Live-line work safety monitoring system based on artificial intelligence

Abstract

The invention discloses a live-line work safety monitoring system based on artificial intelligence, and relates to the technical field of intelligent electric power safety, and the system comprises a data collection module which is used for collecting and preprocessing work data in a live-line work area in real time, and extracting live-line body point cloud data based on the work data; the finite element grid unit is used for dividing all charged bodies into finite element grid units and judging high-temperature charged bodies; and the violation judgment module is used for calculating the electric field intensity value of the position where the operator is located in real time. By combining an electric field, a temperature field and three-dimensional point cloud multi-modal data, comprehensive dynamic perception of an operation area is realized, finite element modeling and dynamic non-uniform heat source distribution analysis are adopted, a high-temperature charged body is accurately identified, the equipment fault early warning capability is improved, action characteristics of an operator are captured through a long-short-term memory network, and the working efficiency is improved. And the illegal behaviors are accurately judged by dynamically combining environmental data, and an optimal operation path is generated in real time by utilizing a three-dimensional grid division and dynamic planning algorithm.

Description

Technical Field

[0001] The present invention relates to the field of intelligent power safety technology, and in particular to a live working safety monitoring system based on artificial intelligence. Background Art

[0002] In recent years, with the rapid development of the power industry, the safety of live working in power transmission and distribution systems has received increasing attention. As an important part of power system maintenance, live working has the advantages of high maintenance efficiency and reduced power outage time. Due to the complex working environment, dense distribution of high-voltage equipment, and the need for operators to operate in dangerous environments such as strong electromagnetic fields and high temperatures, its safety has always been a technical problem that needs to be solved in the industry. Traditional live working safety monitoring mainly relies on manual observation and fixed sensor layout, which is usually difficult to reflect the dynamic environmental changes in the working area in real time and accurately. In recent years, with the rapid development of artificial intelligence (AI) and Internet of Things (IoT) technologies, AI-based safety monitoring systems have gradually been introduced into live working scenarios. Through sensors, point cloud data and deep learning algorithms, intelligent monitoring systems can more efficiently identify the behavior of operators, changes in the surrounding environment, and potential dangerous areas, providing certain technical support for operational safety. However, the existing intelligent monitoring systems still face many shortcomings in practical applications, especially in terms of real-time, accuracy and comprehensiveness in complex dynamic environments, where technical development is not yet mature.

[0003] After searching, in the prior art, China Invention Application No.: 202110355414.6 discloses an active warning and safety monitoring system for live-line work based on precise positioning of artificial intelligence, including an active warning system, a laser radar, an inertial sensor and a camera, and also including a human key point recognition system, a charged body recognition system, a data fusion system and a spatial position judgment system. The camera is electrically connected to the human key point recognition system, the charged body recognition system and the data fusion system respectively, and the laser radar and the inertial sensor are electrically connected to the data fusion system respectively. The output end of the human key point recognition system, the output end of the charged body recognition system and the output end of the data fusion system are respectively connected to the input end of the spatial position judgment system. Compared with the prior art, the present invention can monitor the safety distance and active warning in real time, and prevent or reduce accidents such as personal injury to live-line workers.

