled areas. After all sampling data alignment is completed, establish a horizontal channel mapping to obtain a three-channel alignment data matrix; Based on the response time interval set, by extracting the sampling time points of the voltage channel, temperature channel, and vibration channel in each sampling period, perform unified processing on the time axes of each channel. Select the shortest effective sampling time step of 100 ms as the alignment reference, and reconstruct the time axes of all channels according to this step. Extract the data with valid sampling at the corresponding time points of each channel as the basis for alignment data. If data is missing at a certain time point, fill in the blank through the previous value retention strategy, construct a complete horizontal channel data mapping table, extract example sample data, and construct the following alignment format, as shown in Table 2 specifically: Table 2 Three-channel Sampling Alignment Data Table Channel Type Sampling Time Point 1 Sampling Time Point 2 Sampling Time Point 3 Sampling Value 1 Sampling Value 2 Sampling Value 3 Voltage Channel 1000 1100 1200 220.0 219.8 220.1 Temperature Channel 1000 1100 1200 38.5 38.9 39.1 Vibration Channel 1000 1100 1200 0.062 0.058 0.017

[0024] As shown in Table 2, the sampling data of the three types of channels at three time points have been horizontally aligned through the reference step. Among them, the amplitude change between the three points of the voltage channel is within the range of ±0.3 V, the difference of the temperature channel is concentrated within 0.6 °C, and the vibration channel has an obvious downward trend in numerical value. This alignment data matrix serves as the basic data structure for subsequent behavior determination, and finally a three-channel alignment data matrix is obtained.

[0025] S212: According to the three-channel alignment data matrix, detect whether the amplitude change range within two adjacent sampling segments of the voltage channel is less than the voltage stability threshold, calculate two groups of temperature differences for every three consecutive sampling points in the temperature channel and compare whether their maximum difference is less than or equal to the reference value, extract the current sampling value of the vibration channel and determine whether it is lower than 30% of the average value. Mark the corresponding labels for the channel sampling points that meet the above judgment conditions respectively to generate a label determination sequence; According to the aligned channel data structures in Table 2, perform stability and status recognition and judgment operations on the voltage channel, temperature channel, and vibration channel respectively. First, calculate the difference in the voltage channel sampling values from 1000 ms to 1100 ms as 220.0 V - 219.8 V = 0.2 V, and the difference from 1100 ms to 1200 ms as 219.8 V - 220.1 V = -0.3 V. The absolute values are both less than the voltage stability judgment threshold of 0.5 V, meeting the voltage stagnation judgment condition. Secondly, for the temperature channel sampling point sequence of 38.5 °C, 38.9 °C, and 39.1 °C respectively, calculate the continuous differences as 0.4 °C and 0.2 °C, and the maximum difference is 0.4 °C, which is lower than the temperature passivation reference value of 0.5 °C, judging that temperature passivation is formed. For the vibration channel, extract three point values of 0.062 g, 0.058 g, and 0.017 g respectively. The current point (1200 ms) is 0.017 g, and the percentage obtained by comparing it with the average value of the previous two points of 0.06 g is 28.3%, which is less than the energy decay judgment lower limit of 30%, determining it as a vibration sudden drop. The above judgment process takes the original sampling data as the input item, combines the reference threshold, and screens and determines the status of the current channel at the current time point item by item. After looping through the judgment actions for each time period, record and combine all the status labels that meet the conditions, and finally generate a label judgment sequence.

