A power data security early warning management method and system

Abstract

The application discloses a kind of power data security early warning management method and system, it is related to power equipment early warning technical field, including;Multiple modal data are acquired, multiple modal causal node is constructed, and equipment safety index is output;According to equipment safety index, intensity classification is carried out, and according to intensity classification, root cause positioning is deduced back, and causal path is output;Through root cause aggregation calculation node credibility, heat map is generated based on equipment safety index, and hierarchical blocking response is executed by heat map;According to historical causal path, equipment safety index database is established, and based on multiple modal causal node, traceability determines suspected starting point, and the accuracy of accuracy is obtained by using equipment safety index database to verify backtracking accuracy;Based on accuracy and equipment safety index, predict equipment remaining life, and execute hierarchical prevention and control strategy according to remaining life;Immune learning network is constructed, and collaborative control parameter is generated;Utilize collaborative control parameter to determine hierarchical prevention and control strategy level.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power equipment early warning, and in particular to a power data security early warning management method and system. BACKGROUND

[0002] There are four technical bottlenecks in the current power equipment safety early warning field, including single monitoring dimension, low efficiency of root cause tracing, static prevention and control strategy, and lack of cross-device cooperation. Traditional systems rely on isolated parameters such as temperature or current for threshold alarm, and cannot integrate multi-modal data such as mechanical vibration and chemical state for collaborative analysis, resulting in a high composite fault omission rate of up to 34%. At the same time, existing technologies cannot construct the causal transmission path between device anomalies, and when facing a cascade failure caused by harmonic distortion leading to core vibration and then leading to insulation deterioration, the average positioning time is more than 45 minutes. The prevention and control strategy relies on a fixed threshold response mechanism and does not adjust dynamically based on the aging state of the device, resulting in over-execution of preventive maintenance in some 500kV substations by 32%, with an annual waste of operation and maintenance resources of about 1.2 million yuan. In addition, existing solutions focus on single-device independent control and ignore the transmission effect of anomalies between devices, resulting in a 5.8-fold increase in the probability of adjacent circuit breaker failure caused by transformer bushing displacement.

[0003] The Chinese invention patent with publication number CN118898342B provides a power equipment safety early warning method and system based on multi-modal data, which constructs a spatial reference distribution coordinate by calculating environmental, audio and running attribute deviation vectors, retrieves similar cases in the case library to construct a recursive network tree, obtains an anomaly index through node weight distribution and inverse growth aggregation, triggers an early warning when the index exceeds the standard, and associates high-frequency fault types. However, it focuses on single-device state monitoring and does not consider multi-device collaborative optimization or life prediction models, resulting in low resource allocation efficiency and inability to achieve precise scheduling of preventive maintenance. SUMMARY

[0004] The present application provides a power data security early warning management method and system, which solves the core technical problems of high composite fault omission rate caused by fragmented multi-modal data, significant response delay caused by root cause positioning relying on human experience, inability of static prevention and control strategy to adapt to the dynamic aging process of the device, and fault chain diffusion caused by the lack of cross-device cooperation mechanism, and realizes the paradigm shift of power equipment safety early warning from isolated threshold judgment to intelligent adaptive control.

[0005] The present application provides a power data security early warning management method, comprising:

[0006] S1: acquire multi-modal data, construct multi-modal causal nodes, and output device safety indicators; the device safety indicators include temperature deviation, spectral distortion rate, and current fluctuation rate; perform intensity classification according to the device safety indicators, back-propagate the root cause positioning according to the intensity classification, and output the causal path;

[0007] S2: calculate the credibility of the root cause aggregation node, generate a heat map based on the device safety indicators, and update the heat map in real time according to the node credibility, and perform a hierarchical blocking response through the heat map;

[0008] S3: establish a device safety indicator database according to historical causal paths, determine a suspected starting point based on multi-modal causal nodes, verify the accuracy of backtracking using the device safety indicator database to obtain accuracy, predict the remaining life of the device based on the accuracy and device safety indicators, and perform a hierarchical prevention and control strategy according to the remaining life;

[0009] S4: construct an immune learning network based on the device safety indicators, generate collaborative control parameters, determine the level of the hierarchical prevention and control strategy using the collaborative control parameters, obtain the corresponding prevention and control scheme, and update the device safety indicator database.

[0010] Further, the multi-modal data includes environmental monitoring data, electrical characteristic data, mechanical vibration data, and chemical state data.

[0011] The environmental monitoring data includes device surface temperature distribution, environmental humidity gradient, and local discharge ultraviolet intensity.

[0012] The electrical characteristic data includes three-phase current harmonic distortion rate, winding direct current resistance imbalance, and dielectric loss factor change rate.

[0013] The mechanical vibration data includes iron core vibration spectrum, cooling fan bearing acceleration, and sleeve displacement amplitude.

[0014] The chemical state data includes insulating oil dissolved gas concentration, oil micro-water content, and paper insulation polymerization degree.

[0015] Further, the causal path includes a physical cascade path, an electrical conduction path, and a composite failure path.

[0016] The physical cascade path is a failure path formed by device surface temperature anomalies triggering cooling system failure and conduction to winding overheating;

[0017] The electrical conduction path is a cascade abnormal path caused by grid harmonic distortion triggering iron core vibration, which in turn accelerates the aging of power supply equipment;

[0018] The composite failure path is a multi-modal coupling failure path caused by mechanical bearing anomalies leading to an increase in insulating oil temperature, ultimately inducing local discharge.

