A primary and secondary fusion ring network box state monitoring system based on digital twinning

Abstract

The application discloses a primary and secondary fusion ring network box state monitoring system based on digital twinning, relates to the technical field of ring network box state monitoring, and comprises a primary and secondary acquisition, digital twinning modeling, fusion analysis and output control unit; the fusion analysis unit is integrated with baseline construction, single-parameter pre-judgment, adaptive diagnosis, chain deduction and closed-loop self-evolution modules. The application builds a primary and secondary joint baseline library, completes three-level grading pre-control by means of parameter deviation rate and deterioration slope, dynamically adjusts the fusion weight according to the equipment state dimension, improves the diagnosis accuracy by combining physical constraints and multi-physical field simulation cross verification, deduces the risk along the multi-type fault propagation link, divides the chain fault level by means of cosine similarity, and continuously self-optimizes through the three-layer mechanism of daily baseline, monthly weight and fault trigger link. The application can identify primary and secondary coupling hidden faults, reduce false positives and false negatives, pre-judge faults, prevent and control chain risks, and adapt to the intelligent monitoring of the whole life cycle of the ring network box.

Description

Technical Field

[0001] This invention relates to the field of ring network box condition monitoring technology, and in particular to a primary and secondary fusion ring network box condition monitoring system based on digital twins. Background Technology

[0002] With the advancement of intelligent power distribution network construction, integrated primary and secondary ring main units have become core equipment in urban power distribution network nodes due to their compact structure and high integration. Currently, ring main unit status monitoring generally adopts the traditional solution of discrete sensors + threshold alarms, while some pilot projects have introduced digital twin technology to improve the level of visualization.

[0003] Existing solutions suffer from four progressive technical shortcomings in practical applications: At the data level, primary and secondary data are collected and analyzed independently, making it impossible to identify cross-system hidden faults such as heat conduction from contacts to secondary terminals or transient interference coupling sampling circuits, resulting in a high rate of missed detection for coupled defects; at the diagnostic level, they rely on single parameter thresholds or simple weighted judgments, lacking physical consistency verification, making them susceptible to false alarms due to electromagnetic interference and load fluctuations, and incurring high operation and maintenance verification costs; at the early warning level, they only provide post-event alarms for exceeding limits, lacking the ability to quantitatively assess degradation trends and extrapolate cascading fault propagation, thus failing to support early proactive prevention and control; at the long-term level, baseline thresholds and fusion weights are fixed at the factory, and diagnostic accuracy continuously declines with equipment aging, environmental drift, and component replacement, lacking an autonomous iterative optimization mechanism.

[0004] Existing digital twin monitoring solutions mostly remain at the level of 3D visualization and data display, and have not yet formed a complete technical chain of "primary and secondary joint baseline construction, adaptive fusion diagnosis, chain topology deduction, and hierarchical closed-loop self-evolution", which makes it difficult to meet the actual needs of high reliability and refined monitoring of primary and secondary integrated ring network boxes throughout their entire life cycle. Summary of the Invention

[0005] In view of this, this application provides a primary and secondary integrated ring network box status monitoring system based on digital twins to address the shortcomings of existing technologies.

[0006] The first aspect of this application provides a primary and secondary fusion ring network box status monitoring system based on digital twins, including a primary data acquisition unit, a secondary data acquisition unit, a digital twin modeling unit, a fusion analysis unit, and an output control unit. The fusion analysis unit includes a primary and secondary joint baseline construction module, a single parameter trend prediction module, an adaptive fusion diagnosis module, a cascading fault inference and verification module, and a closed-loop self-evolution module. The primary and secondary joint baseline construction module is used to construct the historical steady-state baseline library, degradation feature library and primary and secondary fault propagation link topology of each monitoring node in the digital twin; The single-parameter trend prediction module is used to calculate in real time the trend deviation rate and degradation trend slope of each primary and secondary monitoring parameter relative to the historical steady-state baseline. When the trend deviation rate or degradation trend slope of any monitoring parameter exceeds the preset prediction threshold, a low confidence abnormality warning is output, and a pre-control command is sent to the output control unit. The adaptive fusion diagnostic module is used to receive the low confidence anomaly warning, determine the equipment state dimension corresponding to the anomaly parameter, dynamically adjust the fusion weight of multi-source data, increase the fusion weight of the first and second monitoring parameters of the corresponding equipment state dimension, and perform physical consistency cross-verification by combining the historical steady-state baseline library, the associated first and second monitoring parameters and the multi-physics field simulation results of the digital twin, and output the fault diagnosis result and fault location information. The chain failure simulation and verification module is used to generate a set of candidate risk nodes by starting from the confirmed fault node and tracing the topology of the primary and secondary fault propagation links. After verification of the conformity of the degradation features, it outputs a chain failure warning. The closed-loop self-evolution module is used to feed back the fault diagnosis results and verified cascading fault data to the digital twin, and synchronously update the parameters of the historical steady-state baseline library, the degradation feature library, the fusion weight model, and the primary and secondary fault propagation link topology.

[0007] In one possible implementation of the first aspect, the primary monitoring parameters acquired by the primary data acquisition unit include primary main circuit electrical operation parameters, primary equipment body status parameters, and box environment parameters; the secondary monitoring parameters acquired by the secondary data acquisition unit include protection action signals of the secondary measurement and control terminal, equipment self-test parameters, operation control events, and power supply status parameters. The monitoring range of the single-parameter trend prediction module covers both primary and secondary monitoring parameters; the primary monitoring parameters are used to directly predict the corresponding primary equipment faults, while the secondary monitoring parameters are used to indirectly infer the latent faults of the primary equipment.

[0008] In one possible implementation of the first aspect, the historical steady-state baseline library stores the baseline values ​​and normal fluctuation ranges of each monitoring parameter under steady-state operating conditions; the degradation feature library stores the historical degradation mode feature parameters corresponding to various typical faults; the trend deviation rate is the change in the deviation between the measured value of the parameter and the baseline value per unit time; the degradation trend slope is the linear regression slope of the parameter sequence within the sliding time window; the equipment status dimension is a monitoring dimension divided according to fault type, including electrical insulation dimension, mechanical performance dimension, thermal state dimension, and secondary circuit reliability dimension; The low-confidence anomaly warning is generated based solely on the trend exceeding limits of a single parameter, while the high-confidence fault diagnosis requires the fusion of multiple parameters and cross-validation of physical consistency.

