A multi-modal partial discharge on-line monitoring method for insulation state

Abstract

The application relates to a multi-modal partial discharge online monitoring method for an insulation state. The method comprises the following steps: collecting, in a cross-modal joint triggering mode, a partial discharge response signal of a GIS insulation state monitoring part to obtain insulation original monitoring data; extracting suspected partial discharge pulse data in the insulation original monitoring data to obtain an insulation event data set; fusing and analyzing electromagnetic propagation response, sound propagation response, mechanical vibration response and cross-modal coupling relationship of a partial discharge event according to the insulation event data set to obtain insulation correlation characteristic data; performing insulation action scope mapping analysis on the partial discharge event according to the insulation correlation characteristic data to obtain insulation action scope data; and analyzing an insulation state evolution process according to the insulation correlation characteristic data and the insulation action scope data to obtain insulation state representation quantities and insulation risk potential values. The method can improve the identification accuracy of a partial discharge event.

Description

Technical Field

[0001] This application relates to the field of smart grid technology, and in particular to an online monitoring method for multimodal partial discharge oriented towards insulation conditions. Background Technology

[0002] Traditional partial discharge monitoring methods mainly include high-frequency current methods, ultra-high frequency methods, ultrasonic methods, transient ground voltage methods, and monitoring methods based on optics or multi-sensor fusion. These methods typically involve deploying sensors on power equipment such as switchgear, transformers, cables, and GIS systems to collect electrical pulses, electromagnetic waves, sound waves, or light signals generated by partial discharges in real time. This allows for online detection, location, and early warning of insulation defects, thereby improving the safety and reliability of equipment operation. However, the monitoring results of traditional technologies are easily affected by strong electromagnetic interference, noise coupling, sensor placement conditions, and differences in equipment structure, leading to insufficient accuracy in identifying partial discharge events. Summary of the Invention

[0003] Therefore, it is necessary to provide a multimodal online monitoring method and device for partial discharge based on insulation state, which can improve the accuracy of partial discharge event identification, in order to address the above-mentioned technical problems.

[0004] Firstly, this application provides a method for online monitoring of multimodal partial discharge in insulation conditions, including: Cross-modal joint triggering acquisition was performed on the partial discharge response signals of the GIS insulation condition monitoring points to obtain the original insulation monitoring data; The suspected partial discharge pulse data in the original insulation monitoring data are extracted to obtain an insulation event data set; Based on the insulation event data set, the electromagnetic propagation response, acoustic propagation response, mechanical vibration response, and cross-modal coupling relationship of the partial discharge events at the GIS insulation status monitoring location are fused and analyzed to obtain insulation correlation feature data; Based on the insulation correlation feature data, insulation domain mapping analysis is performed on the partial discharge events of the GIS insulation status monitoring location to obtain insulation domain data; Based on the insulation correlation feature data and the insulation domain data, the insulation state evolution process of the GIS insulation state monitoring location is analyzed to obtain insulation state characterization quantities and insulation risk potential energy values.

[0005] Secondly, this application also provides a multi-mode partial discharge online monitoring device for insulation conditions, comprising: The data acquisition module is used to perform cross-modal joint triggering acquisition of partial discharge response signals from GIS insulation condition monitoring points to obtain raw insulation monitoring data; The time extraction module is used to extract events from the suspected partial discharge pulse data in the original insulation monitoring data to obtain an insulation event data set; The fusion analysis module is used to perform fusion analysis on the electromagnetic propagation response, acoustic propagation response, mechanical vibration response, and cross-modal coupling relationship of the partial discharge events at the GIS insulation status monitoring location based on the insulation event data set, so as to obtain insulation correlation feature data. The mapping analysis module is used to perform insulation domain mapping analysis on the partial discharge events of the GIS insulation status monitoring location based on the insulation correlation feature data, and obtain insulation domain data. The evolution analysis module is used to analyze the insulation state evolution process of the GIS insulation state monitoring location based on the insulation correlation feature data and the insulation action domain data, and to obtain insulation state characterization quantities and insulation risk potential energy values.

[0006] The aforementioned multimodal online monitoring method and device for insulation conditions, by performing cross-modal joint triggering acquisition of partial discharge response signals from GIS insulation condition monitoring locations, and constructing an insulation event dataset around suspected partial discharge pulse data, and then performing fusion analysis based on the electromagnetic propagation response, acoustic propagation response, mechanical vibration response, and cross-modal coupling relationship of partial discharge events, not only avoids the problems of incomplete partial discharge information representation, insufficient anti-interference capability, and high false alarm rate of traditional single-modal monitoring, but also can more accurately extract the comprehensive response characteristics of partial discharge activity at the event level. Furthermore, by introducing insulation domain mapping analysis, it is no longer limited to detecting or simply locating the partial discharge signal itself, but can identify partial discharge events. This study identifies the area of ​​effect, boundary of influence, and extent of defect expansion of components in GIS insulation systems. This elevates the monitoring results from simply whether a discharge occurs to identifying the specific insulation domain affected by the discharge and how it impacts the insulation boundary. Furthermore, by combining insulation correlation characteristic data and insulation domain data to analyze the insulation state evolution process, a quantitative characterization of the GIS insulation health level, degradation stage, and risk accumulation level can be achieved. This yields insulation state characterization quantities and insulation risk potential values. This improves the accuracy of partial discharge event identification, provides more comprehensive monitoring dimensions, clearer characterization of defect impact range, more reliable judgment of insulation degradation trends, and more forward-looking risk assessment results in online partial discharge monitoring. Ultimately, this provides more effective technical support for condition-based maintenance, early warning classification, and operation and maintenance decisions for GIS equipment. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is an application environment diagram of a multimodal partial discharge online monitoring method for insulation states in one embodiment; Figure 2 This is a flowchart illustrating an online monitoring method for multimodal partial discharge in an insulation state, as shown in one embodiment. Figure 3 This is a structural block diagram of a multimodal partial discharge online monitoring device for insulation states in one embodiment; Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0009] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0010] This application provides an online monitoring method for multimodal partial discharge oriented towards insulation states, which can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on other network servers. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0011] In one exemplary embodiment, such as Figure 2 As shown, a method for online monitoring of multimodal partial discharge in insulation states is provided, which is applied to... Figure 1 Taking the server in the example, the explanation includes the following steps 202 to 210. Wherein:

[0012] Step 202: Perform cross-modal joint triggering acquisition on the partial discharge response signal of the GIS insulation condition monitoring point to obtain the original insulation monitoring data.

[0013] Step 204: Extract events from the suspected partial discharge pulse data in the original insulation monitoring data to obtain an insulation event data set.

[0014] Step 206: Based on the insulation event data set, perform a fusion analysis on the electromagnetic propagation response, acoustic propagation response, mechanical vibration response, and cross-modal coupling relationship of partial discharge events at the GIS insulation condition monitoring location to obtain insulation correlation characteristic data.

[0015] Step 208: Based on the insulation correlation feature data, perform insulation domain mapping analysis on the partial discharge events of the GIS insulation status monitoring location to obtain insulation domain data.

[0016] Step 210: Based on the insulation correlation characteristic data and insulation action domain data, analyze the insulation state evolution process of the GIS insulation state monitoring location to obtain insulation state characterization quantity and insulation risk potential energy value.

[0017] Among them, the GIS insulation status monitoring part refers to the target insulation area or structural part in the GIS equipment that is used to reflect the insulation health status and is suitable for deploying multimodal sensing units to monitor partial discharge activity.

[0018] Among them, the partial discharge response signal refers to the abnormal response signal that can be detected by sensors when a partial discharge occurs in a GIS insulation system in different physical dimensions such as electromagnetic, acoustic, and mechanical vibration.

[0019] Among them, cross-modal joint trigger acquisition refers to a collaborative acquisition method that uses the abnormal response detected by any modality as the trigger point and simultaneously retrieves data from other modalities within the corresponding time window.

[0020] Among them, the original insulation monitoring data refers to the multimodal basic monitoring data obtained through cross-modal joint triggering acquisition that has not yet undergone event-level extraction.

[0021] Among them, suspected partial discharge pulse data refers to data segments in the original insulation monitoring data that have typical transient characteristics of partial discharge and are judged to be likely to correspond to partial discharge activity.

[0022] Event extraction refers to the process of identifying and merging data segments corresponding to a single partial discharge event from the original insulation monitoring data.

[0023] Among them, the insulation event data set refers to a set consisting of multiple event data units corresponding to a single partial discharge event.

[0024] Among them, a partial discharge event refers to a single, independent partial discharge physical process occurring in an insulation system and its corresponding response in each monitoring mode.

[0025] Among them, electromagnetic propagation response refers to the ultra-high frequency or other electromagnetic characteristic response formed when the electromagnetic energy generated by partial discharge propagates in the GIS structure.

[0026] Among them, acoustic propagation response refers to the ultrasonic or acoustic abnormal response formed when the sound waves generated by partial discharge propagate in an insulating medium or structure.

[0027] Among them, mechanical vibration response refers to the mechanical response signal generated by local impact, micro-disturbance or vibration change induced by partial discharge in GIS structure.