[0026] S213: According to the label judgment sequence, map it to the labels of the three categories of voltage stagnation, temperature passivation, and vibration sudden drop respectively, integrate it into a structured recording unit in chronological order, obtain the channel numbers and event timestamps corresponding to each type of behavior, and get the behavior label time series group; According to the judgment results of the behavior labels assigned in the label judgment sequence, extract all the sampling moments and channel information where the behavior status recognition is successful point by point, and map them to a unified structured recording form. Use the three elements of channel number, behavior category, and timestamp to form a recording unit. For example, if the voltage channel at 1100 ms hits an amplitude fluctuation less than the threshold and is recorded as a voltage stagnation behavior, then construct the recorded value V1_S_1100. If the temperature difference of the temperature channel is lower than the reference at 1200 ms and forms temperature passivation, it is recorded as T1_D_1200. If the vibration channel has an energy decrease lower than the average value by 30% at 1200 ms and forms a sudden drop behavior, it is recorded as A1_F_1200. Finally, sort all the records by time and summarize them into a unified sequence set to obtain the behavior label time series group.

[0027] Please refer to Figure 4 , the specific steps of S3 are: S311: Obtain the behavior label, channel number, and timestamp corresponding to each record in the behavior label time series group. Construct a sliding window on the overall time axis according to the time order. Perform time attribution screening on the label records included in each sliding window, extract all label records within the range of this time window, construct a label summary table, remove duplicate behavior labels, and perform unique counting to generate the window label count result; Obtain the behavior label, channel number, and timestamp of each record in the behavior label time series group to construct the basic data structure for sliding window statistics. First, based on the timestamp information of the appearance time of the behavior label, establish a time axis with the time range of the current maximum and minimum timestamps, and divide it into continuous time periods with a window length of 1 second and a step size of 0.5 second to construct the continuous sliding window numbers and their start and end time ranges. From the behavior label time series, screen out the record sets that fall within the current time period within each window range, and map each label record to the window number it belongs to to form the label set corresponding to each sliding window. Remove duplicates from the behavior label content in this set to obtain the unique label type set, and calculate the number of elements in the set as the unique behavior label value under the current window. This operation can be implemented as a sliding traversal process based on the specific sampling period in actual deployment. For example, in a certain system, the sampling start and end times are from 0 second to 5 seconds, and the behavior labels appear at 1.2 seconds, 1.5 seconds, 1.9 seconds, 2.7 seconds, and 2.9 seconds respectively. Then the sliding window slides from the starting point of 0 second to the ending point of 5 seconds to generate a total of 9 window segments. Each segment counts the label records within the window time range and extracts the unique label types respectively. The table is shown as follows: Table 3 Statistical Table of Sliding Window Label Quantities Sliding Window Number Start Time (s) End Time (s) Label Record Quantity Unique Label Type Quantity W1 0.0 1.0 0 0 W2 0.5 1.5 1 1 W3 1.0 2.0 3 2 W4 1.5 2.5 2 2 W5 2.0 3.0 3 2 W6 2.5 3.5 2 1

[0028] As shown in Table 3, by counting the number of non-repeated behavior labels that appear within the sliding window, the changing trend of label aggregation for each time window can be effectively extracted, and finally the window label count result is generated.

[0029] S312: According to the window label count result, screen out all sliding windows with the label quantity greater than or equal to 2, extract the label types and corresponding channel numbers included in the windows, and combine them with the window numbers to form a unique identification structure for all windows that meet the conditions. Integrate information based on the behavior label and channel number to obtain the composite label identification set; According to the window label counting results, filter out all sliding windows with the number of unique labels greater than or equal to 2, establish a combined structure of corresponding window numbers, label types, and channel numbers. Extract the label type and its corresponding channel number information for each record in the filtered windows, and construct a label-channel dual-field combined table. Each combined field is concatenated in the format of "label_channel number", and then combined with the sliding window number to form a three-field structure. The above structure is used for subsequent composite event identification generation. Uniformly assign event record numbers with the sliding window number as the index. For example, in window W3, the labels A1, B1, and B2 belong to channels V1, T1, and T2 respectively. Then the combinations A1_V1, B1_T1, and B2_T2 are constructed respectively, and combined with the window number W3 to form W3_A1V1, W3_B1T1, and W3_B2T2, which are used for subsequent unified number generation of composite events. Construct complete composite label table entries for all windows that meet the conditions to obtain a composite label identification set.

