One-time fusion ring network box fault early warning method and system based on multi-source data fusion

Abstract

The application discloses a primary and secondary fusion ring network box fault early warning method and system based on multi-source data fusion. The method associates device state records with collection link records, and corrects a time reference by using switch operation events; according to a topology state, an operation event, a load and an environmental condition, a health reference is selected, a common change component of a target device and the reference object is deducted, and a corrected deviation representing a local change is obtained; when the corrected deviation reaches a warning starting condition, actual changes of associated measurement positions are verified within an allowed time delay. When the actual changes conform to a fault physical response relationship and a collection link is normal, an anomaly is attributed to a primary device; when a specified associated change is missing and the collection link is abnormal, the anomaly is attributed to a secondary measurement link; and in other cases, rechecking is performed. The system outputs a warning object, a risk level, evidence records and disposal suggestions accordingly.

Description

Technical Field

[0001] This invention relates to the field of power distribution equipment condition monitoring and fault early warning technology, specifically to a method and system for fault early warning of primary and secondary fusion ring network boxes based on multi-source data fusion. Background Technology

[0002] The integrated primary and secondary ring main unit integrates primary switchgear, status monitoring components, power distribution terminals, and communication components, enabling it to generate equipment operation records while simultaneously performing power line switching, isolation, and protection. Because the status of primary equipment, external loads, the internal environment of the cabinet, and the secondary acquisition link interact, the same record changes may correspond to different sources of anomalies.

[0003] Existing fault early warning methods typically employ fixed thresholds, single-channel trends, or direct statistical fusion of multiple types of records. When a ring network box undergoes topology switching, load adjustment, or overall environmental changes, multiple measuring points may shift synchronously, making it difficult for a fixed baseline to distinguish between normal operating condition changes and localized degradation. Furthermore, when records from different sampling frequencies are not aligned with events, it is difficult to maintain the physical order of responses during switching operations.

[0004] Furthermore, sensor drift, sampling sequence interruption, clock skew, or communication anomalies can also create abrupt changes in records. Judging solely by the recorded amplitude can easily lead to attributing secondary measurement link anomalies to primary equipment; conversely, excluding anomalies based solely on data quality may miss actual equipment degradation. Therefore, a fault early warning method is needed that can eliminate common operating condition variations, verify the state response sequence, and distinguish between primary equipment and secondary measurement link anomalies. Summary of the Invention

[0005] To address the issues of synchronous offset of multiple measurement locations in primary and secondary integrated ring main units during topology switching, load adjustment, and environmental changes, and the easy confusion between sensor component, acquisition channel, or time synchronization anomalies and primary equipment degradation, this invention provides a fault early warning method and system for primary and secondary integrated ring main units based on multi-source data fusion. This method does not rely solely on exceeding status value limits; instead, it sequentially completes the establishment of record correspondence, health reference correction under the same operating conditions, verification of fault physical response, and attribution of abnormal objects.

[0006] One aspect of the present invention provides a method for early warning of faults in a primary and secondary fusion ring network box based on multi-source data fusion, comprising the following steps: S1. Obtain the device status records and corresponding acquisition link records of the primary and secondary fusion ring network box, establish the correspondence between the device, measurement location and acquisition link, and use the switch operation event to correct the time base of the record to form a status sequence organized by event or statistical window. S2, determine the current operating condition based on the topology and operation events of the state sequence, and select a health reference from the health reference library that has the same topology and event conditions and similar load and environmental conditions; S3, determine the deviations of the target equipment and the reference object relative to their respective health states, and based on the correspondence between the two deviations during healthy operation, subtract the common change component from the deviation of the target equipment to obtain the correction deviation characterizing the local change. S4, when the correction deviation reaches the warning start condition, verify the actual change direction and occurrence sequence of the associated measurement position within the corresponding allowable time delay according to the predetermined fault physical response relationship; S5. When the actual change conforms to the physical response relationship of the fault and the relevant acquisition link records are normal, the abnormality is attributed to the primary equipment; when the initial abnormality lacks the associated change specified by the physical response relationship of the fault and the completeness, timing continuity or physical correspondence of the corresponding acquisition link records is abnormal, the abnormality is attributed to the secondary measurement link; other cases are marked as pending review. S6 determines the risk level based on the duration of the anomaly, the trend of change, and the role of the monitored component in the current equipment topology, and outputs the warning object, fault category, risk level, status record and data acquisition link record for the attribution conclusion, as well as handling suggestions.

[0007] Specifically, in step S1, each device status record is associated with the device, interval, monitored component, measurement location, and acquisition link. Steady-state records are organized according to a uniform time grid, and action response records use control commands, coil responses, mechanism positions, auxiliary contact displacements, and main circuit state changes as time anchors for the same operational event.

[0008] Let δ_h be the time offset determined by the common event anchor point for the h-th acquisition link, and t_s be the original sampling time. Then the correction time t′ is expressed as: t′=t_s-δ_h.

[0009] When the time offsets given by the shared event anchor points are contradictory, or when the controlled state precedes the control command after correction, the correction result is not adopted, and the corresponding record is marked as a timing anomaly. Similarly, if a record is missing, the sampling sequence number is interrupted, or the record is inconsistent with the corresponding physical state, the record is also marked as a data acquisition link anomaly.

[0010] Specifically, in step S2, topology connectivity and operational event type serve as necessary matching conditions for health references; among health status segments that meet these conditions, they are filtered according to the similarity of load conditions and environmental conditions. Health references are only derived from status segments that have been confirmed to be normal after commissioning, testing, or maintenance; unconfirmed current segments are not used to change the health status.

[0011] Specifically, in step S3, let the target current value of the i-th state variable be x_i(t), the target healthy state be b_i(c), the reference current value be r_i(t), and the reference healthy state be... Let c represent the current operating condition. Then the target deviation Δx_i(t) and the reference deviation Δr_i(t) are expressed as follows: Δx_i(t) = x_i(t) - b_i(c);

[0012] The common variation coefficient γ_i is determined using the target deviation and reference deviation in the healthy operation segment, and the correction deviation ρ_i(t) is obtained according to the following formula: γ_i=Cov(Δx_i,Δr_i) / [Var(Δr_i)+ε_γ]; ρ_i(t)=Δx_i(t)-γ_i·Δr_i(t).

[0013] Where Cov represents covariance, Var represents variance, and ε_γ is a positive number with the same dimensions as Var(Δr_i) to prevent the denominator from being zero. The common variation coefficient is determined only from normal healthy segments recorded by the acquisition link; when the reference state is missing or the common variation coefficient cannot be stably determined, the reference bias is not deducted, and the current result is marked as insufficient reference.

[0014] Specifically, in step S4, the fault physical response relationship is stored according to the monitored component, including the initial abnormality measurement location, associated measurement location, expected change direction, allowable delay, and required associated changes. After the initial abnormality continuously meets the warning initiation conditions, the actual change of the associated measurement location is verified within the corresponding allowable delay. Changes with incorrect direction, exceeding the allowable delay, or abnormal acquisition link are not considered valid responses, and any missing associated changes are retained as response missing records.

[0015] Specifically, in step S5, when the specified associated changes of the initial anomaly and the physical response relationship of the fault both occur, and the relevant acquisition link records are normal, the anomaly is attributed to the primary device; when the initial anomaly occurs, the specified associated changes are missing, and the acquisition link record corresponding to the initial anomaly is abnormal, the anomaly is attributed to the secondary measurement link; when the status of the primary device and the status of the acquisition link are both abnormal, or the existing records cannot form a complete physical response relationship, the result is marked as pending review.