[0009] In one possible implementation of the first aspect, the single-parameter trend prediction module classifies the deviation into first-level, second-level, and third-level deviations according to the excess amplitude of the trend deviation rate and the slope of the deterioration trend, from low to high, and outputs corresponding graded pre-control instructions. A level one deviation corresponds to an increase in the sampling frequency and data reporting frequency of the abnormal nodes; The remote control opening and closing operation and protection setting modification permissions for the corresponding interval of the secondary deviation are locked. Level 3 deviation corresponds to the activation of redundant acquisition channels for the corresponding device status dimension and the output of early risk warnings; All tiered pre-control commands only execute early warning and operation permission management, and do not output a primary circuit trip control signal.

[0010] In one possible implementation of the first aspect, the adaptive fusion diagnostic module dynamically adjusts the fusion weights specifically as follows: Based on the deviation level corresponding to the low confidence anomaly warning, the fusion weight of the primary and secondary monitoring parameters of the corresponding device status dimension is increased by a preset proportional coefficient, and the fusion weight of the other device status dimensions is decreased accordingly. The weight adjustment range is limited to the preset upper and lower limits. The fusion calculation process incorporates primary and secondary inherent physical constraint rules, including current-voltage ratio constraints, opening and closing action timing constraints, and temperature rise-load correspondence constraints. Abnormal data points that do not meet the physical constraint rules are first eliminated, and then weighted fusion calculation is performed.

[0011] In one possible implementation of the first aspect, the primary and secondary fault propagation link topology includes three types of links: A physical propagation path is used to characterize the propagation path of heat conduction, insulation degradation, and electrodynamic mechanical failures within primary equipment. Secondary logic propagation links are used to characterize the propagation paths of electrical signal conduction, protection logic linkage, and power fault propagation within the secondary system. Primary and secondary cross-coupling links are used to characterize the cross-system fault propagation path between primary and secondary devices, including thermal coupling paths, electromagnetic coupling paths, and logical linkage paths.

[0012] In one possible implementation of the first aspect, before performing cross-validation, the adaptive fusion diagnostic module performs timing alignment of primary and secondary multi-source monitoring data based on a unified clock reference, and automatically filters out electromagnetic interference components in the primary monitoring data within the corresponding time window by combining the switching operations, relay actions, and power switching event markers uploaded by the secondary data acquisition unit, before performing fusion calculation and fault determination.

[0013] In one possible implementation of the first aspect, the adaptive fusion diagnostic module calls the multiphysics simulation results of the digital twin for cross-validation specifically as follows: The real-time load, ambient temperature inside the enclosure, and historical degradation characteristics of the equipment are input into the multi-physics coupled reduced-order simulation model built into the digital twin. The multi-physics field includes electric field, thermal field, and stress field. Based on the model, the theoretical reference values ​​and standard residual intervals with a 95% confidence level corresponding to various monitoring parameters under the current operating conditions are output. The residual amplitude and frequency domain harmonic characteristics of the measured monitoring parameters relative to the theoretical reference values ​​are calculated and used as the basis for fault judgment. If the residual amplitude exceeds the standard residual interval and the frequency domain harmonic characteristics match the preset fault template, the equipment is judged to have a fault.

[0014] In one possible implementation of the first aspect, the cascading fault simulation and verification module starts from the diagnosed fault node and performs breadth-first traversal simulation along the primary and secondary fault propagation link topology to generate a set of candidate risk nodes; for each candidate risk node, it extracts three types of feature parameters: historical degradation trend slope, long-term baseline offset, and operating condition sensitivity, and performs cosine similarity matching calculation with the typical feature vector of the simulated fault type; when the matching degree is higher than the first threshold, it outputs a high-risk cascading fault warning; when the matching degree is between the first threshold and the second threshold, it outputs a concern-level warning and upgrades the monitoring level of the node; when the matching degree is lower than the second threshold, it is determined to be low-risk and the warning is excluded.

[0015] In one possible implementation of the first aspect, the closed-loop self-evolutionary module employs a hierarchical update mechanism: Baseline layer updates are performed on a daily basis using steady-state operating data to continuously correct the historical steady-state baselines of each monitoring node. The update process only includes valid steady-state data that is free from faults and operational disturbances. The weight layer is updated monthly, taking into account the fault diagnosis accuracy and false alarm rate of the month, and the gradient descent method is used to optimize the initial value and adjustment range of the fusion weight of each device status dimension. Link layer update is triggered after each confirmation of the actual occurrence of a cascading failure, and corrects the parameters of fault propagation delay, node impact weight and risk judgment threshold within the primary and secondary fault propagation link topology. After equipment maintenance or replacement of core components, manual triggering of baseline reset and rapid calibration of the corresponding sub-module is supported.

[0016] Its beneficial effects are as follows: 1. Construct a unified primary and secondary joint baseline system. Primary parameters directly determine explicit faults, while secondary parameters indirectly infer primary latent faults. This breaks down the data barriers between primary and secondary systems, enabling the identification of cross-system latent defects such as thermal coupling and electromagnetic coupling, and significantly reducing the false negative rate.

[0017] 2. The degree of equipment degradation is quantified by two indicators: the rate of deviation from the trend and the slope of the degradation trend. Three levels of deviation are divided and corresponding differentiated pre-control strategies are adopted. Early warning and control are triggered at the incipient stage of the fault. All pre-control instructions do not involve tripping, thus balancing safety and proactive operation and maintenance.

[0018] 3. The system dynamically adjusts the fusion weights of each state dimension based on the anomaly type, incorporates three types of physical constraints (electrical, temporal, and temperature rise) to eliminate false anomalies, and then combines digital twin multiphysics field reduced-order simulation for cross-validation. This triple guarantee of diagnostic accuracy from the data layer, physical layer, and simulation layer significantly reduces the false alarm rate.

