Visual early warning function improving method

Abstract

The invention provides a visual early warning function improving method. The visual early warning function improving method comprises the steps of S1, unmanned aerial vehicle dynamic path planning and variational method optimization; s2, a distributed sensor data fusion step; s3, reconstructing the intensity of a three-dimensional electric field and analyzing a vector-containing field; s4, identifying a partial discharge mode; s5, a dynamic risk probability assessment step; s6, adjusting a self-adaptive early warning threshold value; s7, a step of rendering and optimizing a visual interface; s8, an emergency response path is generated; s9, a step of constructing a self-organizing communication network; s10, continuously optimizing the efficiency of the system; according to the scheme, a distributed power monitoring network based on an unmanned aerial vehicle cluster is constructed, and intelligent early warning of a power transmission and distribution line is realized through multi-physics field coupling analysis. The system integrates three core modules of electromagnetic field three-dimensional reconstruction, unmanned aerial vehicle dynamic path optimization and abnormal discharge probability prediction, and solves the problems of low spatial resolution, large abnormal positioning error and the like in traditional electric power inspection.

Description

Technical Field

[0001] The present invention specifically relates to a method for improving a visual early warning function. Background Art

[0002] This proposal addresses the technical area of smart grid condition monitoring and fault warning technology, specifically involving three core areas: drone-based power inspections, multi-physics coupling analysis, and power equipment condition assessment. This area integrates power system automation, IoT sensing technology, artificial intelligence algorithms, and high-precision sensor technology. It aims to enhance real-time monitoring of power equipment operating conditions through intelligent means, ensuring safe and stable grid operation. In the power system sector, traditional manual inspections suffer from low efficiency, high risk, and intermittent data collection. With the development of technologies such as ultra-high voltage transmission and the integration of renewable energy, the electromagnetic environment for power equipment has significantly increased in complexity, creating an increasingly urgent need for early warning of potential hazards such as partial discharge and insulation degradation. In recent years, drones, due to their flexible maneuverability and high-resolution data acquisition capabilities, have become a crucial tool for power inspections. However, existing technologies still face three major bottlenecks: spatial limitations: traditional two-dimensional electric field analysis cannot accurately reflect the field strength distortion characteristics in complex three-dimensional space; temporal lag: periodic inspections struggle to capture sudden discharges; and single data dimensions: single sensor data cannot fully characterize the multi-physics coupling state of equipment.

[0003] In summary, this application proposes a method for improving visual warning functions to solve the above problems. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for improving the visual early warning function in view of the shortcomings of the existing technology, which can well solve the above problems.

[0005] To achieve the above requirements, the present invention adopts a technical solution of providing a method for improving the visual early warning function, which comprises the following steps:

[0006] S1: Perform UAV dynamic path planning and optimization steps including variational method, establish an adaptive route planning model based on electromagnetic field strength, use a three-dimensional magnetic flux sensor equipped on the UAV, dynamically adjust the flight trajectory according to the real-time collected electric field strength data, and introduce variational method to optimize the flight path energy consumption:

[0007]

[0008] in:

[0009] α: Field intensity gradient weight coefficient, ranging from 0.6 to 0.8;

[0010] β: motion acceleration penalty factor, ranging from 0.2 to 0.4;

[0011] Time-varying electric field intensity gradient;

[0012] UAV acceleration vector;

[0013] S2: Perform distributed sensor data fusion and deploy a heterogeneous sensor network in the UAV cluster, including: broadband electric field probes, infrared thermal imaging modules, ultraviolet pulse detectors, and microwave radiometers;

[0014] Using DS evidence theory to fuse multi-source data:

[0015]

[0016] in:

[0017] m(A): joint confidence assignment;

[0018] K: conflict factor, when K>0.8, manual review is triggered;

[0019] B, C: different sensor proposition sets;

[0020] S3: Perform three-dimensional electric field intensity reconstruction and vector field analysis. Based on the discrete field intensity data collected by the drone, a three-dimensional vector field reconstruction model is constructed. The unstructured grid finite element method is used to solve the Maxwell equations. The Kriging interpolation algorithm is combined to generate a spatial field intensity distribution cloud map. A dynamic reconstruction threshold is set. When the field intensity gradient change rate exceeds 15kV / (m·s), it is automatically marked as a potential discharge area. The reconstruction accuracy can reach 0.1kV / m, and the spatial resolution is better than 10cm.