[0016] Specifically, in step S6, the risk level is determined based on the duration and trend of the correction deviation and the role of the monitored component in the current topology. Entering a higher risk level requires continuously meeting the corresponding entry conditions, while exiting requires exiting with lower entry conditions. The output record includes the current operating status, health reference source, correction deviation, actual occurrence and missing related changes, the time of each change, data acquisition link status, attribution conclusion, and handling recommendations.

[0017] Health reference maintenance and online early warning are performed separately. Status segments that are confirmed to be normal, have valid data collection records, and show no subsequent abnormalities during the observation period are added to the health reference candidate list; status segments confirmed to be due to primary equipment malfunctions or secondary measurement link malfunctions are not written into the health reference. Health status updates limit the magnitude of single changes and save the confirmation source, the record before the update, and the record after the update.

[0018] Another aspect of the present invention provides a primary and secondary fusion ring network box fault early warning system based on multi-source data fusion, comprising a record association and time correction module, a working condition reference correction module, a physical response verification module, an anomaly attribution determination module, and an early warning output module connected in sequence; The recording association and time correction module receives device status records and acquisition link records, establishes the correspondence between the device, measurement location and acquisition link, corrects the time reference according to the switch operation event, and outputs the status sequence and acquisition link status. The operating condition reference correction module selects a healthy reference based on the current operating conditions, deducts the common variation components of the target equipment and the reference object, and outputs the correction deviation. The physical response verification module generates records of actual changes and missing responses within the allowable time delay based on the correction deviation and the pre-stored fault physical response relationship. The anomaly attribution determination module combines actual changes, missing response records, and acquisition link status to output primary equipment anomaly, secondary measurement link anomaly, or a status pending verification. The early warning output module determines the risk level and generates evidence records based on the duration of the anomaly, the trend of change, and the effect of the equipment topology.

[0019] Specifically, the system also includes a feedback update module and a configuration storage unit; the configuration storage unit stores the correspondence between devices and measurement locations, acquisition link relationships, health references, fault physical response relationships, early warning parameters, and evidence records; the feedback update module associates the inspection results with the evidence records, confirms that normal status segments are used to maintain health references after access review, confirms that abnormal status segments of primary devices are used to verify fault physical response relationships, and confirms that abnormal status segments of secondary measurement links are used to update the maintenance status of the corresponding acquisition links.

[0020] Specifically, the system is set up in the edge computing device of the primary and secondary integrated ring network box; the primary and secondary integrated ring network box also includes primary switchgear, status monitoring components, power distribution terminals and communication components. The status monitoring components are set up corresponding to the monitored components of the primary switchgear. The power distribution terminal receives equipment status records and switch operation events and forms a data acquisition link record. The communication component sends the early warning results to the local alarm interface, remote master station or maintenance terminal and receives the inspection results.

[0021] In summary, compared to the processing method of directly using a fixed threshold, this invention uses the health status of the target device and the reference object as benchmarks, and deducts common loads or environmental changes when the topology and event conditions are consistent, so that local changes of the target device that do not appear in the reference object are retained in the correction deviation; when the reference is insufficient, the deduction is stopped and the risk level is limited, so as to avoid overcorrection caused by unsuitable reference objects.

[0022] This invention further verifies the direction and time delay of associated changes according to the electrical connections, heat conduction, or motion transmission relationships of the monitored components, and uses the data acquisition link records to determine whether abnormal values ​​have reliable physical response support. Thus, the same numerical anomaly can be classified as a primary equipment anomaly, a secondary measurement link anomaly, or pending verification based on the actual response and data acquisition link status, while retaining traceable records. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the primary and secondary fusion ring network box fault early warning method based on multi-source data fusion according to the present invention.

[0024] Figure 2 This is a schematic diagram illustrating the device status record association, event time correction, and acquisition link status determination of the present invention.

[0025] Figure 3 This is a schematic diagram illustrating the working condition health reference matching and correction deviation generation of the present invention.

[0026] Figure 4 This is a schematic diagram illustrating the physical response relationship and actual changes of the fault in this invention.

[0027] Figure 5 This is a schematic diagram illustrating the determination of primary equipment malfunction, secondary measurement link malfunction, and pending verification status in this invention.

[0028] Figure 6 This is a schematic diagram of the deployment of the fault early warning system of the present invention in a primary and secondary fusion ring network box. Detailed Implementation

[0029] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. The described embodiments are used to illustrate the technical solutions of the present invention and are not intended to limit the present invention to implementation with exactly the same equipment structure, measuring components, sampling period, parameter thresholds, or deployment location. Where there is no contradiction, the technical features in the following embodiments can be combined with each other.

[0030] Unless otherwise expressly stated, the term "connection" in this application includes direct connections and indirect connections formed through intermediate components; "correspondence" indicates a pre-configured device, measurement location, event, or data acquisition link relationship between two objects; and "current window" refers to a steady-state statistical window or operational event window that is currently performing early warning processing. Endpoint values ​​representing numerical ranges are included within the corresponding range.

[0031] The status record in this application reflects the operating conditions of a primary switchgear, monitored component, or related conditions. The data acquisition link record reflects the sampling, time, sequence number, quality, and transmission status of the status record from the measurement location to the power distribution terminal or edge computing device. Verification of the data acquisition link record includes the record's completeness, temporal continuity, and consistency with the corresponding physical state. The verification result is not equivalent to a direct judgment of the health status of the primary equipment.

[0032] Example 1

[0033] This embodiment combines Figures 1 to 5 This paper describes a fault early warning method for primary and secondary integrated ring main units based on multi-source data fusion. The primary and secondary integrated ring main unit includes primary switchgear, status monitoring components corresponding to the monitored components, a power distribution terminal, and an edge computing device for performing early warning processing. The edge computing device is located inside the power distribution terminal or connected to the power distribution terminal via a local communication interface.

[0034] This embodiment employs steady-state statistical windows and operational event windows to handle state changes at different time scales. Slow changes such as temperature, humidity, load, and insulation status are recorded using steady-state statistical windows; rapid changes such as control commands, coil responses, mechanism positions, auxiliary contacts, and main circuit status are recorded using operational event windows. Both types of windows retain the correspondence between equipment, intervals, monitored components, measurement locations, and data acquisition links.

[0035] like Figure 1 As shown, this embodiment executes steps S1 to S6. Figure 1Solid arrows indicate the transmission paths of state sequences, health references, correction deviations, actual changes, and attribution results; dashed arrows indicate controlled update paths after maintenance confirmation. Normal state segments are used to maintain health references after access review, while abnormal state segments are used to verify fault physical response relationships or update the maintenance status of the data acquisition link. Unconfirmed segments do not change the online health status.

[0036] Step S1: Associate the equipment status records and the data acquisition link records, and correct the time base of the relevant records according to the operation events. The equipment status records include records reflecting the status of primary switchgear, conductive connection parts, operating mechanisms, and insulation environment; the data acquisition link records include sampling time, arrival time, sampling sequence number, quality flag, time synchronization status, and message transmission status.

[0037] Each device status record is associated with the device identifier, interval identifier, phase or functional unit identifier, monitored component, measurement location, and the data acquisition link from the sensing component to the power distribution terminal point. Through this correspondence, the system can identify the corresponding primary component and secondary acquisition path when a status record is abnormal, and determine whether multiple records reflect the same physical process.