[0028] Among them, cross-modal coupling relationship refers to the correlation in time, intensity and propagation behavior between different modal responses of the same partial discharge event, such as electromagnetic, acoustic and mechanical vibration.

[0029] Fusion analysis refers to the analytical process of jointly processing and comprehensively judging the propagation response characteristics and coupling relationships of different modes.

[0030] Among them, insulation correlation feature data refers to feature data obtained through fusion analysis, which is used to comprehensively characterize the propagation characteristics and cross-modal coupling characteristics of partial discharge events.

[0031] Among them, insulation domain mapping analysis refers to the analysis process of determining the region of action, influence boundary and coupling range of partial discharge events in the insulation system based on insulation correlation characteristic data.

[0032] Among them, insulation domain data refers to data that characterizes the weak insulation region, coupling propagation boundary, and defect influence range corresponding to partial discharge events.

[0033] Among them, the insulation state evolution process refers to the development and change process of the GIS insulation system from initial anomaly to local deterioration and then to risk accumulation.

[0034] Among them, insulation status characterization quantity refers to the parameter used to quantitatively represent the health level, degradation stage or abnormal level of GIS insulation status.

[0035] Among them, the insulation risk potential value refers to the parameter value used to quantitatively represent the cumulative degree of GIS insulation failure risk and the potential development intensity.

[0036] Specifically, multiple sensing units are deployed at the GIS insulation condition monitoring site to collect data for sensing different physical responses. These sensing units include at least a UHF sensor for collecting the electromagnetic propagation response of partial discharge, an ultrasonic sensor for collecting the acoustic propagation response of partial discharge, and a vibration sensor for collecting the mechanical disturbances caused by partial discharge. Each sensing unit continuously monitors the same insulation condition monitoring site and outputs a continuous sampled data stream corresponding to its mode. Instead of a single-channel independent triggering method, a cross-modal joint triggering mechanism is used during data acquisition. When a transient response satisfying a preset anomaly criterion appears in any modal signal, that anomaly response is used as the joint triggering starting point. Simultaneously, synchronous time window data from other modes before and after the occurrence of that anomaly response are retrieved, allowing the data from the electromagnetic, acoustic, and mechanical vibration modes to converge around the same partial discharge physical process. Then, a unified acquisition timestamp, monitoring site identifier, channel identifier, sampling frequency, and trigger sequence number are added to each modal data to form a multimodal raw data package with the same source, window, and event orientation. Ultimately, raw insulation monitoring data reflecting the overall partial discharge response of the GIS insulation condition monitoring site is obtained.

[0037] Pulse detection and anomaly screening are performed on the raw data of each mode in the original insulation monitoring data. In electromagnetic data, anomalous pulses with steep rises and falls and short-term high-energy bursts are identified. In ultrasonic data, anomalous responses with local abrupt changes, envelope peak transitions, or instantaneous acoustic energy increases are identified. In vibration data, signal segments with transient micro-amplitude anomalies, short-term impact characteristics, or abnormal growth of vibration envelopes are identified. Then, using the detected suspected partial discharge pulse data as the core index, event time windows are set before and after their corresponding moments. Electromagnetic, ultrasonic, and mechanical vibration data segments corresponding to the pulse within these time windows are correlated and aggregated. For multiple response segments belonging to the same partial discharge physical process within the same time window, they are merged according to trigger order, mode affiliation, and time overlap to form a single partial discharge event data unit. This process is repeated for multiple partial discharge events, ultimately forming an insulation event data set composed of multiple event data units.

[0038] For each partial discharge event data unit in the insulation event dataset, propagation response features under different modes were extracted. Specifically, pulse peak value, leading-edge steepness, pulse width, energy distribution, bandwidth, arrival time difference, and waveform attenuation features were extracted from electromagnetic data; sound pressure level, envelope shape, propagation delay, attenuation trajectory, wave group distribution, and local resonance features were extracted from ultrasonic data; and vibration peak value, impact duration, vibration envelope variation, spectral perturbation degree, and propagation perturbation features were extracted from mechanical vibration data. After completing the single-mode feature extraction, the consistency and coupling relationship between multi-mode responses were further analyzed. This included analyzing the temporal correspondence between the electromagnetic pulse occurrence time and the ultrasonic burst and vibration perturbation, the correlation between different modal response intensities, the consistency between propagation delay and attenuation patterns, and the coordinated change trend of multi-mode responses in the same event. Based on this, the single-mode propagation response features and cross-modal coupling features were uniformly organized and fused into codes to obtain insulation correlation feature data that characterizes both the intensity of the partial discharge event itself and its propagation path characteristics and multi-physical coupling behavior.

[0039] An insulation domain description system is established for GIS-monitored insulation condition locations. This system includes at least the domain elements of weak insulation areas, coupling propagation boundaries, and defect influence ranges. Then, based on insulation correlation feature data, the combination patterns of partial discharge events in electromagnetic propagation response, acoustic propagation response, mechanical vibration response, and cross-modal coupling relationships are identified. These patterns are then matched with preset insulation domain templates, insulation domain fingerprint databases, or candidate insulation domain sequences. The matching process does not simply output the geometric location of the partial discharge; instead, it determines which type of insulation domain the partial discharge event falls into, which type of weak insulation area it corresponds to, along which type of coupling propagation boundary it extends, and which insulation structural units its influence range covers. For cases where multiple possible domains have overlapping responses, the domain attribution is further identified and constrained by combining the boundary attribution relationships, propagation coupling relationships, and influence range overlap relationships between different domains. After the above mapping analysis, insulation domain data that characterizes the domain area, boundary range, and influence level of a partial discharge event within the GIS insulation condition monitoring location is output.

[0040] Insulation-related characteristic data is used as input information to characterize the current activity intensity, propagation activity, coupling complexity, and anomaly significance of partial discharge events. Insulation domain data is used as input information to characterize the attribution of weak insulation areas, defect influence boundaries, and the extent of influence range. The two are jointly characterized. Then, based on the results of multiple partial discharge events at the same GIS insulation condition monitoring location within a continuous monitoring period, the changing trends of insulation-related characteristic data over time and the migration, expansion, contraction, or aggregation trends of insulation domain data over time are analyzed. This identifies the evolutionary process of insulation status from initial anomaly, local degradation, boundary expansion, and risk enhancement. Simultaneously, the analysis results reflecting the current insulation health level, degradation degree, defect activity level, and insulation boundary disturbance degree are quantified into insulation status characterization quantities. Furthermore, the cumulative degree of risk in different domains, stages, and evolution paths during the insulation status evolution process is quantified, constructing a corresponding risk potential energy expression and outputting the insulation risk potential energy value.

[0041] In the aforementioned multimodal online monitoring method for insulation conditions, cross-modal joint triggering acquisition of partial discharge response signals from GIS insulation condition monitoring locations is performed. An insulation event dataset is constructed around suspected partial discharge pulse data. Then, based on the electromagnetic propagation response, acoustic propagation response, mechanical vibration response, and cross-modal coupling relationship of partial discharge events, a fusion analysis is conducted. This not only avoids the problems of incomplete partial discharge information representation, insufficient anti-interference capability, and high false alarm rate associated with traditional single-modal monitoring, but also more accurately extracts the comprehensive response characteristics of partial discharge activity at the event level. Furthermore, by introducing insulation domain mapping analysis, it is no longer limited to detecting or simply locating the partial discharge signal itself, but can identify partial discharge events. In GIS insulation systems, the effective area, influence boundary, and defect expansion range are defined, thus elevating the monitoring result from whether a discharge occurs to which insulation domain the discharge acts on and how it affects the insulation boundary. Based on this, by combining insulation correlation characteristic data and insulation effective domain data to analyze the insulation state evolution process, it is possible to quantitatively characterize the health level, degradation stage, and risk accumulation level of GIS insulation, thereby obtaining insulation state characterization quantities and insulation risk potential energy values. This can improve the accuracy of partial discharge event identification, provide more comprehensive monitoring dimensions, clearer characterization of defect influence range, more reliable judgment of insulation degradation trend, and more forward-looking risk assessment results in online partial discharge monitoring, providing more effective technical support for condition-based maintenance, early warning classification, and operation and maintenance decisions for GIS equipment.

[0042] In an exemplary embodiment, insulation domain mapping analysis is performed on partial discharge events at GIS insulation condition monitoring locations based on insulation correlation feature data to obtain insulation domain data, including steps 302 to 306. Wherein:

[0043] Step 302: Perform cross-modal response decoupling on the insulation-related feature data to obtain insulation domain fingerprint data.

[0044] Step 304: Based on the insulation domain fingerprint data, perform reverse screening analysis on the preset insulation domain set of the GIS insulation status monitoring location to obtain the candidate insulation domain sequence.

[0045] Step 306: Based on the candidate insulation domain sequence, the insulation coupling boundary of the partial discharge event is folded and mapped to obtain insulation domain data.

[0046] Among them, cross-modal response decoupling refers to the process of separating the mixed response characteristics of partial discharge events in different modes such as electromagnetic, acoustic and mechanical vibration.