[0021] S4: Performing the step of partial discharge pattern recognition and constructing a deep convolutional pulse network for discharge feature extraction, specifically including: performing wavelet packet decomposition on the UV pulse signal, extracting the Mel cepstral coefficients in the time-frequency domain, and constructing a 12-layer residual network for classification;

[0022] A transfer learning mechanism was introduced, with a pre-trained model containing 20,000 sets of discharge samples. This was continuously optimized through federated learning during field deployment, ensuring a classification accuracy of >92%. The system can distinguish: corona discharge with a recognition rate of 95%, surface discharge with a recognition rate of 89%, and internal air gap discharge with a recognition rate of 83%.

[0023] S5: Steps for dynamic risk probability assessment, establishing a dynamic Bayesian network) risk assessment model:

[0024]

[0025] in:

[0026] X t : System state variables at time t;

[0027] Node parent state collection;

[0028] Transition probability matrix dimension: 15×15;

[0029] S6: Steps for adaptively adjusting the warning threshold, using a dynamic programming algorithm to optimize the warning threshold:

[0030]

[0031] in:

[0032] γ: discount factor, set to 0.9;

[0033] P(s′|s,a): state transition probability;

[0034] R(s, a): immediate reward function;

[0035] S7: Optimize the visualization interface rendering process using the WebGL 3D rendering engine to achieve: real-time field strength equipotential surface rendering with an update frequency of 30Hz; simulate the discharge trajectory particle system; overlay the thermal map and topology map; use the octree space segmentation algorithm for LOD control to ensure real-time rendering of tens of millions of data points on a common workstation; the interactive module supports gesture control and VR display, and the delay of dangerous area annotation is less than 50ms;

[0036] S8: Steps for generating emergency response paths. When a fault is confirmed, the system automatically generates a multi-objective optimization path:

[0037]

[0038] in:

[0039] f1: path length;

[0040] f2: obstacle avoidance cost;

[0041] f3: field strength exposure;

[0042] w i : weight coefficient, and ∑w_i=1;

[0043] S9: Steps for constructing a self-organizing communication network. The drone cluster builds a Mesh communication network using the TDMA time slot allocation protocol:

[0044] T cycle =N×T slot +T guard ;

[0045] in:

[0046] N: number of active nodes;

[0047] T slot : 5ms fixed time slot;

[0048] T guard : 0.2ms protection interval;

[0049] S10: Continuously optimize system performance and establish a digital twin simulation platform, including equipment aging models, environmental corrosion models, and material fatigue models. Use Monte Carlo simulation to predict system reliability.

[0050]

[0051] in:

[0052] λ(τ): time-varying failure rate function;

[0053] The simulation sample size is greater than 10^6 times.

[0054] Preferably, in step S1, the energy functional extreme value is solved by the Euler-Lagrange equation to generate the optimal path that meets the minimum energy consumption and maximum field strength sampling density. In actual deployment, a 0.5m differential GPS positioning module needs to be set up to cooperate with the inertial navigation system to achieve centimeter-level trajectory tracking.

[0055] Preferably, in step S2, a three-level confidence evaluation system is established, namely normal, warning, and fault, to achieve multi-dimensional judgment of the equipment status. The system automatically marks data points with a confidence level lower than 0.7, triggering a second review flight of the drone.

[0056] Preferably, the input parameters in step S5 include 12 characteristic quantities such as field strength distortion rate, temperature gradient, humidity change, etc. The system updates the risk probability value every 30 seconds, and starts the early warning protocol when the posterior probability exceeds 0.65.

[0057] Preferably, in step S6, the field strength threshold range is automatically adjusted by combining historical false alarm data and environmental parameters. In thunderstorm weather, the threshold is relaxed by 12-15%; in dry conditions, it is tightened by 8-10%, achieving the goal of a false alarm rate of <3%.

[0058] Preferably, in step S8, the NSGA-II algorithm is used to solve the Pareto optimal solution set, and three alternative paths are output for operation and maintenance personnel to choose. The path planning response time is less than 2 seconds, which is adaptable to complex terrain environments.

[0059] Preferably, an adaptive power control algorithm is designed in step S9, the communication distance is dynamically adjusted to 50-1000m, the packet loss rate is <0.1%, and key data is transmitted using AES-256 encryption to ensure information security.

[0060] The advantages of this method for improving the visual early warning function are as follows:

[0061] 1. Improved spatial resolution: Through three-dimensional field strength reconstruction, the detection accuracy is improved from meter level to centimeter level, and 5mm insulator cracks can be identified.

[0062] 2. Optimized response speed: The warning delay is shortened from 3-5 minutes in traditional solutions to within 8 seconds.