[0030] S313: Based on the composite label identification set, assign numbers to each composite record in accordance with the alarm management specification, combine the time number of the sliding window to construct a unique composite alarm event identification, and file and integrate the number with the start and end times of the window, the involved channels, and the label types, and construct a unified sequence in chronological order to generate a combined trigger window set; According to the composite label identification set, perform a number mapping operation on each record in it, perform a structure encoding on the composite event with reference to the IEC62682 standard. Use the ascending order of the sliding window number as the main sequence of the number, append the behavior type code and the channel number in the combination order for each composite combination as the secondary sequence, and use a unified string structure to concatenate to form a unique event identification. Construct this structure number for all the determined combinations, and record the event identification, the start and end times of the original window time period, the behavior types involved in the combination, and the channel number set into the structured record sequence, and construct a composite alarm event entry table containing four fields: number, time, label, and channel information. For example, under window W3, an ID of "EVT03_A1V1_B1T1_B2T2" is constructed, its corresponding time period is from 1.0 second to 2.0 seconds, the label set is {A1, B1, B2}, and the channel set is {V1, T1, T2}, and finally a combined trigger window set is generated.

[0031] Please refer to Figure 5 , the specific steps of S4 are as follows: S411: According to the sliding window numbers, corresponding channel numbers, and behavior labels recorded in the combined trigger window set, extract all the sliding window numbers and sort them in chronological order. Calculate whether the difference between adjacent numbers in the window number sequence is 1. Continuously merge the window numbers with a difference equal to 1 to construct a set of time-continuous window segment sequences and obtain a continuous window label set; According to the sliding window numbers, channel numbers, and behavior label contents recorded in the combined trigger window set, extract the sequence of sliding window numbers, sort all the numbers in ascending order, and then calculate the interval values between adjacent numbers item by item. If the difference between two adjacent numbers is 1, it is determined that they form a continuous number segment, and the adjacent numbers that meet the conditions are merged into window segment groups in turn. For example, if the number sequence is W2, W3, W4, W6, W7, W9, then the continuous segments are constructed as {W2-W4, W6-W7}. Horizontally pair and extract the channel numbers and behavior label contents involved in each segment, and summarize them into a channel-label combination record table, as shown in Table 4 specifically: Table 4 Continuous Window Channel Label Combination Table Window Segment Number Channel Number Label Type Behavior Times (times) Duration (seconds) W2 - W4 V1 S 6 3 W2 - W4 V2 D 7 3.5 W6 - W7 T1 D 9 4 W6 - W7 T1 F 3 2

[0032] As shown in Table 4, each continuous window segment contains multiple combinations of channels and behavior labels. This data structure serves as the basic input set for subsequent coupling degree calculations, and finally obtains a continuous window label set.