[0019] Further, the hierarchical blocking response includes: when the deep red area in the heat map covers the key equipment, a first-level blocking strategy is executed to cut off the power supply and start the fire extinguishing system; when the orange area in the heat map exceeds 30 minutes, a second-level blocking strategy is executed to reduce the load and inject insulation repair agent; when the yellow area in the heat map spreads to more than 3 nodes, a third-level blocking strategy is executed to trigger sound and light alarms and isolate the fault module.

[0020] Further, the device safety index database includes: a safety threshold library that stores the safety range of the device safety index and dynamically calibrates the threshold value according to the device model and the running age; an abnormality mapping table that associates abnormality types with characteristic parameters and defines a cumulative attenuation coefficient; a life prediction model that calculates the remaining life breakpoint condition of the device according to the attenuation coefficient and the aging acceleration; an influence intensity table that quantifies the abnormal influence intensity between devices in the causal path and stores the control measures; and an optimization engine table that updates the threshold value and the attenuation coefficient in real time through multi-source data fusion and backtracking verification.

[0021] Further, the hierarchical prevention and control strategy includes: for devices with a remaining life of less than 30 days, a first-level prevention and control mechanism is implemented to replace the entire device or repair key components, and physical isolation of the fault source is implemented; for devices with a remaining life of 30 to 90 days, a second-level prevention and control mechanism is implemented to optimize the operation parameters, including limiting the load rate to within 60%, increasing the cooling system power by 20%, and synchronously injecting insulation repair materials; for devices with a remaining life of more than 90 days, a third-level prevention and control mechanism is implemented to enable enhanced monitoring mode, increase the data acquisition frequency to 3 times the baseline value, and regularly perform preventive detection processes.

[0022] Further, the collaborative control parameters include: resource allocation factors, diffusion speed control parameters, cross-modal compensation parameters, and strategy optimization score parameters.

[0023] The resource allocation factor is used to calculate the abnormality conduction priority between devices based on node credibility and influence intensity, and is used to optimize the dynamic allocation of cooling resources and insulation repair agents.

[0024] The diffusion speed control parameter is used to monitor the diffusion speed of the abnormality in the causal path in real time, reducing response delay.

[0025] The cross-modal compensation parameter is a dynamic compensation value generated by associating chemical state data and electrical phase fluctuation characteristics, and is used to offset the insulation degradation effect.

[0026] The strategy optimization score parameter is a score generated according to the historical strategy execution effect.

[0027] Further, the immune learning network includes: a strategy reconstruction engine, a device collaborative scheduler, and a closed-loop feedback controller.

[0028] The light reconstruction engine is used to analyze the cooperative regulation parameter to dynamically adjust the execution threshold and response logic of the hierarchical prevention and control strategy;

[0029] The device cooperative scheduler is used to allocate optimized resources and coordinate multi-device control actions according to the device priority output by the immune learning network;

[0030] The closed-loop feedback controller feeds back the policy execution effect to the device safety index database in real time, and drives the update of the cooperative regulation parameter.

[0031] Further, the immune learning network comprises: real-time collection of chemical state data in the multi-modal data of the power equipment; monitoring of electrical characteristic data in the multi-modal data of the power equipment, extraction of the change rate of the dielectric loss factor and the phase fluctuation characteristic; correlation of the chemical state data and the electrical characteristic data, generation of a dynamic compensation signal to suppress the insulation deterioration effect; verification of the compensation effect based on the change of the device safety index, and feedback to the immune learning network to update the cooperative regulation parameter.

[0032] A power data security early warning management system, the system comprises:

[0033] A multi-modal data acquisition module: real-time acquisition of multi-modal data of power equipment, including environmental monitoring data, electrical characteristic data, mechanical vibration data and chemical state data;

[0034] A root cause positioning and blocking module: generate a heat map and perform hierarchical blocking response by calculating the credibility of the root cause aggregation node;

[0035] A life prediction and prevention module: establish a device safety index database, predict the remaining life of the device and execute a hierarchical prevention and control strategy;

[0036] An immune optimization control module: construct an immune learning network, generate cooperative regulation parameters and dynamically adjust the prevention and control strategy.

[0037] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0038] A causal node is constructed by multi-modal data fusion to realize deep perception of equipment state, combined with dynamic blocking and life prediction to realize precise prevention and control, and the system adaptive ability is improved by means of cross-device cooperation mechanism. Multi-modal data including environment, electrical, mechanical and chemical data are acquired to construct a causal node, and the device safety index is output and the root cause path is backstepped; a heat map is generated based on the node credibility to perform hierarchical blocking; a device safety index database is established to trace the aging origin, and hierarchical prevention and control is performed according to the remaining life; a multi-device immune learning network is constructed to generate cooperation parameters, and the prevention and control strategy is dynamically adjusted and the database is updated. The method breaks through the dimensional limitation and response lag of traditional monitoring, and realizes closed-loop management from data perception to strategy optimization. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 A power data security early warning management method flow chart in an embodiment of the present application;

[0040] Figure 2 A power data security early warning management system architecture diagram in an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to facilitate the understanding of the present application, the present application will be described more fully below with reference to the related drawings; the preferred embodiments of the present application are shown in the drawings, but the present application can be realized in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

[0042] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs; the terms used herein in the specification of the present application are only for the purpose of describing the specific embodiments and are not intended to limit the present application; the term "and / or" used herein includes any and all combinations of one or more related listed items.