[0063] 3. Reduced false alarm rate: By integrating DS evidence theory with the DBN model, the false alarm rate is reduced from 12% to 2.7%.

[0064] 4. Enhanced adaptability: The dynamic threshold mechanism enables the system to operate stably in an environment of -40°C to +70°C. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The same reference numerals are used in these drawings to represent the same or similar parts. The exemplary embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0066] Figure 1 The structural diagram of the method for improving the visual warning function according to an embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0067] In order to make the objectives, technical solutions and advantages of this application clearer, this application is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0068] In the following description, references to "one embodiment," "an embodiment," "an example," "an example," etc. indicate that the embodiment or example described may include certain features, structures, characteristics, properties, elements, or limitations, but not every embodiment or example necessarily includes the certain features, structures, characteristics, properties, elements, or limitations. In addition, repeated use of the phrase "according to one embodiment of the present application" may refer to the same embodiment, but does not necessarily refer to the same embodiment.

[0069] For the sake of simplicity, certain technical features well known to those skilled in the art are omitted in the following description.

[0070] According to an embodiment of the present application, a method for improving a visual warning function is provided, such as Figure 1As shown, the following steps are included:

[0071] S1: Perform UAV dynamic path planning and optimization steps including variational method, establish an adaptive route planning model based on electromagnetic field strength, use a three-dimensional magnetic flux sensor equipped on the UAV, dynamically adjust the flight trajectory according to the real-time collected electric field strength data, and introduce variational method to optimize the flight path energy consumption:

[0072]

[0073] in:

[0074] α: Field intensity gradient weight coefficient, ranging from 0.6 to 0.8;

[0075] β: motion acceleration penalty factor, ranging from 0.2 to 0.4;

[0076] Time-varying electric field intensity gradient;

[0077] UAV acceleration vector;

[0078] S2: Perform distributed sensor data fusion and deploy a heterogeneous sensor network in the UAV cluster, including: broadband electric field probes, infrared thermal imaging modules, ultraviolet pulse detectors, and microwave radiometers;

[0079] Using DS evidence theory to fuse multi-source data:

[0080]

[0081] in:

[0082] m(A): joint confidence assignment;

[0083] K: conflict factor, when K>0.8, manual review is triggered;

[0084] B, C: different sensor proposition sets;

[0085] S3: Perform three-dimensional electric field intensity reconstruction and vector field analysis. Based on the discrete field intensity data collected by the drone, a three-dimensional vector field reconstruction model is constructed. The unstructured grid finite element method is used to solve the Maxwell equations, and the Kriging interpolation algorithm is combined to generate a spatial field intensity distribution cloud map. A dynamic reconstruction threshold is set. When the field intensity gradient change rate exceeds 15kV / (m·s), it is automatically marked as a potential discharge area. The reconstruction accuracy can reach 0.1kV / m, and the spatial resolution is better than 10cm.

[0086] S4: Performing the step of partial discharge pattern recognition and constructing a deep convolutional pulse network for discharge feature extraction, specifically including: performing wavelet packet decomposition on the UV pulse signal, extracting the Mel cepstral coefficients in the time-frequency domain, and constructing a 12-layer residual network for classification;

[0087] A transfer learning mechanism was introduced, with a pre-trained model containing 20,000 sets of discharge samples. This was continuously optimized through federated learning during field deployment, ensuring a classification accuracy of >92%. The system can distinguish: corona discharge with a recognition rate of 95%, surface discharge with a recognition rate of 89%, and internal air gap discharge with a recognition rate of 83%.

[0088] S5: Steps for dynamic risk probability assessment, establishing a dynamic Bayesian network) risk assessment model:

[0089]

[0090] in:

[0091] X t : System state variables at time t;

[0092] Node parent state collection;

[0093] Transition probability matrix dimension: 15×15;

[0094] S6: Steps for adaptively adjusting the warning threshold, using a dynamic programming algorithm to optimize the warning threshold:

[0095]

[0096] in:

[0097] γ: discount factor, set to 0.9;

[0098] P(s′|s,a): state transition probability;

[0099] R(s, a): immediate reward function;

[0100] S7: Optimize the visualization interface rendering process using the WebGL 3D rendering engine to achieve: real-time field strength equipotential surface rendering with an update frequency of 30Hz; simulate the discharge trajectory particle system; overlay the thermal map and topology map; use the octree space segmentation algorithm for LOD control to ensure real-time rendering of tens of millions of data points on a common workstation; the interactive module supports gesture control and VR display, and the delay of dangerous area annotation is less than 50ms;