[0033] S412: Based on the continuous window label set, construct a cross matrix structure with the channel number as the row index and the behavior label as the column index. In each cross cell, count the frequency of the combination of channels and labels in the continuous window segment, and at the same time extract the co-occurrence times of each group of channel-behavior pairs, and use the formula: ; Calculate the behavior coupling degree of channels and labels in the time series window, and determine whether it exceeds the behavior anomaly determination threshold. Screen all channel behavior combinations that meet the coupling strength determination standard to generate a channel-label coupling degree matrix, where, represents the channel and the behavior label coupling degree, represents the channel and the label synchronous co-occurrence time set on the time axis, 、 respectively represent the channel and the label behavior times, 、 represent their durations, 、 represent the behavior frequency (Hz), 、 represent the fluctuation standard deviation (Hz) of this channel and label; Based on the set of consecutive window labels, a two-dimensional cross matrix is constructed, where the channel number is the row index and the behavior label is the column index. In each cross cell, the behavior frequency and its behavior fluctuation amplitude of the corresponding channel and behavior label combination in the consecutive window segment are statistically calculated. Relevant data is extracted from the sample. For example, in the combination of channel V1 and label S, the number of behaviors is 6 times and the duration is 3 seconds, so the behavior frequency is 2 Hz and the standard deviation is 0.8 Hz. The total number of times of label S itself is 5 times, the duration is 2.5 seconds, the frequency is 2 Hz, and the standard deviation is 0.6 Hz. The co-occurrence time is 2 seconds, and it is substituted into the formula for calculation: ; Taking the T1 channel and label D as an example again, the co-occurrence time is 3 s, the behavior frequencies are 9 / 4 = 2.25 Hz and 7 / 3.5 = 2 Hz respectively, and the standard deviations are 1.1 Hz and 0.9 Hz respectively. Then: ; Repeat the above process for all combinations, calculate the coupling degree matrix, and screen and extract the channel-label combinations with coupling degree values greater than or equal to the set threshold to form a set of strong coupling relationships between channels and labels. The setting of this threshold is based on taking the top 30% quantile as the recognition boundary after sorting all the coupling degree calculation results in the sample window segment, and this quantile falls within the interval [2.3, 2.7] in five groups of complete sliding window samples. Therefore, it is set to 2.5. Its value fluctuation is highly correlated with the channel behavior frequency and the label appearance density. The higher the channel frequency or the wider the label coverage, the easier it is for this value to rise. If the frequency is low but the synchronization is strong, it will also form a critical coupling. Finally, a channel-label coupling degree matrix is generated.

[0034] Behavior coupling degree is used to measure the synchronous correlation strength between a specific channel and a specific behavior label in the time series. It reflects whether they frequently appear simultaneously within a unit time, whether the behavior intensity is consistent when they appear, and whether their behavior patterns are stably consistent. Specifically, the higher the coupling degree value, the stronger the collaborative relationship with high time overlap, high behavior frequency, and low volatility between the channel and the label, indicating that this behavior label is very likely to be a true reflection or dominant feature of the current operating state of the channel. On the contrary, if the coupling degree value is low, it may mean that the appearance of this behavior label is an accidental perturbation, with a delay or inconsistency with the behavior of the channel itself. Therefore, this indicator can be used to identify channel-label combinations with stable behavior responses and system linkage characteristics in multi-channel concurrent monitoring scenarios, thus supporting the subsequent establishment of fault area identification or concurrent anomaly triggering mechanisms.

[0035] The formula aims to comprehensively measure the coupling degree between the co-occurrence density, behavior intensity, and fluctuation stability of the channel and the behavior label in time. Among them, the numerator part uses the co-occurrence duration of the channel and the label on the time axis and their respective unit time behavior frequencies , The sum of the two is the product of the overlap degree of the channel and the tag behavior and the density of their respective behaviors within the effective time, reflecting the synchronization strength. The denominator introduces As a joint measure of the fluctuation degree of the two, the total variation degree of the expression channel and the stability of the label behavior is expressed. The larger the degree, the more unstable it is. At the same time, the absolute value of the frequency difference between the two is added. , which is used to punish combinations with inconsistent behavior patterns, and comprehensively forms a ratio relationship of "behavior co-occurrence density to behavior instability and inconsistency", thereby achieving high scores for channel-label combinations with strong synchronization, high intensity but low volatility, and lowering the scores of combinations with large frequency differences or violent fluctuations, ensuring that the coupling index has recognition and discrimination in behavioral state recognition.

[0036] S413: According to the channel label coupling matrix, the corresponding time index is extracted and the continuously distributed segments in the time set are integrated to determine the continuous segment sequence where the abnormal label appears on the time axis, and the start and end time of each segment record and the corresponding channel and label combination are summarized and integrated in time order to obtain the multi-channel concurrent abnormal period;

[0037] The channel label coupling matrix extracts the associated timestamp range for the channel-label combination with a coupling value greater than or equal to 2.5, locates the co-occurrence time segment of each combination in the set, aggregates adjacent time points, and merges continuous time indexes into complete segments. For example, the co-occurrence time points of 1.5s, 2.0s, 2.5s, and 3.0s are merged into the [1.5s-3.0s] segment. The channel numbers and behavior label types involved are then recorded in pairs. All eligible combinations are integrated to construct a record table consisting of channel numbers, label types, and time segments. Finally, the concurrent behavior structure data of all coupling-significant segments are output to obtain multi-channel concurrent abnormal time periods.