[0004] However, the existing live working monitoring technology has obvious deficiencies in adaptability to high-voltage environments, which are specifically manifested in the following aspects: Although some systems have achieved three-dimensional modeling of the working area through point cloud data collection, due to the lack of deep fusion of multimodal data such as electric field and temperature field, the reliability of its recognition results is insufficient in dynamic working scenarios. The high temperature state on the surface of the live body is often an important signal of potential faults, but the existing technology lacks accurate means to determine the distribution of high temperature, and it is difficult to identify equipment hazards in a timely manner. Currently, the analysis of operator actions is mostly based on simple rule matching or traditional computer vision algorithms, which makes it difficult to effectively capture complex dynamic behavior characteristics, especially in high-voltage environments where the electric field strength changes drastically. The accuracy of operator behavior recognition is significantly reduced. The existing technology lacks real-time and dynamic properties in operator path planning, and it is difficult to timely plan the optimal working path and provide accurate safety reminders according to changes in dangerous areas. These deficiencies have seriously restricted the application effect of existing monitoring systems in actual live working scenarios, and have also become an important bottleneck restricting the further improvement of live working safety. Summary of the invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides an artificial intelligence-based live working safety monitoring system to solve the many limitations of existing live working monitoring technologies in terms of accurate judgment of high-temperature live objects, real-time analysis of multimodal data fusion, complex behavior recognition, and dynamic path planning.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] The present invention provides a live working safety monitoring system based on artificial intelligence, which comprises:

[0009] The data acquisition module is used to collect and pre-process the operation data in the live operation area in real time, and extract the point cloud data of the live body based on the operation data.

[0010] A high temperature determination module is used to divide all charged bodies into finite element grid units and determine high temperature charged bodies;

[0011] The violation determination module is used to calculate the electric field strength value of the operator's location in real time, capture the operator's action characteristics through the long short-term memory network, and use the classification model to calculate the probability of violation behavior;

[0012] The path planning module is used to remind operators in real time and dynamically plan the operation path based on the probability of violation;

[0013] The data storage module is used to store and manage all data in a relational database.

[0014] As a preferred solution of the artificial intelligence-based live working safety monitoring system of the present invention, the real-time collection of working data in the live working area includes:

[0015] Install high-voltage electric field sensors and temperature sensors on each live object in the live working area and set the sampling frequency of the sensors. Use a wireless transmission module to transmit the collected electric field data and temperature data to the central control system in real time.

[0016] Start the laser radar and set the scanning frequency of the laser radar. Set the acquisition range to the spherical space of the live working area. The laser radar generates point cloud data in real time through multi-line laser emission and reflection measurement;

[0017] The point cloud data refers to the three-dimensional space coordinates of each reflection point.

[0018] As a preferred solution of the artificial intelligence-based live working safety monitoring system of the present invention, the preprocessing includes:

[0019] Check the integrity of the point cloud data, remove erroneous data during the collection process, calculate the average distance of the field points and remove abnormal points. For the edge area of ​​the live working area, use the density calculation formula, define the point density as the number of field points within the unit spherical radius, detect isolated points through local density changes and remove them.

[0020] The electric field data and temperature data were cleaned and the sliding smoothing filter method was used to smooth the data.

[0021] As a preferred solution of the artificial intelligence-based live working safety monitoring system of the present invention, the extraction of live body point cloud data based on working data includes:

[0022] The point cloud data is aligned and converted into a global coordinate system using a coarse registration method based on feature points and a fine registration method using an iterative closest point algorithm.

[0023] Based on the center position of the live working area, the point cloud data in the global coordinate system is spatially cropped, and the point cloud data centered on the live body is retained;

[0024] Voxel grid downsampling is used to compress point cloud data, and the height distribution of point cloud data is used to identify ground points. The background objects around the charged body are removed through density and curvature analysis to extract the point cloud data of the charged body.

[0025] As a preferred solution of the artificial intelligence-based live working safety monitoring system of the present invention, the method of dividing all live bodies into finite element grid units and determining high-temperature live bodies includes:

[0026] The surface of the geometric shape of the charged body is reconstructed based on the point cloud data to generate the corresponding three-dimensional equipment model. According to the number of charged bodies, the global coordinate system is divided into multiple finite element mesh units using the three-dimensional meshing technology. The type of each finite element mesh unit is set, and physical properties are assigned to all nodes in each finite element mesh unit. The current density value is assigned to the central node of each finite element mesh unit using the inverse distance weighted interpolation method to obtain the current density value of each node in the finite element mesh unit, and the temperature value of the node in each finite element mesh unit is calculated;

