Electric energy data acquisition system and method based on environmental perception

Abstract

The invention discloses an electric energy data acquisition system and method based on environmental perception, and relates to the technical field of electric energy monitoring, and the method comprises the steps: converting original data into a physical field feature vector comprising a heat flow aggregation degree and eddy current intensity; constructing a dynamic heterogeneous graph network based on the physical field feature vectors; solving a continuous region in which a physical field distortion gradient value exceeds a gradient threshold value by utilizing a topological relation of the dynamic heterogeneous graph network; activating the sensors in the continuous areas to form a minimum sensing subgraph; according to the distortion coordinate of the minimum sensing subgraph, associating a target electric energy acquisition point through the current-carrying capability of the conductor; generating an electric energy characteristic dictionary atom based on the Joule's law; generating environment feature dictionary atoms based on a mechanical vibration equation; performing joint sparse coding by using the electric energy feature dictionary atoms and the environment feature dictionary atoms; and separating and reconstructing the electric energy data from the result of the joint sparse coding.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric energy monitoring, in particular to an electric energy data acquisition system and method based on environment perception. BACKGROUND

[0002] In modern power system operation and energy management, accurate and efficient acquisition of electric energy data has a fundamental significance. It is the key support to realize real-time monitoring of power grid state, load prediction analysis, energy efficiency optimization management and fault early warning diagnosis. With the development of smart grid and Internet of Things technology, it is increasingly important to obtain higher quality electric energy data to ensure the safe and stable operation of the power grid and improve energy utilization efficiency.

[0003] However, the traditional electric energy data acquisition method still has room for improvement when faced with complex environments and a large number of monitoring points. The existing method generally uses continuous global or fixed mode data acquisition strategy, which is not efficient in resource-limited scenarios such as communication bandwidth and sensor energy consumption, and it is difficult to balance the real-time monitoring and system economy. How to more intelligently coordinate the perception resources and effectively integrate the environmental information to improve the efficiency of electric energy data acquisition is a direction worth paying attention to in current technology.

[0004] At present, there is no effective solution to the above problems. SUMMARY

[0005] The embodiments of the present application provide an electric energy data acquisition system and method based on environment perception to solve the above technical problems.

[0006] The present application provides an electric energy data acquisition system based on environment perception, comprising:

[0007] A physical field feature vector conversion module is configured to acquire raw data collected by each sensor in a monitoring area in real time, and convert the raw data into a physical field feature vector comprising heat flow concentration and eddy current intensity.

[0008] A heterogeneous graph network construction module is configured to construct a dynamic heterogeneous graph network based on the physical field feature vector.

[0009] A minimum perception subgraph formation module is configured to use the topological relationship of the dynamic heterogeneous graph network to solve the continuous area where the physical field distortion gradient value exceeds the gradient threshold value, and activate the sensors in the continuous area to form a minimum perception subgraph.

[0010] A target electric energy acquisition point association module is configured to associate a target electric energy acquisition point according to the distortion coordinates of the minimum perception subgraph through the conductor current-carrying capacity.

[0011] a joint sparse coding module configured to input the electric energy data of the target electric energy collection point and the environmental data of the minimum perception subgraph into an encoder, and perform: generating electric energy feature dictionary atoms based on Joule's law; generating environmental feature dictionary atoms based on a mechanical vibration equation; and performing joint sparse coding using the electric energy feature dictionary atoms and the environmental feature dictionary atoms;

[0012] an electric energy data reconstruction module configured to separate and reconstruct the electric energy data from the result of joint sparse coding.

[0013] The application provides an electric energy data collection method based on environmental perception, comprising:

[0014] obtaining raw data collected by each sensor in a monitoring area in real time, and converting the raw data into a physical field feature vector comprising heat flow concentration and vortex intensity;

[0015] constructing a dynamic heterogeneous graph network based on the physical field feature vector;

[0016] using the topological relationship of the dynamic heterogeneous graph network to solve a continuous area in which a physical field distortion gradient value exceeds a gradient threshold value, and activating sensors in the continuous area to form a minimum perception subgraph;

[0017] correlating a target electric energy collection point according to a distortion coordinate of the minimum perception subgraph based on conductor current-carrying capacity;

[0018] inputting the electric energy data of the target electric energy collection point and the environmental data of the minimum perception subgraph into an encoder, and performing: generating electric energy feature dictionary atoms based on Joule's law; generating environmental feature dictionary atoms based on a mechanical vibration equation; and performing joint sparse coding using the electric energy feature dictionary atoms and the environmental feature dictionary atoms;

[0019] separating and reconstructing the electric energy data from the result of joint sparse coding.

[0020] According to the embodiments provided in the application, by converting raw sensor data into a feature vector reflecting the state of a physical field (such as heat flow concentration and vortex intensity) and constructing a dynamic heterogeneous graph network based on the feature vector, a key continuous area in which a physical field distortion gradient exceeds a threshold value can be intelligently identified. On this basis, only sensors in the continuous area are activated to form a minimum perception subgraph, and the fixed collection mode of sensors in non-key areas is abandoned. This “on-demand activation and local focus” perception strategy effectively reduces sensor energy consumption and communication bandwidth occupation, significantly improves the overall efficiency of the system in a resource-limited scenario, and optimizes the balance between real-time monitoring and economy.

[0021] The environment data (sensed by sensors in the physical field distortion area) and the associated target electric energy collection point data are processed innovatively by using a joint sparse coding framework. The core of the framework is that: electric energy feature dictionary atoms are generated based on the Joule law, which accurately describes the internal law of electric energy data; environment feature dictionary atoms are generated based on the mechanical vibration equation, which effectively represents the physical nature of environmental interference; and the two types of atoms with clear physical meaning are used for joint sparse coding. This process can deeply mine the structural differences and coupling relationship between electric energy data and environmental interference in the sparse domain, so as to more effectively separate and reconstruct the target electric energy data in the coding result, significantly weaken the interference of complex environmental factors (such as heat, eddy current, vibration) on the interpretation of electric energy data, and improve the accuracy of collected data. The method no longer considers the electric energy data collection in isolation, but closely couples the dynamic perception of the environment physical field state, so that the electric energy data collection has intelligent environmental perception driving, and provides a guarantee for obtaining purer and more reliable electric energy data in a complex interference environment. BRIEF DESCRIPTION OF DRAWINGS

[0022] The accompanying drawings, which are included to provide a further understanding of the embodiments of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and serve to explain the principles of the application, and do not limit the application in any way. In the drawings:

[0023] Figure 1 An optional structure diagram of an environment perception-based electric energy data collection system according to an embodiment of the application;

[0024] Figure 2 An optional flowchart of an environment perception-based electric energy data collection method according to an embodiment of the application;

[0025] Figure 3 An optional flowchart of an environment perception-based electric energy data collection method according to another embodiment of the application;

[0026] Figure 4 An optional structure diagram of an electronic device according to an embodiment of the application.

[0027] The implementation, functional features and advantages of the application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0029] Optionally, as shown in Figure 1 The application provides an environment-aware electric energy data acquisition system, comprising:

[0030] A physical field feature vector conversion module 101 is configured to acquire raw data collected by each sensor in a monitoring area in real time, and convert the raw data into a physical field feature vector comprising heat flow concentration and eddy current intensity.

[0031] A heterogeneous graph network construction module 102 is configured to construct a dynamic heterogeneous graph network based on the physical field feature vector.

[0032] A minimum perception subgraph formation module 103 is configured to solve a continuous area in which a physical field distortion gradient value exceeds a gradient threshold value by using a topological relationship of the dynamic heterogeneous graph network, and form a minimum perception subgraph by activating sensors in the continuous area.

[0033] A target electric energy collection point association module 104 is configured to associate a target electric energy collection point according to a distortion coordinate of the minimum perception subgraph by a conductor current-carrying capacity.

[0034] A joint sparse coding module 105 is configured to input electric energy data of the target electric energy collection point and environment data of the minimum perception subgraph into an encoder, and perform: generating electric energy feature dictionary atoms based on the Joule law; generating environment feature dictionary atoms based on a mechanical vibration equation; and performing joint sparse coding by using the electric energy feature dictionary atoms and the environment feature dictionary atoms.

[0035] An electric energy data reconstruction module 106 is configured to separate and reconstruct the electric energy data from a result of the joint sparse coding.

[0036] Optionally, as shown in Figure 2 The application provides an environment-aware electric energy data acquisition method, comprising:

[0037] S201, acquiring raw data collected by each sensor in a monitoring area in real time, and converting the raw data into a physical field feature vector comprising heat flow concentration and eddy current intensity.

[0038] Each sensor comprises a temperature sensor, a vibration sensor, and a magnetic field sensor.

[0039] The heat flow concentration is obtained by calculating a local area heat accumulation rate (such as a product of a temperature difference change rate of adjacent nodes and a heat conduction coefficient) based on temperature sensor data. The eddy current intensity is obtained by analyzing a magnetic field change frequency and a spatial curl (such as a circulation integral of magnetic field vectors of three adjacent points) based on magnetic field sensor data.

[0040] Exemplarily, a temperature sensor at a certain cable joint measures a temperature rise of 2°C (50% higher than the average) within 1 second, and a magnetic field sensor detects a 100 Hz high-frequency vortex magnetic field. Then the converted feature vector is [heat flow concentration = 0.85, vortex intensity = 1.2] (normalized value).