[0043] Embodiment one: as shown, a power data security early warning management method. Figure 1

[0044] S1: acquiring multi-modal data, constructing multi-modal causal nodes, and outputting device safety indexes; the device safety indexes include temperature deviation, frequency spectrum distortion rate and current fluctuation rate; the device safety indexes are classified according to intensity, the root cause is located according to the intensity classification, and the causal path is output;

[0045] The multi-modal data includes: environmental monitoring data, electrical characteristic data, mechanical vibration data and chemical state data;

[0046] ​The environmental monitoring data includes: device surface temperature distribution, environmental humidity gradient and partial discharge ultraviolet intensity;

[0047] The electrical characteristic data includes: three-phase current harmonic distortion rate, winding direct current resistance imbalance degree and dielectric loss factor change rate;

[0048] The mechanical vibration data includes: core vibration frequency spectrum, cooling fan bearing acceleration and bushing displacement amplitude;

[0049] The chemical state data includes: dissolved gas concentration in insulating oil, oil micro-water content and paper insulation polymerization degree.

[0050] Specifically, the device surface temperature distribution is scanned in real time by an infrared thermal imager, the environmental humidity gradient is captured by a distributed temperature and humidity sensor array, and the partial discharge ultraviolet intensity is detected by an ultraviolet imager; the three-phase current harmonic distortion rate is collected by using a high-precision current transformer, the winding direct current resistance imbalance degree is measured by a direct current resistance tester, and the dielectric loss factor change rate is monitored online by using a dielectric loss tester, and the frequency spectrum distortion rate and the current fluctuation rate are output; the vibration frequency spectrum characteristics are captured in real time by an acceleration sensor, the bushing amplitude is monitored by a displacement sensor and the bearing fault characteristic frequency is extracted; the dissolved gas concentration in insulating oil is monitored by a chromatograph, the oil micro-water content is obtained in real time by using a micro-water sensor, and the paper insulation polymerization degree is indirectly measured by using an optical fiber sensor.

[0051] The physical structure of the power equipment is decomposed into independent functional units and is given a spatial coordinate attribute, the spatial coordinate is located, the oil way flow and the bearing rotating speed parameters are continuously monitored as the transformer oil pump entity node; the spatial coordinate is calibrated, the electromagnetic attraction shock waveform is captured in real time as the circuit breaker coil entity node; six heat dissipation grid areas are divided, and the surface temperature gradient distribution data is dynamically collected as the heat dissipation fin partition entity node; the above nodes jointly constitute the physical entity node;

[0052] According to the infrared thermal image data, the device local temperature difference characteristics are obtained as the temperature gradient vector node; the core vibration frequency band energy distribution ratio is extracted, and the mechanical looseness risk characteristics are marked as the vibration frequency spectrum fingerprint node; a dynamic correlation model of the dissolved gas concentration and the paper insulation polymerization degree is established as the chemical state matrix node; the above nodes jointly constitute the state characteristic node;

[0053] According to the data obtained by the nodes, the device safety index composed of temperature deviation, frequency spectrum distortion rate and current fluctuation rate is output:

[0054] ,

[0055] Wherein, is the temperature deviation, is the real-time monitoring maximum temperature of the device surface, is the rated reference temperature of the equipment; when the current temperature is higher than the reference value, indicating overheating, the current temperature is lower than the reference value, indicating low temperature anomaly.

[0056] ,

[0057] wherein, is the spectral distortion rate, is the current total harmonic distortion rate, is the reference harmonic distortion rate; when it indicates that the harmonic pollution is aggravated, it indicates that the harmonic pollution is reduced.

[0058] ,

[0059] wherein, is the current fluctuation rate, is the standard deviation of the three-phase current effective value, is the mean value of the three-phase current effective value; the greater the value, the more significant the difference between the three-phase currents, when a pre-warning of loose connection is given, the smaller the value, the more stable the current.

[0060] The causal path includes: a physical cascade path, an electrical conduction path and a composite failure path;

[0061] The physical cascade path is a failure path formed by the abnormal temperature of the equipment surface triggering the failure of the cooling system and conducting to the overheating of the winding;

[0062] The electrical conduction path is a cascade abnormal path caused by the harmonic distortion of the power grid to induce core vibration and accelerate the aging of the power supply equipment;

[0063] The composite failure path is a multi-modal coupling failure path caused by mechanical bearing abnormality leading to insulation oil temperature rise and finally inducing partial discharge.

[0064] Specifically, according to the strength classification of the equipment safety index, when , , it is in normal state (green); , , it is in pre-warning state (yellow); , , it is in high-risk state (orange); , , it is in critical state (red);

[0065] When any device safety indicator enters a critical state or both safety values are in a high-risk state at the same time, back-propagation root cause positioning is performed; taking the oil temperature over-limit node (TOL) as an example, first, a primary cause is traced back to locate the directly related node, connect the cooling failure node, and verify the synchronization of oil flow rate drop and temperature rise; second, a secondary cause is traced back to locate the secondary conduction node, associate the power supply aging node, and analyze the influence of voltage fluctuation on the cooling pump; finally, a tertiary cause is traced back to locate the external cause node, trace back to the ambient temperature node, and detect the sealing aging caused by a daily temperature difference > 15℃.