[0027] The finite element mesh unit types include hexahedral units and tetrahedral units;

[0028] The physical data include node coordinates and material parameters;

[0029] The material parameters include thermal conductivity, specific heat capacity and density of the charged body;

[0030] The calculating the temperature value of the node in each finite element mesh unit includes establishing an initial temperature field control equation;

[0031] Dynamic non-uniform heat source distribution modeling is introduced to express the volume heat source intensity of each node as a dynamic distribution related to the current density value, initial temperature value and environmental conditions. The formula is:

[0032]

[0033] in, represents the volume heat source intensity of the g-th node after adjustment, ω0 represents the initial volume heat source intensity, f() represents the exponential function of the nonlinear function, I g represents the current density value of the gth node, T g represents the initial temperature value of the g-th node in the finite element mesh unit, and t represents the current time point;

[0034] The expression of the exponential function of the nonlinear function is:

[0035]

[0036] Among them, b1 represents the adjustment parameter, b2 represents the temperature sensitivity coefficient, b3 represents the amplitude coefficient, and w represents the frequency;

[0037] All temperature values ​​in the charged body obtained by the temperature sensor are converted into absolute temperature, and the thermal radiation term is introduced to describe the heat loss on the surface of the charged body. The improved temperature field control equation is established in combination with the heat source intensity. The formula is:

[0038]

[0039] Where σ represents the Stefan-Bohr slow constant, ∈ represents the surface emissivity, which is dimensionless and determined by the surface material of the charged body, and T a Represents the absolute mean temperature of the charged body, T e Indicates the absolute temperature of the live working area;

[0040] Set boundary conditions and initial conditions, and obtain the temperature value of each node through numerical iteration;

[0041] The initial conditions include setting the initial temperature value and initial volume heat source intensity of each node in the finite element grid unit at the time point t=0;

[0042] The numerical iteration includes using an explicit difference method to perform discrete calculations based on boundary conditions and using Newton iteration to solve the nonlinear term σ∈(T a 4 -T e 4 );

[0043] Set an error threshold. If the temperature change of each node is less than the error threshold in two adjacent iterations, the node stops iterating. Otherwise, it continues iterating. When all nodes stop iterating, the iteration process is completed.

[0044] Set a high temperature safety threshold. If the temperature of a node is higher than the high temperature safety threshold, the node is marked as a high temperature node. Otherwise, no operation is performed.

[0045] A ratio threshold is set. When the ratio of high-temperature nodes in the charged body is higher than the ratio threshold, the charged body is judged to be a high-temperature charged body. Otherwise, no operation is performed.

[0046] As a preferred solution of the artificial intelligence-based live working safety monitoring system of the present invention, the real-time calculation of the electric field strength value at the location of the operator includes:

[0047] The operator area is extracted from the 3D point cloud data generated by the lidar. The 2D detection results are mapped to the 3D point cloud through the image captured by the camera. The region growing algorithm is used to segment the operator area in the 3D point cloud. The algorithm starts from the target seed point and gradually expands to the neighboring points. The operator skeleton is extracted through the segmented point cloud using the PointNet++ model. The joint points of the skeleton point set are tracked frame by frame, and the displacement trajectory of the joint points in the time dimension is calculated to generate a skeleton time series. The skeleton time series includes the 3D spatial coordinates of the operator at each time point.