[0041] In S201, the original sensor data is converted into physical field feature vectors such as heat flow concentration (quantifying heat accumulation trend) and vortex intensity (characterizing electromagnetic loss), breaking through the limitations of traditional static parameters, and providing a calculation primitive with clear physical meaning for environmental perception by extracting the core features of dynamic physical processes (such as temperature rise rate and magnetic field curl).

[0042] S202, constructing a dynamic heterogeneous graph network based on the physical field feature vectors;

[0043] In S202, a topological structure reflecting the correlation of the environmental state of the monitoring area is established (the network nodes represent the sensor positions, and the edges represent the environmental state correlation.

[0044] S203, using the topological relationship of the dynamic heterogeneous graph network, solving the continuous area where the physical field distortion gradient value exceeds the gradient threshold value; activating the sensors in the continuous area to form a minimum perception subgraph;

[0045] Among them, the gradient threshold is a critical value for judging whether the physical field distortion is significant, which is dynamically set according to the safety margin of the conductor material.

[0046] In S203, the significant environmental abnormal area is located through topological analysis, and the sensors in this area are focused (the minimum perception subgraph is the set of sensors in the abnormal area).

[0047] S204, according to the distortion coordinates of the minimum perception subgraph, associating the target electric energy collection point through the conductor current-carrying capacity;

[0048] In S204, the environmental abnormal position is mapped to the electric energy monitoring point in the electrical system that is safely associated with it (the current-carrying capacity is used as the association basis).

[0049] S205, inputting the electric energy data of the target electric energy collection point and the environmental data of the minimum perception subgraph into the encoder, and performing: generating electric energy feature dictionary atoms based on Joule's law; generating environmental feature dictionary atoms based on mechanical vibration equation; and performing joint sparse coding using electric energy feature dictionary atoms and environmental feature dictionary atoms;

[0050] Among them, the electric energy data of the target electric energy collection point includes the current and voltage time series waveforms (such as instantaneous values with a sampling rate of 10 kHz) of the associated collection point.

[0051] The environmental data of the minimum perceptual subgraph includes data collected by sensors (temperature sensors, vibration sensors, and magnetic field sensors) in a continuous area,

[0052] In S205, a separation primitive is constructed based on physical laws (Joule's law / vibration equation) to provide a mathematical tool for mixed signal separation.

[0053] In S206, electrical energy data is separated and reconstructed from the results of joint sparse coding.

[0054] In S206, pure electrical energy data is extracted from the mixed signal by removing environmental interference.

[0055] Further, a dynamic heterogeneous graph network is constructed based on the physical field feature vector, including:

[0056] Each sensor is taken as a node and the physical field feature vector of the node position is associated;

[0057] wherein the node position is the geometric coordinates of the sensor in the monitoring area, used to construct the spatial topology relationship;

[0058] A spatial connection weight between nodes is generated, which decreases exponentially with the increase of the distance between sensors, and the decrease rate is controlled by a pre-set decay coefficient;

[0059] wherein the pre-set decay coefficient is an exponential factor controlling the spatial connection weight decay with distance, determined by historical data fitting. For example, in a cable tunnel, the measured temperature field influence radius is about 5 meters, and the fitted decay coefficient is 0.2. When the distance between two nodes is 3 meters, the spatial weight = e^(-0.2×3)≈0.55.

[0060] A physical field coupling weight between nodes is generated, and when the direction angle between the temperature gradient vector at the first node and the magnetic field gradient vector at the second node is less than 37 degrees, a coupling connection is established between the first node and the second node, and the coupling strength is proportional to the product of the modulus of the temperature gradient vector at the first node and the magnetic field gradient vector at the second node;

[0061] wherein in the electromagnetic-thermal coupling effect, when the direction angle between the temperature gradient and the magnetic field gradient is less than 37°, the ferromagnetic material eddy current heat effect is significantly enhanced to avoid false connection of non-coupling nodes (such as pseudo-association caused by random noise).

[0062] The spatial connection weight and the physical field coupling weight are fused into the final connection weight according to a dynamic blending coefficient, and the dynamic blending coefficient is adjusted according to the sensor distribution density in the monitoring area;

[0063] The sensor-dense area matching coefficient is inclined to the spatial connection weight, and the sensor-sparse area matching coefficient is inclined to the physical field coupling weight. For example, the sensors in the switch cabinet of the transformer substation are dense, and the spatial weight accounts for 90%; the sensors on the overhead line are sparse, and the physical field coupling weight accounts for 70%.

[0064] The connection structure of the dynamic heterogeneous graph network is used to guide the topological traversal path in the physical field distortion gradient calculation process.

[0065] Based on the embodiments provided in the present application, by limiting the sensor types (temperature, vibration, magnetic field) and the specific construction rules of the dynamic heterogeneous graph network (spatial connection weight exponential decay, physical field coupling weight based on temperature / magnetic field gradient direction angle < 37° establishment, dynamic matching coefficient of weight fusion), the accuracy of physical field state representation and the dynamic adaptability of network topology are significantly improved. The spatial weight decay can reflect the natural law of physical quantity decay with distance; the directional limitation (< 37°) of the coupling weight ensures that connections are established only between nodes with strong physical field interaction, effectively filtering noise correlation; the dynamic matching coefficient can automatically adjust the weights of spatial and physical field factors according to the density of sensors, so that the constructed network can more truly reflect the actual propagation and coupling characteristics of the physical field in the monitoring area, laying a solid topological foundation for subsequent accurate identification of distortion areas.

[0066] Further, using the topological relationship of the dynamic heterogeneous graph network, the continuous area with a physical field distortion gradient value exceeding a gradient threshold is calculated; the sensors in the continuous area are activated to form a minimum perception subgraph, including:

[0067] Traverse each node in the dynamic heterogeneous graph network as a current target node;

[0068] Determine a set of adjacent nodes directly connected to the current target node;

[0069] The direct connection is based on the connection relationship in the constructed dynamic heterogeneous graph network. In the dynamic heterogeneous graph network, each sensor represents a node, and whether the nodes are directly connected is determined by the final connection weight. That is, if the final connection weight of two nodes is not zero (i.e., there is an edge), and there is no other node as an intermediate medium between the two nodes (no need to indirectly connect through a third node), it is called direct connection.

[0070] Calculate the gradient vector of the physical field feature vector of the current target node and the physical field feature vector of each adjacent node;

[0071] The physical field feature vector includes heat flow concentration (such as heat flow density per unit area, unit W / m 2 ) and vortex intensity (such as vortex current density, unit A / m 2). The gradient vector is a vector describing the "rate of change" and "direction of change" of the two features in space: for the current target node (such as T3) and the adjacent node (such as M2), the two components of the gradient vector are "heat flow concentration difference" (T3 heat flow concentration - M2 heat flow concentration) and "vortex intensity difference" (T3 vortex intensity - M2 vortex intensity), and the vector direction is determined by the positive and negative and size of the two differences (reflecting the change trend).

[0072] Taking the maximum value of the modulus of all gradient vectors as the physical field distortion gradient value of the current target node;

[0073] Where the modulus of the gradient vector is the size of the vector, and the calculation formula can be analogous to the two-dimensional vector length (sqrt(heat flow concentration difference 2 + vortex intensity difference 2 )). The current target node will form a gradient vector with multiple directly connected adjacent nodes, and the maximum modulus value of these vectors is taken as the physical field distortion gradient value of the node.

[0074] Screening nodes with physical field distortion gradient values exceeding the gradient threshold as candidate distortion nodes.

[0075] In this embodiment, the gradient threshold is a critical value for judging whether a node is a candidate distortion node, which is determined by the physical field stable state of the monitoring area (preset based on historical data or safety standards), and when the physical field distortion gradient value of a node exceeds the gradient threshold, it means that the physical field feature changes dramatically, and there may be distortion.

[0076] For example, the high-power equipment area of the workshop (such as the vicinity of the welding equipment): the physical field background fluctuation is large, and the gradient threshold is set to 6.0 (corresponding to heat flow concentration change > 30W / m 2 or vortex intensity change > 20A / m 2 ); the ordinary lighting power distribution area of the workshop: the physical field is stable, and the gradient threshold is set to 2.5 (corresponding to heat flow concentration change < 10W / m 2 and vortex intensity change < 8A / m 2 , i.e. judged as non-distortion).

[0077] Based on the embodiments provided in the present application, specific implementation schemes for calculating the physical field distortion gradient value are provided (node traversal, neighborhood gradient calculation, and taking the maximum modulus value as the distortion gradient value). This method fully utilizes the topological relationship (adjacent node set) of the dynamic heterogeneous graph network, ensuring that the distortion gradient calculation is based on the spatial change of local physical field features, rather than isolated point values. By screening candidate distortion nodes, the key positions of the physical field state anomaly are preliminarily focused on, providing a reliable target node set for subsequent accurate definition of continuous distortion region, and avoiding resource waste caused by global blind calculation.

[0078] Further, breadth-first traversal is performed along edges with final connection weight greater than a weight threshold value from the candidate distortion node;

[0079] wherein the weight threshold value is a critical value for screening effective connection edges, used to screen edges with high enough connection strength from the adjacent edges of the candidate distortion node as the basis for breadth-first traversal (only traverse along edges with final connection weight greater than the value). Its value is related to the sensor distribution density of the monitoring area (high density, high threshold to avoid redundant paths; low density, low threshold to ensure continuity of traversal).