[0101] S8: Steps for generating emergency response paths. When a fault is confirmed, the system automatically generates a multi-objective optimization path:

[0102]

[0103] in:

[0104] f1: path length;

[0105] f2: obstacle avoidance cost;

[0106] f3: field strength exposure;

[0107] w i : weight coefficient, and ∑w_i=1;

[0108] S9: Steps for constructing a self-organizing communication network. The drone cluster builds a Mesh communication network using the TDMA time slot allocation protocol:

[0109] T cycle =N×T slot +T guard ;

[0110] in:

[0111] N: number of active nodes;

[0112] T slot : 5ms fixed time slot;

[0113] T guard : 0.2ms protection interval;

[0114] S10: Continuously optimize system performance and establish a digital twin simulation platform, including equipment aging models, environmental corrosion models, and material fatigue models. Use Monte Carlo simulation to predict system reliability.

[0115]

[0116] in:

[0117] λ(τ): time-varying failure rate function;

[0118] The simulation sample size is greater than 10^6 times.

[0119] According to one embodiment of the present application, in step S1 of the method for improving the visual warning function, the energy functional extreme value is solved by the Euler-Lagrange equation to generate an optimal path that meets the minimum energy consumption and maximum field strength sampling density. During actual deployment, a 0.5m differential GPS positioning module needs to be set up to cooperate with the inertial navigation system to achieve centimeter-level trajectory tracking.

[0120] According to one embodiment of the present application, in step S2 of the method for improving the visual warning function, a three-level confidence assessment system, namely normal, warning, and fault, is established to achieve multi-dimensional judgment of the equipment status. The system automatically marks data points with a confidence level lower than 0.7, triggering a second review flight of the drone.

[0121] According to one embodiment of the present application, the input parameters in step S5 of the visual warning function improvement method include 12 characteristic quantities such as field strength distortion rate, temperature gradient, and humidity change. The system updates the risk probability value every 30 seconds, and the warning protocol is activated when the posterior probability exceeds 0.65.

[0122] According to one embodiment of the present application, in step S6 of the method for improving the visual warning function, historical false alarm data and environmental parameters are combined to automatically adjust the field strength threshold range. In thunderstorm weather, the threshold is relaxed by 12-15%; in dry conditions, it is tightened by 8-10%, achieving the goal of a false alarm rate of <3%.

[0123] According to one embodiment of the present application, in step S8 of the visual warning function improvement method, the NSGA-II algorithm is used to solve the Pareto optimal solution set, and 3 alternative paths are output for operation and maintenance personnel to choose. The path planning response time is <2 seconds, which can adapt to complex terrain environments.

[0124] According to one embodiment of the present application, an adaptive power control algorithm is designed in step S9 of the visual warning function improvement method, the communication distance is dynamically adjusted to 50-1000m, the packet loss rate is <0.1%, and key data is transmitted using AES-256 encryption to ensure information security.

[0125] The above-described embodiments merely represent several implementations of the present invention. While the descriptions are relatively specific and detailed, they are not to be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the spirit of the present invention, and such modifications and improvements fall within the scope of the present invention. Therefore, the scope of the present invention shall be determined by the claims.