[0048] Convert the three-dimensional space coordinates of the charged body into cylindrical coordinates (r, z y );

[0049] The three-dimensional space coordinate conversion to cylindrical coordinates includes obtaining the three-dimensional space coordinates, length and charge density of the center position of the charged body, taking the three-dimensional space coordinates of the center position of the charged body as the central axis, calculating the shortest radial distance r from the center axis of the charged body in the three-dimensional space coordinates of the operator and obtaining the vertical height z of the three-dimensional space coordinate corresponding to the radial distance y , taking the radial distance and vertical height as cylindrical coordinates;

[0050] According to the current-temperature coupling, the linear charge density is adjusted by the maximum temperature of all nodes in the charged body. The formula is:

[0051] τ(T)=τ0·(1+β·T max ),

[0052] Where τ(T) represents the linear charge density after temperature adjustment, τ0 represents the baseline linear charge density, β represents the temperature influence coefficient, and T max It represents the maximum temperature of all nodes in the charged body;

[0053] Based on the cylindrical coordinates, the equivalent charge method is used to calculate the electric field potential value generated by each charged body. The formula is:

[0054]

[0055] Among them, φ i (r,z y ) represents the i-th charged body in the cylindrical coordinates (r, z y ), L represents the length of the charged body, z s represents the integral variable along the central axis, ε represents the dielectric constant of air, and d represents the integral variable z s Small variables

[0056] The electric field strength value at the operator's location is calculated by the electric field potential value generated by each charged body. The formula is:

[0057]

[0058] Among them, E i represents the electric field strength value of the i-th charged body, A negative value that represents the spatial rate of change of the electric field potential.

[0059] As a preferred solution of the artificial intelligence-based live working safety monitoring system of the present invention, the method of capturing the action characteristics of the operator through the long short-term memory network and using the classification model to calculate the probability of violation behavior includes:

[0060] The generated skeleton time series is input into the long short-term memory network for time series analysis. LSTM captures historical information through memory units, and filters and updates the information through the forget gate, input gate, and output gate to generate action features at each time point.

[0061] The electric field strength value, action characteristics and the temperature values ​​of all nodes in each charged body are fused into comprehensive features through dynamic weighting, and the comprehensive features are input into the classification model to determine the violation behavior. The classification model generates the violation probability through the softmax function.

[0062] As a preferred solution of the artificial intelligence-based live working safety monitoring system of the present invention, the real-time reminder of the operator according to the probability of violation behavior includes:

[0063] A probability threshold is set and the probability of violation generated by the classification model is compared with the probability threshold. If the probability of violation is greater than the probability threshold, all three-dimensional spatial coordinates corresponding to the high-temperature charged body are defined as risk points, and the operator is reminded in real time through the AR device. Otherwise, no operation is performed.

[0064] As a preferred solution of the artificial intelligence-based live working safety monitoring system of the present invention, the dynamic planning operation path includes:

[0065] Divide the live working area into three-dimensional grid units, each of which represents a cubic area;

[0066] All risk points are integrated into a risk coordinate set, and the three-dimensional spatial coordinates of each risk point in the risk coordinate set are traversed and mapped to each three-dimensional grid unit. If the three-dimensional spatial coordinates of the risk point are in the three-dimensional grid unit, the three-dimensional grid unit is marked as inaccessible, and the path is prohibited from passing through the three-dimensional grid unit. Otherwise, the shortest Euclidean distance from the center point of the three-dimensional grid unit to all risk points is calculated, and the shortest Euclidean distance is used as the risk cost of the three-dimensional grid unit. The smaller the shortest Euclidean distance between the three-dimensional grid unit and the risk point, the higher the risk cost.

[0067] According to the target three-dimensional grid unit of the operator, a path cost function of each three-dimensional grid unit is defined;

[0068] The path cost function includes a cumulative cost and an estimated cost;

[0069] The cumulative cost refers to the path cost from the operator's starting point to the current three-dimensional grid unit, including all risk costs of all three-dimensional grid units passed on the path and the Euclidean distance between the center points of all three-dimensional grid units;

[0070] The estimated cost refers to the Euclidean distance from the current grid cell to the target grid cell;

[0071] The D** algorithm is used to expand the three-dimensional grid cell with the minimum total cost, gradually retrieve the optimal path from the starting three-dimensional grid cell to the target three-dimensional grid cell, and generate the optimal path sequence.