[0047] Among them, insulation domain fingerprint data refers to the feature data formed after cross-modal response decoupling, which is used to characterize the domain attribution of partial discharge events.

[0048] Among them, the preset insulation action area set refers to the set of multiple insulation action areas and their corresponding descriptive information that are pre-divided or established for the GIS insulation status monitoring parts.

[0049] Among them, reverse screening analysis refers to the analysis process of gradually eliminating irrelevant domains from the preset set of insulation domains according to the degree of mismatch with the fingerprint data of the insulation domains.

[0050] Among them, the candidate insulation domain sequence refers to the insulation domain sequence that is retained after reverse screening and reduction analysis and sorted according to the probability of domain affiliation.

[0051] Among them, the insulation coupling boundary refers to the boundary region corresponding to the propagation, expansion or mutual influence of partial discharge events between different insulation domains.

[0052] Specifically, the comprehensive characterization information of partial discharge events in electromagnetic propagation response, acoustic propagation response, mechanical vibration response, and cross-modal coupling relationships is identified from the insulation-related feature data. Then, the insulation-related feature data is hierarchically decomposed according to modal origin and coupling level. Feature items reflecting electromagnetic propagation behavior, acoustic propagation behavior, mechanical vibration propagation behavior, and coupling feature items reflecting time synchronization, energy correspondence, and propagation consistency between different modes are separated. Furthermore, the aliasing effects caused by differences in dimensions, signal strength, and propagation paths between different modes are eliminated. After decoupling, the independent response features of each mode and their residual coupling relationships after decoupling are re-encoded to form a combined feature identifier that can characterize the differences in the performance of the partial discharge event under different insulation domains, thus obtaining insulation domain fingerprint data.

[0053] An insulation domain set is pre-constructed for each GIS insulation condition monitoring location. This pre-constructed set includes different weak insulation areas, boundary coupling areas, defect-affected areas, and corresponding domain templates. Insulation domain fingerprint data is then input into a reverse screening and reduction analysis module. Instead of the traditional forward one-to-one matching method, it first removes domains from the pre-constructed set that are clearly incompatible with the current insulation domain fingerprint data or have significant matching deviations. The remaining domains are then progressively shrunk and rearranged in reverse order according to their compatibility with the insulation domain fingerprint data. This gradually compresses the original domain set from a large, multi-branch candidate space into a small number of highly correlated candidate domain paths. Furthermore, the screening results can be ordered by considering the boundary relationships, hierarchical inclusion relationships, and propagation coupling constraints between different pre-constructed insulation domains to form a candidate insulation domain sequence that reflects the possible range and priority of partial discharge events.

[0054] The candidate insulation domain sequence is correlated with the corresponding partial discharge event response, focusing on analyzing the boundary overlap, coupling propagation relationship, and defect influence propagation direction between adjacent candidate insulation domains. Then, the insulation coupling boundary involved in the partial discharge event is folded and mapped. Instead of simply mapping the partial discharge event to a single geometric location point, multiple interconnected boundary units are folded into a unified domain representation structure based on the sequence of domains, coupling strength, and boundary connectivity. Within this structure, the main action area, boundary extension area, and influence coverage area corresponding to the partial discharge event are determined. For cases with boundary intersections, discontinuous boundaries, or competing domains, the folding and mapping results can be further constrained and merged according to the priority order in the candidate insulation domain sequence. Finally, insulation domain data characterizing the weak insulation region, coupling propagation boundary, and defect influence range of the partial discharge event are output.

[0055] In this embodiment, by performing cross-modal response decoupling on insulation-related feature data, the separation degree of partial discharge effect features can be improved; by performing reverse screening analysis on the preset set of insulation action domains, interference from irrelevant action domains can be reduced and the efficiency of candidate action domain screening can be improved; furthermore, by performing folding mapping on the insulation coupling boundary, the range of action of partial discharge events, boundary relationships and the attribution of weak insulation regions can be more clearly characterized, making action domain identification more accurate, adaptable to complex boundaries, and mapping results more complete.

[0056] In an exemplary embodiment, based on the insulation domain fingerprint data, a reverse screening analysis is performed on the preset set of insulation domains for GIS insulation status monitoring locations to obtain a candidate insulation domain sequence, including steps 402 to 408. Wherein:

[0057] Step 402: Perform reverse exclusion coding on each preset insulation domain in the preset insulation domain set to obtain the preset exclusion domain set.

[0058] Step 404: Based on the fingerprint data of the insulation domain, perform fingerprint residual wake-up on the preset rejection domain set to obtain the initial activated domain set.

[0059] Step 406: Perform collapse analysis on the coupling competition relationship between each initial activation scope in the initial activation scope set to obtain a convergent candidate scope set.

[0060] Step 408: Reverse the order of the convergent candidate scope set to obtain the candidate insulating scope sequence.

[0061] Among them, reverse exclusion coding refers to the process of feature encoding a preset insulation domain from the perspective of mismatch or incompatibility.

[0062] Among them, the preset rejection scope set refers to the scope set composed of multiple preset insulation scopes after the reverse rejection coding is completed.

[0063] Fingerprint residual wake-up refers to the process of reactivating relevant domains based on the residual compatibility characteristics between the fingerprint data of the insulating domain and the rejection domain.

[0064] The initial activation scope set refers to the set of scopes that can participate in subsequent competitive analysis after fingerprint residual wake-up.

[0065] Among them, the coupling competition relationship refers to the mutual association and competition formed by different initial activation domains in terms of boundary attribution, propagation path or scope of influence.

[0066] Among them, collapse analysis refers to the analysis process of compressing, merging or reducing multiple scopes that have redundant, overlapping or conflicting relationships.

[0067] Among them, the convergent candidate scope set refers to the set of candidate scopes with clearer boundaries and more stable attribution obtained after collapse analysis.

[0068] Among them, reverse order rearrangement refers to the process of arranging convergent candidate scopes in reverse order according to their explanatory power, attribution priority, or boundary dominance.

[0069] Specifically, for each pre-established insulation domain of the GIS insulation condition monitoring location, its corresponding domain description information is extracted. This domain description information may include the insulation weakness characteristics, boundary distribution characteristics, propagation coupling characteristics, and adjacent or nested relationships with other domains. Then, during the calculation process, instead of directly performing forward matching encoding on each pre-established insulation domain, reverse rejection encoding is performed on each pre-established insulation domain from the perspective of incompatibility with the target partial discharge event. That is, for each pre-established insulation domain, its mismatch characteristics, incompatible boundary characteristics, and rejection constraint characteristics are constructed, and these rejection information are encoded into rejection identifiers that can be used for subsequent screening and reduction judgment. After completing the reverse rejection encoding of all pre-established insulation domains, a set of pre-established rejection domains is formed, consisting of multiple pre-established domain units with rejection attributes.

[0070] Insulation domain fingerprint data is input into the screening model corresponding to the preset rejection domain set. The deviation relationship, residual matching relationship, and local compatibility relationship between the insulation domain fingerprint data and each preset rejection domain in the preset rejection domain set are calculated. For preset rejection domains that are not completely matched overall but still retain some residual similarity in local boundary features, propagation coupling features, or insulation weakness features, they are reactivated through residual wake-up, transforming them from their original rejection state into candidate states that can participate in subsequent competition. That is, by identifying the residual relationship between the insulation domain fingerprint data and rejection features, domains that may have been prematurely eliminated in conventional forward screening but still have explanatory power for the current partial discharge event are awakened, thus forming the initial activated domain set.

[0071] This study identifies overlapping, parallel, and mutually exclusive relationships among different initial active scopes in the initial active scope set in terms of scope, boundary assignment, propagation path, and insulation influence level, and establishes coupling and competitive relationships among these scopes. Scopes that compete but are essentially mergeable within this coupling and competitive relationship are compressed and integrated. Scopes that are clearly enveloped or replaced by other scopes are weakened or eliminated, and scopes with continuous boundary relationships or common propagation directions are converged and merged, thereby reducing redundant and conflicting branches in the candidate scope set. After collapse analysis, the originally scattered, overlapping, or competing initial active scopes are converged into a small number of scope units with clearer boundaries, more stable assignments, and stronger interpretability, resulting in a converged candidate scope set.

[0072] By combining the boundary interpretation capability, defect coverage capability, propagation coupling priority, and rejection release degree of each convergent candidate domain with the fingerprint data of the insulation domain, convergent candidate domains with stronger boundary dominance capability, higher attribution interpretation capability, or larger influence range coverage capability after collapse analysis are prioritized, while convergent candidate domains of the boundary subordinate type, local supplement type, or competitive residual type are placed in the later positions, thus forming a candidate insulation domain sequence with a primary and secondary hierarchical relationship and attribution priority order.

[0073] In this embodiment, by performing reverse exclusion encoding on the preset insulation domains, mismatched domains can be highlighted in advance; then, by fingerprint residual wake-up, domains with local interpretability can be retained; further, by coupling competition relationship collapse analysis, redundant and conflicting domains can be reduced; finally, by reversing the order, a candidate insulation domain sequence with a clearer hierarchical classification can be formed, which can achieve higher screening efficiency, more convergent candidate results, and more reasonable domain sorting.