Claims

1. A method for improving visual early warning function, characterized in that: The steps include: S1: Perform UAV dynamic path planning and optimization steps including variational method, establish an adaptive route planning model based on electromagnetic field strength, use a three-dimensional magnetic flux sensor equipped on the UAV, dynamically adjust the flight trajectory according to the real-time collected electric field strength data, and introduce variational method to optimize the flight path energy consumption: in: α: Field intensity gradient weight coefficient, ranging from 0.6 to 0.8; β: motion acceleration penalty factor, ranging from 0.2 to 0.4; Time-varying electric field intensity gradient; UAV acceleration vector; S2: Perform distributed sensor data fusion and deploy a heterogeneous sensor network in the UAV cluster, including: broadband electric field probes, infrared thermal imaging modules, ultraviolet pulse detectors, and microwave radiometers; Using DS evidence theory to fuse multi-source data: in: m(A): joint confidence assignment; K: conflict factor, when K>0.8, manual review is triggered; B, C: different sensor proposition sets; S3: Perform three-dimensional electric field intensity reconstruction and vector field analysis. Based on the discrete field intensity data collected by the drone, a three-dimensional vector field reconstruction model is constructed. The unstructured grid finite element method is used to solve the Maxwell equations. The Kriging interpolation algorithm is combined to generate a spatial field intensity distribution cloud map. A dynamic reconstruction threshold is set. When the field intensity gradient change rate exceeds 15kV / (m·s), it is automatically marked as a potential discharge area. The reconstruction accuracy can reach 0.1kV / m, and the spatial resolution is better than 10cm. S4: Performing the step of partial discharge pattern recognition and constructing a deep convolutional pulse network for discharge feature extraction, specifically including: performing wavelet packet decomposition on the UV pulse signal, extracting the Mel cepstral coefficients in the time-frequency domain, and constructing a 12-layer residual network for classification; A transfer learning mechanism was introduced, with a pre-trained model containing 20,000 sets of discharge samples. This was continuously optimized through federated learning during field deployment, ensuring a classification accuracy of >92%. The system can distinguish: corona discharge with a recognition rate of 95%, surface discharge with a recognition rate of 89%, and internal air gap discharge with a recognition rate of 83%. S5: Steps for dynamic risk probability assessment, establishing a dynamic Bayesian network) risk assessment model: in: X t : System state variables at time t; Node parent state collection; Transition probability matrix dimension: 15×15; S6: Steps for adaptively adjusting the warning threshold, using a dynamic programming algorithm to optimize the warning threshold: in: γ: discount factor, set to 0.9; P(s′|s,a): state transition probability; R(s, a): instant reward function; S7: Optimize the visualization interface rendering process using the WebGL 3D rendering engine to achieve: real-time field strength equipotential surface rendering with an update frequency of 30Hz; simulate the discharge trajectory particle system; overlay the thermal map and topology map; use the octree space segmentation algorithm for LOD control to ensure real-time rendering of tens of millions of data points on a common workstation; the interactive module supports gesture control and VR display, and the delay of dangerous area annotation is less than 50ms; S8: Steps for generating emergency response paths. When a fault is confirmed, the system automatically generates a multi-objective optimization path: in: f1: path length; f2: obstacle avoidance cost; f3: field strength exposure; w i : weight coefficient, and ∑w_i=1; S9: Steps for constructing a self-organizing communication network. The drone cluster builds a Mesh communication network using the TDMA time slot allocation protocol: T cycle =N×T slot +T guard ; in: N: number of active nodes; T slot : 5ms fixed time slot; T guard : 0.2ms protection interval; S10: Continuously optimize system performance and establish a digital twin simulation platform, including equipment aging models, environmental corrosion models, and material fatigue models. Use Monte Carlo simulation to predict system reliability. R(t)=exp(-∫0tλ(τ)dτ); in: λ(τ): time-varying failure rate function; The simulation sample size is greater than 10^6 times.

2. The method for improving the visual early warning function according to claim 1, characterized in that: In step S1, the energy functional extreme value is solved by the Euler-Lagrange equation to generate the optimal path that meets the minimum energy consumption and maximum field strength sampling density. In actual deployment, a 0.5m differential GPS positioning module needs to be set up to cooperate with the inertial navigation system to achieve centimeter-level trajectory tracking.

3. The method for improving the visual early warning function according to claim 1, characterized in that: In step S2, a three-level confidence assessment system, namely normal, warning, and fault, is established to achieve multi-dimensional judgment of the equipment status. The system automatically marks data points with a confidence level lower than 0.7, triggering a second review flight of the drone.

4. The method for improving the visual early warning function according to claim 1, characterized in that: The input parameters in step S5 include 12 characteristic quantities such as field strength distortion rate, temperature gradient, and humidity change. The system updates the risk probability value every 30 seconds and activates the early warning protocol when the posterior probability exceeds 0.

65.

5. The method for improving the visual early warning function according to claim 1, characterized in that: In step S6, the field strength threshold range is automatically adjusted by combining historical false alarm data with environmental parameters. In thunderstorm weather, the threshold is relaxed by 12-15%; in dry conditions, it is tightened by 8-10%, achieving the goal of a false alarm rate of <3%.

6. The method for improving the visual early warning function according to claim 1, characterized in that: In step S8, the NSGA-II algorithm is used to solve the Pareto optimal solution set and output three alternative paths for operation and maintenance personnel to choose. The path planning response time is less than 2 seconds, which can adapt to complex terrain environments.

7. The method for improving the visual early warning function according to claim 1, characterized in that: In step S9, an adaptive power control algorithm is designed, the communication distance is dynamically adjusted to 50-1000m, the packet loss rate is <0.1%, and key data is transmitted using AES-256 encryption to ensure information security.

Images