[0072] As a preferred solution of the artificial intelligence-based live working safety monitoring system of the present invention, the storing and managing of all data in a relational database includes:

[0073] Choose a relational database to store and manage data and analysis results, design the database table structure to store different types of data, set up regular backup tasks, back up all data in the database, manage permissions for database users and encrypt and store data.

[0074] The beneficial effects of the present invention are as follows: by combining electric field, temperature field and three-dimensional point cloud multimodal data, comprehensive dynamic perception of the working area is achieved, finite element modeling and dynamic non-uniform heat source distribution analysis are used to accurately identify high-temperature charged bodies, and the equipment failure warning capability is improved. The action characteristics of the operators are captured through the long and short-term memory network, and the environmental data is dynamically combined to accurately determine the violations. The optimal working path is generated in real time by using three-dimensional grid division and dynamic programming algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0076] Figure 1 This is a structural schematic diagram of an artificial intelligence-based live working safety monitoring system in Example 1.

[0077] Figure 2 This is a flow chart of an artificial intelligence-based live working safety monitoring system in Example 1. DETAILED DESCRIPTION

[0078] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0079] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0080] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0081] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a live working safety monitoring system based on artificial intelligence, comprising the following steps:

[0082] S1, data acquisition module, used to collect and pre-process the operation data in the live operation area in real time, and extract the point cloud data of the live body based on the operation data;

[0083] Specifically, real-time collection of operation data in live working areas includes:

[0084] Install Hall effect current sensors and temperature sensors on each live object in the live working area (such as cable joints, transformer coils, switch cabinets) and set the sampling frequency of the sensors. Use wireless transmission modules (such as LoRa) to transmit the collected current density and temperature values ​​to the central control system in real time.

[0085] Start the laser radar and set the scanning frequency of the laser radar. Set the acquisition range to the spherical space of the live working area. The laser radar generates point cloud data in real time through multi-line laser emission and reflection measurement;

[0086] The point cloud data refers to the three-dimensional spatial coordinates of each reflection point.

[0087] The present invention realizes the efficient collection and real-time transmission of electric field data and temperature data in the live working area by installing Hall effect current sensors and temperature sensors on the live body and combining them with wireless transmission modules, thereby ensuring the accuracy of the data. The laser radar generates high-precision point cloud data covering the spherical space through multi-line laser scanning, providing three-dimensional modeling support for the working area. This design realizes multi-dimensional dynamic monitoring of the live working environment, provides complete and accurate basic data for subsequent high temperature judgment, violation detection and path planning, and effectively improves the real-time, comprehensiveness and reliability of monitoring.

[0088] Furthermore, preprocessing is performed including:

[0089] Check the integrity of the point cloud data, remove erroneous data during the collection process, calculate the average distance of the field points and remove abnormal points. For the edge area of ​​the live working area, use the density calculation formula, define the point density as the number of field points within the unit spherical radius, detect isolated points through local density changes and remove them.

[0090] The current density and temperature values ​​were cleaned and the data were smoothed using the sliding smoothing filter method.

[0091] The preprocessing step in the present invention significantly improves the reliability of point cloud data by checking the integrity of point cloud data and removing erroneous data. It detects isolated points by calculating the average distance and density of field points, eliminates abnormal points, and optimizes the spatial distribution quality of point cloud data, thereby enhancing the adaptability to the complex environment at the edge of the live working area. It performs sliding smoothing filtering on current density values ​​and temperature values, effectively eliminates sensor noise and data fluctuations, and improves data stability and accuracy. These preprocessing operations provide a more accurate and reliable data basis for subsequent high temperature judgment and violation detection, and significantly improve the overall monitoring performance of the system.

[0092] In addition, the point cloud data of charged bodies extracted based on the operation data includes:

[0093] The point cloud data is aligned and converted into a global coordinate system using a coarse registration method based on feature points and a fine registration method using an iterative closest point algorithm.