[0080] For example, the sensor dense area (such as the inside of the power distribution cabinet, 10 / m 2 ): the weight threshold value is set to 0.3 (only keep edges with high connection strength); the sensor sparse area (such as the cable trench, 2 / m 2 ): the weight threshold value is set to 0.1 (allow to keep more edges to ensure uninterrupted traversal).

[0081] In some embodiments, the candidate distortion node may have multiple directly connected adjacent nodes, but not all edges are meaningful (some edges have too low connection weight, which may be noise or weak association). Therefore, only traverse along edges with final connection weight greater than the weight threshold value to ensure that the traversal path focuses on nodes with tight physical field association.

[0082] Record the change direction of the physical field feature vector of the three consecutive visited nodes during traversal;

[0083] wherein the three visited nodes refer to the three nodes visited consecutively during breadth-first traversal

[0084] Calculate the direction change angle, and if the direction change angle of the three consecutive nodes is less than 15 degrees, mark it as an effective propagation path;

[0085] wherein the direction change angle refers to the angle between the gradient vector direction of the first two nodes and the gradient vector direction of the last two nodes in the three consecutive nodes. For example, the gradient vector direction of node A to B is θ1, and the gradient vector direction of B to C is θ2, then the direction change angle is |θ1-θ2|.

[0086] If the direction change angle of the three consecutive nodes is less than 15 degrees, it means that the change direction of the physical field distortion is stable (without sharp turns), proving that the distortion on this path is "continuous and consistent" (caused by the same reason, such as overheating of the same conductor), rather than scattered random noise, so it is marked as an effective propagation path.

[0087] Aggregate all the nodes covered by the effective propagation paths to form a continuous area;

[0088] verify whether the number of target node pairs satisfying the physical field coupling condition in the continuous region exceeds a set number threshold; wherein, the target node pair includes a temperature sensor node and a magnetic field sensor node;

[0089] wherein, the target node pair refers to a node pair satisfying the physical field coupling weight calculation condition, i.e., the pair composed of the temperature sensor node and the magnetic field sensor node, and the direction included angle between the temperature gradient vector and the magnetic field gradient vector of the two is less than 37 degrees.

[0090] The physical field coupling relationship between the nodes is established through the condition, and the number of such node pairs is verified, in order to ensure that the distortion of the continuous region is indeed caused by the electromagnetic-thermal coupling effect (rather than accidental fluctuations of a single physical field), and to enhance the reliability of the region determination.

[0091] In a specific embodiment, the effective propagation path strength is calculated based on the following formula:

[0092]

[0093] wherein, S path is the effective propagation path strength (dimensionless, 0 to 1), the larger the value, the higher the path reliability; θ1, θ2 are the direction change angles of the three consecutive nodes (unit: °), reflecting the continuity of the physical field characteristic vector change (such as θ1 is the gradient direction included angle from node A to B, and θ2 is the gradient direction included angle from node B to C); w1, w2 are the final connection weights (0 to 1) of the adjacent nodes on the path, which are calculated by the dynamic heterogeneous graph network, reflecting the physical field coupling strength between the nodes; (such as w1 is the final connection weight from node A to B, and w2 is the final connection weight from node B to C); k dens is a sensor distribution density correction factor (dimensionless, 0.7 to 1.3), taking 1.2 in a dense sensor area (>5 / m2) and 0.8 in a sparse sensor area (<2 / m2), to adapt to different deployment scenarios; d avg is the average distance (unit: m) between the nodes on the path, such as the distance between nodes A and B is 0.3m, and the distance between nodes B and C is 0.5m, then d avg = 0.4m; L0 is a reference distance (unit: m), taking the average installation distance of the sensors in the monitoring region (such as 0.5m in the power distribution area of the workshop), to determine the attenuation effect of the distance on the path reliability.

[0094] It should be noted that θ1, θ2 are the direction change angles of the three consecutive nodes, and since the effective propagation path needs to satisfy: the direction change angles of the three consecutive nodes are all less than 15 degrees, i.e., θ1<15° and θ2<15°, thus

[0095] It should be further noted that w1, w2 are the final connection weights of adjacent nodes on the effective propagation path (reflecting the strength of physical field coupling between nodes, with a value range of 0 to 1). The meaning of taking the average of the connection weights of two adjacent nodes is to take the average of the coupling strength of the two adjacent edges on the path, so as to comprehensively reflect the overall correlation degree of the path segment.

[0096] Since w1, w2 are both normalized weights with a value range of 0 to 1, the average of w1 and w2 is still within the range of 0 to 1, which can reasonably quantify the average coupling strength of the path. If the number of target node pairs in the continuous region that meet the physical field coupling condition exceeds the set number threshold, all sensors in the continuous region are activated to form a minimum perception subgraph.

[0097] The set number threshold refers to the minimum number of target node pairs in the continuous region that meet the physical field coupling condition, which is used to verify the effectiveness of the region (to ensure that there is sufficient electromagnetic-thermal coupling relationship in the region, and to prove that the distortion is systematic).

[0098] For example, the sensor-dense power distribution cabinet region (containing 10 temperature nodes and 8 magnetic field nodes): the set number threshold is 8 pairs (at least 8 coupled node pairs are required to ensure that the region has strong correlation); the sensor-sparse cable trench region (containing 3 temperature nodes and 2 magnetic field nodes): the set number threshold is 2 pairs (adapted to the scene with few nodes to avoid false judgment).

[0099] Based on the embodiments provided in the present application, the specific logic of continuous region identification and minimum perception subgraph activation is defined (breadth-first search, effective propagation path determination, node aggregation, and physical field coupling verification). The breadth-first search ensures that the exploration is performed on the path with strong physical field correlation based on the weight threshold. The determination standard of the effective propagation path (the direction change angle of the continuous three points is less than 15°) ensures that the identified path has the direction consistency of the physical field distortion propagation, effectively excluding the pseudo-continuous region formed by random fluctuations or isolated noise points. After the aggregation of the nodes forms the continuous region, the number of temperature-magnetic field sensor node pairs in the region that meet the coupling condition is further verified to confirm that the region distortion is a real anomaly jointly dominated by electromagnetic-thermal effect, rather than a single factor interference, thereby ensuring that the minimum perception subgraph formed by activation has high target specificity and physical correlation credibility.

[0100] Further, according to the distortion coordinates of the minimum perception subgraph, the target power collection points are associated through the conductor current-carrying capacity, including:

[0101] locating and monitoring the set of power collection points electrically connected to the conductors in the region;

[0102] extracting temperature data of all temperature sensors in the minimum perceptual subgraph;

[0103] calculating deviation values of the temperature data from the area average temperature;

[0104] screening temperature sensors with deviation values greater than three times the standard deviation to form a high-temperature cluster;

[0105] calculating weighted spatial barycentric coordinates of the high-temperature cluster, with the weight being the temperature deviation value;

[0106] outputting the weighted spatial barycentric coordinates as the heat aggregation core coordinates, and obtaining material parameters of conductors associated with the heat aggregation core coordinates;

[0107] screening multiple power collection points closest to the heat aggregation core coordinates in an electrically connected subnetwork; wherein the electrically connected subnetwork is divided according to circuit breaker tripping logic;

[0108] In some embodiments of the present application, three power collection points closest to the heat aggregation core coordinates can be screened;

[0109] calling the conductor resistance temperature coefficient and the safe current carrying limit value;

[0110] For each of the multiple power collection points, obtaining the real-time current value of the point; according to the conductor resistance temperature coefficient, the safe current carrying limit value and the real-time temperature of the conductor where the point is located, calculating the dynamic safe current carrying limit value of the point; and calculating the real-time current carrying margin of the point according to the real-time current value and the dynamic safe current carrying limit value;

[0111] selecting the collection point with the smallest real-time current carrying margin as the associated target.

[0112] In one embodiment, taking substation cable joint monitoring as an example: the minimum perceptual subgraph detects temperature data [85℃, 92℃, 78℃, 45℃] at the joint, and the area average temperature is 65℃ (standard deviation 15℃). Screening high-temperature sensors with deviation > 45℃ (85℃, 92℃), calculating the heat aggregation core coordinates at 2cm east of the joint with the deviation value as the weight, and associating the copper core cable (α = 0.00393). Selecting the three closest points in the electrically connected subnetwork divided by the circuit breaker: point A (0.5m), point B (1.2m), and point C (2.0m). Calculating the dynamic safe current carrying limit value: when the temperature of point A is 92℃, the limit value is corrected from 900A to 0; when the temperature of point B is 85℃, the limit value is 760A; and when the temperature of point C is 78℃, the limit value is 820A. Combined with the real-time current (810A for point A, 750A for point B, and 700A for point C), the current carrying margin is calculated to be -810A, 10A, and 120A respectively. Selecting the point A with the smallest margin (negative value indicating serious overload) as the associated target, triggering deep collection and alarm of the power data of the point. This embodiment verifies the ability of the method to accurately locate the most dangerous node in a high-temperature overload scenario.