[0019] 4. Establish three types of fault propagation topologies: primary physical, secondary logical, and primary-secondary cross-coupling. Starting from the confirmed node, perform a breadth-first traversal of candidate risk nodes. Determine cascading risks by classifying the cosine similarity of the three types of features to support proactive prevention and control and avoid fault escalation.

[0020] 5. The baseline is updated daily, the weights are optimized and integrated monthly, and the fault-triggered correction propagation link is updated independently without interference. Manual recalibration is supported after equipment maintenance or component replacement to ensure that the system's accuracy does not decrease over long-term operation. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0022] Figure 1 This is a block diagram of a primary and secondary fusion ring network box status monitoring system based on digital twins, provided in an embodiment of this application. Figure 2 This is a schematic diagram of the internal module composition of the fusion analysis unit provided in the embodiments of this application; Figure 3 This is a flowchart of the data processing of the adaptive fusion diagnostic module provided in the embodiments of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0025] Example

[0026] like Figure 1 As shown, the system of the present invention is deployed in an outdoor 10kV primary and secondary fusion ring network box, and the whole system consists of a primary data acquisition unit, a secondary data acquisition unit, a digital twin modeling unit, a fusion analysis unit and an output control unit.

[0027] The primary data acquisition unit is deployed in each primary interval of the ring network box, and its hardware composition and sampling parameters are as follows (for illustrative purposes only): Three-phase protection-grade current transformer (CT): ratio 600 / 5, accuracy class 0.5, sampling frequency default 1kHz, which can be increased to 5kHz after abnormal triggering; Three-phase voltage transformer (PT): 10kV / 100V ratio, 0.2 accuracy, synchronous sampling; Passive wireless surface acoustic wave (SAW) temperature sensor: installed at the A / B / C three-phase incoming contact, outgoing contact and busbar connection, a total of 9 temperature measuring points, temperature measuring range 40℃~150℃, accuracy ±0.5℃, reporting cycle default 1 minute; Ultra-high frequency (UHF) partial discharge sensor: installed on the inner wall of the switch cabinet, detection frequency band 300MHz~1500MHz, sensitivity ≤1pC; Temperature and humidity transmitter inside the enclosure: measuring range 20℃~80℃ / 0~100%RH, accuracy ±0.3℃ / ±2%RH.

[0028] The monitoring parameters are divided into three categories: Primary main circuit electrical operating parameters: three-phase phase current / / Three-phase line voltage / / Zero-sequence current Zero-sequence voltage Active power P, reactive power Q, power factor ; Primary equipment body status parameters: 9-channel contact / busbar temperature, partial discharge amplitude and discharge count, switch energy storage status, contact stroke, and opening / closing coil current; Internal environmental parameters: internal temperature, relative humidity, and condensation status.

[0029] The secondary data acquisition unit interfaces with the built-in protection and control terminal (DTU) of the ring main unit via the IEC 61850 MMS protocol to acquire four types of secondary parameters: Protection action signals: Overcurrent I / II / III stage action indicator, zero sequence protection action indicator, over-temperature alarm indicator, device lockout signal, with millisecond-level SOE time stamp; Equipment self-test parameters: AD sampling channel deviation value, device internal temperature, power module voltage, communication link status, and memory verification status; Operation control events: remote control opening / closing commands, protection setting modification, soft and hard pressure plate activation / deactivation, device reset events; Power status parameters: DC operating bus voltage, UPS battery voltage, remaining battery capacity, power module alarms.

[0030] The digital twin modeling unit is deployed on the distribution network edge gateway (ARM Cortex-A72, 4GB memory) or the regional master station server. It constructs a geometric model based on the factory-installed 3D CAD model of the ring network box, and overlays equipment nameplate parameters and material properties to build a multiphysics simulation model. It incorporates a three-physics coupled reduced-order simulation model of electric field, thermal field, and stress field. The original finite element model is reduced in order using the intrinsic orthogonal decomposition (POD) method, with a basis vector number of 20. Single-step simulation time is <80ms, and the relative error is <3%, meeting online real-time requirements.

[0031] The fusion analysis unit is the core of the system, deployed within the edge computing gateway. It adopts a modular software architecture, with five internal functional modules interacting with data via shared memory and a message queue, achieving an end-to-end processing latency of <500ms. The five modules are: primary and secondary joint baseline construction module, single-parameter trend prediction module, adaptive fusion diagnosis module, cascading fault inference and verification module, and closed-loop self-evolution module.

[0032] The output control unit connects to the ring network box's measurement and control I / O circuit and the distribution network operation and maintenance cloud platform, executes pre-control commands such as sampling frequency adjustment, operation permission locking, and redundant channel switching, and sends hierarchical early warning and fault diagnosis information through the MQTT protocol.

[0033] Furthermore, the various modules of the fusion analysis unit and their working principles are explained in detail, such as... Figure 2 As shown, it includes: 1. Principles and Examples of Constructing a Joint Primary and Secondary Baseline Library Before the system is put into operation, the initial baseline calibration is performed by the primary and secondary joint baseline library construction module, and after the system is put into operation, the closed-loop self-evolution module continuously updates the baseline.

[0034] 1.1 Construction of Historical Steady-State Baseline Library Steady-state determination conditions (all three conditions must be met simultaneously): Condition 1: No protection action, no remote control action, and no device self-test alarm in the past 30 minutes.

[0035] Condition 2: Load current fluctuation range ≤ ±10% of rated current.

[0036] Condition 3: Ambient temperature change rate ≤ 1℃ / 10min.

[0037] Baseline calculation method: For 7 consecutive days of data that meet steady-state conditions, robust statistical methods are used to calculate the baseline values ​​of each parameter. Benchmark value = median of the data series (to avoid outliers from skewing the data). The normal fluctuation range is [P2.5, P97.5], which is the 2.5% quantile to the 97.5% quantile of the data series.

