Power transmission channel cooperative networking observation method based on multi-source data fusion

Abstract

The application discloses a power transmission channel cooperative networking type observation method based on multi-source data fusion and relates to the technical field of power system safety. The method comprises the following steps: matching radar echo and geographic information data coordinates, combining real-time weather micro-disturbance, and performing dynamic pre-distortion compensation to obtain a three-dimensional flow field; generating a scanning priority sequence based on the energy transmission topological relationship between convective vortex elements; controlling multiple networking radars to perform scanning according to the sequence to obtain vertical structure data; constructing a spatial acceleration structure for the data, and reconstructing a three-dimensional wind field parameter through a reverse projection sampling algorithm. The application is used to solve the problems of detection positioning distortion under strong convective weather, lack of evolution prediction for multi-target cooperative scheduling, and low calculation efficiency of high-resolution data reconstruction. The application improves the spatial positioning accuracy and system response real-time performance of extreme weather early warning of the power transmission channel.

Description

Technical Field

[0001] This invention relates to the field of power system security technology, and more specifically, to a collaborative network-based observation method for power transmission channels based on multi-source data fusion. Background Technology

[0002] Utilizing multi-source detection data from multiple weather radars, radiosonde stations, and satellites for coordinated observation of power transmission channels is a primary technical means of capturing the physical damage caused to towers and cables by extreme weather events such as severe convection and hail.

[0003] Currently, mainstream solutions in the industry typically employ multi-radar network detection combined with spatial alignment technology from Geographic Information Systems (GIS). At the data processing level, coordinate transformation is often performed using time synchronization algorithms based on a fixed frame rate and a standard atmospheric refraction physical model. For observation task allocation, centralized control nodes are primarily used for serial scheduling. In the 3D wind field reconstruction stage, forward-spraying interpolation algorithms are commonly used to map polar coordinate base data to a Cartesian coordinate grid.

[0004] However, existing technologies still have significant limitations in terms of observation accuracy and real-time performance under extreme weather conditions. First, the sampling rates of heterogeneous detection devices vary greatly, and traditional frame alignment methods are prone to producing severe inter-frame spatiotemporal ghosting when processing rapidly evolving meteorological cells. Second, standard atmospheric refraction models cannot dynamically compensate for nonlinear refraction distortion caused by local meteorological perturbations (abrupt changes in temperature, humidity, and pressure gradients), leading to deviations in 3D spatial positioning. Third, existing cell identification logic often treats each convective vortex element in isolation, ignoring the topological evolutionary relationships of thermodynamic energy transfer between cells, making it difficult to accurately assess the probability of large-scale eruptions. Finally, centralized scheduling strategies are prone to computational bottlenecks and strategy deadlocks in multi-target concurrent scenarios, and traditional 3D reconstruction interpolation algorithms suffer from severe memory mutex contention under multi-threaded high-resolution processing, causing the system to easily trigger memory overflows or response delays during large-scale computations, making it difficult to meet the urgent need for high spatiotemporal resolution observations in power transmission channels. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a collaborative network observation method for power transmission channels based on multi-source data fusion. This method addresses the problems of nonlinear deviation of radar detection coordinates, lack of evolution prediction in multi-target collaborative scheduling, and computational performance bottlenecks in high-resolution three-dimensional wind field reconstruction under strong convective conditions by using real-time dynamic compensation for meteorological micro-disturbances in the troposphere, generation of topological sequences for energy transfer of convective vortex elements, and reverse projection sampling reconstruction based on spatial acceleration structures.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A collaborative network-based monitoring method for power transmission channels based on multi-source data fusion includes the following steps: Radar echo data and geographic information data are matched for coordinates, and dynamic pre-distortion compensation is performed in conjunction with real-time meteorological perturbation parameters to obtain three-dimensional flow field data. Based on the three-dimensional flow field data, multiple convective vortex elements are identified, and a scanning priority sequence is generated according to the energy transfer topology relationship between each convective vortex element. Multiple networked radars are controlled to perform scanning according to the scanning priority sequence to obtain convective vertical structure data. A spatial acceleration structure is constructed from the convective vertical structure data, and three-dimensional wind field parameters are reconstructed through a back-projection sampling algorithm.

[0007] In a preferred embodiment, before performing coordinate matching between radar echo data and geographic information data, the process includes: converting numerical change events in radiosonde data, radar echo data, and satellite observation data with different sampling rates into asynchronously triggered pulse spike signals; performing time-domain integration on the pulse spike signals to eliminate inter-frame spatiotemporal ghosting and generate a continuous time-domain meteorological feature stream.

[0008] In a preferred embodiment, the step of converting numerical change events in the radiosonde data, radar echo data, and satellite observation data with different sampling rates into asynchronously triggered pulse spike signals includes: setting independent data change thresholds for each device that collects the radiosonde data, radar echo data, and satellite observation data; monitoring the observation values ​​of each device in real time, and generating a pulse spike signal at the corresponding timestamp in the continuous time domain when the change amplitude of the observation value exceeds the corresponding data change threshold; and inputting the pulse spike signal into the neurons of a spiking neural network to update the neuron membrane potential to characterize the dynamic evolution of meteorological characteristics.

[0009] In a preferred embodiment, the real-time meteorological perturbation parameters include temperature perturbation data, humidity perturbation data, and air pressure perturbation data. The step of combining the real-time meteorological perturbation parameters to perform dynamic pre-distortion compensation to obtain three-dimensional flow field data includes: establishing a unified data format compatible with the geographic information data and radar echo data, and containing a three-dimensional mapping grid; extracting the temperature perturbation data, humidity perturbation data, and air pressure perturbation data, and calculating the real-time nonlinear refractive index; generating a spatial phase conjugate matrix based on the real-time nonlinear refractive index, and performing dynamic distortion pre-distortion on the three-dimensional mapping grid in the unified data format to compensate for coordinate offsets caused by abrupt changes in atmospheric refraction.