[0038] When the condition monitoring components and power distribution terminals use different time bases, the common changes of the same operational event in different records are used as time anchors. Let the time offset determined by the h-th acquisition link based on the common event anchor be δ_h, and the original sampling time be t_s, then the correction time is t′=t_s-δ_h. The time offset is determined using the median or truncated average of multiple common anchors.

[0039] If, after correction, a control command is later than the change in the controlled state, the offsets given by different common anchor points contradict each other, or the sequence of events does not conform to the switching action process, the corresponding record will be marked as a timing anomaly. Action response records maintain their chronological relationship relative to the event start point, and missing records will not be replaced by interpolation results to determine whether a response occurred.

[0040] Steady-state records are organized according to a uniform time grid. Records with a frequency higher than the uniform grid frequency form representative values ​​and changes within the grid window, while records with a frequency lower than the uniform grid frequency retain the most recent valid value and data age. If a record is missing beyond the allowable range, the sampling sequence number is interrupted, the corrected time sequence is abnormal, or the record is inconsistent with the corresponding physical state, the acquisition link status is marked as abnormal; all other records are marked as valid.

[0041] Figure 2The data path for step S1 is shown. Device status records and acquisition link records enter the device-measurement location-acquisition link association processing, while switch operation events enter time correction processing. The two results are organized into a state sequence after passing through a steady-state segment and an action response segment. Record integrity, sequence number continuity, event order, and physical correspondence form separate acquisition link states, which are not merged with primary device status values ​​to form new state variables.

[0042] Step S2: Determine the current operating condition based on the topology status and operation events, and select a comparable health reference from the health reference library. The current operating condition in the steady-state window includes power supply direction, bus connection relationship, switch position, load conditions, and environmental conditions; the operation event window also includes operation direction, pre-operation status, and post-operation status.

[0043] The health reference library is established from confirmed normal status segments obtained during factory testing, handover testing, initial operation, and post-maintenance checks. Each status segment stores the corresponding operating condition, data acquisition link status, formation time, and confirmation status. Unconfirmed current segments, primary equipment malfunction segments, and data acquisition link malfunction segments are not considered as health references.

[0044] Candidate health segments first satisfy the conditions of consistent topological connectivity and operational event type, and then are sorted according to load conditions, environmental conditions, acquisition link status, and formation time. The selected segments determine the health status of the target device under the current operating conditions, and reference objects under the same power supply and environmental conditions determine the health status and current changes of the reference objects. A correspondence is established between the measurement location, dimensions, and physical meaning of the reference objects and the target devices.

[0045] Step S3: Calculate the deviations of the target device and the reference object relative to their respective health states. Let the current target value of the i-th state variable be x_i(t), the target health state be b_i(c), the current reference value be r_i(t), and the reference health state be... Then the target deviation and reference deviation are respectively: Δx_i(t) = x_i(t) - b_i(c);

[0046] The common variation coefficient γ_i is estimated using normal healthy segments recorded by the acquisition link, and the correction bias ρ_i(t) is obtained according to the following formula: γ_i=Cov(Δx_i,Δr_i) / [Var(Δr_i)+ε_γ]; ρ_i(t)=Δx_i(t)-γ_i·Δr_i(t).

[0047] Where Cov represents covariance, Var represents variance, and ε_γ is a positive number with the same dimensions as Var(Δr_i) and to prevent the denominator from being zero. The direction and range of the common variation coefficients correspond to the physical action direction of the corresponding state quantities. When the current record of the reference object is missing or the common variation coefficients cannot be stably determined, the current window does not deduct the reference deviation, but uses the target deviation as the correction deviation and marks it as insufficient reference.

[0048] Taking the temperature of conductive connection parts as an example, an increase in load or a rise in the ambient temperature inside the cabinet will cause both the target connection part and the reference connection part to heat up simultaneously. After subtracting their respective health conditions, the reference deviation is used to estimate the common temperature rise. The additional temperature rise of the target connection part due to changes in contact conditions is not present in the reference object and is therefore retained in the correction deviation.

[0049] Figure 3 This diagram illustrates the relationship between health reference selection and deviation generation. Topology and event conditions are used to exclude incomparable segments, while load and environmental conditions are used to select close health references. The target device and reference object each exhibit deviations relative to their own health status, and the common variation components are determined by the correspondence during healthy operation. When references are insufficient, the target deviation is used, and a degradation marker is generated.

[0050] Step S4: For the correction deviation that meets the warning initiation conditions, verify the actual changes in the associated measurement positions according to the fault physical response relationship. For each fault physical response relationship, save the initial abnormal measurement position, associated measurement position, expected change direction, allowable time delay, and necessary associated changes according to the monitored component.

[0051] For overheating at conductive connection points, the initial anomaly is a continuous increase in the temperature rise correction deviation at the target connection point. Associated changes include subsequent temperature rise at adjacent locations within the same phase and an increase in the temperature difference between the target phase and other phases. For anomalies in the switching mechanism, the initial anomaly is a shift in the coil response phase relative to the healthy action record. Associated changes include delays in the mechanism position, auxiliary contacts, and main circuit state relative to the event anchor point.

[0052] After an initial anomaly continuously meets the warning initiation conditions, the system searches for related changes only within the corresponding allowable time delay. Changes whose direction matches expectations, whose sequence of occurrence conforms to the physical process, and whose corresponding data acquisition record is valid are considered valid responses; changes exceeding the allowable time delay, whose direction is reversed, or whose data acquisition record is abnormal are not considered valid responses. Related changes that do not appear by the end of the allowable time delay are recorded as missing response records.

[0053] Figure 4This diagram illustrates the physical response relationship of a fault, including the initial abnormal measurement location, associated measurement locations, expected direction of change, sequence of occurrence, and allowable delay. It also shows the process of verifying the actual change and generating an actual response record or a missing response record after the correction deviation reaches the warning initiation condition. The acquisition link status is used to determine whether a record can serve as a valid response, without altering the predetermined physical response direction.

[0054] Step S5: Determine the abnormal object based on the actual changes, missing response records, and acquisition link status. If both the initial abnormality and the specified associated changes occur, and the relevant acquisition link records are valid, the abnormality is attributed to the primary device. If the initial abnormality occurs, the specified associated changes are missing, and the completeness, temporal continuity, or physical correspondence of the acquisition link corresponding to the initial abnormality is abnormal, the abnormality is attributed to the secondary measurement link.

[0055] When both the device status and the data acquisition link status are abnormal, the existing measurement configuration cannot obtain the associated changes specified in the physical response relationship of the fault, or the direction and timing of the initial abnormality and the associated changes cannot form a complete relationship, the result is marked as pending review. The pending review status retains the existing records and indicates the measurement locations or data acquisition links that need to be checked.

[0056] Figure 5 Three categories of branches are shown. When the actual change conforms to the specified direction and sequence and the data acquisition link is normal, it enters the primary equipment anomaly branch; when the specified associated change is missing and the initial anomaly corresponds to an anomaly in the data acquisition link, it enters the secondary measurement link anomaly branch; all other records enter the pending review branch. Each of the three branches generates evidence records including status records, occurrence time, missing response, and data acquisition link status.