[0094] Based on the center position of the live working area, the point cloud data in the global coordinate system is spatially cropped, and the point cloud data centered on the live body is retained;

[0095] Voxel grid downsampling is used to compress point cloud data, and the height distribution of point cloud data is used to identify ground points. The background objects around the charged body are removed through density and curvature analysis to extract the point cloud data of the charged body.

[0096] The present invention uses a coarse registration method based on feature points and a fine registration method using an iterative closest point algorithm to accurately align point cloud data and perform global coordinate conversion, performs spatial cropping based on the center of the live working area, effectively retains core point cloud data related to the live body, compresses the data volume through voxel grid downsampling, identifies ground points in combination with height distribution analysis, and accurately removes background objects using density and curvature analysis, ultimately extracting high-quality point cloud data of the live body. This method significantly improves the accuracy and efficiency of point cloud processing, reduces the interference of redundant data, and provides an accurate and reliable three-dimensional model foundation for subsequent high-temperature judgment and risk analysis.

[0097] S2, a high temperature determination module, used to divide all charged bodies into finite element grid units and determine high temperature charged bodies;

[0098] Specifically, all charged bodies are divided into finite element mesh units and the high-temperature charged bodies are determined to include:

[0099] The surface of the geometric shape of the charged body is reconstructed based on the point cloud data to generate the corresponding three-dimensional device model. According to the number of charged bodies, the global coordinate system is divided into multiple finite element mesh units using the three-dimensional mesh division technology (each finite element mesh unit corresponds to a three-dimensional device model of a charged body). The type of each finite element mesh unit is set, and physical properties are assigned to all nodes in each finite element mesh unit. The number of nodes is determined according to the size of the charged body and the density of the divided mesh. The current density value is assigned to the central node of each finite element mesh unit using the inverse distance weighted interpolation method to obtain the current density value of each node in the finite element mesh unit, and the temperature value of the node in each finite element mesh unit is calculated;

[0100] The finite element mesh unit types include hexahedral units and tetrahedral units. The hexahedral units are used for the division of regular areas to ensure high calculation accuracy. The tetrahedral units are suitable for the division of complex or irregular areas. The high temperature gradient area is divided into fine mesh units (such as a mesh size of 0.01m), and the low temperature gradient area is divided into coarse mesh units (such as a mesh size of 0.05m).

[0101] The physical data include node coordinates and material parameters;

[0102] The material parameters include thermal conductivity, specific heat capacity and density of the charged body;

[0103] The calculation of the temperature value of each node in the finite element mesh unit includes establishing an initial temperature field control equation, which is:

[0104]

[0105] Where ρ represents density in kg / m 3 , cc represents specific heat capacity, the unit is J / kg·K, Represents partial derivative, which is used to describe the rate of change of a physical quantity with respect to another variable. is the Laplace operator of temperature, T g represents the initial temperature value of the g-th node in the finite element mesh unit, λ represents the thermal conductivity, the unit is W / m·K, ω0 represents the initial volume heat source intensity, the unit is W / m3;

[0106] The temperature Laplace operator represents the curvature of the temperature field in space. If a node in the finite element grid unit is higher than the surrounding area, the Laplace operator is negative, indicating that heat is transferred from the point to the surrounding area. Otherwise, the Laplace operator is positive, indicating that heat is transferred from the surrounding area to the point.

[0107] Traditional temperature field modeling mostly assumes that the heat source is evenly distributed, ignoring the dynamic characteristics of the heat generation process and operating conditions inside the equipment. This leads to deviations from actual results when the heat source distribution is complex (such as uneven current density or significant differences in material properties).