[0074] In an exemplary embodiment, a collapse analysis is performed on the coupling competition relationships between the initial activation scopes in the initial activation scope set to obtain a convergent candidate scope set, including steps 502 to 510. Wherein:

[0075] Step 502: Competitive pre-freeze each of the initial active scopes in the initial active scope set to obtain a pre-frozen scope pair set.

[0076] Step 504: Based on the pre-frozen scope pair set, flip the coupling competition edges between each initially activated scope to obtain the competition-flipped scope graph.

[0077] Step 506: Based on the competitive flipping scope graph, extract the dominant chain from the local competitive paths corresponding to each initial activated scope to obtain the dominant competitive chain set.

[0078] Step 508: Perform a compression and collapse analysis on the scopes in the dominant competitive chain set where conflicts occur between different dominant competitive chains in terms of scope affiliation and coupling boundaries to obtain a set of collapsed intermediate scopes.

[0079] Step 510: Converge the collapsed intermediate scope set to obtain a convergent candidate scope set.

[0080] Among them, competition pre-freezing refers to the process of pre-locking the competition status of scopes with potential conflict relationships before the formal competition adjudication.

[0081] Among them, the pre-frozen scope pair set refers to the set consisting of multiple scope pairing relationships that have undergone competitive pre-freezing.

[0082] Among them, the coupling competition edge refers to the connection edge between different initial activation scopes that has a low degree of coupling but still has a competitive relationship.

[0083] Among them, "flipping" refers to the process of adjusting the original competitive direction or priority relationship of the coupled competitive edge.

[0084] Among them, the competitive flipped scope graph refers to a graph structure with the initial active scope as nodes and the flipped competitive relationships as edges.

[0085] Among them, a local competitive path refers to a local competitive link formed by extending a certain initial activated scope along the competitive relationship in the competitive flip scope graph.

[0086] Among them, dominant chain extraction refers to the process of extracting chain segments that can represent the main attribution trend and competitive direction from local competitive paths.

[0087] The dominant competitive chain set refers to a collection of multiple dominant competitive chains extracted from the dominant chain.

[0088] Among them, domain attribution refers to the correspondence between a partial discharge event and a certain insulation action area or a certain candidate domain.

[0089] Among them, a conflict on the coupling boundary refers to the overlap or competition between the scopes in different dominant competitive chains in terms of boundary coverage, boundary interpretation, or boundary dominance.

[0090] Among them, compression and collapse analysis refers to the analysis process of compressing, merging or reducing multiple scopes that have conflicting or redundant relationships.

[0091] The collapse intermediate action domain set refers to the set of intermediate action domain units formed after compression and collapse analysis.

[0092] Convergence encapsulation refers to the process of uniformly organizing and grouping the collapsed intermediate scope units to form stable candidate results.

[0093] Specifically, the attribution features, boundary features, propagation coupling features, and influence range features corresponding to each initial activated scope are extracted from the initial activated scope set. For any two initial activated scopes, pairings are established to determine if there is overlap in scope, boundary contact, cross propagation path, or competition in influence levels, thus creating initial competitive relationship pairs. After forming initial competitive relationship pairs, real-time competition adjudication is not initiated directly. Instead, scope pairs with significant competitive tendencies or potential conflict relationships are temporarily frozen into pre-frozen relationships in the current competition state, boundary contact state, attribution conflict state, and coupling association state. These scope pairs are retained as priority objects in subsequent analysis, thus pre-extracting and freezing the complex competitive relationships that were originally scattered throughout the initial activated scope set into a structured set of pre-frozen scope pairs.

[0094] Using a pre-frozen set of scope pairs as input, the strength of competing edges between these pairs is determined. Specifically, the degree of boundary overlap, propagation coupling strength, degree of ownership conflict, and scope coverage of each competing edge are analyzed. For competing edges determined to be weakly coupled, their original competing directions are no longer maintained; instead, edge flipping is performed. This involves changing the original direction of the competing relationship from one scope to another, or adjusting a scope that was originally in a subordinate competitive position to a priority interpretation node, thereby reorganizing the competitive topology between the initially activated scopes. After flipping all coupled competing edges, the initially activated scopes are used as nodes, and the flipped competing relationships are used as edges to construct a competing flipped scope graph.

[0095] For each initially activated scope node in the competitive flipping scope graph, a corresponding local competitive path is searched along its connected competitive edges. The relative relationships of nodes on this path in terms of boundary dominance, attribution stability, propagation continuity, and influence coverage are analyzed. For scope nodes with continuous dominance characteristics within the same path, they are connected in series according to the direction of competitive relationship and the path extension order. Dominant chains that represent the main competitive trends and attribution directions are extracted from each local competitive path. When there are multiple intersecting paths or multi-branch competitive paths, path segments with stronger dominance, higher boundary interpretation ability, or more obvious coupling and connectivity are preferentially retained and organized into corresponding dominant competitive chains, resulting in a set of dominant competitive chains composed of multiple dominant competitive chains.

[0096] Cross-comparison of different dominant competitive chains within the dominant competitive chain set identifies conflicting scopes that simultaneously compete for the same insulating region in terms of scope attribution, simultaneously cover the same boundary unit at the coupling boundary, or form mutually crowding relationships in the propagation path. These conflicting scopes undergo compression and collapse analysis, which, considering the chain-level priority, boundary dominance, scope coverage, and coupling continuity of their respective dominant competitive chains, compresses, merges, weakens, or replaces the conflicting scopes. Multiple conflicting scopes with similar interpretations are grouped into a smaller number of intermediate scope units, and conflicting nodes with insufficient boundary support or poor attribution stability are removed. After compression and collapse, the originally repetitive or conflicting scope relationships between different dominant competitive chains are compressed into a more concise and unified intermediate expression, resulting in a collapsed intermediate scope set.

[0097] The intermediate action domain units in the collapsed intermediate action domain set are uniformly organized and encapsulated. Their corresponding attribution, boundary, coupling, and priority information are centrally summarized and then converged according to the stability of the action domain, the integrity of its boundaries, and its ability to explain partial discharge events. Intermediate action domain units that can collectively characterize the same weak insulation region or the same coupled boundary cluster are further encapsulated into the same candidate action domain object. Intermediate action domain units that still have slight differences but are generally consistent in their attribution are retained as different sub-level representations under the same candidate action domain object. Finally, the multiple candidate action domain objects after uniform encapsulation are output as a converged candidate action domain set.

[0098] In this embodiment, by pre-freezing the initial active scope, potential competitive relationships can be locked first; by flipping the coupled competitive edges, the competitive direction between scopes can be reconstructed; further, by extracting the dominant chain and squeezing and collapsing the conflict scope, the main belonging path can be highlighted and the inter-chain conflict can be compressed; finally, by convergence encapsulation, a convergence candidate scope set with more stable boundaries and more concentrated belonging can be formed, which makes the competitive relationship clearer, the conflict scope compressed more fully, and the candidate scope results more stable.

[0099] In an exemplary embodiment, the insulation coupling boundary of a partial discharge event is folded and mapped according to the candidate insulation domain sequence to obtain insulation domain data, including steps 602 to 608. Wherein:

[0100] Step 602: Perform boundary suspension processing on the candidate insulation domain sequence to obtain the set of boundary fragments to be mapped.

[0101] Step 604: Based on the set of boundary fragments to be mapped and the sequence of candidate insulation domains, borrow projection is performed on the boundary pointing relationship of partial discharge events to obtain the boundary projection domain network.

[0102] Step 606: Perform folding and closure analysis on the broken coupled boundaries in the boundary projection domain network to obtain the closed boundary domain set.

[0103] Step 608: Based on the set of closed boundary domains, encapsulate and map the domain attribution of partial discharge events to obtain insulation domain data.

[0104] Boundary suspension refers to the process of temporarily separating the boundary information in the candidate insulation domain from the original global domain expression.

[0105] The set of boundary segments to be mapped refers to a collection of multiple boundary segments that have been processed by boundary suspension and can independently participate in subsequent mapping.

[0106] Among them, the boundary pointing relationship refers to the extension direction and belonging relationship of the boundary segment corresponding to the partial discharge event between different candidate insulation domains.

[0107] Borrowed projection refers to a process that allows boundary segments to temporarily borrow the boundary positions or pointing relationships of adjacent candidate insulating domains for mapping.

[0108] Among them, the boundary projection domain network refers to the network structure formed by the projection connection relationship between boundary segments and candidate insulating domains.

[0109] Among them, fold closure analysis refers to the analysis process of merging and completing the broken coupled boundaries in the boundary projection domain network to form a continuous boundary domain.

[0110] The closed boundary domain set refers to the set of multiple boundary domains with continuous boundaries and well-defined domain characteristics obtained after folding and closing analysis.

[0111] Encapsulation mapping refers to the process of organizing and outputting the attribution relationship between partial discharge events and closed boundary domains, as well as related boundary information, as insulation domain data.