[0057] Step S6: Determine the risk level based on the duration and trend of the correction deviation and the role of the monitored component in the current topology. Higher risk levels use continuous window entry conditions, with exit conditions lower than entry conditions, to reduce repeated changes in results near the boundary. Events requiring rapid processing, such as inconsistencies between the main circuit state of the switch and control requirements, use an event fast channel, but time correction and data acquisition link verification records are still maintained.

[0058] The early warning results include the warning object, fault category, risk level, evidence record, and handling recommendations. Handling recommendations for primary equipment anomalies point to the corresponding primary component and retest items; handling recommendations for secondary measurement link anomalies point to sensing components, data acquisition channels, power distribution terminal point tables, time synchronization, or communication connections; the pending verification status lists any related changes that have not yet been obtained and the locations that need to be verified.

[0059] The warning results are confirmed after on-site inspection, maintenance, or observation. Status segments confirmed to be normal and with valid data acquisition link records are added to the health reference candidate list; status segments confirming primary equipment anomalies are used to verify the corresponding physical response relationship of the fault; status segments confirming secondary measurement link anomalies are used to update the maintenance status of the corresponding data acquisition link. Health reference updates limit the magnitude of single changes and save the source of confirmation and records before and after the update.

[0060] This embodiment maintains the correspondence between equipment, measurement locations, and data acquisition links on a unified time base. It reduces the impact of shared load and environmental changes by using a shared operating condition health reference. Furthermore, it distinguishes between primary equipment anomalies and secondary measurement link anomalies by associating the actual direction of change and latency of the measurement locations. All of the above processing is traceable to the original records and does not rely on black-box scoring that does not specify input-output relationships.

[0061] Example 2

[0062] This embodiment combines Figure 3 The specific implementation of health reference matching and common change correction is explained. This embodiment receives the state sequence and acquisition link status formed in step S1, and outputs correction deviation and reference deficiency markers.

[0063] The health reference library stores device type, device identifier, topology status, operation event, load conditions, environmental conditions, segment start and end time, acquisition link status, and health confirmation status according to status segments. Action status segments also save the operation direction, pre-operation status, and post-operation status, ensuring that the opening and closing processes are matched separately.

[0064] The initial state fragments in the health reference library come from records confirmed to be normal during factory testing, handover testing, or initial commissioning. During operation, only state fragments that have been inspected, confirmed to be normal, and have passed the observation period are received. If the target equipment's own health records are insufficient, records from equipment of the same type and at the same station with the same topological role can be used, but a correspondence must be established between the dimensions, phases, and physical meanings of the target measurement location and the reference measurement location.

[0065] When the processing window arrives, the current operating condition is first determined based on the target state sequence. Power direction, bus connection relationship, switch position, and operation event type serve as necessary filtering conditions; load conditions and environmental conditions are used to determine the degree of proximity among candidate segments with consistent necessary conditions. If a candidate segment lacks necessary filtering conditions or the acquisition link record is abnormal, it will not be included in the health reference.

[0066] Candidate segments were sorted according to differences in load conditions, environmental conditions, acquisition link status, and formation time. Load conditions were represented by the equivalent load related to the heating or stress of the monitored component within the processing window, and environmental conditions were represented by the ambient temperature inside the cabinet or the equivalent environmental state formed by temperature, humidity, and dew point margin. Conditions of different dimensions were normalized according to their respective allowable comparison ranges before sorting.

[0067] The target health status is determined by the selected health segments of the target device; the reference health status and the reference current status come from the same reference object. The target device and the reference object each deviate from their own health status, avoiding inherent differences caused by installation location, sensor zero point, and heat dissipation conditions from entering into common change correction.

[0068] Common variation coefficients are determined using healthy segments of the acquisition link that are normal and not confirmed to be abnormal. Isolated offset records are excluded before calculation, and the direction of the coefficients is checked to ensure they conform to the corresponding physical relationship. If the estimated result is opposite to the direction of the physical effect or the variation between consecutive healthy windows exceeds the allowable range, the previous valid coefficient is used, and the reference validity is reduced.

[0069] The target deviation, reference deviation, common variation coefficient, and correction deviation are calculated according to the formulas described in Example 1. When load or environmental changes act simultaneously on the target equipment and the reference object, changes in the same direction enter the common variation component; local changes of the target equipment itself do not appear in the reference object and are therefore retained in the correction deviation.

[0070] When there are multiple available reference objects in the health reference library, they are sorted according to the similarity of operating conditions and the status of the acquisition link. The reference object that is ranked first and whose current acquisition link record is normal is selected to perform common change correction. The selected reference object and its screening conditions are saved in the evidence record. No competitive warning results are generated for different reference objects.

[0071] Reference validity is determined based on the number of healthy segments meeting the necessary criteria, the proximity of the load and environment, the completeness of the reference current state, and the stability of the common variation coefficients. When reference validity is insufficient, the corrected deviation is set equal to the target deviation, and the reason for insufficient reference is recorded in the evidence log; insufficient reference results do not directly enter the highest risk level.

[0072] Health reference library updates and online matching are performed separately. The absence of an alert in the current window does not equate to a confirmed health status; faulty segments, segments with abnormal acquisition links, segments with incomplete operating conditions, and segments with insufficient reference validity are not included in health status updates.

[0073] Through the above processing, the health reference sources, screening criteria, the respective health statuses of the target and reference, the common variation coefficients, and the reasons for insufficient reference are all recorded in the evidence record. This allows verification that a particular common variation originates from comparable health statuses, rather than from arbitrary historical fragments or unverified current records.

[0074] Example 3

[0075] This embodiment combines Figure 4 and Figure 5 This embodiment explains the relationship between the physical response to the fault, the verification of actual changes, and the attribution method between primary equipment anomalies and secondary measurement link anomalies. This embodiment receives the correction deviations generated in Embodiment 2 and retains the corresponding equipment, monitored component, measurement location, acquisition link, and occurrence time for each deviation.

[0076] The physical response relationships of faults are stored according to fault categories, and the initial abnormal measurement location, associated measurement locations, expected change directions of each measurement location, allowable time delays between adjacent changes, and necessary associated changes are recorded for each monitored component. These relationships are determined by the equipment's electrical connections, heat conduction paths, transmission sequence of actuating mechanisms, and factory or field test results.

[0077] The initial anomaly measurement location is used to determine the verification starting point. When the correction deviation continuously reaches the warning initiation condition along the expected direction, the time of the first occurrence of the condition is recorded, and the corresponding allowable delay is enabled. A single isolated spike is not considered an initial anomaly; records of abnormal acquisition link status are retained but cannot be used to initiate device anomaly attribution on a separate basis.

[0078] Correlation measurement locations are used to verify whether the expected physical results occur after the initial anomaly. Actual changes must simultaneously meet the following conditions: direction matches the expectation, occurrence time is within the allowable delay, continuous duration meets the corresponding requirements, and the data acquisition link records are valid. Changes in the same direction occurring after the allowable delay are saved as relevant records but are not considered valid responses to the current fault's physical response relationship.

[0079] Figure 4 The physical response relationship of the fault is shown using the initial abnormal measurement location and associated measurement locations. Associated measurement locations are used to express the directional and time delay constraints formed by the electrical connection, heat conduction, or motion transmission of the equipment, and do not limit different faults to having the same number of measurement locations; the associated changes that need to occur in the specific fault relationship are determined by the monitored component and the actual measurement configuration.