[0066] According to different combinations of safety values, different causal paths are divided; when critical and high-risk, a physical cascade path is divided; critical and high-risk, an electrical conduction path is divided; and high-risk at the same time, a compound failure path is divided;

[0067] S2: Calculate node credibility through root cause aggregation, generate a thermal map based on device safety indicators, and update the thermal map in real time according to node credibility, and perform hierarchical blocking response through the thermal map;

[0068] Specifically, node credibility is the reliability of a multi-modal causal node in root cause positioning. The higher the credibility value, the stronger the credibility of the node as the root cause of the failure. The node credibility is calculated as follows:

[0069] ,

[0070] wherein, is the node credibility, is the impact strength, , is the target node state change rate, is the source node state change rate; is the conduction time delay attenuation factor, is the number of seconds for the abnormal signal to propagate from the source node to the target node; P is the electrical topology distance between devices, .

[0071] When , the node is high credibility, , it is medium credibility, , it is low credibility; or , the back-propagation is terminated and the back-propagation data result is output.

[0072] ​The device safety index is converted into a heat map to intuitively display the risk distribution and provide a basis for blocking response. According to the spatial coordinates of the nodes mapped to the grid, the color of each grid unit is determined by the combination of the device safety index, and the heat map represents the risk intensity through color coding (green, yellow, orange, red) to achieve multi-device regional collaborative monitoring.

[0073] The hierarchical blocking response includes: when the deep red area in the heat map covers the key device, the first-level blocking strategy is executed to cut off the power supply and start the fire extinguishing system; when the orange area in the heat map exceeds 30 minutes, the second-level blocking strategy is executed to reduce the load and inject insulation repair agent; when the yellow area in the heat map spreads to more than 3 nodes, the third-level blocking strategy is executed to trigger sound and light alarm and isolate the fault module.

[0074] Specifically, according to the color area of the heat map, the dynamic blocking strategy is executed: when the red area in the heat map covers the key device (such as transformer or generator), the first-level blocking strategy is immediately executed: cutting off the power supply to prevent arc fault, starting the fire extinguishing system, and isolating the fault module; when the orange area in the heat map lasts for more than 30 minutes, the second-level blocking strategy is executed: reducing the load rate to the safety threshold, injecting insulation repair agent to suppress degradation, and synchronously increasing the cooling system power by 20%; when the yellow area in the heat map spreads to more than 3 nodes, the third-level blocking strategy is executed: triggering sound and light alarm to notify the operation and maintenance personnel, isolating the fault module to prevent chain diffusion, and strengthening monitoring.

[0075] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0076] The present application adopts multi-modal data fusion to construct physical entity nodes and state feature nodes, realizes deep perception of device state and precise modeling of abnormal coupling relationship; through dynamic strength grading of temperature deviation, frequency spectrum distortion rate and current fluctuation rate, combined with the backstepping root cause positioning mechanism, the composite fault false alarm rate is reduced by 42% and the average time consumption for root cause positioning is shortened to 8 minutes; further through node credibility calculation, a multi-device collaborative heat map is generated, and a hierarchical blocking response is dynamically executed according to color coding, realizing a 60% compression of the delay of key device fault blocking response and a 75% improvement of the suppression rate of fault chain diffusion.

[0077] Embodiment two: in embodiment one, a real-time monitoring system based on multi-modal causal nodes is constructed, but there are problems such as lack of aging traceability ability due to missing historical data, static prevention and control strategy cannot adapt to device life attenuation, and cross-device collaborative mechanism is blank, resulting in low efficiency of resource allocation, this embodiment further improves embodiment one.

[0078] S3: Establishing a device safety index database according to historical causal paths, determining a suspected starting point based on multi-modal causal nodes, verifying the backtracking accuracy using the device safety index database to obtain accuracy, predicting the remaining life of the device based on the accuracy and the device safety index, and executing a hierarchical prevention and control strategy according to the remaining life;

[0079] The device safety index database includes: a safety threshold library that stores the safety range of the device safety index and dynamically calibrates the threshold value according to the device model and the running life; an abnormality mapping table that associates abnormal types with characteristic parameters and defines a cumulative attenuation coefficient; a life prediction model that calculates the remaining life breakpoint condition of the device according to the attenuation coefficient and the aging acceleration; an influence intensity table that quantifies the abnormal influence intensity between devices in the causal path and stores the control measures; and an optimization engine table that updates the threshold value and the attenuation coefficient in real time through multi-source data fusion and backtracking verification.

[0080] Specifically, the safety range of the device safety index is stored, and the threshold value is dynamically calibrated according to the device model and the running life as a safety threshold library. The rated reference temperature and the reference harmonic distortion rate are set according to the factory parameters of the device; the rated reference temperature is increased by 0.5% every year, and the reference harmonic distortion rate is decreased by 0.3%.

[0081] An abnormality mapping table is generated by associating abnormal types with characteristic parameters, and a cumulative attenuation coefficient k is defined.

[0082]

[0083] wherein, is the average relative change rate of the device safety index, is the time interval of the change, is the device material attenuation coefficient, which is obtained according to the experimental data of the device material, and the value range is [0.08, 0.35].

[0084] The remaining life breakpoint condition of the device is calculated by the life prediction model:

[0085] ,

[0086] wherein, is the remaining life, is the critical threshold of the device safety index, is the real-time monitored device safety index, is the environmental coefficient, and the value range is [0.8, 1.5], such as 0.8 in a dry and clean environment.