[0010] In a preferred embodiment, the calculation of the real-time nonlinear refractive index and the generation of a spatial phase conjugate matrix based on the real-time nonlinear refractive index include: calculating spatial temperature gradient, humidity gradient, and pressure gradient data based on the temperature perturbation data, humidity perturbation data, and pressure perturbation data; converting the temperature gradient, humidity gradient, and pressure gradient data into a three-dimensional spatial refractive index perturbation field based on an atmospheric refractive perturbation physical model; performing path integration on the three-dimensional spatial refractive index perturbation field along the propagation path of the radar beam with the radar source as the origin to calculate the spatial phase delay, and then constructing a spatial phase conjugate matrix for inverse distortion of the three-dimensional mapping mesh.

[0011] In a preferred embodiment, generating a scanning priority sequence based on the energy transfer topology between each convective vortex element includes: identifying each convective vortex element within the radar detection range as a node in the topological network, setting the atmospheric moisture transport and wind shear paths between each convective vortex element as edges, and constructing a dynamic topological graph; parameterizing meteorological thermodynamic energy transfer as the infection rate between nodes in the topological graph using an infectious disease dynamics network model; calculating the probability weight of large-scale strong convective outbreaks induced by the convective vortex element corresponding to each node based on the infection rate, and generating the scanning priority sequence by sorting them in descending order.

[0012] In a preferred embodiment, the step of parameterizing meteorological thermodynamic energy transfer into the transmission rate between nodes in the topology graph through the infectious disease dynamics network model includes: calculating the wind shear intensity and water vapor flux divergence between convective vortex elements corresponding to adjacent nodes in the dynamic topology graph based on the three-dimensional flow field data; determining whether the wind shear intensity and water vapor flux divergence satisfy a preset energy interaction condition; if so, establishing directed edges representing energy transfer channels between adjacent nodes; and normalizing the water vapor flux divergence and wind shear intensity corresponding to the directed edges, mapping them to the transmission rate corresponding to the transformation from susceptible state nodes to infected state nodes in the infectious disease dynamics network model.

[0013] In a preferred embodiment, controlling multiple networked radars to perform scanning according to the scanning priority sequence includes: employing a decentralized multi-agent collaborative allocation strategy, configuring an independent agent node for each radar, and setting the task reward weight corresponding to each scanning task in the scanning priority sequence; each agent node of the networked radar independently initiates bidding for each scanning task based on its own mechanical rotation cost, observation blind zone, and expected observation reward; and conducting multiple rapid bidding processes through a non-cooperative game algorithm until the network composed of all agent nodes reaches a Nash equilibrium state, at which point each networked radar executes the corresponding scanning task according to the task assignment at the time of reaching the Nash equilibrium state.

[0014] In a preferred embodiment, constructing a spatial acceleration structure from the convective vertical structure data and reconstructing the three-dimensional wind field parameters using a back-projection sampling algorithm includes: extracting polar coordinate radar base data from the convective vertical structure data and constructing the polar coordinate radar base data into a hierarchical bounding box tree structure; using each grid point on a preset three-dimensional rectangular target grid as a ray starting point, emitting back-acquisition rays towards the hierarchical bounding box tree structure; and collecting and fusing effective meteorological values ​​around each grid point through intersection operations between the back-acquisition rays and the hierarchical bounding box tree structure to generate the corresponding three-dimensional wind field parameters.

[0015] In a preferred embodiment, effective meteorological data around each grid point are collected and fused through the intersection operation between the reverse acquisition ray and the hierarchical bounding box tree structure. This includes: calculating the intersection state between the reverse acquisition ray and the axial bounding boxes at each level in the hierarchical bounding box tree structure; when the reverse acquisition ray passes through the bounding box of a leaf node, acquiring the polar coordinate radar base data points contained in that leaf node; using a three-dimensional spatial distance weighting algorithm, with the Euclidean distance from the ray's starting point to each polar coordinate radar base data point as the weight, performing inverse distance weighted fusion calculation on the extracted effective meteorological data, and writing the fusion result into the corresponding grid point of the three-dimensional rectangular target grid in a lock-free concurrent manner.

[0016] The technical effects and advantages of this invention's multi-source data fusion-based collaborative network-based observation method for power transmission channels are as follows: This invention effectively eliminates positioning errors caused by atmospheric nonlinear refraction under strong convection conditions by combining real-time meteorological perturbation parameters with dynamic pre-distortion compensation, thereby improving the spatial accuracy of three-dimensional flow field data from the ground up. By utilizing the energy transfer topology relationship between each convective vortex element to generate a scanning priority sequence, it realizes the transformation from traditional isolated single-unit observation to intelligent collaborative scheduling based on evolutionary trends, ensuring accurate capture of high-threat meteorological targets. Combined with a spatial acceleration structure and a back-projection sampling algorithm, it significantly reduces the computational overhead of multi-source data fusion and reconstruction while solving the response lag problem of high-resolution gridded processing, greatly enhancing the real-time performance and detection accuracy of power transmission channel observation tasks under complex meteorological backgrounds. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the process of a collaborative network-based observation method for power transmission channels based on multi-source data fusion, provided in an embodiment of the present invention.

[0018] Figure 2 The graph showing the fitting relationship between the convection vortex element infection rate and the actual outbreak probability is provided for an embodiment of the present invention.

[0019] Figure 3The global comprehensive payoff convergence curve for multi-agent non-cooperative game provided in this embodiment of the invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1, Figure 1 This invention presents a collaborative network-based observation method for power transmission channels based on multi-source data fusion, comprising the following steps: S1, coordinate matching of radar echo data and geographic information data, and dynamic pre-distortion compensation by combining real-time meteorological perturbation parameters, to obtain three-dimensional flow field data.

[0022] In this embodiment, the specific implementation of step S1 is as follows: First, multi-source meteorological data is collected and preprocessed at the lowest level. Specifically, before matching the radar echo data with geographic information data, the following steps are taken: converting numerical change events in radiosonde data, radar echo data, and satellite observation data with different sampling rates into asynchronously triggered pulse spike signals; performing time-domain integration on the pulse spike signals to eliminate inter-frame spatiotemporal ghosting and generate a continuous time-domain meteorological feature stream.