[0080] After the correction deviation reaches the warning initiation condition, according to Figure 4 The verification process shown searches for actual changes within the allowed time delay, checks the direction, sequence, and status of the data acquisition link, and outputs actual response records, missing response records, and reasons for pending verification. This process does not perform numerical interpolation for missing actual changes.

[0081] For heating at conductive connection points, the initial abnormal measurement location is the temperature measurement location of the target connection point. Associated measurement locations include adjacent temperature measurement locations within the same phase and temperature measurement locations at connections with other phases. After the initial temperature rise correction deviation continues to increase, the adjacent locations within the same phase heat up within the allowable time delay of heat conduction, and the temperature difference between the target phase and other phases widens, constituting an actual change consistent with a local heat source.

[0082] If only the target temperature measurement record shows a transitionless jump, and there are no corresponding changes in adjacent temperature measurement locations in the same phase, other phase connection parts, and synchronous loads, and at the same time there is an interruption in the sampling sequence number, time deviation, or abnormal correspondence of measurement points in the target temperature measurement channel, then a complete physical response of heating at the conductive connection part will not be formed, and the abnormality will be attributed to the target temperature measurement acquisition link.

[0083] For abnormalities in the switch operating mechanism, the control command serves as the event starting point, the coil response as the initial abnormality measurement position, and the mechanism position, auxiliary contacts, and main circuit status as associated measurement positions set according to the action transmission sequence. After the coil response deviates, the mechanism position, auxiliary contacts, and main circuit status are sequentially delayed within their respective allowable time delays, constituting the actual change in the mechanism abnormality on the primary equipment side.

[0084] If the coil response, mechanism position, and main circuit status are all within the healthy delay range, and only the auxiliary contact record uploaded to the power distribution terminal is delayed, and the remote signaling message has a time deviation or sampling sequence interruption, then the actual operation process of the primary equipment does not support mechanism abnormality, and the abnormality belongs to the acquisition link corresponding to the auxiliary contact.

[0085] Anomaly attribution is performed in the following order: First, check whether the actual changes specified in the physical response relationship of the fault are complete; then check the status of the acquisition links that formed each change. If the actual changes are complete and the acquisition link records are valid, output a primary device anomaly; if the specified associated changes are missing and the acquisition link corresponding to the initial anomaly is abnormal, output a secondary measurement link anomaly; if both types of states occur simultaneously or the existing records are insufficient, output "pending verification".

[0086] Figure 5 Two judgment conditions form three branches. The first judgment condition is whether the actual direction of change, the order of occurrence, and the allowable delay conform to the physical response relationship of the fault; the second judgment condition is whether the relevant data acquisition link records are normal. The branch results, together with the duration of the anomaly and the trend of change, are included in the risk classification and form evidence records.

[0087] The primary equipment anomaly evidence record includes the initial anomaly measurement location, associated measurement locations, directions of change, occurrence time, and acquisition link status; the secondary measurement link anomaly evidence record includes the initial anomaly record, missing associated changes as specified, and anomalies in record completeness, sampling sequence number, time synchronization, or physical correspondence; the record to be reviewed explains the current conflict and items that need to be supplemented for verification.

[0088] The maintenance of fault physical response relationships is carried out in a controlled manner. After on-site inspection confirms a primary equipment anomaly, it can be verified whether the expected direction and allowable delay cover the actual response. When a secondary measurement link anomaly is confirmed, only the maintenance status of the corresponding acquisition link is updated, and the fault physical response relationship of the primary equipment is not changed due to the anomaly segment. The updated relationship is reviewed and released for subsequent processing, and historical versions are retained.

[0089] This embodiment converts numerical exceedances into verifiable physical response judgments by clearly defining the measurement location, direction of change, allowable delay, and acquisition link status. Primary equipment anomalies require mutual support from actual state changes, while secondary measurement link anomalies require both missing response and acquisition link anomalies to occur simultaneously, thus avoiding the substitution of multi-layered custom scoring for the physical relationships between equipment.

[0090] Example 4

[0091] This embodiment combines Figure 6 The system used to implement the aforementioned fault early warning method is described below. This system includes a recording correlation and time correction module, a working condition reference correction module, a physical response verification module, an anomaly attribution determination module, and an early warning output module, as well as a feedback update module and a configuration storage unit.

[0092] The record association and time correction module receives device status records and acquisition link records. This module completes record association based on the correspondence table of device, interval, monitored component, measurement location, and acquisition link, and corrects time offset based on common anchor points in switch operation events. It outputs a status sequence and acquisition link status organized according to a steady-state statistical window or operation event window.

[0093] The operating condition reference correction module receives state sequences, topology states, operating events, load conditions, and environmental conditions, and reads the confirmed health reference from the configuration storage unit. This module first performs topology and event necessity screening, then selects health states with similar load and environmental conditions, calculates the deviations of the target device and the reference object relative to their respective health states, and outputs the correction deviations.

[0094] The physical response verification module reads the fault physical response relationship corresponding to the monitored component from the configuration storage unit. When the correction deviation reaches the warning start condition, the module establishes a verification record including the start time, the measurement location to be verified, and the allowable end time; after the new status record arrives, the actual changes and missing response records are updated according to the expected direction, allowable delay, and acquisition link status.

[0095] The anomaly attribution determination module receives actual changes, missing response records, and acquisition link status. When the actual change matches the physical response relationship of the fault and the relevant records are valid, it outputs a primary device anomaly; when the specified associated change is missing and the acquisition link corresponding to the initial anomaly is abnormal, it outputs a secondary measurement link anomaly; when the primary device status conflicts with the acquisition link status or there are insufficient records, it outputs a status pending review.

[0096] The early warning output module receives the anomaly attribution, correction deviation duration, change trend, and device topology effect. It determines the risk level according to the continuous window entry conditions and hysteresis exit conditions, and combines the health reference source, initial anomaly, actual change, response missing, occurrence time, data acquisition link status, and handling suggestions into an evidence record.

[0097] The feedback update module receives inspection results returned by the maintenance terminal or remote master station and associates the confirmed results with the original evidence records using an early warning record identifier. Status segments confirming normal status are entered into the health reference access review; status segments confirming primary equipment anomalies are used to verify the physical response relationship of the fault; and status segments confirming secondary measurement link anomalies are used to update the maintenance status of the corresponding acquisition link.

[0098] The configuration storage unit stores the device-to-measurement location correspondence table, data acquisition link relationships, health references, fault physical response relationships, processing windows, early warning initiation conditions, allowable latency, risk level conditions, evidence records, and feedback review records. Both health references and fault physical response relationships have version and release status settings; unreviewed content is not included in online early warnings.

[0099] In terms of physical deployment, the system includes primary switchgear, condition monitoring components, power distribution terminals, edge computing devices, and communication components. The condition monitoring components are configured to correspond to the monitored parts of the primary switchgear, generating records of temperature, switch actions, insulation environment, and other equipment status. The power distribution terminals receive the status records and switch operation events, and simultaneously generate acquisition link records related to sampling, time synchronization, point metering, and message transmission.

[0100] The edge computing device includes a processor, memory, and a local communication interface. The recording correlation and time correction module runs continuously as a receiving task, while the operating condition reference correction module operates according to a steady-state window or an operational event window. The physical response verification module saves the current verification status after the initial anomaly occurs and outputs the results after the allowed delay expires, indicating whether the actual change is complete, the response is missing, or the recording is insufficient.