[0038] See also Figure 6 , the specific steps of S5 are: S511: according to the channel numbers involved in the multi-channel concurrent abnormal period, count the number of behavior labels corresponding to each channel number in the covered time segment, extract the label frequency information of the channel in the concurrent abnormal segment, calculate the sum of the number of label records in each channel respectively, and obtain the channel label frequency data; Extract the sliding time intervals corresponding to each channel number based on the channel numbers involved in the multi-channel concurrent abnormal period, and perform duration statistics on them. The statistical content needs to call the original behavior label record sequence within the abnormal period for each channel number, extract the number of records in the sequence as the label frequency value, and establish a one-to-one mapping relationship between the frequency value and the channel number. When executing this process, the number of label frames recorded per second based on the sliding window number needs to be accumulated. For example, there are 6 label records for channel G1 between abnormal segments W1 to W5, 9 for channel G2, and 4 for channel G3. Record the frequency data with the channel number as the index. If there are multiple window segments corresponding to the same channel number, their frequencies need to be accumulated. For example, if G1 appears in two segments W1-W3 and W6-W7, with frequency values of 3 and 5 respectively, then the total frequency value is 8. Finally, structure the above frequency values into key-value pairs, establish the corresponding mapping structure between the channel number and the total frequency of behavior labels, and obtain the channel label frequency data.

[0039] S512: According to the channel label frequency data, combine the total time length of each channel within the concurrent abnormal segment, the number of types of historical normal behavior labels, and the number of abnormal co-occurring labels, and use the formula: ; Calculate the abnormal contribution frequency of the channel to the current concurrent abnormal segment, and statistically obtain the abnormal contribution quantity. Among them, represents the abnormal contribution frequency (Hz) of channel to the abnormal segment, represents channel and label 's co-occurrence time (seconds) within the abnormal segment, represents channel 's coverage time (seconds) in this segment, represents channel and label 's co-occurrence frequency per unit time (Hz), represents label 's recognition frequency per unit time (Hz) in the overall segment, represents channel 's number of types of historical normal behavior labels, represents channel 's number of types of historical abnormal co-occurring labels, is the total number of labels participating in the calculation; According to the channel label frequency data, extract the co-occurrence frequency, co-occurrence time, channel duration, label duration, total label frequency, number of historical normal label types, and number of abnormal co-occurrence label types of each channel and its corresponding label within the section. Construct a time-frequency feature table for each channel and its associated labels. Taking channel G1 as an example, its co-occurrence time with label K1 is 2.0 seconds, the duration of the channel itself is 5.0 seconds, the frequency is 8 times, and the unit frequency is 1.6 Hz. At the same time, label K1 appears 12 times in this section, its duration is 6.0 seconds, and the frequency is 2 Hz. Substitute the data of G1 into the formula for calculation as follows: ; Substitute the data of G2 into the formula for calculation as follows: ; Substitute the data of G3 into the formula for calculation as follows: ; Thus, the abnormal contribution frequency values of each channel are G1 = 0.16 Hz, G2 = 0.32 Hz, and G3 = 0.23 Hz. The relevant parameters are shown in Table 5: Table 5 Calculation Parameters Table of Abnormal Contribution Channel Number Label Number Co - occurrence Duration (seconds) Channel Duration (seconds) Label Duration (seconds) Co - occurrence Frequency (times) Total Label Frequency (times) Normal Label Types Abnormal Label Types G1 K1 2.0 5.0 6.0 8 12 4 3 G2 K2 3.5 7.0 6.5 10 15 3 2 G3 K3 1.5 4.0 5.0 5 7 2 1

[0040] As shown in Table 5, all parameters are from the sliding window sampling process and behavior record statistics. The frequency is calculated as the frequency divided by the duration. The total label frequency needs to be aggregated and counted among all channels to obtain the abnormal contribution quantity.