[0112] Specifically, boundary description information corresponding to each candidate insulation domain is extracted from the candidate insulation domain sequence. This boundary description information may include the domain edge position, boundary extension direction, contact relationship between adjacent domains, and boundary coverage. Instead of directly using the entire boundary of each candidate insulation domain for mapping, the boundary units originally attached to the complete candidate insulation domain are temporarily separated from the overall candidate insulation domain representation. These separated units are then split and recombined according to boundary continuity, boundary direction consistency, and boundary coupling correlation to form multiple boundary segments that can independently participate in subsequent mapping. For cases where multiple candidate insulation domains have overlapping, intersecting, or nested boundaries, the relevant boundary segments are further separated, marked, and suspended to prevent them from being pre-fixed and assigned to a single domain, resulting in a set of boundary segments to be mapped.

[0113] The propagation direction of partial discharge events, boundary continuity relationships, and potential directional relationships between domains are analyzed together with the set of boundary segments to be mapped and the sequence of candidate insulation domains. Then, based on the potential directional relationships, each boundary segment to be mapped is allowed to temporarily borrow the boundary position relationship, coverage relationship, or propagation direction relationship of adjacent candidate insulation domains while maintaining its original boundary attributes, thus establishing a projection relationship between the boundary segment and multiple candidate insulation domains. Based on the projection relationship, the boundary segments that were originally scattered in the candidate insulation domain sequence are connected according to the boundary direction logic, forming a network structure composed of the projection relationships between boundary segments and candidate insulation domains. The projection connection relationships between each boundary segment to be mapped, between boundary segments and candidate insulation domains, and between different candidate insulation domains are uniformly organized to obtain the boundary projection domain network.

[0114] The boundary projection domain network is traversed to identify broken coupling boundaries with boundary interruptions, discontinuities, misalignments, or missing coverage. Then, based on the candidate insulation domains on both sides of the broken boundary, the projection relationship between boundary segments, the propagation direction, and the continuity of boundary coupling, logically connected but structurally separated boundary segments are merged using boundary folding. The coupling boundaries that originally did not form complete loops or complete coverage areas are completed into closed or quasi-closed boundary domains using boundary closure. For multiple broken coupling boundaries pointing to the same weak insulation region or the same propagation extension region, the relevant broken boundaries can be further folded into a single boundary domain object, reconstructing a set of closed boundary domains.

[0115] Based on the completeness of the corresponding closed boundary domain set, the number of candidate insulation domains covered, the consistency of boundary orientation, and the ability to interpret the propagation behavior of partial discharge events, it is determined which closed boundary domain a partial discharge event should preferentially belong to, or which closed boundary domains with primary and secondary hierarchical relationships it should simultaneously belong to. Then, the attribution results are encapsulated and mapped to uniformly organize the information on weak insulation areas, coupling boundaries, scope of action, and attribution priority under the same closed boundary domain, forming a data structure that can comprehensively characterize the area of ​​action, boundary range, and influence level of a partial discharge event in the GIS insulation condition monitoring location, thus obtaining insulation domain data.

[0116] In this embodiment, by performing boundary suspension processing on the candidate insulation domain sequence, the boundary information originally fixed in the domain can be separated; then, by borrowing projection, a flexible association between the boundary fragment and the candidate insulation domain can be established; further, by performing folding and closure analysis on the broken coupling boundary, the continuity and completeness of the boundary expression can be improved; finally, by encapsulation mapping, insulation domain data with clearer attribution relationships can be formed, which enables more detailed boundary characterization, stronger adaptability of broken boundaries, and more complete expression of domain attribution.

[0117] In an exemplary embodiment, based on the set of boundary fragments to be mapped and the sequence of candidate insulating domains, borrowing projection is performed on the boundary pointing relationship of partial discharge events to obtain a boundary projection domain network, including steps 702 to 708. Wherein:

[0118] Step 702: Perform offset processing on each boundary segment in the set of boundary segments to be mapped to obtain a set of boundary borrowed segments.

[0119] Step 704: Analyze the misalignment relationship between the boundary borrowing fragment set and the candidate insulation domain sequence to obtain the initial set of linked domain pairs.

[0120] Step 706: Based on the initial set of attached domain pairs, perform torsional compensation on the boundary pointing offset relationship of the partial discharge event to obtain the pointing compensation domain chain set.

[0121] Step 708: Perform textured networking on the chain set pointing to the compensation scope to obtain the boundary projection scope network.

[0122] Among them, the yielding offset processing refers to the process of adjusting the position representation, orientation representation or attachment priority of the boundary segment to reserve space for subsequent borrowing mapping.

[0123] Among them, the boundary borrowing fragment set refers to the set of boundary fragments that have the ability to be attached across scopes after being processed by yielding and biasing.

[0124] Among them, misaligned connection refers to the relationship between the boundary borrowed segment and the candidate insulation domain, which is not a standard corresponding position but can still form an effective connection.

[0125] The initial attachment scope pair set refers to the scope pair set formed by the misaligned attachment relationship between the boundary borrow segment and the candidate insulation scope.

[0126] Among them, the boundary pointing offset relationship refers to the relationship in which the extension direction, connection direction or belonging direction of the boundary segment between candidate insulating domains deviates from the original propagation logic.

[0127] Torsional compensation refers to the process of correcting the boundary pointing offset relationship to restore the reasonable boundary connection direction and propagation link.

[0128] Among them, the scope chain set for compensation refers to a set of multiple scope chains with continuous boundary pointing relationships formed after torsion compensation.

[0129] Among them, textured networking refers to the process of weaving multiple scope chains into a network structure according to the boundary orientation, inter-chain connection relationship and scope succession hierarchy.

[0130] Specifically, the boundary position features, extension direction features, boundary length features, and contact relationship features with adjacent candidate insulation domains are extracted from each boundary segment in the set of boundary segments to be mapped. Then, based on the spatial adjacency relationship between the boundary segments and adjacent candidate insulation domains, the boundary extension trend, and the potential ownership competition relationship, the position representation, direction representation, or attachment priority of the boundary segments are adaptively biased and adjusted to reserve borrowing space for cross-domain mapping without destroying the original boundary attributes. For boundary segments with overlapping boundaries, misaligned boundaries, or boundary competition pointing, they are preferentially adjusted to an intermediate state where they can borrow space from adjacent candidate insulation domains, resulting in a set of boundary borrowing segments.

[0131] A matching analysis is performed on the relative positional relationships, boundary adjacency relationships, propagation direction consistency, and domain coverage correlation between each boundary borrow segment in the boundary borrow segment set and each candidate insulation domain in the candidate insulation domain sequence. During the analysis, the focus is on identifying misaligned connections between boundary borrow segments and candidate insulation domains. That is, although a boundary borrow segment is not located at the standard boundary position of a candidate insulation domain, it can still form an effective connection with that candidate insulation domain in terms of boundary extension direction, propagation connection, or domain continuation logic. For the identified effective misaligned connections, the corresponding boundary borrow segments are paired with candidate insulation domains to form multiple initial connected domain pairs. These domain pairs are recorded and organized according to connection strength, boundary continuity, and propagation interpretation capability, ultimately resulting in an initial set of connected domain pairs.

[0132] Based on the initial set of attached scope pairs, the boundary pointing offset of the boundary borrowing segments in each attached scope pair is identified. This offset can include situations where the boundary pointing is inconsistent with the original propagation direction, the boundary extension direction deviates from the dominant scope direction, or there is a deflection in the attachment of the boundary segment among multiple candidate insulating scopes. Then, based on the attachment relationship, propagation pointing logic, and scope sequence order corresponding to the boundary borrowing segments, the connection direction of the boundary borrowing segments between candidate insulating scopes is corrected, interrupted boundary propagation links are repaired, and scope pointing deviations caused by misaligned attachments are corrected. This allows multiple initial attached scope pairs to be sequentially connected along a more reasonable boundary propagation direction. Therefore, the originally discrete initial attached scope pairs are organized into a scope chain structure with continuous boundary pointing logic, ultimately resulting in a pointing-compensated scope chain set.

[0133] This process extracts the chain nodes, chain directions, inter-chain connections, and boundary continuation relationships from each scope chain in the target compensation scope chain set. Based on whether there are shared boundaries, intersecting directions, overlapping paths, or parallel propagation among the scope chains, a textured connection relationship is established between different scope chains. For scope chains with similar boundary texture features, similar propagation direction features, and compatible connection relationships, multiple scope chains are woven into a unified network structure according to boundary direction and scope continuation hierarchy, forming ordered connections, intersections, and extensions between different chain segments. Simultaneously, isolated chain segments are supplemented, duplicate chain segments are merged, and locally conflicting chain segments are textured and adjusted to ensure that the final network structure has good continuity and expressive integrity, forming a boundary projection scope network.

[0134] In this embodiment, by performing a yielding offset processing on the boundary segment to be mapped, the adaptability of the boundary segment to adjacent scopes can be improved; further, through misalignment connection analysis, a non-fixed correspondence between the boundary segment and the candidate insulating scope can be established; further, through torsional compensation of the boundary pointing offset relationship, the deviation of the boundary propagation direction can be corrected and the link continuity can be enhanced; finally, through textured networking processing, a more complete boundary projection scope network can be formed, making the boundary mapping more flexible, the cross-scope connection capability stronger, and the scope network expression more continuous.