[0038] For example, the steady-state data statistics of the A-phase incoming contact of a certain ring network box during the initial operation period are as follows: the median is 41.8℃, P2.5=35.2℃, P97.5=48.6℃, so the baseline reference value is recorded as 42℃, and the normal fluctuation range is [35℃, 49℃].

[0039] 1.2 Construction of Degradation Feature Library The degradation feature library stores the degradation mode feature vectors of 12 typical faults. Each fault feature vector contains 10 dimensions: degradation trend slope, baseline offset rate, operating condition sensitivity coefficient, harmonic distortion rate, proportion of 3rd / 5th / 7th harmonics, peak factor, impulse factor, and margin factor.

[0040] The initial feature library is derived from the statistical average of historical failure cases of the same model of equipment, and is subsequently supplemented and corrected by the closed-loop self-evolution module as real failure cases are encountered.

[0041] 1.3 Initialization of Primary and Secondary Fault Propagation Link Topology The topology is represented by a directed weighted graph G = (V, E, W): Node set V: includes primary equipment nodes (contacts, insulators, busbars, operating mechanisms), secondary equipment nodes (protection devices, terminal blocks, power modules, sampling circuits), and coupling nodes (CT secondary windings, wiring terminals, grounding blocks), for a total of 32 nodes; Edge set E: Represents the direction of fault propagation, with a total of 47 directed edges; Weight W: Each edge corresponds to three parameters—propagation delay. (Unit: hour), Influence Weight (0~1) Fault propagation probability (0~1).

[0042] The initial weights are assigned based on the equipment structure design parameters and industry experience, and are subsequently corrected by the closed-loop self-evolution module based on real cascading failure data.

[0043] 1.4 Device Status Dimension Definition and Initial Weight Configuration The equipment status dimension is divided into four categories of monitoring and analysis dimensions based on the fault mechanism. Each dimension is associated with a set of primary and secondary monitoring parameters, as detailed in Table 1.

[0044] Table 1 - Device Status Dimension Definition and Weight Configuration Table:

[0045] The sum of the weights of the four dimensions is always equal to 1. The quantum weights of each parameter under each dimension are initially allocated according to the information entropy method to distinguish the contribution of each parameter to the fault of that dimension.

[0046] 2. Single-parameter trend prediction algorithm and three-level hierarchical control The single-parameter trend prediction module independently calculates two trend indicators for each monitored parameter. If either indicator exceeds the limit, a low-confidence warning is triggered.

[0047] 2.1 Calculation of Trend Deviation Rate Definition: Rate of deviation from trend It represents the change in parameter deviation value per unit time, reflecting the degree of acceleration of degradation.

[0048] Calculation formula: in, for Real-time measured value relative baseline reference value The absolute deviation; To calculate the time window, we use 1 hour.

[0049] Predicted threshold: Temperature parameters Current-related parameters Voltage parameters .

[0050] 2.2 Calculation of the slope of the degradation trend Definition: Slope of the degradation trend This represents the slope of the linear regression of the parameter sequence within the sliding time window, reflecting the overall trend of long-term degradation.

[0051] Calculation method: Take the time interval series of the past T=24 hours. The least squares method is used to fit the straight line. slope This is the slope of the degradation trend, and the formula is as follows: The slope of the degradation trend can be calculated using the slope formula of univariate linear regression (least squares method). .

[0052] Predicted threshold: Temperature parameters Partial discharge parameters .

[0053] 2.3 Low Confidence Anomaly Warning Judgment when or When the error occurs, a low-confidence anomaly warning is output, and the anomaly parameter number, the corresponding equipment status dimension, and the over-limit amplitude are marked.

[0054] Low confidence warnings only indicate an abnormal trend in a single parameter dimension, which is likely to be affected by interference factors. Therefore, the confidence level is low and it is not directly determined as a device failure. It must be cross-validated by the adaptive fusion diagnostic module before a diagnosis can be made.

[0055] 2.4 Three-level deviation classification judgment and pre-control strategy The over-limit ratio is calculated based on the ratio of the over-limit amplitude to the predicted threshold. Divide into three levels of deviation: Level 1 deviation, exceeding the limit range The corresponding pre-control instructions are: increase the sampling frequency of abnormal nodes to 10Hz and shorten the data reporting cycle to 1 minute; Level 2 deviation, exceeding the limit range The corresponding pre-control instruction is: based on the first level, lock the remote control opening and closing permissions and protection setting modification permissions of the corresponding bay; Level 3 deviation, exceeding the limit range of multiplier The corresponding pre-control instruction is: based on the second level, activate the redundant collection channel for the corresponding status dimension and push early risk SMS messages.

[0056] Access control principle: The output control unit sends a "remote operation lockout" command to the protection and control terminal, setting the operation lockout soft pressure plate. All remote control opening and closing commands and setting value modification commands are intercepted and rejected by the terminal until the warning is lifted or manual unlocking is required.

[0057] Redundant channel activation principle: When the thermal state dimension is abnormal, the backup infrared temperature measurement channel is activated; when the electrical insulation dimension is abnormal, the high-frequency current (HFCT) partial discharge auxiliary channel is activated; when the secondary circuit is abnormal, the backup communication link is activated. Redundant channel data is synchronously accessed to the fusion diagnostic module to improve diagnostic data redundancy.

[0058] Important constraint: All pre-control commands are warning-level control measures, and no main circuit trip signal is output. Trip protection is still executed independently by the ring main unit's native protection device to avoid unplanned power outages caused by misjudgment by the monitoring system.

[0059] 3. Adaptive fusion diagnostic algorithm and physical consistency verification The adaptive fusion diagnostic module is the core diagnostic component of the system. Upon receiving a low-confidence warning, it executes a seven-step process: "time sequence alignment → interference filtering → weight adjustment → physical constraints → weighted fusion → simulation verification → fault diagnosis." Figure 3 As shown.

[0060] 3.1 Timing Alignment and Electromagnetic Interference Filtering Unified clock reference: The entire system adopts the IEEE 1588 PTP precise time protocol for synchronization. Both primary sample data and secondary event data are timestamped at the nanosecond level, with a time synchronization accuracy of ≤1. .