[0023] Furthermore, regarding the conversion of numerical change events in the radiosonde data, radar echo data, and satellite observation data with different sampling rates into asynchronously triggered pulse spike signals, the specific scheme of this embodiment includes: setting independent data change thresholds for each meteorological detection device that collects the radiosonde data, radar echo data, and satellite observation data; monitoring the observed values ​​of each meteorological detection device in real time; when the change amplitude of the observed value exceeds the corresponding data change threshold, generating a pulse spike signal at the corresponding timestamp in the continuous time domain; inputting the pulse spike signal into the neurons of a spiking neural network, and updating the neuron membrane potential to characterize the dynamic evolution of meteorological characteristics. In the actual hardware and algorithm deployment, based on the physical detection characteristics of different devices, a change amplitude threshold of 0.1°C is set for the temperature data collected by the radiosonde, a change amplitude threshold of 2.0 dBZ is set for the reflectivity factor change amplitude of the radar echo data, and a change amplitude threshold of 0.5 K is set for the brightness temperature data observed by the satellite microwave radiometer. This process employs a leakage current integral triggering model to integrate and accumulate discrete cross-modal asynchronous pulses over time. Simultaneously, during the gaps without pulse input, the neuron potential exhibits natural exponential decay, thereby accurately mapping the continuous physical process of dynamic generation and dissipation of the meteorological flow field.

[0024] The formula for calculating the membrane potential of neurons in the spiking neural network is as follows: (1) In the formula, Characterizing time The neuronal membrane potential, after being dimensionlessly processed, is used to characterize the transient activation intensity of meteorological features; is the membrane time constant, with the dimension of seconds, representing the exponential decay period of meteorological state memory; The resting potential of the neuron serves as the lower limit of its decay. For the first The connection weights of each synapse correspond to the reliability confidence levels of different meteorological data sources; Indicates the first Each weather detection source at any time The input pulse spike signal is composed of discrete Dirac functions. In this embodiment, the connection weights... The data is normalized and allocated based on the reciprocal of the root mean square error (RMSE) of the most recent calibration of each detection device. This is used to automatically adjust the contribution of data sources with different reliability confidence levels to the evolution of meteorological characteristics through an algorithm.

[0025] After generating the feature flow, spatial compensation is performed. The real-time meteorological perturbation parameters include temperature perturbation data, humidity perturbation data, and air pressure perturbation data. The dynamic pre-distortion compensation based on the real-time meteorological perturbation parameters to obtain three-dimensional flow field data includes: establishing a unified data format compatible with the geographic information data and the radar echo data, and containing a three-dimensional mapping grid; extracting the temperature perturbation data, humidity perturbation data, and air pressure perturbation data, and calculating the real-time nonlinear refractive index; generating a spatial phase conjugate matrix based on the real-time nonlinear refractive index, and performing dynamic distortion pre-distortion on the three-dimensional mapping grid in the unified data format to compensate for coordinate offsets caused by abrupt changes in atmospheric refraction. The unified data format adopts the Earth-centered equidistant tangent coordinate system (ENU coordinate system), where the horizontal resolution of the three-dimensional mapping grid is set to 1 km, and the vertical resolution is set to 0.5 km. During coordinate matching, the original base data in the radar polar coordinate system is extracted, and the reflectivity values ​​in the polar coordinate system are projected and transformed to the three-dimensional mapping grid in the ENU coordinate system using the 3D Barnes interpolation algorithm to complete the initial alignment of spatial dimensions.

[0026] Specifically, the calculation of the real-time nonlinear refractive index and the generation of a spatial phase conjugate matrix based on the real-time nonlinear refractive index include: calculating spatial temperature gradient, humidity gradient, and pressure gradient data based on the temperature perturbation data, humidity perturbation data, and pressure perturbation data; converting the temperature gradient, humidity gradient, and pressure gradient data into a three-dimensional spatial refractive index perturbation field based on an atmospheric refractive perturbation physical model; performing path integration along the propagation path of the radar beam with the radar source as the origin to calculate the spatial phase delay, and then constructing the spatial phase conjugate matrix for the inverse distortion of the three-dimensional mapping mesh. In this process, an adaptive optics deformable mirror mechanism is introduced, treating the three-dimensional propagation path of the radar beam in the troposphere as a process of traversing a non-uniform dielectric constant medium. The sudden drop in temperature, intense accumulation of water vapor, and pressure disturbances caused by localized convective weather directly lead to nonlinear spatial distortion of the atmospheric refractive index. By calculating the three-dimensional gradient of the above-mentioned micro-meteorological parameters and performing integral calculations along the electromagnetic wave ray path, the wavefront phase difference caused by the abrupt change in refraction can be reconstructed. Then, its conjugate matrix is ​​generated to apply reverse elastic stretching and twisting to the three-dimensional mapping mesh, thereby achieving pixel-level flexible coordinate pre-calibration.

[0027] The formulas for calculating the three-dimensional spatial refractive index perturbation field and spatial phase delay are as follows: (2) (3) In the formula, Located in a three-dimensional rectangular coordinate system The refractive index perturbation value at that point is a dimensionless parameter; , and These represent the temperature gradient, humidity gradient, and air pressure gradient vectors in the corresponding spatial coordinate system, respectively. , and These are the atmospheric temperature, humidity, and pressure refractive index conversion sensitivity constant coefficients related to the band; The distance from the origin of the transmitter along the radar detection path is The spatial phase delay at a given point, in radians; The operating wavelength for transmitting beams in radar equipment; Let be a one-dimensional integral path variable, representing the integral from the radar transmission origin 0 along the beam propagation trajectory to the distance. place, This represents the refractive index perturbation value at the corresponding coordinate along the beam propagation path. In this embodiment, the coefficient... , and The value is taken with reference to the Smith-Weintraub constant under the standard atmospheric model. For example, in the microwave band (X / S band). The value is approximately , The value is approximately , The value is approximately .