[0135] In an exemplary embodiment, based on insulation correlation characteristic data and insulation domain data, the insulation state evolution process of the GIS insulation state monitoring location is analyzed to obtain insulation state characterization quantities and insulation risk potential energy values, including steps 802 to 808. Wherein:

[0136] Step 802: Perform state folding on the insulation correlation feature data and insulation domain data to obtain the insulation state folding characterization set; Step 804: Based on the insulation state folding characterization set, the insulation state evolution chain of the GIS insulation state monitoring location is de-wound and unfolded to obtain the insulation evolution path set; Step 806: Based on the set of insulation evolution paths, perform potential energy accumulation analysis on the risk accumulation relationship during the insulation state evolution process to obtain the risk potential energy distribution set; Step 808: Perform dual-quantity encapsulation on the insulation state folding characterization set and the risk potential energy distribution set to obtain the insulation state characterization quantity and the insulation risk potential energy value.

[0137] State folding refers to the process of compressing multidimensional state information from insulation-related feature data and insulation domain data into a unified state expression.

[0138] Among them, the insulation state folding characterization set refers to the set of characterization units formed after state folding, which is used to characterize the current insulation state characteristics.

[0139] Among them, the insulation state evolution chain refers to the state connection link of the GIS insulation state monitoring part as it develops from initial anomaly to continuous deterioration and increased risk.

[0140] Among them, decoupling refers to the process of reversing the analysis from the folded state representation and recovering the evolution process of the insulating state.

[0141] Among them, the insulation evolution path set refers to the set of multiple insulation state evolution paths obtained after dewinding and unfolding.

[0142] Among them, risk accumulation relationship refers to the correlation relationship in which risk gradually accumulates during the evolution of insulation state as state changes, boundary expansion and domain migration occur.

[0143] Among them, potential energy accumulation analysis refers to the analytical process of characterizing the cumulative intensity, distribution trend and transmission relationship of risks during the evolution of insulation state using potential energy.

[0144] Among them, the risk potential energy distribution set refers to the data set obtained after potential energy accumulation analysis, which is used to characterize the distribution of risk in different evolution paths and stages.

[0145] Among them, dual-quantity encapsulation refers to the process of unifying and outputting the characterization results of the state dimension and the characterization results of the risk dimension into insulation state characterization quantities and insulation risk potential energy values.

[0146] Specifically, insulation correlation feature data and insulation domain data are used as joint inputs. Insulation correlation feature data characterizes the comprehensive response characteristics of partial discharge events in terms of electromagnetic propagation, acoustic propagation, mechanical vibration, and cross-modal coupling. Insulation domain data characterizes the weak insulation regions, coupling propagation boundaries, and defect influence ranges corresponding to partial discharge events. The two types of data are correlated, aligned, and jointly characterized, binding the correlation feature information and domain information for the same GIS insulation status monitoring location, time period, and partial discharge event. After binding, multiple characterization information items, originally scattered across different dimensions and data structures but collectively reflecting insulation health status, boundary disturbance degree, defect expansion degree, and partial discharge activity level, are compressed into several folded characterization units with unified state meaning. The hierarchical, correlation, and strength relationships between different folded characterization units are preserved, forming an insulation state folded characterization set.

[0147] This study identifies the connections between folded representation units in the insulation state folded representation set in terms of temporal continuity, state progression, boundary expansion, and risk correlation, and establishes a folded representation of the initial insulation state evolution chain based on this. From the folded representation results, the potential evolutionary path of the insulation state from initial local anomalies, boundary disturbances, and expansion of the domain to increased risk is analyzed in reverse. The multi-stage state change information, previously compressed into the same folded representation unit, is unfolded hierarchically, and the sequential connections and path transfer relationships between each stage are restored. For cases with the possibility of multi-branch evolution, multiple evolutionary branches can be unfolded and differentiated in parallel according to the state weights, boundary continuity, and domain evolution intensity in the insulation state folded representation set, thus forming multiple path units that can reflect different insulation degradation directions and evolutionary trends. Finally, these path units are unified to obtain the insulation evolution path set.

[0148] This study analyzes the local discharge activity, boundary disturbance, domain expansion, and state degradation rate of each evolution path in the insulation evolution path set at different evolution stages, and identifies the formation, growth, and transmission relationships of risks along each path. The risk contributions at different stages and along different paths are cumulatively calculated based on their state progression, boundary influence, and domain expansion intensity, so that risk is no longer expressed as an anomalous quantity at a single moment, but rather as potential energy gradually accumulated along the insulation state evolution process. For cases where different evolution paths intersect, run parallel, or bifurcate, the study further analyzes their risk superposition, transmission, and local amplification relationships, mapping these relationships together into a corresponding risk potential energy distribution structure, ultimately forming a risk potential energy distribution set.

[0149] An insulation state folding characterization set is used to reflect the current state characteristics and degradation stage of the GIS insulation state monitoring site, while a risk potential energy distribution set is used to reflect the risk accumulation degree and potential development trend under the corresponding insulation state. Then, core quantitative information for the final output is extracted from both the state and risk dimensions, and a correspondence is established between the insulation state folding characterization set and the risk potential energy distribution set, so that the degree of change in insulation state and the accumulation level of risk potential energy can be jointly reflected in a unified result. For example, characterization information reflecting the insulation health level, boundary disturbance degree, local degradation degree, and evolution stage can be encapsulated as insulation state characterization quantities, while characterization information reflecting risk accumulation intensity, risk distribution density, path expansion trend, and potential failure tendency can be encapsulated as insulation risk potential energy values. The resulting insulation state characterization quantities are used to quantitatively represent the current insulation state level of the GIS insulation state monitoring site, and the resulting insulation risk potential energy values ​​are used to quantitatively represent the risk accumulation intensity and potential development degree corresponding to that monitoring site.

[0150] In this embodiment, by performing state folding on insulation-related feature data and insulation domain data, scattered multidimensional insulation information can be compressed into a unified state expression; then, by de-coiling and unfolding the insulation state evolution chain, the evolution path of insulation degradation can be more clearly recovered; further, by performing potential energy accumulation analysis on the risk accumulation relationship, the cumulative trend of risk along the evolution process can be characterized; finally, through dual-quantity encapsulation, insulation state characterization quantity and insulation risk potential energy value can be output simultaneously, making the state expression more concentrated, the evolution path identification clearer, and the risk quantification results more comprehensive.

[0151] In an exemplary embodiment, the insulation state evolution chain of the GIS insulation state monitoring location is de-wound and unfolded according to the insulation state folding characterization set to obtain the insulation evolution path set, including steps 902 to 908. Wherein:

[0152] Step 902: Perform evolution chain suspension processing on the insulation state folding characterization set to obtain the set of chain segments to be unfolded.

[0153] Step 904: Based on the set of segments to be unfolded, awaken the folding remnants in the insulation state folding characterization set to obtain the activated evolution segment set.

[0154] Step 906: Based on the set of activated evolution chain segments, the insulation state evolution chain is reverse-connected and expanded to obtain the set of evolution path segments.

[0155] Step 908: Encapsulate the evolution path fragment set into path clusters to obtain the insulation evolution path set.

[0156] Among them, the evolution chain suspension treatment refers to the process of temporarily releasing the original fixed connection relationship in the insulation state folding characterization group and releasing potential chain segments.

[0157] Among them, the set of segments to be expanded refers to the set of segments that can independently participate in subsequent evolution recovery after the evolution chain suspension process.

[0158] Among them, folding traces refer to the stage transition information, path branch information, or boundary change traces that are compressed and preserved in the folding representation during the state folding process.

[0159] Among them, "awakening" refers to the process of releasing folded remnants from their latent state into explicit chain segment information that can participate in the construction of the evolutionary chain.

[0160] Among them, the activated evolutionary chain set refers to the set of evolutionary chain segments that are awakened by the folded remnants and contain more complete connection information.

[0161] Among them, reverse expansion refers to the process of recovering the connection relationship of the preceding chain segment from the already revealed state result or higher-order state unit.

[0162] Among them, the evolution path fragment set refers to the set of path fragments that have a clear evolutionary sequence logic obtained after inverse expansion.

[0163] Path cluster encapsulation refers to the process of merging and organizing path segments with similar evolutionary directions or the same degeneracy main line into path clusters.

[0164] Specifically, the state hierarchy information, temporal correlation information, domain migration information, and risk indication information corresponding to each folded representation unit in the insulation state folded representation set are extracted. Then, the original fixed chain connection relationship between each folded representation unit is temporarily released, and the evolutionary segments that may be contained in each folded representation unit are independently separated, so that they exist in a state to be unfolded. For folded representation units containing multi-stage state information, composite boundary change information, or multi-domain migration information, they are further decomposed into multiple chain segment expression units that can independently participate in subsequent unfolding according to the state progression characteristics and boundary change characteristics, resulting in a set of chain segments to be unfolded.