[0041] The abnormal contribution frequency is used to measure the degree of aggregation of a certain channel to the abnormal behavior pattern during the multi-channel concurrent abnormal period, specifically reflecting the co-occurrence density, frequency deviation degree, and its behavior response complexity between the channel and multiple abnormal labels per unit time. The higher this index, the more significant the co-temporal distribution between the channel and abnormal labels in the abnormal section, and the stronger the deviation of its behavior frequency from the average behavior pattern of the labels, indicating that the channel not only participates in the abnormal process but also has dominant behavior characteristics in the time-frequency distribution. At the same time, the abnormal contribution frequency also combines the number of label types carried by the channel in historical behaviors to dynamically adjust the amplitude of contribution assessment and avoid misjudgment of channels with complex label structures. Therefore, this index combines the three aspects of current behavior density, frequency difference, and historical label distribution and is the key calculation basis for identifying the dominant abnormal channel.

[0042] The formula is constructed based on the behavior coordination characteristics of the channel and the label within the abnormal section, where the numerator part is obtained by multiplying the co-occurrence time of the channel and the label by the total time of the channel Perform ratio processing to obtain the normalized occurrence ratio of the label in the channel, and then compare it with the channel unit time frequency Average frequency of labels The absolute value of the sum and difference between them is added to characterize the frequency matching strength and difference sensitivity. This composite feature is used as the coupling effect of the label on the channel to form the time-frequency coupling term in a multiplicative manner, and then for all labels Sum and form a channel The multi-label coupling aggregation behavior is expressed by; the denominator uses the number of normal labels in the channel history The number of historical abnormal labels The square root of the sum is intended to reflect the structural complexity of the channel label behavior. The square root operation slows down its growth, which plays a role in normalization and suppressing the influence of historical characteristics, thereby ensuring that the current coupling strength of the channel is the dominant indicator. The structure as a whole controls the label coverage in the channel through time normalization, guides the offset strength evaluation of the channel label pair with the frequency matching degree, and then makes relative corrections based on the historical behavior background to form a stable and dynamically balanced abnormal contribution frequency measurement indicator.

[0043] S513: According to the abnormal contribution quantity, filter all channel numbers whose abnormal contribution quantity is greater than or equal to 2, construct a corresponding switch identification port mapping table, mark all qualified channels as dominant state numbers, load the dominant state numbers and synchronously output them to the switch state identification port, and issue a switch fault warning message; According to the abnormal contribution amount, the corresponding The value is screened to determine whether it meets the abnormal dominant channel standard and the threshold benchmark is set as This value is obtained by comprehensively counting the occurrence frequency and label coverage in 18 groups of channel fault reproduction experiments. It falls at the critical turning point of the interval [0.2, 0.3]. The average value is selected as the setting benchmark. Combined with the above calculation, the G1 value is 0.16Hz, G2 is 0.32Hz, and G3 is 0.23Hz. Only G2 meets the set conditions. The channel number G2 is screened out, and the corresponding mapping table between it and the control port is established. The state signal path is configured, and the channel number G2 is marked as the dominant channel. The state identification signal is sent to the downstream identification module and written into the event stream recording module to obtain the switch fault warning information.