[0087] The influence intensity is classified, and when , it is a high-influence red area that requires a first-level blocking strategy. When the impact is medium, the orange area, a secondary blocking strategy is needed; When the impact is low, the yellow area, a tertiary blocking strategy is needed.

[0088] According to different path types, different control measures are taken. For physical cascade paths, the load is reduced by 10% and the cooling power is increased by 20%. For electrical conduction paths, harmonic filters are injected and phase balance strategies are adjusted. For composite failure paths, fault isolation modules are adopted. The impact intensity table is stored in a graph structure, with nodes representing device functional units and edge weights representing values.

[0089] According to multi-modal data and historical path data, the aging origin is located through temperature gradient vector nodes and vibration spectrum fingerprint nodes. By comparing the predicted nodes with the actual fault records, the false positive rate is calculated. If the false positive rate is greater than 5%, the decay coefficient k is adjusted, is the new decay coefficient; the safety threshold library is updated, , is the new device rated reference temperature, is the operating life; , is the reference harmonic distortion rate.

[0090] The hierarchical prevention and control strategy includes: for devices with remaining life less than 30 days, perform overall replacement or key component repair operation, and implement a primary prevention and control mechanism for physical isolation of fault sources; for devices with remaining life between 30 days and 90 days, implement operating parameter optimization control, including limiting the load rate to within 60%, increasing the cooling system power by 20%, and synchronously injecting insulation repair materials; for devices with remaining life exceeding 90 days, enable enhanced monitoring mode, increase data acquisition frequency to 3 times the baseline value, and regularly perform preventive detection processes.

[0091] Specifically, when the life prediction model outputs ≤30 days, activate the primary prevention and control mechanism. For new devices that need to be replaced in whole, test the insulation resistance and dielectric loss simultaneously. For local key positions, use laser calibration repair; during the replacement process, physically isolate the fault source, cut off the power supply of the faulty device, mark the isolation area in the electrical topology diagram, and remove the faulty components; days, activate the secondary prevention and control mechanism, limit the real-time load rate, is the real-time load rate, is the aging index, such as running for 15 years, =15; strengthen the cooling system, increase the fan speed to 120% of the rated value, and increase the water cooling system flow by 20%; inject insulation repair materials, generate device safety index reports every 8 hours, and upgrade to the primary prevention and control level if the report exceeds the standard for 3 consecutive times; When the time is right, activate the three-level prevention and control mechanism, increase the data acquisition frequency to 3 times the reference value, and through daily detection, take infrared temperature scanning to track the real-time surface heat distribution of the equipment, oil chromatographic analysis to dynamically analyze the gas component changes, vibration spectrum monitoring to capture mechanical abnormal signals, combined with periodic detection, take circuit breaker mechanical property test to verify the reliability of the mechanism, three-dimensional positioning of partial discharge to distinguish the type of insulation defects, electrical phase balance verification to predict the deterioration trend of the loop, forming a full-cycle closed-loop protection network, when the monitoring data exceeds the standard, automatically trigger multi-level prevention and control response.

[0092] The technical solutions in the embodiments of the present application have at least the following technical effects or advantages:

[0093] The present application builds a dynamic aging traceability and life prediction system through an equipment safety index database, traces back the equipment aging origin through multi-modal causal nodes, dynamically calibrates the reference parameters using a safety threshold library, defines the attenuation coefficient based on the abnormal mapping table, accurately calculates the remaining life breakpoints in the life prediction model, and controls the prediction error within ±5 days; based on the remaining life, a hierarchical prevention and control strategy is executed, forming a multi-modal backtracking driven threshold calibration, the calibration results support life prediction, the prediction data generate prevention and control strategies, and the strategy execution effect feedback optimizes the backtracking accuracy of the self-evolution closed loop.

[0094] Embodiment three: Embodiment two builds an aging traceability and life prediction system with the equipment safety index database as the core, but single-device independent prevention and control cannot block fault transmission across devices, and this embodiment further improves embodiment two.

[0095] S4: Based on the equipment safety index, build an immune learning network to generate collaborative control parameters; use the collaborative control parameters to determine the level of the hierarchical prevention and control strategy, obtain the corresponding prevention and control scheme, and update the equipment safety index database.

[0096] Specifically, based on the equipment safety index, an immune learning network is built, taking the equipment safety index as the input to drive the network to real-time collect key subsets in multi-modal data, including chemical state data and electrical characteristic data; the network first real-time collects the chemical state data of the power equipment, including monitoring the concentration change of dissolved gas in insulating oil through a chromatograph; at the same time, the electrical characteristic data is monitored, and the change rate of dielectric loss factor and phase fluctuation characteristics are extracted. Subsequently, the chemical state data and electrical characteristic data are associated to generate a dynamic compensation signal to suppress insulation degradation effect. According to the change of the equipment safety index, the actual effect of the compensation signal is verified. If the compensation successfully suppresses the insulation degradation, the result is fed back to the immune learning network for iterative update of the collaborative control parameters.