[0113] Based on the embodiments provided in the present application, precise association is realized through the synergistic mechanism of thermal aggregation core positioning and dynamic current carrying capacity evaluation. First, based on the temperature sensor data in the minimum perception subgraph, temperature sensors with a deviation value greater than three times the standard deviation are screened to form a high temperature cluster, and non-significant temperature rise interference is excluded; the spatial barycenter coordinates are calculated by weighted temperature deviation value to lock the thermal aggregation core area, ensuring the accuracy of physical positioning. Then, the three electric energy collection points closest to the thermal aggregation core are selected in the electrically connected subnetwork (divided according to the circuit breaker tripping logic), taking into account spatial proximity and electrical relevance.

[0114] For each candidate point, the dynamic safety current carrying limit is calculated: the conductor resistance temperature coefficient and the safety current carrying limit are called, combined with the real-time temperature of the conductor at the point, to correct the current carrying capacity threshold suitable for the actual working condition; then the current carrying margin (difference between dynamic limit and actual current) is calculated through the real-time current value, and finally the point with the smallest margin is selected as the association target. The environmental anomaly (overheated core) is directly mapped to the electrically most vulnerable node (such as the overload point with a current carrying margin tending to zero or negative value), avoiding misjudgment caused by the split analysis of temperature and current data in traditional methods, and optimizing the calculation efficiency by limiting the number of candidate points.

[0115] Further, in the step of calling the conductor resistance temperature coefficient and the safety current carrying limit, mechanical deformation compensation and electromagnetic skin effect analysis are performed, as shown in Figure 3 , specifically including:

[0116] S301, based on the mechanical vibration frequency spectrum of the conductor collected by the vibration sensor in the minimum perception subgraph, identifying the characteristic harmonic component formed by the combined action of electromagnetic force and mechanical stress;

[0117] In some embodiments, the mechanical vibration time domain signal of the conductor is obtained from the vibration sensor of the minimum perception subgraph, which is converted into a frequency spectrum graph (horizontal axis for frequency, vertical axis for amplitude) through fast Fourier transform. The harmonics generated by electromagnetic force are usually related to the current frequency (such as the 100Hz double frequency of 50Hz fundamental wave due to the periodic electromagnetic repulsion of alternating current), and the harmonics generated by mechanical stress are related to the thermal expansion and contraction of the conductor and the structure resonance (such as the 30Hz natural frequency harmonic caused by temperature change). By comparing the reference frequency spectrum of pure electromagnetic force (without temperature stress) and pure mechanical stress (without current), the frequency component with significantly enhanced amplitude (such as the superimposed harmonic of 100Hz and 30Hz) under the combined action of the two forces is screened out, which is the characteristic harmonic component.

[0118] In S301, the key signals of electromagnetic-thermal-mechanical multi-physical field coupling are accurately captured by identifying the characteristic harmonics that work together, which are the "comprehensive fingerprints" of the interaction between the conductor's current-carrying state (electromagnetic force) and environmental stress (mechanical stress), providing multi-dimensional input for subsequent microscopic deformation inversion and avoiding the one-sidedness of single physical field analysis.

[0119] In S302, the microscopic deformation distribution pattern of the conductor near the thermal aggregation core coordinates is inverted according to the amplitude-frequency characteristics and phase difference of the characteristic harmonic components.

[0120] In some embodiments, the amplitude-frequency characteristics refer to the variation law of the amplitude of the characteristic harmonics with frequency (such as 100Hz harmonic amplitude 0.05mm, 30Hz harmonic amplitude 0.03mm), and the phase difference refers to the delay of different characteristic harmonics in time (such as 100Hz harmonic leading 30Hz harmonic by 10ms). By establishing a "characteristic harmonic-deformation" mapping model (which can be trained based on finite element simulation data), the amplitude-frequency characteristics are converted into deformation amplitude (the larger the amplitude, the more intense the deformation), and the phase difference is converted into deformation direction (phase advance corresponds to stretching in a certain direction, and phase lag corresponds to compression), and finally the microscopic deformation distribution of the conductor around the thermal aggregation core coordinates is inverted (such as local stretching 0.02mm due to electromagnetic force and compression 0.01mm due to thermal stress).

[0121] In S302, instead of directly relying on contact strain sensors to measure deformation (which is susceptible to environmental interference and has limited installation), non-invasive microscopic deformation monitoring is achieved through "amplitude-frequency + phase" two-dimensional inversion of vibration spectrum. This reverse logic from signal to physical state not only reduces hardware cost, but also captures subtle deformations inside the conductor that are difficult to measure directly (such as micron-level bending), providing accurate deformation basis for subsequent coordinate compensation and parameter correction.

[0122] In S303, the thermal aggregation core coordinates are compensated for three-dimensional spatial shift using the microscopic deformation distribution pattern, the deformation compensation heat source coordinates are generated, and the equivalent cross-sectional area change rate of the conductor at the deformation compensation heat source coordinates is simultaneously corrected.

[0123] In some embodiments, the microscopic deformation distribution pattern shows that the thermal aggregation core coordinates (originally calculated as X=10m, Y=2m, Z=0.5m) have shifted due to deformation (such as stretching 0.05m along the X-axis and bending 0.02m along the Z-axis), so after three-dimensional spatial shift compensation, the deformation compensation heat source coordinates are corrected to (10.05m, 2m, 0.52m). At the same time, according to the stretching / compression degree of the conductor in the deformation pattern (such as local elongation leading to a smaller cross-sectional area), the equivalent cross-sectional area change rate is calculated (such as the original cross-sectional area 100mm 2 , now 95mm 2 , change rate -5%).

[0124] In S303, the thermal aggregation core coordinate is the key reference point for subsequent current carrying capacity analysis, and the conductor deformation will cause the actual position to deviate from the original calculated position, and the equivalent cross-sectional area will also change due to deformation (affecting the current carrying capacity). This step eliminates the interference of deformation on physical parameters through coordinate compensation and cross-sectional area correction, ensuring that the subsequent resistivity calculation and current carrying capacity analysis based on the coordinate are established on the real physical state, avoiding virtual errors caused by deformation.

[0125] In S304, the real-time temperature data of the temperature sensor at the deformation compensation heat source coordinate (such as 10.05m) is extracted, and the dynamic resistivity under the influence of deformation is calculated combined with the resistance temperature characteristic of the conductor material.

[0126] In some embodiments, the real-time temperature (such as 85℃) of the temperature sensor (such as T5) corresponding to the deformation compensation heat source coordinate (such as 10.05m) is obtained, and the resistance temperature characteristic formula of the conductor material (such as copper) is called. At the same time, the equivalent cross-sectional area change rate (such as -5%) obtained in the third step is introduced, because the change of cross-sectional area will affect the current density, and then indirectly affect the resistance (under the same current, the cross-sectional area becomes smaller, the current density increases, and the resistance appears equivalent increase), and finally the dynamic resistivity containing the influence of deformation (such as 3% higher than the resistivity considering only temperature) is calculated.

[0127] In S304, the dynamic resistivity is calculated by coupling the two parameters of "temperature + deformation", so that the resistance parameter is more consistent with the real state of the conductor, providing high-precision basic data for subsequent skin effect analysis and current carrying capacity evaluation, which is a key link to realize the "environmental perception-accurate collection of electric energy data" closed loop.

[0128] Based on the embodiments provided in the present application, the preliminary processing of mechanical deformation compensation and electromagnetic skin effect analysis (identification of characteristic harmonic, inversion of deformation mode, spatial offset compensation, equivalent cross-sectional area correction, dynamic resistivity calculation) is integrated in the step of calling conductor parameters. By analyzing the characteristic harmonic components in the vibration frequency spectrum (electromagnetic force and mechanical stress jointly acting), the micro deformation distribution mode of the conductor near the thermal aggregation core is inverted. According to this, the thermal source coordinate is compensated in three-dimensional space (the deformation compensation heat source coordinate is generated) and the equivalent cross-sectional area change rate of the conductor at this position is corrected at the same time, which significantly improves the positioning accuracy of the heat source and the authenticity of the geometric state representation of the conductor. Combined with temperature data, the dynamic resistivity under the influence of deformation is calculated, which provides more accurate conductor state parameters for subsequent accurate analysis of skin effect and current carrying capacity, and weakens the interference of mechanical deformation on electrical parameter evaluation.

[0129] Further, the method further comprises:

[0130] The current harmonic spectrum monitored by the target electric energy collection point is obtained, and the dominant harmonic frequency associated with the frequency of the characteristic harmonic component of vibration is selected.

[0131] In some embodiments, the current change data over a period of time (such as 1000 current values recorded per second) is collected by the current sensor of the target power collection point, and then the current harmonic spectrum is obtained by signal processing methods (such as converting the change of current over time into different frequency current components, similar to decomposing sound into different tones).

[0132] Meanwhile, from the identified vibration characteristic harmonic components (i.e. vibration frequencies caused by electromagnetic force and mechanical stress together, such as 100Hz, 200Hz), the current harmonics with the same frequency or multiple relationship are found (such as the vibration characteristic harmonic is 100Hz, and the 100Hz harmonic in the current spectrum is selected; the vibration characteristic harmonic is 200Hz, and the 200Hz harmonic in the current spectrum is selected), and these selected current harmonics are the dominant harmonic frequencies.

[0133] The harmonic frequency of the current and the characteristic harmonic frequency of the vibration are inherently related, such as the electromagnetic force generated by alternating current changes with the current frequency, and in turn induces conductor vibration, and the frequencies of the two are often synchronized. By selecting the associated dominant harmonic frequency, the core frequency of the "current-vibration" coupling can be accurately locked, and the target for subsequent analysis of the special influence of high-frequency current on the conductor (such as skin effect) can be found, and deviation caused by analyzing irrelevant frequencies can be avoided.