[0061] Timing alignment method: Primary and secondary event data with different sampling rates are uniformly resampled to a 100ms time granularity. Primary sampled data is resampled using linear interpolation; secondary event data is mapped to the corresponding time slice using nearest neighbor matching.

[0062] Electromagnetic interference filtering rules: When the secondary side sends the "switch opening and closing" relay action "power switching" event flag, it automatically locks the time window ±200ms before and after the event time; Band-stop filtering (stopband 100kHz~1MHz) is performed on the primary current, voltage and partial discharge data in the window to remove high-frequency electromagnetic interference components generated by the operation transient. Filtered data is marked as "interference removed" and proceeds to subsequent fusion calculations; data outside the window is directly fused and calculated.

[0063] 3.2 Device Status Dimension Identification and Dynamic Weight Adjustment Dimension identification: Mapping abnormal parameters to the corresponding device status dimension based on their category. For example: excessive contact temperature → thermal status dimension; excessive partial discharge amplitude → electrical insulation dimension; abnormal secondary power supply voltage → secondary circuit reliability dimension.

[0064] Weight adaptive adjustment method: Let the first The initial weights for each dimension are: The abnormal deviation level is L (Level 1 L=1, Level 2 L=2, Level 3 L=3), and the proportionality coefficient is... (Adjustment range per level), then the adjusted weight of the abnormal dimension is: in This is the upper limit for the weight of a single dimension; the weights of other non-abnormal dimensions are adjusted downwards proportionally.

[0065] For example: all four dimensions have an initial weight of 0.25, and the hot state dimension triggers a second-order deviation (L=2): The target weight for the heat dimension is 0.25 × (1 + 0.2 × 2) = 0.35 (not exceeding the upper limit of 0.6). The total weight of the other three dimensions should be 0.65, and each dimension should be adjusted proportionally: 0.25 × (0.65 / 0.75) ≈ 0.217, with a sum of 1.

[0066] Weight upper and lower limits constraint: The weight of a single dimension is at most 0.6 and at least 0.05, ensuring that no dimension is completely ignored and avoiding misjudgment caused by a single dimension abnormality dominating the diagnostic results.

[0067] 3.3 Verification of Primary and Secondary Inherent Physical Constraints Before the fusion calculation, three physical constraint checks are performed on all data points to be fused. Data points that do not meet the checks are marked as abnormal data points and removed.

[0068] Constraint 1: Current-voltage ratio constraint The measured transformation ratio of the three-phase current and line voltage in the same interval shall not deviate from the nominal transformation ratio of the CT / PT by more than ±5%, that is: If the conditions are not met, the CT / PT sampling circuit is deemed abnormal, and the data point is discarded.

[0069] Constraint 2: Timing constraints for opening and closing operations After the tripping command is issued, the auxiliary contact change time should be within the range of 50ms to 100ms; after the closing command is issued, the auxiliary contact change time should be within the range of 60ms to 120ms. Data exceeding the timing range is marked as an operating mechanism anomaly and is not included in the steady-state fusion calculation.

[0070] Constraint 3: Temperature rise-load correspondence constraint Under normal operating conditions, the contact temperature rise is approximately proportional to the square of the load current: If the load current decreases but the contact temperature rises abnormally (current decrease > 10% and temperature rise > 2°C), it is marked as a suspected deterioration data point. Such data is not removed, but its sub-weight in the thermal state dimension is increased, because it may reflect the actual increase in contact resistance deterioration.

[0071] 3.4 Multi-dimensional weighted fusion calculation Data subject to physical constraints enters the weighted fusion layer, employing a layered weighted fusion architecture: First layer: Dimensional intrinsic parameter quantum weight fusion to obtain the fault membership degree of each dimension. ; The second layer involves weight fusion across dimensions to obtain the comprehensive fault membership degree. .

[0072] Intra-dimensional fusion uses a weighted average method: in, For dimension Next Sub-weights of each parameter The standard deviation of the baseline fluctuation of this parameter is given by the fractional expression, which is the normalized deviation.

[0073] Interdimensional fusion: , when When the fault determination threshold is set to 3.0 by default, a preliminary fault is determined to exist, and the process proceeds to the next step of multiphysics simulation cross-validation.

[0074] 3.5 Cross-validation of Multiphysics Simulation Input to the reduced-order simulation model: real-time load current , ambient temperature inside the box Equipment service life (Deterioration correction factor).

[0075] Simulation output: Theoretical baseline values ​​of each monitored parameter under the current operating condition. and the standard residual interval at the 95% confidence level. .

[0076] Residual calculation: , These are the measured values ​​of each monitoring parameter.

[0077] Frequency domain harmonic feature extraction: Perform a 256-point FFT transformation on the residual sequence to extract the proportion of the amplitude of the 3rd, 5th, and 7th harmonics to the total harmonic distortion (THD), and construct the harmonic feature vector H=[h3,h5,h7].

[0078] Dual judgment conditions (must be met simultaneously): Condition A: (The residual exceeds the 95% confidence interval); Condition B: The cosine similarity between the harmonic feature vector H and the corresponding fault template in the degradation feature library is >0.8.

[0079] When both conditions are met, the fault is diagnosed as having high confidence, and the fault type, fault location node number, and fault confidence score are output.

[0080] Through a triple judgment mechanism of "data layer weighted fusion + physical layer constraint screening + simulation layer cross-verification", the fault diagnosis accuracy of this system is significantly improved compared with the single threshold alarm scheme, while the false alarm rate is also significantly reduced.

[0081] 4. Chain Fault Inference and Feature Matching Algorithm 4.1 Detailed Explanation of Primary and Secondary Fault Propagation Link Topology and Three Types of Links The topological directed graph contains 47 propagation edges, belonging to three types of links: (1) One physical propagation link (18 links) Fault propagation characterized by purely physical mechanisms within primary equipment, such as: Heat conduction path: A-phase contact overheating → A-phase busbar temperature rise → B-phase adjacent contact temperature rise → cabinet ambient temperature rise; Insulation degradation path: contact surface oxidation → partial discharge → surface flashover → phase-to-phase insulation breakdown; Electromechanical path: short-circuit impact current → electrodynamic deformation of contacts → increased contact resistance → intensified contact heating.