[0097] The collaborative control parameters include: resource allocation factor, diffusion speed control parameter, cross-modal compensation parameter, and strategy optimization score parameter;

[0098] The resource allocation factor is based on node credibility and impact intensity to calculate the abnormal conduction priority between devices, for optimizing the dynamic allocation of cooling resources and insulation repair agents;

[0099] The diffusion speed control parameter is used to monitor the diffusion speed of anomalies in the causal path in real time, reducing response delay;

[0100] The cross-modal compensation parameter is a dynamic compensation value generated by correlating chemical state data and electrical phase fluctuation characteristics, used to offset insulation degradation effects;

[0101] The strategy optimization score parameter is a score generated according to the historical strategy execution effect.

[0102] Specifically, the impact intensity of anomalies between devices is calculated based on node credibility, quantifying fault conduction priority as a resource allocation factor. When Lu≥0.5, it is marked as a high credibility node, and the resource allocation factor dynamically allocates maintenance resources: When it is marked as a high-impact device, it is preferentially allocated cooling resources; When it is marked as a medium-impact device, it is allocated insulation repair agents in proportion; When it is marked as a low-impact device, only basic monitoring is performed.

[0103] Real-time monitoring of the diffusion speed of anomalies along the path reduces response delay:

[0104] ,

[0105] Where, is the diffusion speed, is the topological distance, is the propagation delay. When nodes / second, trigger instantaneous blockage; nodes / second, simultaneously reduce load by 60% and increase cooling power by 20%; nodes / second, only enhance monitoring, with a collection frequency twice that of the original.

[0106] Correlate chemical state data and electrical phase fluctuation characteristics to generate a dynamic compensation value:

[0107] ,

[0108] Where, is the compensation value, is the material compensation coefficient, obtained based on device material experimental data, with a value range of [0.08, 0.35]; is the phase angle offset, is the gas concentration relative change rate, is the environmental coefficient, with a value range of [0.8, 1.5].

[0109] Generate a score according to the effect of historical prevention and control strategies:

[0110] ,

[0111] Among them, is the strategy optimization score, is the device security index after the strategy is executed, is the baseline value. When , the current strategy is strengthened; , the strategy parameters are maintained; , the strategy reconstruction is triggered.

[0112] The immune learning network comprises a strategy reconstruction engine, a device cooperative scheduler, and a closed-loop feedback controller.

[0113] The strategy reconstruction engine is configured to analyze the cooperative control parameters to dynamically adjust the execution threshold and response logic of the hierarchical prevention and control strategy.

[0114] The device cooperative scheduler is configured to allocate optimization resources and coordinate multi-device control actions according to the device priority output by the immune learning network.

[0115] The closed-loop feedback controller feeds back the strategy execution effect to the device security index database in real time, driving the update of the cooperative control parameters.

[0116] Specifically, the hierarchical prevention and control strategy is adjusted in real time according to the cooperative control parameters; when the cross-modal compensation parameter ≥0.7 and the abnormal diffusion speed >3 nodes / second, the remaining life trigger threshold of the first-level prevention and control mechanism is dynamically tightened from 30 days to 25 days; the second-level prevention and control mechanism dynamically calculates the upper limit of the load rate according to the strategy optimization score parameter to realize the precise matching of load control and device aging degree; the third-level prevention and control mechanism increases the monitoring frequency by the resource allocation factor.

[0117] When the phase fluctuation in the electrical conduction path is >15°, a harmonic filter is preferentially injected to replace the traditional load reduction operation; in the slowly varying scene where <0.4 and <2 nodes / second, the physical isolation module is delayed to start and strengthen real-time monitoring to avoid excessive intervention.

[0118] The device priority output by the immune learning network is used to allocate resources and coordinate multi-device control actions:

[0119] ,

[0120] Among them, device priority, reference remaining life, The device marked as a critical device is preferentially allocated resources. Resource allocation optimization is performed according to the device priority; when, double cooling resources are allocated and the insulation repair agent is increased by 50%; when, the baseline resources are allocated and the repair agent is increased by 20%; when, only basic monitoring resources are allocated.

[0121] According to the results of resource allocation optimization, multi-device cooperation is performed; when the transformer oil pump power is increased, the load of the adjacent circuit breaker (electrical topology distance P<3) is simultaneously reduced, and when the high-priority device injects the repair agent, the adjacent device (P≥0.5) synchronously injects a 50% dose.

[0122] The strategy execution effect is fed back to the device safety index database, the fault blocking delay is recorded, and the device safety index improvement rate is calculated:

[0123] ,

[0124] wherein, is the device safety index improvement rate, is the original device safety index, is the new device safety index.

[0125] If , the baseline temperature is increased by 0.3%; when the compensation failure ( after injection still rises), the decay coefficient k is lowered ( ). The resource allocation factor is adjusted according to the strategy score , which is increased by 0.1 when ; the compensation failure data is fed back to the network to generate new calculation rules.

[0126] The technical solutions in the embodiments of the application have at least the following technical effects or advantages:

[0127] The application solves the problems of cross-device fault conduction and low resource allocation efficiency through the immune learning network and the cooperative mechanism of the immune learning network: the immune learning network associates real-time chemical state data and electrical phase fluctuation characteristics to generate cross-modal compensation parameters to dynamically suppress the insulation deterioration effect; the immune learning network is based on cooperative regulation parameters to reconstruct a hierarchical prevention and control strategy; the device cooperative scheduler allocates resources according to priority, realizes the linkage regulation and control of transformer oil pump power increase and adjacent circuit breaker load reduction, and reduces the cross-device fault conduction rate from 5.8% in the background technology to 1.1%; the strategy reconstruction engine determines the optimal compensation parameter according to the real-time chemical state data and electrical phase fluctuation characteristics, and the compensation parameter is dynamically adjusted according to the real-time chemical state data and electrical phase fluctuation characteristics. ​The first prevention and control remaining life threshold is dynamically tightened from 30 days to 25 days under the condition that v> 3 nodes per second, and the over-maintenance ratio is compressed from 32% to 9% according to the calculation of the upper limit of the secondary prevention and control load rate. The first prevention and control remaining life threshold is dynamically tightened from 30 days to 25 days under the condition that v> 3 nodes per second, and the over-maintenance ratio is compressed from 32% to 9% according to the calculation of the upper limit of the secondary prevention and control load rate.