[0165] Based on the set of segments to be unfolded, fold remnants that are still implicitly present but not yet explicitly released in the insulation state fold representation set are identified. These fold remnants can be stage transition information, path branch information, boundary expansion traces, or risk progression traces compressed within the same fold representation unit. Then, based on the mapping relationship between each segment to be unfolded and the original fold representation unit, it is analyzed which fold remnants have potential continuation, complement, or enhancement relationships with the current segment to be unfolded. These fold remnants are then awakened, transforming them from implicit states into explicit segment information that can participate in the construction of the evolutionary chain. For cases where multiple segments point to the same fold remnant, the remnant awakening intensity can be adjusted based on state continuity, domain consistency, and risk progression to control the release priority of the remnants, thus obtaining an activated evolutionary segment set.

[0166] For each segment in the activated evolution chain set, we analyze its corresponding state starting point, state ending point, boundary evolution direction, domain migration direction, and risk progression direction. Based on this, starting from the currently manifested state result or higher-order state unit, we trace back its preceding and basic chains to gradually recover the potential connections between different stages. For cases where multiple activated evolution chains have sequential, branching, or parallel evolutionary relationships, we perform reverse continuation based on the rationality of state progression, boundary continuity, and consistency of risk transmission, constructing sets of evolutionary path segments that can independently express the evolution trajectory of the insulating state. After reverse continuation and expansion, the originally scattered segments in the activated evolution chain set are restored to a path segment structure with clear sequential logic and evolutionary direction, thus obtaining the evolutionary path segment set.

[0167] The evolution path fragment set is organized and grouped as a whole, with path fragments having the same initial state characteristics, similar evolution directions, similar boundary expansion methods, or the same risk progression trend grouped into the same path cluster. Then, the path fragments within each path cluster are sequentially organized, hierarchically merged, and their continuity is verified, so that multiple path fragments belonging to the same insulation degradation direction can be uniformly encapsulated into a relatively complete insulation evolution path. For path fragments with bifurcated, parallel, or locally overlapping relationships, their branch structure or hierarchical relationship is preserved within the path cluster to ensure that the encapsulation result reflects both the overall evolutionary main line and the local evolutionary branches. All encapsulated path clusters are output uniformly to form an insulation evolution path set.

[0168] In this embodiment, by suspending the evolution chain of the insulation state folding characterization set, the originally compressed state chain segments can be released from the fixed connection relationship; by waking up the folding remnants, the evolution information hidden by the folding can be supplemented; further, by reverse connection unfolding, the connection relationship between the front and back of the insulation state evolution chain can be restored; finally, by path cluster encapsulation, a more clearly structured insulation evolution path set can be formed. Therefore, the evolution information is released more fully, the path recovery capability is stronger, and the expression of the evolution results is more hierarchical.

[0169] In an exemplary embodiment, based on the set of insulation evolution paths, a potential energy accumulation analysis is performed on the risk accumulation relationship during the insulation state evolution process to obtain a risk potential energy distribution set, including steps 1002 to 1010. Wherein:

[0170] Step 1002: Perform path potential difference suspension processing on the insulation evolution path set to obtain the potential difference segment set to be accumulated.

[0171] Step 1004: Based on the set of potential difference segments to be accumulated, the residual voltage of the insulation evolution path set is activated to obtain the set of activated accumulation segments.

[0172] Step 1006: Based on the activated accumulation segment set, perform a risk accumulation slope transformation analysis on the insulation state evolution process to obtain the risk transformation segment set.

[0173] Step 1008: Based on the risk transition segment set, the aggregation relationship between potential energy slots is merged to obtain the potential energy slot set.

[0174] Step 1010: Distribute and encapsulate the potential energy slot set to obtain the risk potential energy distribution set.

[0175] Among them, path potential difference suspension processing refers to the process of separating the state differences and risk differences between adjacent stages in the insulation evolution path from the complete path and temporarily retaining them independently.

[0176] Among them, the set of potential difference segments to be clustered refers to the set of multiple potential difference segments that can be used for subsequent risk clustering analysis after path potential difference suspension processing.

[0177] Among them, evolutionary residual voltage refers to the amount of risk or influence that is not released in time during the insulation state evolution process and continues to remain in subsequent stages.

[0178] Among them, the activated accumulation segment set refers to the accumulation segment set that can participate in the risk accumulation calculation after being awakened by evolutionary residual pressure.

[0179] Among them, the risk accumulation slope refers to the degree to which the risk increases, changes, or accelerates as the insulation state evolves.

[0180] Turning point analysis refers to the analytical process of identifying the location and process at which the slope of risk accumulation changes, abruptly changes, or intensifies.

[0181] Among them, the risk transition segment set refers to the set of multiple risk segments that represent the changes in the pace of risk accumulation, obtained after transition analysis.

[0182] Among them, the accumulation relationship between potential energy channels refers to the correlation formed between different risk potential energy concentration areas in terms of risk transmission, superposition or convergence.

[0183] Among them, aggregation refers to the process of merging and organizing multiple potential energy slots that are closely related or continuously transmitted to each other.

[0184] Among them, the potential energy groove set refers to the set of potential energy grooves formed after aggregation, which is used to characterize the main risk accumulation area.

[0185] Distributed encapsulation refers to the process of organizing and outputting the distribution relationship of potential energy slots on different paths, stages and domains as a risk potential energy distribution set.

[0186] Specifically, the state change amplitude, boundary expansion degree, domain migration degree, and risk change intensity of each insulation evolution path in the insulation evolution path set at different evolution stages are extracted. Then, the state drop, risk drop, and boundary influence drop between adjacent evolution stages in each insulation evolution path are separated to form multiple path segment expression units with independent potential difference meanings. These path segments are temporarily suspended from the original complete path structure as potential difference segments that can be analyzed separately. For cases where there are multiple abrupt, gradual, or alternating evolutions in the same evolution path, it is further subdivided and labeled according to the potential difference change direction, potential difference change amplitude, and stage progression relationship to obtain the potential difference segment set to be aggregated.

[0187] Based on the set of potential difference segments to be accumulated, the evolving residual voltages that are implicitly present but not yet explicitly expressed in the insulation evolution path set are identified. Evolving residual voltages can be understood as risk remnants, boundary disturbance margins, or domain expansion margins that are not immediately released during the insulation state evolution process but continue to be retained in subsequent stages. Then, based on the above results, the correspondence between each potential difference segment to be accumulated and the evolving residual voltage is analyzed to identify which potential difference segments can trigger residual voltage release, inheritance, or amplification. The corresponding evolving residual voltages are then awakened, transforming them from their original implicit accumulation state into explicit risk segments that can participate in subsequent risk accumulation calculations. For cases where multiple insulation evolution paths share the same type of residual voltage source, the degree of residual voltage awakening can be adjusted according to path continuity, stage progression, and domain consistency to obtain an activated accumulation segment set.

[0188] For each activated accumulation segment in the activated accumulation segment set, the risk growth trend, boundary disturbance growth trend, domain expansion rate, and state degradation acceleration degree during the insulation state evolution process are analyzed, and corresponding risk accumulation slopes are established based on these changing relationships. Furthermore, key slope inflection points are identified based on each risk accumulation slope, indicating a shift from slow to rapid risk accumulation, from local growth to overall expansion, or from single-path transmission to multi-path diffusion. Based on these inflection points, the accumulation segments are re-segmented and reorganized. In cases with multiple adjacent slope segments or slope abrupt change segments, they can be further classified according to the slope turning direction, turning intensity, and duration to group accumulation segments with similar risk transformation characteristics into similar transition segments, resulting in a risk transition segment set.

[0189] Based on the risk concentration areas, risk trough areas, and risk progression channels corresponding to each risk transition segment in the risk transition segment set, the potential energy groove structure of different risk transition segments in the insulation state evolution process is identified. The potential energy groove can represent a relatively concentrated interval of risk in a certain stage, path, or domain. Then, it is analyzed whether there are continuous accumulation relationships, mutual conduction relationships, parallel superposition relationships, or local confluence relationships among different potential energy grooves. Multiple potential energy grooves that are closely related, have continuous risk transmission, or have successive evolution stages are merged and organized, transforming them from scattered local groove units into groove group structures with overall accumulation significance. For potential energy grooves with intersecting risks in different evolution paths, they can also be merged or hierarchically divided according to risk transmission priority and accumulation intensity to obtain a potential energy groove set.

[0190] Information on risk intensity, distribution range, path affiliation, stage affiliation, and clustering hierarchy for each potential energy slot in the potential energy slot set is extracted. Then, the potential energy slots are distributed and organized according to insulation evolution path, insulation state stage, and influence range, and their distribution relationships between different paths, stages, and domains are uniformly expressed. For adjacent, partially overlapping, or continuously transmitted potential energy slots, their connection and hierarchical relationships are further preserved during the encapsulation process to ensure that the final result reflects both the local risk characteristics of individual potential energy slots and the overall distribution pattern among multiple potential energy slots. Finally, the uniformly organized potential energy slot results are output as a risk potential energy distribution set.