[0082] (2) Secondary logic propagation links (15 links) Characterizing fault propagation in the interaction between electrical signals and logic within a secondary system, such as: Power supply propagation path: DC power module failure → protection device power failure → multi-interval protection failure → communication interruption; Sampling linkage path: CT secondary circuit open circuit → abnormal sampling value → protection misjudgment → false trip signal; Cascaded action path: Single bay protection action → Interlocking logic trigger → Interlocking of adjacent bays.

[0083] (3) Primary and secondary cross-coupled links (14 links) Characterizing primary and secondary cross-system coupling fault propagation, such as: Thermal coupling path: primary current-carrying contact heats up → terminal heat conduction → secondary terminal block aging → poor contact in secondary circuit; Electromagnetic coupling path: primary short circuit transient → CT secondary side induced high voltage → secondary sampling AD chip damage → sampling deviation; Logical linkage path: Secondary protection malfunction → Primary switch malfunction → Downstream bay power loss → Expanding the power outage area.

[0084] 4.2 Breadth-first traversal deduction Starting with the confirmed faulty node, perform a breadth-first search (BFS) traversal along the directed edges of the topology: The traversal depth is limited to 3 levels (to avoid overly large deduction ranges); If the propagation probability of each edge is less than 0.1, it is pruned and not included in the candidate set; Output the set of candidate risk nodes in descending order of propagation probability.

[0085] 4.3 Verification of Deterioration Feature Conformity (Cosine Similarity Matching) For each candidate risk node, extract three-dimensional features to construct a feature vector.

[0086] Historical degradation trend slope (linear regression slope over the past 7 days, normalized to [0,1]); Long-term baseline offset (the percentage shift of the mean baseline over the past 30 days relative to the initial baseline, normalized). Operating condition sensitivity (parameter change / load change, reflecting the equipment's sensitivity to load fluctuations, normalized).

[0087] candidate node feature vectors Typical feature vectors corresponding to the fault types in the degradation feature library Calculate cosine similarity: The similarity value range is: A higher value indicates a higher degree of matching in the degradation pattern.

[0088] Level 3 matching determination (first threshold) Second threshold ): High-risk cascading failure warning, red alert, push notification to maintenance personnel for on-site verification within 2 hours; Level 2 alert, yellow warning: the sampling frequency of this node is increased to 5 minutes, and it is included in the key observation list; Low-risk cases are excluded, no warnings are triggered, and normal monitoring frequency is maintained.

[0089] For example: If an overheating fault is diagnosed in the A-phase contact, the candidate node "secondary A-phase terminal" is deduced, with the feature vector F = [0.72, 0.68, 0.55], and the thermal coupling degradation template. = [0.8, 0.75, 0.6], the cosine similarity is approximately 0.992 > 0.75, which is considered a high-risk cascading failure warning.

[0090] 5. The principle of the three-layer closed-loop self-evolution mechanism The closed-loop self-evolution module adopts a three-layer independent update mechanism. The update cycles, triggering conditions, and optimization goals of the three layers are different and do not interfere with each other.

[0091] 5.1 Baseline Layer Update - Daily Rolling Correction Triggering cycle: Automatically triggered at 02:00 every day (during off-peak hours).

[0092] Update method: Exponentially weighted moving average (EWMA) rolling update. Let the mean of the effective steady-state data for the day be... The historical baseline value is Smoothing coefficient (New data weighted 10%, historical data weighted 90%) , Data inclusion criteria: Only valid steady-state data that meets the three conditions of "no fault alarms, no operational events, and stable load" on the same day will be included. Data from abnormal days will be skipped and not updated to avoid baseline drift caused by single-day interference.

[0093] 5.2 Weight Layer Update - Monthly Gradient Descent Optimization Trigger cycle: Automatically triggered at midnight on the 1st of each month. Optimization goal: Maximize fault diagnosis accuracy Minimize the false alarm rate The objective function is: , in, , The penalty for false alarms is higher, which is in line with the relevant safety principles of the power system.

[0094] Optimization method: Batch Gradient Descent. Initial weights are set using four dimensions. To optimize the variables, the weights are iteratively updated using all early warning and verification samples from the previous month as the training set: , Learning rate The iteration termination condition is two consecutive rounds. The absolute value of the difference is less than The maximum iteration can be 500 rounds. After the iteration converges, the initial weights of each dimension and the weight adjustment range of each level of deviation are updated simultaneously.

[0095] 5.3 Link Layer Update - Fault Trigger Correction Triggering conditions: The cascading fault warning is triggered immediately after on-site verification by maintenance personnel that the fault has actually occurred.

[0096] Propagation delay correction: Based on the actual time it takes for a fault to propagate from the starting node to the target node, the propagation delay parameter of the corresponding edge is corrected using the moving average method; Impact weight adjustment: Adjust the impact weight of the corresponding edge proportionally (10%~30%) according to the actual severity of the fault. Node threshold correction: Fine-tune the first threshold based on the feature similarity distribution of real-world cases. With the second threshold (The adjustment range for a single instance shall not exceed ±0.05).

[0097] 5.4 Manual Reset and Rapid Calibration Triggering scenarios: after equipment overhaul, replacement of core components (switch body, CT / PT, protection device), or complete replacement of ring main unit.

[0098] Rapid calibration process: Manually trigger the "baseline reset" command to clear the historical baseline and degradation records of the corresponding node; After the equipment has been running stably under power for 4 hours, it automatically collects 4 hours of steady-state data to complete the initial baseline reconstruction. The degradation feature library and propagation topology inherit the common template of the same model of equipment, and will be gradually modified to be personalized after data is accumulated. After calibration, the system automatically switches to normal monitoring mode.