[0044] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for fault warning of switchgear based on multi-source data fusion, characterized in that, It includes the following steps: S1: Obtain the first timestamps corresponding to the state transition points in the voltage channel, temperature channel, and vibration channel of the switchgear, sort them in chronological order, and obtain a set of response time intervals; S2: Based on the set of response time intervals, horizontally align the sampled data, detect whether there are situations of voltage stagnation, temperature passivation, and vibration sudden drop labels, and record the corresponding timestamps to obtain a group of behavior label time series; S3: According to the group of behavior label time series, construct a sliding window, count the number of non-repeated behavior labels that appear in each sliding window. When the number of labels is greater than or equal to 2, combine the corresponding sliding window number, label type, and corresponding channel number, and uniformly number them as the composite alarm event ID to obtain a set of combined trigger windows; S4: According to the set of combined trigger windows, perform cross-matrix analysis on the marked corresponding channel numbers and corresponding behavior labels, and extract all time periods during which abnormal labels continuously appear on the time axis to obtain multi-channel concurrent abnormal time periods; S5: According to the channel numbers involved in the multi-channel concurrent abnormal time periods, calculate the single-channel abnormal contribution frequency, mark the abnormal dominant channel, and send a switch failure warning message.

2. The method for early warning of switchgear faults based on multi-source data fusion according to claim 1, characterized in that, The set of response time intervals includes the channel response sequence, timestamp pairing structure, and time interval arrangement sequence. The group of behavior label time series includes voltage stagnation labels, temperature passivation labels, and vibration sudden drop labels. The set of combined trigger windows includes sliding window numbers, composite alarm event IDs, and label type matching results. The multi-channel concurrent abnormal time periods include channel number corresponding matrix analysis results, abnormal label continuous sections, and abnormal label cross-mapping relationships. The switch failure warning message includes behavior label frequency distribution records, channel number and abnormality correlation records, and switch failure warning real-time records; Voltage stagnation: The amplitude change of the voltage channel within two adjacent cycles is less than the voltage stability threshold; Temperature passivation: The numerical difference of the temperature channel in three consecutive sampling points is less than or equal to the reference value; Vibration sudden drop: There is vibration energy attenuation in the vibration channel; Abnormal dominant channel: A channel with a contribution quantity greater than or equal to 2.

3. The method for early warning of switchgear faults based on multi-source data fusion according to claim 1, characterized in that The specific steps for obtaining the set of response time intervals are as follows: S111: Obtain the original signal sequences monitored in the voltage channel, temperature channel, and vibration channel of the switchgear, identify and extract the first timestamp values corresponding to the state transition points, and record them as the transition start time points of each channel respectively, forming a corresponding mapping set composed of channel numbers and timestamp values to obtain a set of channel first-hop times; S112: According to the set of channel first-hop times, sort all channel numbers in ascending order according to the corresponding first-hop timestamp values, record the original sequence numbers corresponding to the channel numbers after ascending sorting as the current sorting sequence numbers, and construct a set of paired value pairs between the sorting sequence numbers and the channel first-hop times to obtain a group of time series mapping data; S113: Based on the time series mapping data group, calculate the response time difference between adjacent channels based on the timestamp values between adjacent sorting serial numbers, sort the time intervals between all channel sorting serial numbers in sequence and incorporate them into a unified set to obtain the response time interval set.

4. The method for early warning of switchgear faults based on multi-source data fusion according to claim 1, wherein The specific steps for obtaining the behavior label time series group are as follows: S211: Based on the response time interval set, perform unified horizontal alignment processing on the sampling time axes of the voltage channel, temperature channel, and vibration channel. Use the minimum time step as the reference step distance to construct a unified sampling time axis, extract the sampling amplitude records at the corresponding time points in each channel, fill the unsampled areas using the previous value retention method to complete data alignment. After all the sampling data is aligned, establish a horizontal channel mapping to obtain a three-channel aligned data matrix; S212: According to the three-channel aligned data matrix, detect whether the amplitude change range within two adjacent cycle sampling segments in the voltage channel is less than the voltage stability threshold, calculate the temperature difference between two groups for every three consecutive sampling points in the temperature channel and compare whether the maximum difference is less than or equal to the reference value, extract the current sampling value in the vibration channel and determine whether it is lower than 30% of the mean value. Mark the corresponding tags for the channel sampling points that meet the above judgment conditions respectively to generate a tag determination sequence; S213: According to the tag determination sequence, map it to the tags of the three categories of voltage stagnation, temperature passivation, and vibration sudden drop respectively, integrate it into a structured record unit in chronological order, obtain the channel numbers and event timestamps corresponding to each type of behavior, and obtain the behavior label time series group.