[0028] The steps for constructing the spatial phase conjugate matrix for inverse warping of the 3D mapped mesh are as follows: 1) The calculated spatial phase delay Converted into equivalent radar line-of-sight path offset The conversion relationship is as follows: .

[0029] 2) Combine the azimuth angle of the current radar beam in the polar coordinate system and pitch angle The line-of-sight path offset Orthogonally decomposed into three-dimensional coordinate distortion components in the ENU coordinate system: horizontal distortion component and and vertical distortion components .

[0030] 3) Construct the spatial phase conjugate matrix. In this embodiment, the matrix is ​​represented as a fourth-order homogeneous transformation matrix containing a reverse displacement vector, and its translation vector part is... .

[0031] When performing dynamic distortion pre-distortion on the 3D mapped mesh in the unified data format, the system extracts the coordinates of each original mesh point of the 3D mapped mesh in the ENU coordinate system. And transform it into homogeneous coordinate form, through the spatial phase conjugate matrix. Matrix multiplication is performed, which involves applying reverse displacement compensation to each grid point affected by refraction. Geometrically, this operation manifests as a reverse elastic stretching of the 3D mesh in the refractive perturbation region, thereby achieving pixel-level precise reconstruction and correction of coordinate shifts caused by abrupt atmospheric refraction.

[0032] This step, by employing a spiking neural network and a deformable mirror mechanism, solves the problem of spatiotemporal ghosting of multi-source heterogeneous data and coordinate shifts caused by nonlinear refraction under severe convective weather, providing a high-precision foundation for subsequent identification.

[0033] To verify the effectiveness of the above-mentioned dynamic pre-distortion compensation mechanism under high-gradient meteorological environment, this embodiment conducted multiple coordinate space mapping comparison tests under different meteorological perturbation intensities. The comparison results of the core perturbation parameters and compensation errors under different working conditions are shown in Table 1.

[0034] Table 1

[0035] Table 1 objectively records the extreme values ​​of the physical gradient threshold under three typical meteorological evolution conditions. The test data fully demonstrates that under extreme perturbation conditions such as squall line passage, due to the severe spatial non-uniformity of the atmospheric dielectric constant, the root mean square error of the three-dimensional coordinates of the traditional static geometric matching algorithm surges to the tens of meters level, which can no longer meet the requirements of high-precision networking. However, after using the deformable mirror mechanism and spatial phase conjugate matrix described in this invention for flexible stretching compensation, the coordinate mapping error can still be forced to converge and stabilize at the sub-meter level under extreme conditions, proving the robustness of the pre-distortion algorithm with extremely significant data differences.

[0036] S2, based on the three-dimensional flow field data, identify multiple convective vortex elements, and generate a scanning priority sequence according to the energy transfer topology relationship between each convective vortex element.

[0037] In this embodiment, generating a scanning priority sequence based on the energy transfer topology between each convective vortex element includes: identifying each convective vortex element within the radar detection range as a node in the topological network. The "convective vortex element" corresponds to a "convective cell" in meteorological detection. Specifically, this is achieved by extracting data from the three-dimensional flow field where the vertical vorticity is greater than... Furthermore, the geometric center of a connected region with a three-dimensional reflectivity greater than 35 dBZ is defined as a topological network node. Atmospheric moisture transport and wind shear paths between the convective vortex elements are set as edges to construct a dynamic topological graph. Using an infectious disease dynamics network model, meteorological thermodynamic energy transfer is parameterized as the infection rate between nodes in the dynamic topological graph network. Based on the infection rate, the probability weights for inducing large-scale strong convective outbreaks corresponding to each node are calculated, and the scanning priority sequence is generated by sorting them in descending order. In this embodiment, the "wind shear path" does not refer to a specific physical entity trajectory, but rather to the momentum and energy transport correlation between adjacent convective vortex element nodes that satisfies a preset physical dynamic threshold. Specifically, the method for constructing the dynamic topological graph is as follows: calculating the wind shear intensity and water vapor flux divergence between adjacent nodes, and establishing a directed edge between the two nodes if and only if a preset energy interaction condition is met. The scanning priority sequence is specifically represented as an ordered task list containing the center three-dimensional coordinates of each convective vortex element and the corresponding probability weights. The formula for calculating the probability weights for inducing large-scale strong convective outbreaks is as follows: (4) in Pointing to a node The set of all directed edges, To the topology node To topology nodes The rate of transmission of meteorological thermodynamic energy.

[0038] In the topology evolution analysis process, the step of parameterizing meteorological thermodynamic energy transfer into the transmission rate between network nodes using the infectious disease dynamics network model includes: calculating the wind shear intensity and water vapor flux divergence between convective vortices corresponding to adjacent nodes in the dynamic topology graph based on the three-dimensional flow field data; determining whether the wind shear intensity and water vapor flux divergence satisfy preset energy interaction conditions; if so, establishing directed edges representing energy transfer channels between adjacent nodes; and normalizing the water vapor flux divergence and wind shear intensity corresponding to the directed edges, mapping them to the transmission rate corresponding to the transformation from susceptible to infected nodes in the infectious disease dynamics network model. In this transformation mechanism, the meteorological evolution process is deeply mapped to the physical boundary of the SEIR infectious disease network model. Static convective vortices without significant convection but with water vapor accumulation are defined as susceptible (S) nodes; convective vortices disturbed by surrounding wind shear and experiencing intense water vapor flux convergence but not yet triggering heavy precipitation are defined as exposed (E) nodes; core strong convective vortices that have experienced strong convective outbursts and have high radar echo reflectivity are defined as infected (I) nodes; and vortices where precipitation has ended, energy has dissipated, and reflectivity is below a threshold are defined as removed (R) nodes, which will no longer participate in the infection process during the current observation period. To accurately define energy transfer channels, the preset energy interaction condition is set to a stringent physical and dynamic threshold: when the wind shear intensity between two nodes is greater than or equal to... And the water vapor flux divergence is less than or equal to At that time, it was determined that there was substantial cross-domain transport of water vapor and momentum, thus establishing a topological directed edge. Subsequently, the wind field shear intensity and water vapor flux divergence that satisfy the interaction conditions were nonlinearly compressed and normalized using a nonlinear sigmoid activation function, and smoothly mapped to the probability space in the (0,1) interval. This was used to quantify the transmission rate of susceptible state nodes affected by the thermodynamic transport of adjacent infected state nodes, thereby causing convective evolution and transfer.