[0099] 6. Complete Workflow Example The contacts of the A-phase incoming line of a certain 10kV outdoor integrated primary and secondary ring network box gradually deteriorate due to increased contact resistance. The complete system workflow is as follows: Phase 1: Trend Judgment The slope of the temperature degradation trend of phase A contact reached 0.35℃ / h, exceeding the threshold of 0.3℃ / h, and the over-limit ratio was [missing information]. This triggers a Level 1 low-confidence deviation warning; Pre-control instructions: Increase the contact temperature sampling frequency from 1 minute to 10 seconds, and shorten the reporting cycle to 1 minute.

[0100] Phase 2: Upgraded Warning The degradation trend slope increased to 0.8℃ / h, the trend deviation rate reached 4.2℃ / h, and the over-limit ratio... It has been upgraded to Level 3 deviation; Pre-control instructions: Lock remote control access for this interval, activate the backup infrared temperature measurement redundancy channel, and send early risk SMS messages.

[0101] Phase 3: Convergence Diagnosis The adaptive fusion diagnostic module receives the early warning, identifies the abnormal dimension as the thermal state dimension, and increases the weight by 60% to 0.4; After timing alignment, recent switching operation interference data were removed. Physical constraint verification confirmed that the temperature rise-load relationship was abnormal (the temperature rose even though the load did not increase), and it was marked as a suspected degradation. Using digital twin multiphysics simulation: Under the current load of 400A and ambient temperature of 30℃, the theoretical contact temperature should be 45℃, but the actual measured temperature is 58℃, with a residual of 13℃, which is far beyond the 95% confidence interval of ±2℃; the residual harmonic characteristics have a similarity of 0.89 to the "increased contact resistance" template. High-confidence fault diagnosis: A-phase incoming line contact failure and overheating fault, confidence level 92%.

[0102] Phase 4: Chain Reaction Starting from "A-phase incoming contact", perform a BFS traversal of 3 layers. Candidate nodes include A-phase secondary wiring terminals and adjacent B-phase contacts. The cosine similarity between the feature vector of phase A secondary terminal and the thermal coupling template is 0.82 > 0.75, indicating a high-risk cascading fault warning. The similarity of phase B contacts is 0.61, triggering a warning at the level of concern and upgrading the monitoring level.

[0103] Phase 5: Self-evolutionary update On-site maintenance confirmed the overheating of contacts and aging of terminals, confirming the cascading fault was indeed present. Link layer update: Corrects the conduction delay and impact weight of the contact → terminal thermal coupling link; Next month's weight layer update: Optimize the initial weight of the hot state dimension based on this case; The baseline layer is continuously updated daily to track new baselines after equipment maintenance.

[0104] This invention constructs a four-layer closed-loop architecture of "prediction, diagnosis, deduction, and evolution": The bottom layer establishes three basic libraries—steady-state baseline, degradation characteristics, and propagation topology—through a primary and secondary joint baseline construction module; the first layer, single-parameter trend prediction, uses dual indicators of trend deviation rate and degradation slope for quantitative monitoring, provides three-level graded early warning, and issues non-tripping pre-control commands to achieve early detection; the second layer, adaptive fusion diagnosis, dynamically adjusts fusion weights according to the anomaly dimension, filters out pseudo-data through time-series interference removal and physical constraints, and combines digital twin multi-physics simulation cross-validation to output high-confidence fault location, achieving accurate diagnosis; the third layer, cascading fault deduction, traverses along three types of propagation links with breadth-first priority, and determines cascading risks through cosine similarity feature matching to achieve anti-propagation; the top-level closed-loop self-evolution updates the baseline, weights, and topology parameters using a three-layer mechanism of daily / monthly / event, achieving long-term optimization. The entire system breaks down the barriers between primary and secondary data, integrating the dual advantages of data-driven and physical simulation, significantly improving the monitoring reliability and proactive operation and maintenance of primary and secondary fusion ring network boxes.

[0105] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computing software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0106] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0107] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A primary and secondary fusion ring network box status monitoring system based on digital twins, comprising a primary data acquisition unit, a secondary data acquisition unit, a digital twin modeling unit, a fusion analysis unit, and an output control unit, characterized in that, The fusion analysis unit includes a primary and secondary joint baseline construction module, a single parameter trend prediction module, an adaptive fusion diagnosis module, a cascading fault inference and verification module, and a closed-loop self-evolution module. The primary and secondary joint baseline construction module is used to construct the historical steady-state baseline library, degradation feature library and primary and secondary fault propagation link topology of each monitoring node in the digital twin; The single-parameter trend prediction module is used to calculate in real time the trend deviation rate and degradation trend slope of each primary and secondary monitoring parameter relative to the historical steady-state baseline. When the trend deviation rate or degradation trend slope of any monitoring parameter exceeds the preset prediction threshold, a low confidence abnormality warning is output, and a pre-control command is sent to the output control unit. The single-parameter trend prediction module classifies the deviation into first-level, second-level, and third-level deviations according to the excess amplitude of the trend deviation rate and the slope of the degradation trend, from low to high, and outputs corresponding graded pre-control instructions. A level one deviation corresponds to an increase in the sampling frequency and data reporting frequency of the abnormal nodes; The remote control opening and closing operation and protection setting modification permissions for the corresponding interval of the secondary deviation are locked. Level 3 deviation corresponds to the activation of redundant acquisition channels for the corresponding device status dimension and the output of early risk warnings; All hierarchical pre-control commands only perform early warning and operation permission control, and do not output a primary circuit trip control signal; The adaptive fusion diagnostic module is used to receive the low confidence anomaly warning, determine the equipment state dimension corresponding to the anomaly parameter, dynamically adjust the fusion weight of multi-source data, increase the fusion weight of the first and second monitoring parameters of the corresponding equipment state dimension, and perform physical consistency cross-verification by combining the historical steady-state baseline library, the associated first and second monitoring parameters and the multi-physics field simulation results of the digital twin, and output the fault diagnosis result and fault location information. The chain failure simulation and verification module is used to generate a set of candidate risk nodes by starting from the confirmed fault node and tracing the topology of the primary and secondary fault propagation links. After verification of the conformity of the degradation features, it outputs a chain failure warning. The closed-loop self-evolution module is used to feed back the fault diagnosis results and verified cascading fault data to the digital twin, and synchronously update the parameters of the historical steady-state baseline library, the degradation feature library, the fusion weight model, and the primary and secondary fault propagation link topology.