[0101] The communication component sends the warning target and risk level to the local alarm interface, uploads the warning summary and evidence index to the remote master station, and allows the maintenance terminal to read relevant status records and data acquisition link records based on the evidence index. When communication is interrupted, the edge computing device continues to execute the warning and save the evidence records; after communication is restored, the data is retransmitted according to the formation time.

[0102] Figure 6 This diagram illustrates the deployment relationships between primary switchgear, condition monitoring components, power distribution terminals, edge computing modules, configuration storage units, and alarm and maintenance interfaces. Solid arrows indicate the transmission relationships between status records, processing results, configuration data, and early warning results, while dashed arrows indicate the controlled update relationships between inspection result feedback, health references, and physical fault responses.

[0103] Figure 6 The record association and time correction module, operating condition reference correction module, physical response verification module, anomaly attribution determination module, and early warning output module are connected in sequence; the configuration storage unit provides health reference to the operating condition reference correction module and provides fault physical response relationship to the physical response verification module; the feedback update module receives the maintenance confirmation result and performs controlled maintenance.

[0104] When the above modules are set up on the same edge computing device, data is transmitted through internal records with device identifiers, measurement location identifiers, acquisition link identifiers, and timestamps. When they are distributed and set up in power distribution terminals and station terminals, the sender retains the same identifier and transmits the correction deviation and acquisition link status, so that the attribution of anomalies can still be traced back to the original record.

[0105] This embodiment achieves joint verification of device status and data acquisition link through corresponding settings of module inputs, outputs, interfaces, and data streams. Each module does not merely list functions, but sequentially completes time synchronization recording, operating condition correction, physical response verification, anomaly identification, and evidence output.

[0106] Example 5

[0107] This embodiment combines Figure 6This paper uses a four-bay integrated primary and secondary ring main unit for a 10kV distribution line as an example to illustrate the on-site technical requirements, hardware configuration, parameter settings, fault early warning process, and exemplary processing results. The following parameters are used to explain the calculation and execution methods of the technical solution in this application and can be adjusted according to the equipment's rated parameters, monitoring component configuration, and on-site operating procedures. The state values ​​and calculation results in the examples are used to illustrate the relationships between intermediate quantities and are not intended as on-site statistical test data or to limit the protection scope.

[0108] The ring main unit in this embodiment includes one incoming line bay, two outgoing line bays, and one transformer feeder bay. The rated voltage of the primary equipment is 12kV, and the rated current is 630A. The two outgoing line bays are connected to the same busbar and have the same switchgear and cable connection structure. The first outgoing line bay serves as the target bay, and the second outgoing line bay serves as the reference bay for the same station. The target bay and the reference bay have the same power flow direction during normal operation and can be affected by both changes in busbar load and changes in the internal environment of the unit.

[0109] During on-site implementation, status records must be linked to specific equipment, intervals, phases, components, and acquisition channels. Distribution terminals must store sampling time, arrival time, quality indicators, and sampling sequence numbers. The time deviation of steady-state measurement channels must be controlled within 2 seconds, and the time resolution of switch operation event records must be no greater than 1 ms. Edge computing devices must continue local recording and early warning processing for at least 72 hours when remote communication is interrupted; after communication is restored, evidence records must be retransmitted according to the early warning record identifier and formation time. Version identifiers must be set for health references, fault physical response relationships, and operating parameters; parameter updates must not overwrite currently executing historical versions.

[0110] The hardware configuration of this embodiment is shown in the table below.

[0111]

[0112] The temperature measurement component at the connection points has one temperature measurement point at each of the three-phase cable connection points in the target bay (A, B, and C), and a response temperature measurement point is set at the insulation support location adjacent to the connection point of phase A in the target bay. The reference bay uses the same three-phase temperature measurement arrangement. Each temperature measurement point is assigned a measurement point identifier in the distribution terminal point table, and the three phases are not combined into a single temperature channel. The control commands, coil responses, auxiliary contacts, and main circuit status of the switch operating mechanism are each assigned an event identifier, and are associated as an operating event each time the switch is opened or closed.

[0113] The edge computing device employs a 5-minute steady-state statistical window with a 1-minute sliding step. Current records are used within the window to calculate the average load rate; connection point temperatures are used to calculate the window median and slope of change; and environmental records are used to calculate the average temperature and dew point margin. Switching operations utilize an event window, saving the coil response and state changes from 1 second before the control command to 5 seconds after. A continuous absence of a steady-state channel for more than two output cycles is recorded as a missing state. If a necessary anchor point is missing from an action event, no interpolation is performed; instead, the missing response state is retained.

[0114] This embodiment determines the validity of a measurement record based on the completeness of the record, the continuity of the sampling sequence number, the corrected time sequence, and the consistency between the record and the corresponding physical state. If there are no missing channels, interrupted sequence numbers, or time anomalies, and the measurement point, phase, and monitored component correspond correctly, the record is marked as valid. If any of the above anomalies occur, the status value and the acquisition link record are retained, but a separate equipment anomaly warning is not generated for that channel.

[0115] The health reference database is initialized using state segments that have been confirmed to be normal after equipment commissioning and on-site inspection. Subsequent data will only be received from segments that have been confirmed to be normal through maintenance or observation periods. This embodiment requires that the target segment and candidate segments have the same topological state, the same switch location, and the same power flow direction. They are then sorted according to differences in load rate, differences in cabinet ambient temperature, valid recorded status, and formation time.

[0116] Each time, a target health status is formed from the top-ranked and valid health segments, and a corresponding health segment from the same station reference interval is selected to form a reference health status. The current status of the reference object is formed by the in-phase temperature measurement record and load record of the current window of the second outgoing interval; if the number of candidates that meet the necessary conditions is insufficient, the current window is marked as insufficient reference.

[0117] The common variation coefficient γ_i is estimated using the most recent 30 valid health windows, requiring both the target channel and the reference channel involved in the estimation to have valid records. For the temperature channel at the connection point, the allowable range for γ_i is set to 0.5–1.5; if the estimated value exceeds this range, the previous valid value is used and the reference is marked as insufficient. In this example window, the γ_i for the temperature channel at the connection point of phase A is 0.80. If there are fewer than 3 healthy segments meeting the necessary conditions, two consecutive steady-state windows are missing from the reference current state, or γ_i is invalid for three consecutive estimation periods, the common variation component is not deducted, and the risk level is limited to warning or pending review.

[0118] This embodiment describes the physical response relationship for the overheating of conductive connection parts in the target interval configuration. The initial anomaly is a continuous increase in the temperature rise correction deviation at the connection part of target phase A. The warning initiation condition is that the correction deviation reaches 3°C and remains there for three consecutive windows. The first associated change is that the temperature rise correction deviation at the insulation support location adjacent to phase A reaches 1.5°C and remains there for two consecutive windows within 5–20 minutes after the initial anomaly occurs. The second associated change is that the temperature difference correction value between the connection parts of target phase A and phases B and C reaches 2°C and remains there for two consecutive windows within 0–10 minutes after the initial anomaly occurs. Both associated changes serve as the actual response required for attributing the overheating of conductive connection parts.