[0039] The formula for calculating the transmission rate between network nodes in the infectious disease dynamics network model is as follows: (5) In the formula, To the topology node To topology nodes The infectivity rate for transferring meteorological thermodynamic energy is a dimensionless probability value. For nodes With nodes The wind shear intensity between them, with dimensions of ; For nodes To the node The transported water vapor flux divergence, with dimensions of Here, a negative value is used to characterize the strong convergence properties of water vapor; , is the sensitivity weighting coefficient for wind shear intensity, dimensionless; , is a dimensionless sensitivity weighting coefficient for the degree of water vapor convergence; This is a bias constant used to adjust the baseline environmental tolerance threshold for strong convection triggering. In this embodiment, The value is 1.2. The value is 0.8. The value is set to -5.0 to ensure that the infection rate is within the linear variation range of the Sigmoid function under standard atmospheric conditions.

[0040] It should be noted that the wind shear intensity mentioned above... With water vapor flux divergence These are all conventional physical parameters in the field of meteorological dynamics, and their specific values ​​are calculated based on the accurate three-dimensional flow field data (including three-dimensional wind speed components and humidity perturbation data) obtained in step S1. Specifically, for adjacent nodes i and j in the topology graph, It is obtained by calculating the ratio of the magnitude of the three-dimensional wind speed vector difference between two nodes to the straight-line distance between them in space (i.e., the velocity space gradient); and This is based on the water vapor flux vector field of the region where the two nodes are located. Its spatial divergence is calculated using the standard finite difference algorithm. Based on the acquired three-dimensional flow field base data and conventional meteorological physics formulas, those skilled in the art can unambiguously calculate the above two parameters and then substitute them into the infectious disease dynamics network model of this application.

[0041] This step, by introducing an epidemiological network model to quantify meteorological thermodynamic energy transfer, solves the technical problem that traditional isolated cell assessments are prone to missing cells with high potential for large-scale outbreaks, and significantly improves the foresight and global perspective of networked radar observation and scheduling.

[0042] To visually demonstrate the advantages of the aforementioned infectious disease dynamics network model in nonlinear fitting for meteorological thermodynamic energy transfer prediction, offline simulation playback was performed using five-year historical radar-based data on the evolution of severe convection. Key state parameters during the dynamic topology evolution process were extracted, and a scatter plot of the nonlinear fitting between the topological node infection rate and the actual outbreak probability of severe convection was plotted, as shown below. Figure 2 As shown.

[0043] Figure 2The horizontal axis represents the preprocessed and fused joint characteristic value of wind field shear and water vapor divergence, while the vertical axis represents the actual frequency of convective vortex elements (susceptible nodes) evolving into heavy precipitation centers within the next half hour due to energy transport from adjacent strong echo centers (infected nodes). The distribution trend of a large number of sample points in the figure clearly shows that when the joint energy interaction characteristic value exceeds a specific physical critical point, the infection rate calculated based on the aforementioned Sigmoid activation function can rapidly approach the actual outbreak probability, exhibiting a highly consistent "S"-shaped step-up characteristic. This rigorous interdisciplinary nonlinear mapping effectively filters out the base noise under low-energy disturbance backgrounds, avoiding the false alarms and missed alarms that are easily generated by linear extrapolation models in the initial stage of convection.

[0044] S3 controls multiple networked radars to perform scanning according to the scanning priority sequence to acquire convective vertical structure data.

[0045] In this embodiment, controlling multiple networked radars to perform scanning according to the scanning priority sequence includes: employing a decentralized multi-agent collaborative allocation strategy to configure an independent intelligent agent node for each of the multiple networked radars, and assigning task reward weights to each scanning task in the scanning priority sequence; each intelligent agent node of the networked radar independently initiates bidding for each scanning task based on its own mechanical rotation cost, observation blind zone, and expected observation reward; multiple rapid bidding processes are conducted through a non-cooperative game algorithm until the network composed of all intelligent agent nodes reaches a Nash equilibrium state, and each networked radar executes the corresponding scanning task according to the task assignment at the time of reaching the Nash equilibrium state.

[0046] In this decentralized, non-cooperative game mechanism, the servo control unit of each radar is abstracted as a rational economic agent model with independent decision-making capabilities. Each intelligent agent node performs independent calculations based only on local information, thus avoiding the communication delays and single-point failure risks of centralized control nodes. During the bidding process, the agent node first parses the scanning priority sequence, evaluates the three-dimensional coordinates of the target space grid, and, combined with the current elevation and azimuth angles of the radar antenna array, calculates the energy consumption and time delay of the mechanical servo motors transferring to the target airspace, quantifying them as mechanical rotation costs. Simultaneously, based on the radar's own geographical obstruction three-dimensional model and the minimum elevation angle constraint, it determines whether the target falls within the radar's effective line-of-sight, imposing a significant penalty for tasks located in blind spots or obscured by terrain. After multiple rounds of iterative bidding and strategy updates, the network state converges to a Nash equilibrium. Specifically, the intelligent agent nodes adopt an optimal feedback strategy for iteration; that is, in each round of bidding, each agent node calculates its own comprehensive benefit. Maximization strategy The strategy is executed in the next round until the change in strategy is below a preset threshold for two consecutive rounds. In this stable state, the decision mechanism strictly follows game theory principles: if any single radar intelligent agent node unilaterally changes its current scanning task allocation strategy (i.e., abandons the task under the current Nash equilibrium and bids for other tasks), its overall gain will be strictly reduced or remain unchanged. When all nodes meet this condition, the network determines that a Nash equilibrium has been reached, and each node then solidifies its bidding results, driving the underlying servo motors to execute the corresponding beam scanning actions. The convective vertical structure data specifically refers to the radar base data acquired by each network radar after executing the scanning task, including reflectivity factor, radial velocity, and spectral width in the corresponding target area polar coordinate system.