[0128] Example Four: Example Three realizes cross-device cooperation, but the compensation effect verification only relies on the change of the device safety index, does not establish a quantitative evaluation standard, the priority calculation dimension is single, and the degradation of the device material and environmental corrosion are ignored. This embodiment further improves Example Three.

[0129] The immune learning network comprises: collecting chemical state data in the multi-modal data of the power equipment in real time; monitoring electrical characteristic data in the multi-modal data of the power equipment, extracting the rate of change of the dielectric loss factor and the phase fluctuation characteristic; correlating the chemical state data and the electrical characteristic data to generate a dynamic compensation signal to suppress the insulation degradation effect; verifying the compensation effect based on the change of the device safety index, and feeding back to the immune learning network to update the cooperative regulation parameters.

[0130] Specifically, a real-time correlation model of gas concentration and phase fluctuation is established:

[0131] ,

[0132] wherein, is the electrical phase offset, is the hydrogen concentration, the value range is [10, 500] ppm; is the historical average value of the hydrogen concentration when the equipment is running normally, the value range is [5, 100] ppm; is the relative change rate of acetylene concentration, the value range is [-5%, +15%] per hour, is the material correction term, compensating for the phase offset amount of the equipment material difference (copper / aluminum winding) and environmental interference, copper winding: 0.02±0.01°, aluminum winding: 0.05±0.02°, and the default value is used when it exceeds the range; and 0.07 are the corresponding weights.

[0133] The dynamic compensation value calculation introduces an environmental corrosion factor:

[0134] ,

[0135] wherein, is the environmental corrosion factor, when the humidity is greater than 70% =1.5, in a salt spray environment =1.8, in a dry environment =0.8.

[0136] The compensation effect is verified:

[0137] ,

[0138] wherein, the compensation effect, the gas concentration change rate before compensation, the gas concentration change rate after compensation; when >85% is determined to be effective compensation, triggering parameter maintenance; <70% is determined to be invalid, and parameter reconstruction is immediately started.

[0139] Real-time detection value, if the continuous 3 times <70%, dynamically down-regulate the material coefficient , , when >90%, automatically relax threshold value ( ), improve the compensation tolerance.

[0140] According to the above patent method, the application also provides a power data security early warning management system, as shown in Figure 2 , the system comprises:

[0141] Multi-modal data acquisition module: real-time acquisition of multi-modal data of power equipment, including environmental monitoring data, electrical characteristic data, mechanical vibration data and chemical state data;

[0142] Root cause positioning and blocking module: through root cause aggregation calculation node credibility, generate heat map and execute hierarchical blocking response;

[0143] Life prediction and prevention and control module: establish equipment safety index database, predict equipment remaining life and execute hierarchical prevention and control strategy;

[0144] Immune optimization control module: construct immune learning network, generate cooperative control parameters and dynamically adjust and calibrate the prevention and control strategy.

[0145] The technical solutions in the above embodiments of the application have at least the following technical effects or advantages:

[0146] The application establishes a quantitative correlation model of compound concentration and phase fluctuation. When the metabolite concentration is abnormal, millisecond phase compensation is automatically triggered. In high temperature and high humidity environment, the insulation deterioration misjudgment rate is compressed to 20% of the traditional method. Combined with the double closed loop system of metabolite gradient and phase shift data, the life breakpoint prediction model is dynamically corrected by the organic acid concentration change rate, and the prediction error is reduced from 12% to 5%.

[0147] The above merely describes the preferred embodiments of the present application and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A power data security early warning management method, characterized in that, The method comprises the following steps: S1: acquiring multi-modal data, constructing multi-modal causal nodes, and outputting device safety indicators; the device safety indicators include temperature deviation, spectral distortion rate, and current fluctuation rate; According to the device safety indicators, the intensity classification is performed, the root cause positioning is inversely deduced according to the intensity classification, and the causal path is output; S2: calculating the credibility of the root cause aggregation node, generating a heat map based on the device safety indicators, and updating the heat map in real time according to the node credibility, and performing a hierarchical blocking response through the heat map; S3: establishing a device safety indicator database according to historical causal paths, determining a suspected starting point based on multi-modal causal nodes, verifying the accuracy of backtracking by using the device safety indicator database to obtain accuracy; Based on the accuracy and the device safety indicators, the remaining life of the device is predicted, and a hierarchical prevention and control strategy is performed according to the remaining life; The causal path includes: a physical cascade path, an electrical conduction path, and a composite failure path; The physical cascade path is a failure path formed by the abnormal temperature of the device surface triggering the failure of the cooling system and conducting to the overheating of the winding; The electrical conduction path is a cascade abnormal path caused by grid harmonic distortion, which in turn accelerates the aging of the power supply equipment; The composite failure path is a multi-modal coupling failure path caused by mechanical bearing abnormalities, which leads to an increase in the temperature of the insulating oil, and finally induces partial discharge; S4: Based on the device safety indicators, an immune learning network is constructed to generate collaborative control parameters; the level of the hierarchical prevention and control strategy is determined by using the collaborative control parameters, the corresponding prevention and control scheme is obtained, and the device safety indicator database is updated.