[0191] In this embodiment, by performing path potential difference suspension processing on the insulation evolution path set, the risk potential difference between different evolution stages can be extracted independently; then, by waking up the evolution residual voltage, the potential risk residue in subsequent stages can be released; further, by performing a reversal analysis on the risk accumulation slope, the key change stage from slow increase to sudden increase in risk can be identified; subsequently, by merging the accumulation relationship between potential energy grooves, a more concentrated risk accumulation region can be formed; finally, by distributed encapsulation, a risk potential energy distribution set with clearer hierarchy can be obtained, which enables a more detailed characterization of the risk accumulation process, more accurate identification of key risk turning points, and a more systematic expression of risk distribution.

[0192] Based on the same inventive concept, this application also provides an online monitoring device for multimodal partial discharge in insulation states, used to implement the above-mentioned online monitoring method for multimodal partial discharge in insulation states. For example... Figure 3 As shown, it includes: a data acquisition module, a time extraction module, a fusion analysis module, a mapping analysis module, and an evolution analysis module. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the online monitoring device for multimodal partial discharge in insulation state provided below can be found in the limitations of the online monitoring method for multimodal partial discharge in insulation state described above, and will not be repeated here.

[0193] The modules in the aforementioned multimodal partial discharge online monitoring device for insulation states can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0194] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4As shown. This computer device includes a processor, memory, input / output interfaces (I / O), and communication interfaces.

[0195] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0196] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0197] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0198] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to perform the steps in the above-described method embodiments.

[0199] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0200] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods.

[0201] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0202] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for online monitoring of multimodal partial discharge in insulation conditions, characterized in that, The method includes: Cross-modal joint triggering acquisition was performed on the partial discharge response signals of the GIS insulation condition monitoring points to obtain the original insulation monitoring data; The suspected partial discharge pulse data in the original insulation monitoring data are extracted to obtain an insulation event data set; Based on the insulation event data set, the electromagnetic propagation response, acoustic propagation response, mechanical vibration response, and cross-modal coupling relationship of the partial discharge events at the GIS insulation status monitoring location are fused and analyzed to obtain insulation correlation feature data; Based on the insulation correlation feature data, insulation domain mapping analysis is performed on the partial discharge events of the GIS insulation status monitoring location to obtain insulation domain data; Based on the insulation correlation feature data and the insulation domain data, the insulation state evolution process of the GIS insulation state monitoring location is analyzed to obtain insulation state characterization quantities and insulation risk potential energy values.

2. The method according to claim 1, characterized in that, The step involves performing insulation domain mapping analysis on partial discharge events at the GIS insulation status monitoring location based on the insulation correlation feature data to obtain insulation domain data, including: Cross-modal response decoupling is performed on the insulation-related feature data to obtain insulation domain fingerprint data; Based on the insulation domain fingerprint data, a reverse screening and reduction analysis is performed on the preset insulation domain set of the GIS insulation status monitoring location to obtain a candidate insulation domain sequence. Based on the candidate insulation domain sequence, the insulation coupling boundary of the partial discharge event is folded and mapped to obtain the insulation domain data.

3. The method according to claim 2, characterized in that, The step involves performing a reverse screening analysis on the preset set of insulation domains for the GIS insulation status monitoring location based on the insulation domain fingerprint data to obtain a candidate insulation domain sequence, including: Perform reverse rejection encoding on each preset insulation domain in the preset insulation domain set to obtain a preset rejection domain set; Based on the insulation domain fingerprint data, fingerprint residual wake-up is performed on the preset rejection domain set to obtain an initial activation domain set; A collapse analysis is performed on the coupling competition relationship between each initial activation scope in the initial activation scope set to obtain a convergent candidate scope set; The convergent candidate scope set is reversed and rearranged to obtain the candidate insulating scope sequence.

4. The method according to claim 3, characterized in that, The process involves collapsing the coupling competition relationships among the initial activation scopes in the initial activation scope set to obtain a convergent candidate scope set, including: Competitive pre-freezing is performed on each of the initial activation scopes in the initial activation scope set to obtain a pre-frozen scope pair set; Based on the pre-frozen scope pair set, the coupling competition edges between each of the initial activated scopes are flipped to obtain a competition-flipped scope graph; Based on the competitive flipping scope graph, the dominant chain is extracted from the local competitive paths corresponding to each initial activation scope to obtain a set of dominant competitive chains. A squeezing and collapse analysis is performed on the scopes in the dominant competition chain set where conflicts occur between different dominant competition chains at scope affiliation and coupling boundaries to obtain a set of collapsed intermediate scopes. The collapsed intermediate scope set is converged and encapsulated to obtain the converged candidate scope set.

5. The method according to claim 2, characterized in that, The step of folding and mapping the insulation coupling boundary of the partial discharge event according to the candidate insulation domain sequence to obtain the insulation domain data includes: The candidate insulating domain sequence is subjected to boundary suspension processing to obtain a set of boundary fragments to be mapped; Based on the set of boundary fragments to be mapped and the sequence of candidate insulating domains, the boundary pointing relationship of the partial discharge event is borrowed and projected to obtain the boundary projection domain network. A folding and closure analysis is performed on the broken coupling boundaries in the boundary projection domain network to obtain a set of closed boundary domains; Based on the set of closed boundary domains, the scope of the partial discharge event is encapsulated and mapped to obtain the insulation scope data.

6. The method according to claim 5, characterized in that, The step of performing borrowed projection on the boundary pointing relationship of the partial discharge event based on the set of boundary fragments to be mapped and the candidate insulation domain sequence to obtain the boundary projection domain network includes: Each boundary segment in the set of boundary segments to be mapped is subjected to a yield offset process to obtain a set of boundary borrow segments. The misalignment relationship between the boundary borrowing fragment set and the candidate insulation domain sequence is analyzed to obtain the initial set of attached domain pairs; Based on the initial set of attached action domains, the boundary pointing offset relationship of the partial discharge event is tortuously compensated to obtain a set of pointing compensation action domain chains. The set of pointing compensation scope chains is textured and networked to obtain the boundary projection scope network.

7. The method according to claim 1, characterized in that, The process of analyzing the insulation state evolution of the GIS insulation state monitoring location based on the insulation correlation feature data and the insulation domain data, to obtain insulation state characterization quantities and insulation risk potential energy values, includes: The insulation correlation feature data and the insulation domain data are folded to obtain an insulation state folding characterization set. Based on the insulation state folding characterization set, the insulation state evolution chain of the GIS insulation state monitoring location is de-wound and unfolded to obtain the insulation evolution path set; Based on the set of insulation evolution paths, a potential energy accumulation analysis is performed on the risk accumulation relationship during the insulation state evolution process to obtain a risk potential energy distribution set; The insulation state folding characterization set and the risk potential energy distribution set are encapsulated in a dual-quantity manner to obtain the insulation state characterization quantity and the insulation risk potential energy value.

8. The method according to claim 7, characterized in that, The insulation state evolution chain of the GIS insulation state monitoring location is de-wound and unfolded according to the insulation state folding characterization set to obtain an insulation evolution path set, including: The insulation state folding characterization set is subjected to evolution chain suspension processing to obtain the set of chain segments to be unfolded; Based on the set of segments to be unfolded, the folding remnants in the set of insulation state folding characterizations are awakened to obtain the set of activated evolution segments. Based on the set of activated evolution chain segments, the insulation state evolution chain is reverse-connected and expanded to obtain a set of evolution path segments; The evolution path fragment set is encapsulated into path clusters to obtain the insulation evolution path set.

9. The method according to claim 7, characterized in that, The step involves performing potential energy accumulation analysis on the risk accumulation relationship during the insulation state evolution process based on the insulation evolution path set, to obtain a risk potential energy distribution set, including: The path potential difference suspension process is applied to the insulation evolution path set to obtain the potential difference segment set to be accumulated; Based on the set of potential difference segments to be accumulated, the residual voltage of the insulation evolution path set is awakened to obtain the activated accumulation segment set; Based on the set of activated accumulation segments, the risk accumulation slope in the insulation state evolution process is transformed and analyzed to obtain the set of risk transformation segments; Based on the risk transition segment set, the aggregation relationship between potential energy slots is merged to obtain the potential energy slot set; The potential energy slot set is distributed and encapsulated to obtain the risk potential energy distribution set.

10. A multi-mode partial discharge online monitoring device for insulation states, characterized in that, The device includes: The data acquisition module is used to perform cross-modal joint triggering acquisition of partial discharge response signals from GIS insulation condition monitoring points to obtain raw insulation monitoring data; The time extraction module is used to extract events from the suspected partial discharge pulse data in the original insulation monitoring data to obtain an insulation event data set; The fusion analysis module is used to perform fusion analysis on the electromagnetic propagation response, acoustic propagation response, mechanical vibration response, and cross-modal coupling relationship of the partial discharge events at the GIS insulation status monitoring location based on the insulation event data set, so as to obtain insulation correlation feature data. The mapping analysis module is used to perform insulation domain mapping analysis on the partial discharge events of the GIS insulation status monitoring location based on the insulation correlation feature data, and obtain insulation domain data. The evolution analysis module is used to analyze the insulation state evolution process of the GIS insulation state monitoring location based on the insulation correlation feature data and the insulation action domain data, and to obtain insulation state characterization quantities and insulation risk potential energy values.

Images