[0047] The calculation formula for the comprehensive revenue function of the network radar intelligent agent node described in each part is as follows: (6) In the formula, For the first The networked radar adopts its own strategies And the remaining radars adopt a strategy set The overall benefits at that time; To be assigned to scanning tasks The task reward weight is positively correlated with the probability weight in the scan priority sequence generated in the preceding steps; For the first Radar for mission The expected observational benefit is calculated as follows: (7) in For the first Radar Center to Scanning Mission Euclidean distance of the target region Atmospheric path attenuation factor; and They represent the first The center of the radar's antenna array was moved to the mission area. The increments in azimuth and pitch angles required to traverse the airspace, measured in radians. ); and These are the azimuth and pitch angle mechanical rotation cost penalty coefficients for the corresponding underlying servo motors; in this embodiment... and The setting is based on the maximum rotational speed of the radar servo system. For example, for a conventional X-band phased array radar, the azimuth mechanical loss is used as the benchmark. The value is set to 0.5 / rad; considering that the pitch servo motor needs to overcome a larger gravitational torque or have a longer mechanical settling time when frequently switching pitch angles, its penalty weight is relatively high. The value can be set to 0.8 / rad. As a penalty for blind spots in observation, when the task... In the first When the radar is in a geographically obstructed or detection blind zone, this term takes a penalty constant that approaches positive infinity. When within the effective observation range, this value is [value]. .

[0048] This step effectively avoids the computational bottleneck and radar beam local optimum deadlock trap when high-density multi-target concurrency is used instead of centralized scheduling by adopting multi-agent non-cooperative game theory, and realizes the global adaptive optimal configuration of observation resources of multiple radars.

[0049] To further evaluate the dynamic response time and network convergence stability of the decentralized multi-agent collaborative allocation strategy, high-frequency sampling and recording were performed on the global utility value evolution trajectory of multiple radars concurrently bidding in a scenario of sudden high-density strong convective cluster outbreaks within the region. The game-theoretic convergence simulation curves are shown below. Figure 2 As shown.

[0050] Figure 2 This study demonstrates the bidding dynamics of intelligent agent nodes in a high-conflict task allocation scenario involving multiple networked radars with severe overlap in observation areas and blind spots. Initially, the curve shows a dramatic downward oscillation in the overall system payoff due to nodes spontaneously vying for high-yield scanning tasks based on local optima, frequently triggering mechanical rotation and blind spot penalties. However, under the penalty constraints and payoff drive of the non-cooperative game theory algorithm, the overall global payoff curve quickly smooths out after only a few rapid strategy adjustments and bidding iterations, and robustly converges to the globally optimal utility asymptote. This convergence state precisely corresponds to the network Nash equilibrium solution, and experimental data proves that this mechanism can completely resolve high-density beam spatial conflicts within a millisecond time window.

[0051] S4. Construct a spatial acceleration structure from the convective vertical structure data, and reconstruct the three-dimensional wind field parameters using a back-projection sampling algorithm.

[0052] In this embodiment, S4 includes: extracting polar coordinate radar base data from the convective vertical structure data, constructing the polar coordinate radar base data into a hierarchical bounding box tree structure; using each grid point on a preset three-dimensional rectangular target grid as the ray starting point, emitting a reverse acquisition ray towards the hierarchical bounding box tree structure; and through the intersection operation between the reverse acquisition ray and the hierarchical bounding box tree structure, concurrently acquiring and fusing effective meteorological values ​​around each grid point in multiple threads to generate the corresponding three-dimensional wind field parameters.

[0053] Specifically, in order to solve the spatial indexing problem between the divergent distribution of radar database data in polar coordinates and the uniform distribution of grid points in Cartesian coordinates, this embodiment first uses a spatial recursive partitioning algorithm to encapsulate each radar volume scan sampling unit (Range Bin) located in polar coordinates in an axially consistent bounding box (AABB) according to its spatial extension boundary.

[0054] Specifically, in three-dimensional space, distance Azimuth and pitch angle Definition. First, the eight geometric vertices of the sampling unit are transformed from the polar coordinate system to the ENU rectangular coordinate system using the coordinate transformation formula: (8) In the formula, , and These are the coordinate components in the ENU Cartesian coordinate system. , , Here, , and These represent the initial detection distance, initial azimuth angle, and initial elevation angle of the sampling unit in polar coordinate space, respectively. , and These correspond to the radar equipment's length resolution, azimuth resolution, and elevation step resolution, respectively. The resulting 8 sets of coordinate points are generated through combination. Subsequently, the extreme values ​​of the point set along the three axes are calculated to obtain the parameters of the minimum bounding rectangle (AABB) in the ENU coordinate system: (9) In the formula, and Let represent the minimum and maximum geometric vertex coordinates of the AABB bounding box in a Cartesian coordinate system, respectively. These two points uniquely determine the axially aligned cubic spatial envelope. Based on this, a hierarchical bounding box (BVH) tree structure is constructed from the bottom up. The spatial acceleration structure is specifically represented by this hierarchical bounding box (BVH) tree structure. During the construction process, the surface area heuristic algorithm (SAH) is used for spatial segmentation, and its cost function is... The calculation formula is as follows: (10) In the formula, The cost of traversing internal nodes, The cost of finding the intersection between the ray and the sampling unit. =The surface area of ​​the parent node, and These are the surface areas of the left and right child nodes after the partitioning, respectively. and These represent the number of sampling units contained in the left and right child nodes, respectively. The cost function is minimized. Determine the optimal splitting plane. This spatial acceleration structure transforms the complex global search into an efficient tree-like hierarchical culling, allowing spatial regions that do not contain the target grid points to be quickly filtered out early in the reconstruction process.