[0134] The dynamic resistivity, the corrected equivalent cross-sectional area change rate, the dominant harmonic frequency, and the conductor permeability are input into the skin effect depth calculation model to solve the equivalent skin depth of the conductor under the coupling state of mechanical deformation and high-frequency current;

[0135] In some embodiments, the skin effect depth calculation model can calculate the thickness of the surface layer (i.e. skin depth) through which the high-frequency current flows in the conductor. Specifically, the larger the dynamic resistivity and the higher the dominant harmonic frequency, the smaller the skin depth (the current is more concentrated on the surface); the higher the conductor permeability, the smaller the skin depth; and the equivalent cross-sectional area change rate will affect the actual conduction space of the conductor (such as the cross-sectional area becomes smaller, and the effect of current concentration is more obvious). Through the comprehensive consideration of these factors, the "equivalent skin depth" is finally obtained.

[0136] Traditional calculation of skin effect only considers the current frequency and resistance at room temperature, ignoring the mechanical deformation (such as bending, stretching) of the conductor that may occur. This step takes into account both mechanical deformation (through equivalent cross-sectional area change rate) and dynamic resistance (including temperature effect), and calculates the skin depth that can more truly reflect the current distribution state of the conductor under actual working conditions (both high-frequency current and deformation), providing a key basis for subsequent judgment of the current carrying capacity of the conductor. Based on the equivalent skin depth and the corrected equivalent cross-sectional area change rate, the effective current carrying capacity of the conductor under harmonic working conditions is calculated;

[0137] In a specific embodiment, the expression of the skin effect depth calculation model includes:

[0138]

[0139] where δ eq is the equivalent skin depth (unit: m), reflecting the thickness of the surface layer through which high-frequency current concentrates in the conductor; ρ d is the dynamic resistivity (unit: Ω·m), including temperature and mechanical deformation effects (such as copper at 80℃ + 5% deformation is about 2.3×10 -8 Ω·m); f h is the dominant harmonic frequency (unit: Hz); μ is the magnetic permeability of the conductor (unit: (Ω·s) / m), a material inherent property (copper takes 4π×10 -7 H / m, iron takes 2×10 -4 H / m, 1H=1Ω·s (Ohm·second); k def is the deformation enhancement factor (0.3 to 0.8), taking 0.6 when the deformation rate is >5%, quantifying the intensification effect of deformation on current concentration; ε S is the equivalent cross-sectional area change rate, such as -0.08 indicating a 8% reduction in cross-sectional area; k temp is the temperature enhancement factor (unit: ℃ -1 , taking values from 0.01 to 0.03), increasing by 0.005 for every 10℃ increase in the core temperature difference (such as 40℃ temperature difference taking 0.02); ΔT core is the temperature difference between the heat aggregation core and the environment (unit: ℃), obtained by the difference between the temperature at the heat aggregation core coordinate and the average temperature of the region.

[0140] It should be noted that f h μ is the core influencing factor of skin effect, which is the phenomenon that high-frequency current generates eddy current in the conductor due to electromagnetic induction, resulting in the concentration of current to the surface of the conductor. Its essence is directly related to the current frequency and the magnetic permeability of the conductor.

[0141] Specifically, f hThe higher the frequency (unit Hz, i.e. 1 / s) is, the stronger the electromagnetic induction effect is, the more significant the "crowding out" effect of eddy current on current is, and the smaller the skin depth is. μ (unit (Ω·s) / m) is a parameter representing the magnetization ability of the conductor. The larger μ is, the stronger the induced eddy current generated by the magnetic field in the conductor is, the easier the current is to be limited on the surface, and the smaller the skin depth is. The product of f h μ comprehensively quantifies the influence of "high-frequency electromagnetic induction strength" on current distribution, and is a core variable of the classical skin effect theory, with clear physical meaning.

[0142] In some embodiments, the equivalent skin depth determines the area actually flowed by the high-frequency current in the conductor (for example, the skin depth is small, and the current only flows in a very thin layer on the surface of the conductor); and the "modified equivalent cross-sectional area change rate" reflects the actual size of the cross section of the conductor (for example, the cross section becomes small, and the space capable of accommodating the current is less).

[0143] When calculating the effective current carrying capacity, the effective area actually flowed by the current is calculated according to the equivalent skin depth (for example, the total cross section of the conductor is 100 square millimeters, but the skin depth is small, and the effective area is only 80 square millimeters), the modified equivalent cross-sectional area is combined (for example, due to deformation, the total cross section becomes 90 square millimeters, and the effective area is adjusted to 72 square millimeters accordingly), and finally the maximum current that the conductor can carry under the current harmonic working condition is calculated according to the maximum current density that the conductor material can withstand (for example, 5 amperes of current can be flowed per square millimeter), which is the effective current carrying capacity.

[0144] The current carrying capacity of the conductor is not only determined by the size of the cross section itself, but also affected by the characteristic that the high-frequency current is "concentrated on the surface" - even if the cross section is large enough, if the high-frequency current only flows on the surface layer, the surface layer may be overheated due to the over-density of the current. The effective current carrying capacity calculated by combining the skin depth and the actual cross-sectional area can truly reflect the actual carrying limit of the conductor under high-frequency harmonics, avoiding the misjudgment caused by only looking at the total cross-sectional size (for example, the total cross-sectional size is large enough, but the surface layer is already overloaded)

[0145] According to the effective current carrying capacity and the conductor allowable temperature rise threshold, the safety current carrying limit is dynamically updated;

[0146] In some embodiments, the conductor allowable temperature rise threshold is a value that the temperature of the conductor can be increased by at most compared with the ambient temperature when the conductor is in normal operation (for example, the ambient temperature is 30℃, the maximum temperature of the conductor is 70℃, and the allowable temperature rise threshold is 40℃).

[0147] During the calculation, the conductor's heat generated under the current is first estimated according to the obtained current-carrying capacity (the greater the current, the more heat, and the more obvious the temperature rise), and then it is determined whether the temperature rise corresponding to the heat exceeds the allowed temperature rise threshold: if it does not exceed, it means that the current current-carrying capacity is safe, and it can be used as the "safe current-carrying limit"; if it exceeds, it means that running under this current-carrying capacity will overheat, and the safe current-carrying limit needs to be reduced (for example, from 400 amperes to 350 amperes) until the corresponding temperature rise is within the allowed range. In this way, the safe current-carrying limit is dynamically adjusted according to the actual working conditions.

[0148] The traditional safe current-carrying limit is a fixed value (such as the "rated current-carrying capacity" marked in the manual), but in practice, the conductor's heat dissipation conditions, whether there are high-frequency harmonics, whether there is deformation, etc. will affect the current it can withstand. This step makes the safe current-carrying limit "alive" and dynamically adjusts it according to the actual temperature rise that the conductor can withstand, which can both avoid false positives caused by fixed limits (such as being limited when actually able to carry more current) and prevent overloading and overheating, and better meet the actual needs of engineering.

[0149] The real-time current-carrying margin is calculated again using the updated safe current-carrying limit.

[0150] In some embodiments, the real-time current-carrying margin is the additional current that the conductor can currently carry. When calculating, the updated safe current-carrying limit (such as 350 amperes) is subtracted from the actual current monitored by the target power collection point (such as 300 amperes), and the difference (50 amperes) is the real-time current-carrying margin. If the actual current is 360 amperes, then the margin is -10 amperes, indicating that the safe limit has been exceeded and there is an overload risk.

[0151] The real-time current-carrying margin directly reflects the conductor's "remaining carrying space" at the moment. Using the updated safe current-carrying limit for calculation ensures that this "remaining space" is based on the real value of the current working conditions (including high-frequency harmonics, deformation, temperature, etc.), rather than the theoretical value based on a fixed limit. For example, if the conductor's safe current-carrying limit is reduced due to high-frequency harmonics, even if the actual current does not exceed the original fixed limit, the margin may be negative, thereby timely warning of the risk of overload.

[0152] According to the real-time current-carrying margin recalculated, the selection result of the associated target power collection point is verified and corrected.

[0153] In some embodiments, the real-time current-carrying margin of the previously selected target power collection point (such as point A) is first checked: if the margin is positive and reasonable (such as 50 amperes), it means that the current state of this point is stable and suitable for being a collection point; if the margin is negative (such as -10 amperes), it means that this point is already overloaded, and the data collected may be affected by abnormal conditions, and needs to be reselected.

[0154] When reselecting, find other collection points (such as point B) that are closer to the heat aggregation core coordinates in the electrical communication subnetwork (the area divided according to the circuit breaker tripping logic) where the collection point is located, calculate the real-time current carrying margin (such as 20 amperes) thereof, and if the margin of point B is reasonable, correct the target collection point to point B; if the margins of all collection points in the subnetwork are negative, it indicates that the entire area may be overloaded, and it is necessary to expand the range to reselect a suitable collection point.

[0155] The selection of the target power collection point directly affects the accuracy of subsequent power data. If the original collection point is overloaded, the data thereof may contain abnormal interference (such as measurement deviation caused by overheating). The collection point is verified and corrected through the real-time current carrying margin, which can ensure that the power data used for encoding reconstruction in the subsequent stage comes from a “state healthy” collection point, thereby providing a guarantee for the reliability of the final data and avoiding errors in the entire analysis result caused by improper selection of the collection point.