[0119] In this embodiment, when both associated changes occur and the relevant records are valid, the anomaly is attributed to the primary equipment; when both associated changes are missing and the acquisition link record corresponding to the initial anomaly is abnormal, the anomaly is attributed to the secondary measurement link; when only one associated change occurs, the primary equipment status and the acquisition link status are both abnormal, or the reference is insufficient, the result is marked as pending verification. A level-of-concern warning requires the initial anomaly to meet the warning initiation condition for three consecutive steady-state windows and the occurrence of the first associated change; a level-severe warning requires a correction deviation of 8°C and the occurrence of two associated changes. Events where the exit condition is lower than the corresponding entry condition, involving inconsistencies between the switch main circuit status and control requirements, utilize an independent fast channel.

[0120] The following example illustrates the specific execution process using the gradual heating of the cable connection point of phase A in the target bay as an example. Step S1 receives the three-phase current, target and reference temperature measurement records, cabinet environment records, and acquisition link records, establishing a correspondence between the target phase A connection point, the adjacent insulation support location, and the reference bay phase A connection point. Within the six example windows, there are no missing records or interrupted sampling sequences in any of the relevant channels. The corrected time sequence conforms to the chronological order of the operation events, and the correspondence between the measurement point identification, phase, and monitored component is correct. Therefore, the above records are marked as valid.

[0121] Step S2 identifies that both the target interval and the reference interval are on the same bus, in the closed position, and in a positive power flow state. The target load rate for the six windows is 72%–79%, and the ambient temperature is 31.5℃–34.5℃. Segments with similar load rates and ambient temperatures are selected from the healthy reference library to form a healthy state. Step S3 calculates the correction deviation of the target phase A connection location according to γ_i = 0.80. An exemplary intermediate result is shown in the table below.

[0122]

[0123] As shown in the table, both the target actual temperature and the reference actual temperature increase with load and environment, and the reference deviation is used to represent the common change. The correction deviation of target phase A increases from 1.80℃ to 7.84℃, reaching the warning initiation condition at W3 and remaining thereafter until W6. The conventional high-temperature alarm threshold set in this scenario is 70℃. The target actual temperature at W6 is 64.3℃, which has not yet triggered the conventional high-temperature alarm, but the correction deviation already reflects the additional temperature rise of the target connection part relative to the same operating condition and reference interval.

[0124] Step S4 records that the connection point of target phase A reaches the warning initiation condition at W3. The temperature rise correction deviation of the adjacent insulation support position is 1.2℃ at W4, 1.8℃ at W5, and 2.4℃ at W6, satisfying the first correlation change condition from W5 onwards; the temperature difference correction value between target phase A and phases B and C is 2.3℃ at W4, 3.1℃ at W5, and 3.8℃ at W6, satisfying the second correlation change condition from W4 onwards. The directions of the two correlation changes are consistent with the heat conduction relationship after the formation of the local heat source, and the occurrence time is within the specified allowable time delay. There are no missing records, interrupted serial numbers, or time anomalies in the relevant temperature measurement channels, thus forming a complete actual change record.

[0125] Step S5 verifies that the initial anomaly, the first associated change, and the second associated change have all occurred, and the relevant acquisition link records are valid. Therefore, the warning target is determined to be the primary equipment-side anomaly at the A-phase cable connection point of the target bay, rather than the environmental temperature rise or the temperature measurement acquisition link anomaly.

[0126] Step S6 detects that the initial anomaly continuously meets the entry conditions for the concern level, and both associated changes have occurred, outputting a concern level warning. Evidence records include the target and reference health status, γ_i, correction deviation, two actual changes, the occurrence time of each change, and the status of the relevant acquisition link in six windows. The recommended action is to perform infrared retesting on the connection point of target phase A cable, check the connection tightness, and measure the loop resistance when operational conditions permit. If the subsequent correction deviation reaches 8°C and the two associated changes continue, the warning enters the severe level; if the correction deviation continuously falls below the exit conditions after maintenance, the warning exits.

[0127] To illustrate the handling of secondary measurement link anomalies, this embodiment also provides an example of a temperature measurement channel anomaly. The temperature at the target phase B connection point jumps from 51℃ to 82℃ between two adjacent 10-second records, while the load rate and ambient temperature remain essentially unchanged. Adjacent response temperature measurement locations and other thermal status records for phase B show no temperature increase. The distribution terminal records indicate that this temperature measurement channel has an interrupted sampling sequence and an abnormal measurement point correspondence; the corrected time sequence also does not meet the valid recording conditions.

[0128] Although the temperature of phase B reached the high-temperature alarm condition, no adjacent thermal response or continuous change in inter-phase temperature difference occurred within the allowable time delay, and the acquisition link record corresponding to the initial anomaly was abnormal. Therefore, no overheating warning for the primary equipment of phase B was output, but an abnormality warning for the temperature measurement acquisition link of phase B was output. Evidence records indicate isolated temperature jumps, missing specified correlation changes, interrupted sampling sequence numbers, and abnormal correspondence between measurement points. The recommended course of action is to check the temperature measurement components, acquisition terminals, channel mapping, and time status of the power distribution terminal.

[0129] The processing results of the two examples above are shown in the table below.

[0130]

[0131] This table reflects the judgment relationships obtained by processing the example states according to the parameters of this embodiment. For phase A gradual heating, the common temperature rise of the target and reference is subtracted, and the retained local additional temperature rise, together with the two associated changes, supports the primary equipment anomaly; for phase B isolated jump, although the single value exceeds the high temperature alarm condition, the specified associated change is missing and the acquisition link record is abnormal, so the warning is attributed to the secondary measurement link. The values ​​in the two examples are used to illustrate the recalculated processing procedure and do not represent the field statistical performance or fixed warning effect.

[0132] After on-site inspection, the maintenance terminal returns the confirmation result with the original warning record identifier. When an abnormal contact condition is confirmed at the connection point of phase A, the corresponding status segment is written into the verification record of the heating relationship of the conductive connection point, but not into the health reference library; after maintenance is restored, segments that are observed continuously for 7 days and whose relevant status is normal can be included in the health reference candidate review. When the abnormality of phase B is confirmed to be caused by the temperature measuring component or channel mapping, the maintenance status of the acquisition link is updated, and the temperature segment during the abnormal period does not participate in the health status and common change coefficient update. All updates retain the operation time, confirmation source, version before update, and version after update.

[0133] This embodiment utilizes monitoring components, distribution terminals, edge computing devices, and communication interfaces suitable for 10kV integrated primary and secondary ring network boxes. Figure 6 The deployment is shown and executed in the following order: record association and time correction, operating condition reference correction, fault physical response verification, anomaly attribution, and early warning output. The example parameters enable those skilled in the art to configure the processing window, health reference, common variation coefficient, associated measurement location, allowable delay, and risk level conditions, and to adjust them according to the equipment's rated parameters and field operating procedures.

[0134] The technical solution of the present invention has been described above in conjunction with specific embodiments. The above-mentioned equipment configuration, processing window, health reference conditions, common variation coefficient range, early warning start conditions, and allowable delay are used to illustrate the implementation process of the present invention; under the conditions of maintaining the record correspondence, common variation correction under the same working conditions, fault physical response verification, and anomaly attribution relationship, corresponding adjustments can be made according to the ring network box equipment type, measurement configuration, and operating procedures.

[0135] Based on the disclosure of this invention, those skilled in the art can make adaptive adjustments to specific parameters according to the equipment measurement configuration and operating procedures. However, the corresponding adjustments should maintain the correspondence between the equipment status record and the acquisition link record, the correction for changes in the same working conditions, the verification of physical response to faults, and the technical relationship of the attribution of abnormalities between the primary equipment and the secondary measurement link.