[0055] Furthermore, the step of concurrently collecting and fusing effective meteorological data around each grid point through the intersection operation between the reverse acquisition ray and the hierarchical bounding box tree structure includes: calculating the intersection state of the reverse acquisition ray with the axial bounding boxes at each level in the hierarchical bounding box tree structure. In this embodiment, the Slab (split-axis projection) algorithm is used to calculate the intersection state. Let the parametric equation of the reverse acquisition ray be... ,in Let these coordinates be the starting coordinates of the target grid points. Let be the ray direction vector; for any axial bounding box (AABB), its ray direction vector is... The range of the axial direction is The time parameters for the ray entering and leaving the axial plane are... and The calculation formula is as follows: (11) Similarly, calculate shaft and shaft Define the time it takes for a ray to enter the bounding box. and departure time for: In the formula, and The starting coordinates and direction vectors are respectively in The components of the axis are calculated similarly for the other axes; if and only if and When the reverse acquisition ray intersects with the axial bounding box of that level, it is determined that the ray intersects with the axial bounding box of that level; if they do not intersect, all child nodes under that branch are immediately removed.

[0056] In this embodiment, the reverse acquisition ray is centered on the target grid point and has a preset neighborhood search radius. Using a detection length of 2.0 km as an example, all polar coordinate radar base data points within this radius are quickly located by intersecting with the hierarchical bounding box (BVH). When the reverse acquisition ray passes through the bounding box of a leaf node, the polar coordinate radar base data points contained within that leaf node are acquired. A three-dimensional spatial distance weighting algorithm is used, with the Euclidean distance from the ray's origin to each polar coordinate radar base data point as the weight, to perform inverse distance weighted fusion calculation on the extracted effective meteorological values. The fusion result is then written into the corresponding grid points of the three-dimensional rectangular target grid in a lock-free concurrent manner.

[0057] At the underlying storage and parallel computing architecture level, this embodiment employs a reverse mapping logic that differs significantly from traditional forward projection interpolation algorithms. In traditional methods, multiple divergent radar sampling points compete to write to the same target grid memory address, leading to severe thread mutex lock contention. In contrast, in this embodiment's reverse projection sampling, each thread is assigned to a unique 3D rectangular target grid point. This thread independently emits rays and retrieves surrounding radar base data for computation. In actual deployment, this step utilizes a GPU-accelerated architecture, mapping each grid point of the 3D target grid to a thread within a CUDA kernel function, enabling parallel reconstruction of massive grid points. Since each grid point corresponds to a unique memory write address, the threads are completely decoupled in memory write operations, achieving naturally lock-free concurrency and significantly improving the throughput of large-scale meteorological data processing.

[0058] The calculation formula for the three-dimensional spatial inverse distance weighted fusion is as follows: (12) (13) In the formula, Meteorological feature values, such as three-dimensional wind field components or reflectivity intensity, are reconstructed from target grid points in a rectangular coordinate system. The first one retrieved by reverse acquisition ray Observed values ​​of polar coordinate radar base data points within a neighborhood; Coordinates of the target grid point Coordinates of polar coordinate radar base data points The three-dimensional Euclidean distance between them; This is the distance decay power parameter, usually taken as 2, used to adjust the smoothness of spatial weights; The total number of valid neighborhood data points selected by the intersection operation of the ray.

[0059] This step completely eliminates the multi-threaded memory bandwidth blocking problem caused by the parallel 3D gridding of multi-source radar data by introducing back projection and hierarchical bounding box acceleration structures in the ray tracing architecture of computer graphics, and achieves an order-of-magnitude improvement in the computational efficiency of severe convective weather observation data.

[0060] Regarding the performance of the underlying computer system bus and concurrent video memory in 3D reconstruction computation, this embodiment built a heterogeneous computing test platform and conducted extreme multi-threaded concurrent stress tests using Cartesian target meshes of different magnitudes. The aim was to quantitatively compare the underlying hardware resource consumption of the traditional forward-spraying interpolation algorithm and the backward-projection sampling algorithm of this invention. Performance comparison parameters for different acceleration architectures are shown in Table 2.

[0061] Table 2

[0062] Table 2 details the measured performance of the two spatial retrieval computing frameworks in terms of mutex lock contention frequency and global latency under the high computational pressure of an exponentially increasing number of 3D spatial grid nodes. Benchmark tests show that as grid density increases, the number of memory write conflicts in the traditional forward mapping algorithm deteriorates catastrophically, even leading to memory overflow (OOM) and thread crashes at ultra-high resolutions. Conversely, thanks to the precise spatial culling of the hierarchical bounding box (BVH) tree and the independent memory addressing mechanism of reverse ray casting, this invention maintains zero memory write conflicts at any grid scale. The measured latency data confirms that the lock-free concurrent architecture breaks through the memory bandwidth bottleneck of the traditional von Neumann architecture when processing massive divergent meteorological base data, demonstrating extremely high engineering application value.

[0063] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0064] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0065] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0066] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0067] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0068] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A collaborative network-based observation method for power transmission channels based on multi-source data fusion, characterized in that, Includes the following steps: The radar echo data and geographic information data are matched for coordinates, and dynamic pre-distortion compensation is performed by combining real-time meteorological perturbation parameters to obtain three-dimensional flow field data. Based on the three-dimensional flow field data, multiple convective vortex elements are identified, and a scanning priority sequence is generated according to the energy transfer topology relationship between each convective vortex element. Multiple networked radars are controlled to perform scanning according to the scanning priority sequence to acquire convective vertical structure data; A spatial acceleration structure is constructed from the convective vertical structure data, and the three-dimensional wind field parameters are reconstructed using a back-projection sampling algorithm.

2. The method for collaborative network-based observation of power transmission channels based on multi-source data fusion according to claim 1, characterized in that, Before performing coordinate matching between radar echo data and geographic information data, the following steps are included: Numerical change events in sounding data, radar echo data, and satellite observation data with different sampling rates are converted into asynchronously triggered pulse spike signals. The pulse spike signal is integrated in the time domain to eliminate inter-frame spatiotemporal ghosting and generate a meteorological feature stream in the continuous time domain.