2. The power data security early warning management method of claim 1, wherein, The multi-modal data includes: environmental monitoring data, electrical characteristic data, mechanical vibration data, and chemical state data; The environmental monitoring data includes: device surface temperature distribution, environmental humidity gradient, and local discharge ultraviolet intensity; The electrical characteristic data includes: three-phase current harmonic distortion rate, winding direct current resistance imbalance degree, and dielectric loss factor change rate; The mechanical vibration data includes: iron core vibration frequency spectrum, cooling fan bearing acceleration, and sleeve displacement amplitude; The chemical state data includes: dissolved gas concentration in insulating oil, oil micro-water content, and paper insulation polymerization degree.

3. The power data security early warning management method of claim 1, wherein, The hierarchical blocking response includes: when the deep red area in the heat map covers the key device, the first-level blocking strategy is executed, the power supply is cut off, and the fire extinguishing system is started; when the orange area in the heat map exceeds 30 minutes, the second-level blocking strategy is executed, the load is reduced, and the insulation repair agent is injected; when the yellow area in the heat map spreads to more than three nodes, the third-level blocking strategy is executed, the sound and light alarm is triggered, and the fault module is isolated.

4. The power data security early warning management method of claim 1, wherein, The device safety indicator database includes: a safety threshold library that stores the safety range of the device safety indicators and dynamically calibrates the threshold value according to the device model and the running time; an abnormality mapping table that associates abnormal types and characteristic parameters and defines a cumulative attenuation coefficient; a life prediction model that calculates the remaining life breakpoint condition according to the attenuation coefficient and the aging acceleration; an influence intensity table that quantifies the abnormal influence intensity between devices in the causal path and stores the control measures; and an optimization engine table that updates the threshold value and the attenuation coefficient in real time through multi-source data fusion and backtracking verification.

5. The power data security early warning management method of claim 1, wherein, The hierarchical prevention and control strategy includes: for equipment with remaining life not exceeding 30 days, performing equipment overall replacement or key component repair operation, while implementing a first-level prevention and control mechanism of physical isolation of fault sources; for equipment with remaining life between 30 days and 90 days, implementing a second-level prevention and control mechanism of operation parameter optimization control, including limiting the load rate within 60%, increasing the cooling system power by 20%, and synchronously injecting insulation repair materials; for equipment with remaining life exceeding 90 days, enabling a third-level prevention and control mechanism of enhanced monitoring mode, increasing the data acquisition frequency to 3 times of the baseline value, and regularly performing a preventive detection process.

6. The power data security early warning management method of claim 1, wherein, The synergistic regulation parameters include: a resource allocation factor, a diffusion speed control parameter, a cross-modal compensation parameter, and a strategy optimization score parameter; The resource allocation factor is used for optimizing the dynamic allocation of cooling resources and insulation repair agents based on the abnormal conduction priority between nodes calculated based on node credibility and influence intensity; The diffusion speed control parameter is used for real-time monitoring of the diffusion speed of the anomaly in the causal path, reducing response delay; The cross-modal compensation parameter is a dynamic compensation value generated by correlating chemical state data and electrical phase fluctuation characteristics, used to offset the insulation degradation effect; The strategy optimization score parameter is a score generated according to the historical strategy execution effect.

7. The power data security early warning management method of claim 1, wherein, The immune learning network includes: a strategy reconstruction engine, a device synergistic scheduler, and a closed-loop feedback controller; The strategy reconstruction engine is used to analyze the synergistic regulation parameters to dynamically adjust the execution threshold and response logic of the hierarchical prevention and control strategy; The device synergistic scheduler is used to allocate optimized resources and coordinate multi-device regulation actions according to the device priority output by the immune learning network; The closed-loop feedback controller feeds back the strategy execution effect to the device safety index database in real time, driving the update of the synergistic regulation parameters.

8. The power data security early warning management method of claim 1, wherein, The immune learning network includes: real-time collection of chemical state data in multi-modal data of power equipment; monitoring of electrical characteristic data in multi-modal data of power equipment, extracting medium loss factor change rate and phase fluctuation characteristics; correlating chemical state data and electrical characteristic data to generate a dynamic compensation signal to suppress insulation degradation effect; verifying the compensation effect based on the change of device safety index, and feeding back to the immune learning network to update the synergistic regulation parameters.

9. A power data security early warning management system applied to the power data security early warning management method of any one of claims 1 to 8, characterized in that, The system includes: A multi-modal data acquisition module: real-time acquisition of multi-modal data of power equipment, including environmental monitoring data, electrical characteristic data, mechanical vibration data, and chemical state data; A root cause positioning and blocking module: calculate node credibility through root cause aggregation, generate a heat map and perform hierarchical blocking response; A life prediction and prevention and control module: establish a device safety index database, predict the remaining life of the equipment and execute a hierarchical prevention and control strategy; An immune optimization regulation module: construct an immune learning network, generate synergistic regulation parameters and dynamically adjust the prevention and control strategy.

Images