[0156] Based on the embodiments provided in the present application, the skin effect analysis and dynamic updating of the safe current carrying limit value (harmonic spectrum correlation, skin depth calculation, effective current carrying capacity calculation, safe limit value updating, current carrying margin recalculation, target point verification and correction) are deepened. The dominant harmonic frequency associated with the vibration characteristic frequency in the current harmonic spectrum of the power collection point is selected, the dynamic resistivity and the corrected cross-sectional area are input into the skin effect depth calculation model, and the equivalent skin depth under the coupling state of mechanical deformation and high-frequency current is solved. Based on this, the effective current carrying capacity of the conductor under the harmonic working condition is calculated, and the safe current carrying limit value is dynamically updated according to the capacity and the allowable temperature rise threshold. The recalculation of the current carrying margin using the updated limit value and the verification / correction of the selection result of the associated target point make the association of the target power collection point more accurate and reliable, fully consider the complex influence of the combined action of high-frequency harmonic current and conductor mechanical deformation on the actual current carrying capacity of the conductor, and improve the real-time and accuracy of the safety evaluation.

[0157] Further, the power feature dictionary atom is generated based on the Joule law, and the environment feature dictionary atom is generated based on the mechanical vibration equation; the power feature dictionary atom and the environment feature dictionary atom are jointly sparse coded, including:

[0158] The heat generation of the conductor per unit time is calculated as the basic power feature according to the Joule law for the current and voltage time series data of the target power collection point;

[0159] The equivalent cross-sectional area change rate and the dynamic resistivity of the minimum perception subgraph are fused to compensate the basic power feature for the conductor deformation parameter, and the power feature dictionary atom is generated;

[0160] In some embodiments, the current and voltage time series data refers to the current and voltage records of the target power collection point over time (such as recording once per second, resulting in a series of current and voltage values). According to Joule's law, heat is generated when current flows through a conductor, and the more current, the higher the voltage, the longer the time, the more heat generated.

[0161] When calculating the heat generated per unit time, a fixed time (such as 1 second) is selected, and the average current and average voltage in this time are used to calculate the total heat generated by the conductor in this time, and then divided by the time (1 second) to get the heat generated per second (such as 500 joules of heat per second). This heat generated per second is used as the basic power feature because it directly reflects how much electrical energy is converted into heat energy in the conductor.

[0162] The equivalent cross-sectional area change rate is the proportion of the change in the cross-sectional size of the conductor due to deformation (such as bending, stretching) (such as a 8% reduction in cross-sectional area); The "dynamic resistivity" is the actual value of the conductor resistance considering temperature and deformation (such as 10% larger than at room temperature).

[0163] The heat generated by the conductor is not only related to the current and voltage, but also related to the resistance and cross-sectional size: the larger the resistance and the smaller the cross-sectional size, the more heat generated when the current flows. Therefore, when compensating the basic power feature, these two parameters need to be combined - if the equivalent cross-sectional area decreases, it means that the space through which the current flows is more crowded, and the actual heat generated will be more than the basic feature calculated; If the dynamic resistivity increases, it will also cause the actual heat generation to increase. By combining these two factors, the basic power feature is adjusted (such as the original basic feature is 500 joules / second, and after compensation it becomes 550 joules / second), to get a more accurate power feature.

[0164] Deformation of the conductor (such as slight bending during installation, thermal expansion and contraction during operation) will change its electrical properties and affect the heat generated. The traditional calculation of heat generation ignores the influence of deformation, which may lead to a mismatch between the power feature and the actual situation. This step compensates for the deformation parameter to allow the power feature to truly reflect the energy state of the conductor under complex working conditions of "both power transmission and deformation", providing more reliable basic data for subsequent analysis.

[0165] Extract the time-domain vibration waveform of the vibration sensor in the minimum perceptual subgraph, and solve the mechanical energy transmission efficiency based on the mechanical vibration equation;

[0166] In some embodiments, the time-domain vibration waveform refers to a curve recorded by the vibration sensor, in which the amplitude of the conductor vibration changes over time (e.g., the amplitude is large at a certain time and small at a certain time). The mechanical vibration equation is a formula for describing the vibration law of an object, which can reflect how the vibration energy is transmitted from one point to another (e.g., the greater the input force, the greater the vibration amplitude; there will be energy loss in the transmission process, and the amplitude will decrease).

[0167] In combination with the topological connection strength of the dynamic heterogeneous graph network on the effective propagation path, the environmental mechanical vibration energy is corrected for network propagation loss to generate environmental feature dictionary atoms;

[0168] In some embodiments, the topological connection strength of the dynamic heterogeneous graph network refers to the tightness of the connection between nodes (sensors) in the network (e.g., a high connection strength indicates that the physical field correlation at the positions of the two sensors is strong, and the vibration energy transmission loss is small); the effective propagation path refers to the path identified in claim 5 for continuous propagation of physical field distortion (e.g., the path from sensor A to B to C).

[0169] When correcting the environmental mechanical vibration energy, the energy loss is calculated according to the topological connection strength between each node along the effective propagation path (low connection strength, high loss; high connection strength, low loss). For example, when the vibration energy is transmitted from A to B, the connection strength is 0.8 (loss of 20%), and when it is transmitted from B to C, the connection strength is 0.6 (loss of 40%). Therefore, the total loss from A to C is 20% + 40%, and the original vibration energy after such loss correction is the actual vibration energy propagated to each node. These corrected vibration energies are used as "environmental feature dictionary atoms" (i.e., the basic unit of environmental features).

[0170] The vibration energy will be lost due to distance, obstacles, etc. during propagation, and the topological connection strength of the dynamic heterogeneous graph can accurately reflect the size of such loss (nodes with close correlation have small loss). Through this correction, the environmental feature dictionary atoms not only contain the vibration energy itself, but also incorporate spatial propagation information, which can more accurately reflect the actual state of the environmental features at different positions and avoid distortion of the environmental features due to neglecting the propagation loss.

[0171] An intersection mapping relationship between the electrical energy feature dictionary atoms and the environmental feature dictionary atoms is established, and an atom association is established when the difference between the harmonic frequency in the electrical energy feature dictionary atom and the resonance frequency in the environmental feature dictionary atom is less than a set difference threshold;

[0172] In some embodiments, the "electric energy feature dictionary atom" is a compensated electric energy feature base unit (e.g., containing a harmonic frequency of 100 Hz) obtained in step 2; and the "environment feature dictionary atom" is a loss-corrected environment feature base unit (e.g., containing a resonance frequency of 102 Hz, at which the conductor vibrates most violently) obtained in step 4.

[0173] The difference threshold is a pre-set frequency difference (e.g., 5 Hz, 10 Hz). When the difference between the harmonic frequency of the electric energy feature (100 Hz) and the resonance frequency of the environment feature (102 Hz) is less than the threshold (5 Hz), it is considered that the two atoms are inherently related (e.g., the current harmonic of 100 Hz triggers the vibration resonance of 102 Hz), thereby establishing the atom association. The interaction between electric energy and environment is often reflected in the association of frequencies (e.g., a current of a specific frequency can trigger a vibration of a specific frequency). By establishing the atom association through the frequency difference, the inherent relationship can be accurately captured, and the originally independent electric energy feature and environment feature can be "bound" to provide a structured basis for subsequent joint analysis, avoiding confusion between unrelated electric energy and environment features during analysis.

[0174] A joint sparse coding objective function is constructed, and a sparse coefficient coupling constraint term of the associated atom pair is added to the objective function, which forces the difference between the sparse coefficients of the associated atom pair to be less than a set tolerance threshold.

[0175] In a specific embodiment, the expression of the joint sparse coding objective function is:

[0176]

[0177] wherein J * is the minimum objective function value after optimization, and the numerical size of J * directly reflects the coding quality: in a certain power distribution monitoring scenario, when J * <0.5 (unit A 2 +(m / s) 2 ), it is determined that the coding result meets the engineering requirements (reconstruction error <5%, sparsity >80%, correlation coefficient deviation <0.1), without the need for secondary optimization; α is the electric energy feature sparse coefficient (k x 1 vector), and the elements in α represent the combination weights of each atom in D; β is the environment feature sparse coefficient (l x 1 vector), and the elements in β represent the combination weights of each atom in E; D is the electric energy feature dictionary (n x k matrix, unit A), which is composed of k electric energy feature atoms (each atom is an n x 1 vector); E is the environment feature dictionary (m x l matrix, unit m / s), which is composed of l environment feature atoms;

[0178] X is the real power data (e.g. the current sampling sequence in a certain 10 minutes, dimension n x 1, unit A); Y is the real environment data (e.g. the vibration speed sampling sequence in the corresponding time period, dimension m x 1, unit m / s); Ω is the index set of the associated atomic pairs, recording the physically meaningful atomic pairs; for example, the atomic (index 3) corresponding to the 100 Hz harmonic in the power feature and the atomic (index 5) corresponding to the 100 Hz resonance in the environment vibration form a pair (3, 5), Ω = {(3, 5), (7, 9),...}; α Ω is the power coefficient associated with the environment feature extracted from α (for example, the first pair in Ω has the index (3, 5), then α Ω contains α3); β Ω is the environment coefficient associated with the power feature extracted from β (corresponding to the above example, β Ω contains β5); γ1 is the sparse weight (value 0.2) for balancing the "reconstruction accuracy" and "sparsity". The larger γ1 is, the more emphasis on sparsity (possibly at the expense of part of the reconstruction accuracy); otherwise, more emphasis on reconstruction accuracy; γ2 is the association weight (value 0.15) for controlling the strength of the association constraint, the larger γ2 is, the more α Ω and β Ω are close (for example, the 100 Hz harmonic coefficient and the resonance coefficient are almost equal), which strengthens the coupling relationship between the power and the environment features; ||.||1 represents the L1 norm (also known as Manhattan norm); represents the square of the L2 norm (also known as the square of the Euclidean norm).