3. The method for collaborative network-based observation of power transmission channels based on multi-source data fusion according to claim 2, characterized in that, The process of converting numerical change events in the sounding data, radar echo data, and satellite observation data with different sampling rates into asynchronously triggered pulse spike signals includes: For each device that collects the radiosonde data, radar echo data, and satellite observation data, an independent data change threshold is set. Real-time monitoring of the observed values ​​of each device; when the change in the observed value exceeds the corresponding data change threshold, a pulse spike signal is generated at the corresponding timestamp in the continuous time domain. The pulse spike signal is input into the neurons of the spiking neural network, and the dynamic evolution of meteorological characteristics is characterized by updating the neuron membrane potential.

4. The method for collaborative network-based observation of power transmission channels based on multi-source data fusion according to claim 1, characterized in that, The real-time meteorological perturbation parameters include temperature perturbation data, humidity perturbation data, and air pressure perturbation data; The dynamic pre-distortion compensation, combined with the real-time meteorological perturbation parameters, yields three-dimensional flow field data, including: Establish a unified data format that is compatible with the geographic information data and radar echo data, and includes a three-dimensional mapping grid; Extract the temperature perturbation data, humidity perturbation data, and air pressure perturbation data, and calculate the real-time nonlinear refractive index; Based on the real-time nonlinear refractive index, a spatial phase conjugate matrix is ​​generated, and the three-dimensional mapping mesh in the unified data format is dynamically distorted to compensate for the coordinate offset caused by abrupt atmospheric refraction.

5. The method for collaborative network-based observation of power transmission channels based on multi-source data fusion according to claim 4, characterized in that, The calculation of the real-time nonlinear refractive index and the generation of the spatial phase conjugate matrix based on the real-time nonlinear refractive index include: Based on the temperature perturbation data, humidity perturbation data, and air pressure perturbation data, calculate the spatial temperature gradient, humidity gradient, and air pressure gradient data. Based on the atmospheric refraction perturbation physical model, the temperature gradient, humidity gradient and air pressure gradient data are transformed into a three-dimensional spatial refractive index perturbation field. Taking the radar source as the origin, the path integral of the three-dimensional spatial refractive index perturbation field is performed along the propagation path of the radar beam to calculate the spatial phase delay, and then a spatial phase conjugate matrix for inverse twisting of the three-dimensional mapping mesh is constructed.

6. The method for collaborative network-based observation of power transmission channels based on multi-source data fusion according to claim 1, characterized in that, The step of generating a scanning priority sequence based on the energy transfer topology relationship between each convective vortex element includes: Each convective vortex element within the radar detection range is identified as a node in the topological network, and the atmospheric moisture transport and wind field shear paths between the convective vortex elements are set as edges to construct a dynamic topological graph. The meteorological thermodynamic energy transfer parameter is converted into the infection rate between nodes in the topological graph using an infectious disease dynamics network model. The probability weights of large-scale strong convective outbreaks induced by the convective vortex elements corresponding to each node are calculated based on the infection rate, and the scanning priority sequence is generated by sorting them in descending order.

7. The method for collaborative network-based observation of power transmission channels based on multi-source data fusion according to claim 6, characterized in that, The method of parameterizing meteorological thermodynamic energy transfer into the transmission rate between nodes in a topological graph through an infectious disease dynamics network model includes: Based on the three-dimensional flow field data, calculate the wind shear intensity and water vapor flux divergence between the convective vortex elements corresponding to adjacent nodes in the dynamic topology diagram. Determine whether the wind field shear intensity and water vapor flux divergence meet the preset energy interaction conditions. If so, establish a directed edge representing an energy transfer channel between adjacent nodes. The water vapor flux divergence and wind field shear intensity corresponding to the directed edges are normalized and mapped to the infection rate corresponding to the transformation of susceptible state nodes to infected state nodes in the infectious disease dynamics network model.

8. The method for collaborative network-based observation of power transmission channels based on multi-source data fusion according to claim 1, characterized in that, The control of multiple networked radars to perform scanning according to the scanning priority sequence includes: A decentralized multi-agent collaborative allocation strategy is adopted, with each radar configured with an independent agent node, and the task reward weight corresponding to each scanning task in the scanning priority sequence is set. Each network radar's agent node independently submits bids for each scanning task based on its own mechanical rotation costs, observation blind spots, and expected observation benefits; Multiple rapid bidding processes are conducted using a non-cooperative game theory algorithm until the network composed of all agent nodes reaches a Nash equilibrium state. Each network radar then executes the corresponding scanning task according to the task assignment at the time the Nash equilibrium state is reached.

9. The method for collaborative network-based observation of power transmission channels based on multi-source data fusion according to claim 1, characterized in that, A spatial acceleration structure is constructed from the convective vertical structure data, and the three-dimensional wind field parameters are reconstructed using a back-projection sampling algorithm, including: Extract the polar coordinate radar base data from the convective vertical structure data, and construct the polar coordinate radar base data into a hierarchical bounding box tree structure; Using each grid point on the preset three-dimensional rectangular target grid as the ray starting point, a reverse acquisition ray is emitted towards the hierarchical bounding box tree structure; By performing the intersection operation between the reverse acquisition ray and the hierarchical bounding box tree structure, the effective meteorological data around each grid point are collected and fused to generate the corresponding three-dimensional wind field parameters.

10. The method for collaborative network-based observation of power transmission channels based on multi-source data fusion according to claim 1, characterized in that, By performing intersection operations between the reverse acquisition rays and the hierarchical bounding box tree structure, effective meteorological data around each grid point are acquired and fused, including: Calculate the intersection state between the reverse acquisition ray and the axial bounding boxes at each level in the hierarchical bounding box tree structure; When the reverse acquisition ray passes through the bounding box of the leaf node, the polar coordinate radar base data points contained within the leaf node are acquired. A three-dimensional spatial distance weighting algorithm is adopted, using the Euclidean distance from the ray starting point to each polar coordinate radar base data point as the weight, to perform inverse distance weighted fusion calculation on the extracted effective meteorological values, and the fusion result is written into the corresponding grid points of the three-dimensional rectangular target grid in a lock-free concurrent manner.

Images