2. The primary and secondary fusion ring network box status monitoring system based on digital twins according to claim 1, characterized in that, The primary monitoring parameters acquired by the primary data acquisition unit include primary main circuit electrical operation parameters, primary equipment body status parameters, and internal environment parameters; the secondary monitoring parameters acquired by the secondary data acquisition unit include protection action signals from the secondary measurement and control terminal, equipment self-test parameters, operation control events, and power status parameters. The monitoring range of the single-parameter trend prediction module covers both primary and secondary monitoring parameters; the primary monitoring parameters are used to directly predict the corresponding primary equipment faults, while the secondary monitoring parameters are used to indirectly infer the latent faults of the primary equipment.

3. The primary and secondary fusion ring network box status monitoring system based on digital twins according to claim 1, characterized in that, The historical steady-state baseline library stores the baseline values ​​and normal fluctuation ranges of each monitoring parameter under steady-state operating conditions; the degradation feature library stores the historical degradation mode feature parameters corresponding to various typical faults. The trend deviation rate is the change in the deviation between the measured value of the parameter and the baseline value per unit time; the degradation trend slope is the linear regression slope of the parameter sequence within the sliding time window; the equipment status dimension is a monitoring dimension divided according to the fault type, including electrical insulation dimension, mechanical performance dimension, thermal status dimension and secondary circuit reliability dimension. The low-confidence anomaly warning is generated based solely on the trend exceeding limits of a single parameter, while the high-confidence fault diagnosis requires the fusion of multiple parameters and cross-validation of physical consistency.

4. The primary and secondary fusion ring network box status monitoring system based on digital twins according to claim 1, characterized in that, The adaptive fusion diagnostic module dynamically adjusts the fusion weights as follows: Based on the deviation level corresponding to the low confidence anomaly warning, the fusion weight of the primary and secondary monitoring parameters of the corresponding device status dimension is increased by a preset proportional coefficient, and the fusion weight of the other device status dimensions is decreased accordingly. The weight adjustment range is limited to the preset upper and lower limits. The fusion calculation process incorporates primary and secondary inherent physical constraint rules, including current-voltage ratio constraints, opening and closing action timing constraints, and temperature rise-load correspondence constraints. Abnormal data points that do not meet the physical constraint rules are first eliminated, and then weighted fusion calculation is performed.

5. The primary and secondary fusion ring network box status monitoring system based on digital twins according to claim 1, characterized in that, The primary and secondary fault propagation link topology includes three types of links: A physical propagation path is used to characterize the propagation path of heat conduction, insulation degradation, and electrodynamic mechanical failures within primary equipment. Secondary logic propagation links are used to characterize the propagation paths of electrical signal conduction, protection logic linkage, and power fault propagation within the secondary system. Primary and secondary cross-coupling links are used to characterize the cross-system fault propagation path between primary and secondary devices, including thermal coupling paths, electromagnetic coupling paths, and logical linkage paths.

6. The primary and secondary fusion ring network box status monitoring system based on digital twin as described in claim 1, characterized in that, Before performing cross-validation, the adaptive fusion diagnostic module aligns the timing of primary and secondary multi-source monitoring data based on a unified clock reference. It also automatically filters out electromagnetic interference components in the primary monitoring data within the corresponding time window by combining the switching operations, relay actions, and power switching event markers uploaded by the secondary data acquisition unit. Finally, it performs fusion calculations and fault determination.

7. The primary and secondary fusion ring network box status monitoring system based on digital twin as described in claim 1, characterized in that, The adaptive fusion diagnostic module calls the multiphysics simulation results of the digital twin for cross-validation in the following way: The real-time load, the ambient temperature inside the chamber, and the historical degradation characteristics of the equipment are input into the multi-physics coupled reduced-order simulation model built into the digital twin. The multi-physics includes electric field, thermal field, and stress field. Based on the model, the theoretical benchmark values ​​and standard residual intervals with a 95% confidence level corresponding to various monitoring parameters under the current operating conditions are output. The residual amplitude and frequency domain harmonic characteristics of the measured monitoring parameters relative to the theoretical reference value are calculated and used as the basis for fault judgment. If the residual amplitude exceeds the standard residual range and the frequency domain harmonic characteristics match the preset fault template, the equipment is judged to be faulty.

8. The primary and secondary fusion ring network box status monitoring system based on digital twins according to claim 1, characterized in that, The chain fault simulation and verification module starts from the confirmed fault node and performs a breadth-first traversal simulation along the primary and secondary fault propagation link topology to generate a set of candidate risk nodes. For each candidate risk node, three types of feature parameters are extracted: historical degradation trend slope, long-term baseline offset, and operating condition sensitivity. These parameters are then matched with the typical feature vectors of the simulated fault type using cosine similarity calculation. When the matching degree is higher than the first threshold, a high-risk chain fault warning is output. When the matching degree is between the first and second thresholds, a warning of concern level is output and the monitoring level of the node is upgraded. When the matching degree is lower than the second threshold, the node is determined to be low-risk and the warning is excluded.

9. The primary and secondary fusion ring network box status monitoring system based on digital twins according to claim 1, characterized in that, The closed-loop self-evolution module adopts a hierarchical update mechanism: Baseline layer updates are performed on a daily basis using steady-state operating data to continuously correct the historical steady-state baselines of each monitoring node. The update process only includes valid steady-state data that is free from faults and operational disturbances. The weight layer is updated monthly, taking into account the fault diagnosis accuracy and false alarm rate of the month, and the gradient descent method is used to optimize the initial value and adjustment range of the fusion weight of each device status dimension. Link layer update is triggered after each confirmation of the actual occurrence of a cascading failure, and corrects the parameters of fault propagation delay, node impact weight and risk judgment threshold within the primary and secondary fault propagation link topology. After equipment maintenance or replacement of core components, manual triggering of baseline reset and rapid calibration of the corresponding sub-module is supported.

Images