[0179] It should be noted that the matrix D (n x k) is multiplied by the vector α (k x 1), and the result Dα has a dimension of n x 1 (vector), which is completely consistent with the dimension of X (n x 1), so X-Dα is a vector subtraction (corresponding element subtraction);

[0180] Similarly, the matrix E (m x l) is multiplied by the vector β (l x 1), and the result Eβ has a dimension of m x 1 (vector), which is consistent with the dimension of Y (m x 1), and Y-Eβ is also a legal vector subtraction.

[0181] It should be explained that α is a k x 1 sparse coefficient vector, and β is a l x 1 sparse coefficient vector, so ||α||1 and ||β||1 are the L1 norms (Manhattan norms) of the vectors, which quantify the sum of the absolute values of the vector elements.

[0182] The dimension of X-Dα is n x 1 (vector), and the dimension of Y-Eβ is m x 1 (vector), so 、 is the square of the L2 norm of the vector (the square of the Euclidean norm), which quantifies the sum of the squares of the vector elements.

[0183] The classical definitions of L1 and L2 norms apply to vector spaces, where all norms are applied to vectors.

[0184] In some embodiments, joint sparse coding is a data processing method for extracting key information from a large number of features (such as finding the core features that best represent complex power and environmental data from them); the "objective function" is the rule that guides this extraction process, ensuring that the extracted information is both concise and accurately reflects the original data.

[0185] Sparse coefficients refer to the "importance" of each feature during coding (for example, a large coefficient indicates that this feature is more critical); and the associated atomic pair is the power and environmental feature atom that establishes the connection. Adding the "sparse coefficient coupling constraint term" means that the importance of the associated atomic pair cannot differ greatly - for example, if the sparse coefficient of the power atom is 0.8, the sparse coefficient of the environmental atom must be between 0.7 and 0.9. This ensures that the two atoms associated with each other are "strong and weak" during coding, and the specific value of the tolerance threshold set in this embodiment is not limited.

[0186] Without coupling constraints, joint coding may result in a situation where "power features are important but environmental features are ignored", which is not consistent with reality (as the two are associated). By constraining the sparse coefficients of the associated atomic pair, the importance of the power and environmental features in the coding result can be ensured to be consistent, reflecting their true coupling relationship, avoiding the coding result from deviating from physical laws, and improving the accuracy of subsequent data reconstruction.

[0187] The updated safe current carrying limit is converted into an amplitude regularization factor, which is added to the boundary constraint condition of the objective function.

[0188] In some embodiments, the updated safe current carrying limit is the maximum safe carrying current of the conductor after dynamic adjustment (such as 350 amperes); and the amplitude regularization factor is a parameter used to limit the size of the data (for example, the larger the factor, the more severe the penalty for data exceeding the limit).

[0189] When converting the safe current carrying limit into a regularization factor, the smaller the limit (indicating that the conductor can carry less current), the larger the regularization factor. After adding this factor to the boundary constraint condition of the objective function, when the amplitude of the power data (such as the current size) in the result of joint sparse coding exceeds the safe current carrying limit, the regularization factor will "punish" this result, forcing the coding process to adjust, and the final coding result will not have the amplitude of the power data exceeding the safe current carrying limit.

[0190] Based on the embodiments provided in the present application, the specific implementation of joint sparse coding is refined (electric energy feature compensation generation, environment feature correction generation, cross mapping association, coupling constraint term, current carrying limit regularization). Based on the calculation of the basic electric energy feature (heat generation) based on the Joule law, the compensation is performed by fusing the conductor deformation parameters (equivalent cross-sectional area change rate, dynamic resistivity), and the generated electric energy feature dictionary atom more truly reflects the electric energy loss characteristics of the deformed conductor. For environmental vibration data, based on the mechanical vibration equation to solve the mechanical energy transmission efficiency, combined with the dynamic graph network topology connection strength to correct the propagation loss, the generated environment feature dictionary atom more accurately represents the transmission and attenuation of vibration energy in the network. The cross mapping relationship (difference < threshold value when associated) between the electric energy atom (harmonic frequency) and the environment atom (resonant frequency) is established, and the sparse coefficient coupling constraint term (the difference between the coefficients does not exceed the tolerance) of the associated atom pair is added in the joint sparse coding objective function. From a mathematical point of view, the synergy of physically associated atoms in sparse representation is strengthened, and the accuracy of feature separation is improved. The updated safe current carrying limit is converted into an amplitude regularization factor to add boundary constraints, so that the coding process naturally meets the electrical boundary conditions of the safe operation of the conductor.

[0191] Further, separate and reconstruct the electric energy data from the results of joint sparse coding, including:

[0192] Parse the coefficient matrix of joint sparse coding, and separate the electric energy coefficient subset according to the attribute labels of the electric energy feature dictionary atom and the environment feature dictionary atom;

[0193] Use the linear combination of the electric energy coefficient subset on the electric energy feature dictionary to preliminarily reconstruct the current and voltage waveforms;

[0194] Extract the vibration spectrum features related to the conductor current carrying capacity in the environment feature dictionary atom;

[0195] According to the high-frequency attenuation characteristics in the vibration spectrum features, the skin effect distortion correction is performed on the preliminarily reconstructed waveforms;

[0196] Use the updated safe current carrying limit as the amplitude constraint boundary to perform amplitude normalization processing on the corrected waveforms;

[0197] Output the final electric energy data reconstruction result after the skin effect correction and current carrying capacity constraint.

[0198] Based on the embodiments provided in the present application, the enhancement steps (coefficient separation, preliminary reconstruction, skin effect correction, current carrying capacity constraint, final output) of the electric energy data separation and reconstruction are clarified. After separating the coefficients and preliminarily reconstructing the waveform, the vibration spectrum features (especially the high-frequency attenuation characteristics) related to the current carrying capacity of the conductor in the environmental feature atoms are extracted, and the preliminarily reconstructed electric energy waveform (current / voltage) is corrected for skin effect distortion based on the vibration spectrum features, so as to effectively compensate for the waveform distortion caused by the skin effect of the high-frequency current. Further, the updated safe current limit is used as the amplitude constraint boundary for amplitude normalization processing, so as to ensure that the finally reconstructed electric energy data not only has more accurate waveform form (after skin correction), but also strictly conforms to the safe current carrying capacity boundary of the conductor in the current state (considering deformation and skin effect), so that the finally reconstructed electric energy data has more pure, more reliable and engineering safety constraint results.

[0199] It should be noted that, in the present application, the embodiments implemented on the electric energy data acquisition system based on environmental perception can be mutually referenced with the embodiments implemented on the electric energy data acquisition method based on environmental perception, and the present application will not be repeated here.

[0200] According to another aspect of the embodiments of the present application, an electronic device for implementing the above-mentioned xx method is also provided, which can be Figure 4 a terminal device or a server. The present embodiment is described by taking the electronic device as a server. As shown in Figure 4 The electronic device includes a memory 402, a processor 404 and a transmission device 406, the memory 402 stores a computer program, and the processor 404 is configured to execute the steps in any of the above-mentioned method embodiments by the computer program.

[0201] Optionally, in the present embodiment, the above-mentioned electronic device can be located in at least one network device of a plurality of network devices of a computer network.

[0202] Optionally, the transmission device 406 is used to receive or send data via a network. The above-mentioned network can include a wired network and a wireless network. In one example, the transmission device 406 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices and routers through a network cable, so as to communicate with the Internet or a local area network. In one example, the transmission device 406 is a radio frequency (Radio Frequency, RF) module, which is used to communicate with the Internet in a wireless manner.

[0203] In addition, the electronic device further includes a display 408 for displaying the target identification character contained in the identity of the identified target object, and a connection bus 410 for connecting each module component in the electronic device.

[0204] The above merely describes the preferred embodiments of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation, or direct or indirect application in other related technical fields, based on the content of the present application specification and drawings, are also included in the patent protection scope of the present application.

Claims

1. A power data acquisition system based on environmental perception, characterized in that, include: The physical field feature vector conversion module is used to acquire raw data collected by various sensors within the monitoring area in real time and convert the raw data into physical field feature vectors that include heat flux concentration and eddy current intensity. A heterogeneous graph network construction module is used to construct a dynamic heterogeneous graph network based on the physical field feature vector; The minimum sensing subgraph forming module is used to utilize the topological relationship of the dynamic heterogeneous graph network to solve the continuous region where the physical field distortion gradient value exceeds the gradient threshold; and to activate the sensors in the continuous region to form the minimum sensing subgraph. The target power acquisition point association module is used to associate the target power acquisition point with the conductor current carrying capacity based on the distortion coordinates of the minimum sensing sub-map. The joint sparse coding module is used to input the power data of the target power acquisition point and the environmental data of the minimum sensing sub-map into the encoder, and perform the following: generating power feature dictionary atoms based on Joule's law; generating environmental feature dictionary atoms based on mechanical vibration equations; and performing joint sparse coding using the power feature dictionary atoms and the environmental feature atoms. The power data reconstruction module is used to separate and reconstruct the power data from the results of joint sparse coding.