Claims

1. A method for early warning of faults in a ring network box based on multi-source data fusion, characterized in that, include: S1. Obtain the device status records and corresponding acquisition link records of the primary and secondary fusion ring network box, establish the correspondence between the device, measurement location and acquisition link, and use the switch operation event to correct the time base of the record to form a status sequence organized by event or statistical window. S2, determine the current operating condition based on the topology state and operation events of the state sequence, and select a health reference from the health reference library that has the same topology and event conditions and similar load and environmental conditions; S3, determine the deviations of the target equipment and the reference object relative to their respective health states, and based on the correspondence between the two deviations during healthy operation, subtract the common change component from the deviation of the target equipment to obtain the correction deviation characterizing the local change. S4, when the correction deviation reaches the warning start condition, according to the predetermined fault physical response relationship, verify the actual change direction and occurrence sequence of the associated measurement position within the corresponding allowable time delay; S5, when the actual change conforms to the fault physical response relationship and the relevant acquisition link records are normal, the anomaly is attributed to the primary device; When an initial anomaly occurs, the associated changes specified in the physical response relationship of the fault are missing, and the record integrity, temporal continuity, or physical correspondence of the corresponding acquisition link is abnormal, the anomaly will be attributed to the secondary measurement link; other cases will be marked as pending review. S6 determines the risk level based on the duration of the anomaly, the trend of change, and the role of the monitored component in the current equipment topology, and outputs the warning object, fault category, risk level, status record and data acquisition link record for the attribution conclusion, as well as handling suggestions.

2. The method for early warning of faults in a ring network box based on multi-source data fusion according to claim 1, characterized in that, In step S1, each device status record is associated with the device identifier, interval identifier, monitored component, measurement location, and acquisition link. Steady-state records are organized according to a unified time grid. Action response records use control commands, coil responses, mechanism positions, auxiliary contact displacements, and main circuit state changes as time anchors for the same operation event. The validity of a record is determined based on record missingness, sampling sequence continuity, corrected event sequence, and the correspondence between the record and the physical state. Invalid records are marked as acquisition link abnormalities and are not used as a separate basis for determining a device abnormality.

3. The method for early warning of primary and secondary fusion ring network box faults based on multi-source data fusion according to claim 1, characterized in that, In step S2, the topology connection relationship and operation event type are used as necessary matching conditions for health reference. From the state segments that have been confirmed to be healthy and whose link records are normal, the health status of the target device, the health status of the reference object, and the status of the reference object in the current window are determined according to the differences in load conditions, environmental conditions, and formation time. When there are not enough healthy references that meet the conditions, the common change component deduction is not performed, and the risk level of the output is limited.

4. The method for early warning of primary and secondary fusion ring network box faults based on multi-source data fusion according to claim 1, characterized in that, In step S3, the target deviation of the current state of the target device relative to its health state under the same operating conditions and the reference deviation of the current state of the reference object relative to its health state under the same operating conditions are calculated respectively; the common variation coefficient is determined by the correspondence between the target deviation and the reference deviation in the health operation segment, the common variation component is represented by the product of the common variation coefficient and the reference deviation, and the common variation component is subtracted from the target deviation. When the common variation coefficient cannot be stably determined by the current health reference, the previous effective coefficient is used and the current result is marked as insufficient reference.

5. The method for early warning of primary and secondary fusion ring network box faults based on multi-source data fusion according to claim 1, characterized in that, In step S4, the fault physical response relationship record includes the initial abnormal measurement location of the monitored component, the associated measurement locations that have an electrical connection, thermal conduction, or motion transmission relationship with the initial abnormal measurement location, the expected change direction of each measurement location, and the allowable time delay between adjacent changes. After the initial abnormality continuously meets the warning initiation conditions, only the associated changes that occur within the allowable time delay, whose change direction meets expectations, and whose acquisition link records are normal are confirmed as actual responses. The associated changes that do not occur are retained as response missing records.

6. The method for early warning of primary and secondary fusion ring network box faults based on multi-source data fusion according to claim 5, characterized in that, In step S5, when the specified associated changes of the initial anomaly and the physical response relationship of the fault both occur and the relevant records are valid, the primary equipment anomaly attribution is output; when the initial anomaly occurs, the specified associated changes are missing, and the acquisition link corresponding to the initial anomaly has missing records, interrupted sampling sequence numbers, abnormal time sequence, or inconsistent records with physical status, the secondary measurement link anomaly attribution is output; when the initial anomaly and the associated changes cannot form a complete physical response relationship, or when the primary equipment status and the acquisition link status are both abnormal, the status pending review is output.

7. The method for early warning of faults in a ring network box based on multi-source data fusion according to claim 6, characterized in that, In step S6, the entry conditions for the risk level are higher than the corresponding exit conditions, and the risk level is increased after the entry conditions are met in the continuous processing window; the warning results record the current operating conditions, health reference sources, correction deviations, existing and missing related changes, the time of occurrence of each change, the status of the data acquisition link, and the attribution of the anomaly. Status segments that are confirmed to be normal, have valid data collection links, and show no subsequent abnormalities during the observation period are included in the health reference candidate. When updating the health status, the update range is limited and the records before and after the update are saved.

8. A primary and secondary fusion ring network box fault early warning system based on multi-source data fusion, characterized in that, It includes a record association and time correction module, a working condition reference correction module, a physical response verification module, an anomaly attribution determination module, and an early warning output module, which are connected in sequence. The recording association and time correction module receives device status records and acquisition link records, establishes the correspondence between the device, measurement location and acquisition link, corrects the time reference according to the switch operation event, and outputs the status sequence and acquisition link status. The operating condition reference correction module selects a healthy reference based on the current operating condition, deducts the common variation components of the target equipment and the reference object, and outputs the correction deviation. The physical response verification module generates records of actual changes and missing responses within the allowable time delay based on the correction deviation and the pre-stored fault physical response relationship. The anomaly attribution determination module combines actual changes, missing response records, and acquisition link status to output primary equipment anomaly, secondary measurement link anomaly, or a status pending verification. The early warning output module determines the risk level and generates evidence records based on the duration of the anomaly, the trend of change, and the effect of the equipment topology.

9. The primary and secondary fusion ring network box fault early warning system based on multi-source data fusion according to claim 8, characterized in that, It also includes a feedback update module and a configuration storage unit; the configuration storage unit stores the correspondence between devices and measurement locations, acquisition link relationships, health references, fault physical response relationships, early warning parameters, and evidence records; the feedback update module associates the inspection results with the evidence records, confirms that normal status segments are used to maintain health references after access review, confirms that status segments of primary equipment abnormalities are used to verify fault physical response relationships, and confirms that status segments of secondary measurement links abnormalities are used to update the maintenance status of the corresponding acquisition links.

10. The primary and secondary fusion ring network box fault early warning system based on multi-source data fusion according to claim 9, characterized in that, The fault early warning system is set in the edge computing device of the primary and secondary integrated ring network box; the primary and secondary integrated ring network box also includes primary switchgear, status monitoring component, power distribution terminal and communication component. The status monitoring component is set to correspond to the monitored part of the primary switchgear. The power distribution terminal receives equipment status records and switch operation events and forms a data acquisition link record. The communication component sends the early warning results to the local alarm interface, remote master station or maintenance terminal and receives the inspection results.

Images