5. The fault warning method for switchgear based on multi-source data fusion according to claim 1, wherein The specific steps for obtaining the combined trigger window set are as follows: S311: Obtain the behavior labels, channel numbers, and timestamps corresponding to each record in the behavior label time series group. Construct a sliding window on the overall time axis according to the time sequence, perform time attribution screening on the label records included in each sliding window, extract all the label records within the range of this time window and construct a label summary table, perform unique counting after removing duplicate behavior labels to generate a window label counting result; S312: According to the window label counting result, screen all the sliding windows with the number of labels greater than or equal to 2, extract the label types and corresponding channel numbers included in the window, and combine them with the window number. Form a unique identification structure for all the windows that meet the conditions, and perform information integration based on the behavior labels and channel numbers to obtain a composite label identification set; S313: Based on the composite label identification set, allocate numbers to each composite record according to the alarm management specification, combine the time number of the sliding window to construct a unique composite alarm event identification, and file and integrate the number with the start and end times of the window, the involved channels, and the label types, and construct a unified sequence in chronological order to generate a combined trigger window set.

6. The method for early warning of switchgear faults based on multi-source data fusion according to claim 1, wherein, The specific steps for obtaining the multi-channel concurrent abnormal period are as follows: S411: Extract all the sliding window numbers sorted in chronological order according to the sliding window numbers, corresponding channel numbers, and behavior labels recorded in the combined trigger window set. Calculate whether the difference between adjacent numbers in the window number sequence is 1. Continuously merge the window numbers with a difference equal to 1 to construct a set of sequences of continuously time-ordered window segments, and obtain a set of continuous window labels; S412: Based on the set of continuous window labels, construct a cross-matrix structure with the channel number as the row index and the behavior label as the column index. In each cross-cell, count the frequency of the combination of the channel and the label in the continuous window segment. At the same time, extract the co-occurrence times of each group of channel-behavior pairs, calculate the behavior coupling degree between the channel and the label in the time-series window, and determine whether it exceeds the behavior anomaly determination threshold. Screen all channel-behavior combinations that meet the coupling strength determination standard to generate a channel-label coupling degree matrix; S413: According to the channel-label coupling degree matrix, extract the corresponding time index and integrate the continuously distributed sections in the time set to determine the sequence of consecutive sections where the abnormal label appears on the time axis. Summarize the start and end times, corresponding channels, and label combinations recorded in each section, and integrate them in chronological order to obtain the multi-channel concurrent anomaly time period.

7. The method for fault early warning of switchgear based on multi-source data fusion according to claim 1, characterized in that The specific steps for obtaining the switch fault warning information are as follows: S511: According to the channel numbers involved in the multi-channel concurrent anomaly time period, count the number of behavior labels corresponding to each channel number within the covered time section, extract the label frequency information of the channel in the concurrent anomaly section, calculate the sum of the label records in each channel respectively, and obtain the channel label frequency data; S512: According to the channel label frequency data, combine the total time length of each channel in the concurrent anomaly section, the number of types of historical normal behavior labels, and the number of abnormally co-occurring labels, calculate the abnormal contribution frequency of the channel to the current concurrent anomaly section, and statistically obtain the abnormal contribution quantity; S513: According to the abnormal contribution quantity, screen all channel numbers with an abnormal contribution quantity greater than or equal to 2, construct a mapping table with the corresponding switch identification ports, mark all qualified channels as the dominant state numbers, load the dominant state numbers and synchronously output them to the switch state identification ports, and issue a switch fault warning information.

Images