2. A method for acquiring electrical energy data based on environmental perception, characterized in that, include: The system acquires raw data collected by various sensors within the monitoring area in real time and converts the raw data into physical field feature vectors that include heat flux concentration and eddy current intensity. A dynamic heterogeneous graph network is constructed based on the physical field feature vectors; Using the topological relationship of the dynamic heterogeneous graph network, the continuous region where the physical field distortion gradient value exceeds the gradient threshold is solved; the sensors in the continuous region are activated to form a minimum sensing subgraph. Based on the distorted coordinates of the minimum sensing sub-graph, the target power acquisition point is associated with the conductor's current-carrying capacity; The power data from the target power acquisition point and the environmental data from the minimum sensing sub-map are input into the encoder, and the following steps are performed: generating power feature dictionary atoms based on Joule's law; generating environmental feature dictionary atoms based on mechanical vibration equations; and performing joint sparse coding using the power feature dictionary atoms and the environmental feature atoms. The electrical energy data is separated and reconstructed from the results of joint sparse coding.

3. The method for acquiring electrical energy data based on environmental perception according to claim 2, characterized in that, The sensors include a temperature sensor, a vibration sensor, and a magnetic field sensor; the construction of a dynamic heterogeneous graph network based on the physical field feature vector includes: Treat each sensor as a node and associate it with the physical field feature vector of that node's location; Spatial connection weights are generated between nodes, and these spatial connection weights decrease exponentially as the sensor spacing increases, with the rate of decrease controlled by a pre-calibrated attenuation coefficient. The physical field coupling weight between the nodes is generated. When the angle between the temperature gradient vector at the first node and the magnetic field gradient vector at the second node is less than 37 degrees, a coupling connection is established between the first node and the second node. The coupling strength is proportional to the product of the magnitudes of the temperature gradient vector at the first node and the magnetic field gradient vector at the second node. The spatial connection weight and the physical field coupling weight are fused together according to a dynamic ratio coefficient to form the final connection weight. The dynamic ratio coefficient is adjusted according to the sensor distribution density within the monitoring area. The connection structure of the dynamic heterogeneous graph network is used to guide the topology traversal path in the process of solving the physical field distortion gradient.

4. The method for acquiring electrical energy data based on environmental perception according to claim 3, characterized in that, The topological relationship of the dynamic heterogeneous graph network is used to solve the continuous region where the physical field distortion gradient value exceeds the gradient threshold. Activating sensors within the continuous region to form a minimal sensing sub-map includes: Each node in the dynamic heterogeneous graph network is traversed as the current target node; Determine the set of adjacent nodes that are directly connected to the current target node; Calculate the gradient vector of the physical field feature vector of the current target node and the physical field feature vector of each neighboring node; Take the maximum value of the magnitude of all gradient vectors as the physical field distortion gradient value of the current target node; Nodes whose physical field distortion gradient values ​​exceed the gradient threshold are selected as candidate distortion nodes.

5. The method for acquiring electrical energy data based on environmental perception according to claim 4, characterized in that, Starting from the candidate distorted node, perform a breadth-first traversal along the edges whose final connection weight is greater than the weight threshold value; During the traversal, record the direction of change of the physical field feature vectors of three consecutive visited nodes; Calculate the direction change angle. If the direction change angle of three consecutive nodes is less than 15 degrees, then it is marked as a valid propagation path. The nodes covered by all valid propagation paths are aggregated to form the continuous region; Verify whether the number of target node pairs satisfying the physical field coupling condition within the continuous region exceeds a set threshold; wherein, the target node pair includes a temperature sensor node and a magnetic field sensor node; If the number of target node pairs satisfying the physical field coupling condition in the continuous region exceeds the set threshold, then all sensors in the continuous region are activated to form a minimum sensing subgraph.

6. The method for acquiring electrical energy data based on environmental perception according to claim 5, characterized in that, The step of associating the target power acquisition point with the conductor's current-carrying capacity based on the distortion coordinates of the minimum sensing sub-map includes: Locate the set of power collection points that are electrically connected to the conductors of the monitoring area; Extract temperature data from all temperature sensors within the minimum sensing sub-map; Calculate the deviation between the temperature data and the regional average temperature; Temperature sensors with a deviation value greater than three standard deviations are selected to form a high-temperature cluster. Calculate the weighted spatial centroid coordinates of the high-temperature cluster, with the weight being the temperature deviation value; The weighted spatial centroid coordinates are output as the thermal accumulation core coordinates, and the material parameters of the conductor associated with the thermal accumulation core coordinates are obtained. In the electrical connectivity subnet, select multiple power collection points that are closest to the coordinates of the heat accumulation core; wherein, the electrical connectivity subnet is divided according to the circuit breaker tripping logic; Invoke the conductor's temperature coefficient of resistance and safe current carrying limit; For each of the plurality of power acquisition points, the real-time current value of that point is obtained; the dynamic safe current carrying limit of that point is calculated based on the conductor resistance temperature coefficient, the safe current carrying limit, and the real-time temperature of the conductor at that point; and the real-time current carrying margin of that point is calculated based on the real-time current value and the dynamic safe current carrying limit. The sampling point with the smallest real-time current carrying capacity is selected as the correlation target.

7. The method for acquiring electrical energy data based on environmental perception according to claim 6, characterized in that, The step of invoking the conductor resistance temperature coefficient and safe current carrying limit incorporates mechanical deformation compensation and electromagnetic skin effect analysis, specifically including: Based on the conductor mechanical vibration spectrum collected by the vibration sensor in the minimum sensing sub-graph, the characteristic harmonic components formed by the combined action of electromagnetic force and mechanical stress are identified. Based on the amplitude-frequency characteristics and phase difference of the characteristic harmonic components, the microscopic deformation distribution pattern of the conductor near the coordinates of the thermal accumulation core is inverted. Using the microscopic deformation distribution pattern, the coordinates of the heat accumulation core are compensated for in three-dimensional space to generate the coordinates of the deformation-compensated heat source, and the equivalent cross-sectional area change rate of the conductor at the coordinates of the deformation-compensated heat source is corrected simultaneously. Real-time temperature data from the temperature sensor at the coordinates of the deformation compensation heat source is extracted, and the dynamic resistivity under the influence of deformation is calculated by combining the resistance-temperature characteristics of the conductor material.

8. The method for acquiring electrical energy data based on environmental perception according to claim 7, characterized in that, The method further includes: Obtain the current harmonic spectrum monitored at the target power acquisition point, and select the dominant harmonic frequency associated with the frequency of the vibration characteristic harmonic component; The dynamic resistivity, the corrected equivalent cross-sectional area change rate, the dominant harmonic frequency, and the conductor permeability are input into the skin effect depth calculation model to solve the equivalent skin depth of the conductor under mechanical deformation and high-frequency current coupling. Based on the equivalent skin depth and the corrected rate of change of equivalent cross-sectional area, the effective current carrying capacity of the conductor under harmonic conditions is calculated. The safe current carrying limit is dynamically updated based on the effective current carrying capacity and the conductor's allowable temperature rise threshold. Recalculate the real-time current margin using the updated safe current carrying limits; Based on the recalculated real-time current carrying capacity, the selection results of the associated target power acquisition points are verified and corrected.

9. The method for acquiring electrical energy data based on environmental perception according to claim 7, characterized in that, The generation of electrical energy characteristic dictionary atoms based on Joule's law; the generation of environmental characteristic dictionary atoms based on mechanical vibration equations; Joint sparse coding using the electrical energy feature dictionary atoms and the environmental feature dictionary atoms includes: Based on the current and voltage time-series data of the target power acquisition point, the heat generated per unit time in the conductor is calculated according to Joule's law as the basic power characteristic. By integrating the equivalent cross-sectional area change rate of the minimum perceptual subgraph and the dynamic resistivity, conductor deformation parameter compensation is performed on the basic electrical energy characteristics to generate the electrical energy characteristic dictionary atoms; Extract the time-domain vibration waveform of the vibration sensor in the minimum sensing sub-graph, and calculate the mechanical energy transfer efficiency based on the mechanical vibration equation; By combining the topological connection strength of the dynamic heterogeneous graph network on the effective propagation path, network propagation loss correction is performed on the environmental mechanical vibration energy to generate the environmental feature dictionary atoms.

10. The method for acquiring electrical energy data based on environmental perception according to claim 9, characterized in that, The method further includes: Establish a cross-mapping relationship between the atoms in the electrical energy feature dictionary and the atoms in the environmental feature dictionary. When the difference between the harmonic frequency in the atoms in the electrical energy feature dictionary and the resonant frequency in the atoms in the environmental feature dictionary is less than a set difference threshold, establish an atom association. Construct a joint sparse coding objective function, and add a sparse coefficient coupling constraint term for associated atom pairs to the objective function. The sparse coefficient coupling constraint term forces the difference in sparse coefficients of associated atom pairs to not exceed a set tolerance threshold. The updated safe current carrying limit is converted into an amplitude regularization factor and added to the boundary constraints of the objective function.

Images