[0108] Dynamic non-uniform heat source distribution modeling is introduced to express the volume heat source intensity of each node as a dynamic distribution related to the current density value, initial temperature value and environmental conditions. The formula is:

[0109]

[0110] in, represents the volume heat source intensity of the g-th node after adjustment, ω0 represents the initial volume heat source intensity, f() represents the exponential function of the nonlinear function, I g represents the current density value of the gth node, T g represents the initial temperature value of the g-th node in the finite element mesh unit, and t represents the current time point;

[0111] The expression of the exponential function of the nonlinear function is:

[0112]

[0113] Among them, b1 represents the adjustment parameter, which is obtained by least squares fitting, b2 represents the temperature sensitivity coefficient, which is obtained by regression analysis of experimental data, b3 represents the amplitude coefficient, and w represents the frequency. The amplitude coefficient and frequency are obtained by fast Fourier transform and are used to control the dynamic fluctuation of time to heat source;

[0114] Set the maximum value of the volume heat source intensity to prevent the heat source intensity from increasing rapidly under high current or high temperature conditions;

[0115] The dynamic non-uniform heat source distribution modeling reflects the dynamic changes of the internal heat source of the charged body with the working conditions, accurately describes complex phenomena such as local overheating and high-power areas, and is particularly suitable for scenarios where the internal heat source of power equipment (such as transformers, insulators, and conductors) is unevenly distributed, solving the problem of insufficient accuracy caused by the assumption of uniform heat source in existing models.

[0116] All temperature values ​​in the charged body obtained by the temperature sensor are converted into absolute temperature, and the thermal radiation term is introduced to describe the heat loss on the surface of the charged body. The improved temperature field control equation is established in combination with the heat source intensity. The formula is:

[0117]

[0118] Where ρ represents density, cc represents specific heat capacity, represents the partial derivative, λ represents the thermal conductivity, is the Laplace operator of temperature, T g represents the initial temperature of the gth node in the finite element mesh unit, σ represents the Stefan-Bohr slow constant, ∈ represents the surface emissivity, T a Represents the absolute mean temperature of the charged body, T e Indicates the absolute temperature of the live working area;

[0119] Set boundary conditions and initial conditions, and obtain the temperature value of each node through numerical iteration;

[0120] The boundary conditions are set to accurately simulate the temperature field distribution of the charged body, and the boundary conditions are selected by combining the thermal conductivity boundary conditions and the thermal radiation boundary conditions to fully reflect the heat exchange characteristics between the surface of the charged body and the surrounding environment;

[0121] The initial conditions include setting the initial temperature value and initial volume heat source intensity of each node in the finite element grid unit at the time point t=0;

[0122] The numerical iteration includes using an explicit difference method to perform discrete calculations based on boundary conditions and using Newton iteration to solve the nonlinear term σ∈(T a 4 -T e 4 ) to avoid the low solution efficiency caused by directly substituting numerical solutions;

[0123] Set an error threshold. If the temperature change of each node is less than the error threshold in two adjacent iterations, the node stops iterating. Otherwise, it continues iterating. When all nodes stop iterating, the iteration process is completed.

[0124] According to the material characteristics of the charged body, a high temperature safety threshold is set. If the temperature value of the node is higher than the high temperature safety threshold, the node is marked as a high temperature node, otherwise, no operation is performed;

[0125] A ratio threshold is set. When the ratio of high-temperature nodes in the charged body is higher than the ratio threshold, the charged body is judged to be a high-temperature charged body. Otherwise, no operation is performed.

[0126] The present invention divides all charged bodies into finite element grid units, generates a three-dimensional equipment model based on point cloud data, introduces dynamic non-uniform heat source distribution modeling and improved temperature field control equations, accurately calculates the temperature distribution of high-temperature charged bodies, and allocates current density values, thermal radiation terms and dynamic heat source distribution characteristics by combining the inverse distance weighted interpolation method. It can accurately describe the complex heat generation and transfer process inside the charged body, especially the phenomena of local overheating and high-power areas. The high and low temperature gradient areas are refined by hexahedron and tetrahedron grids, which significantly improves the modeling accuracy. At the same time, the nonlinear terms are solved by the Newton iteration method to improve the calculation efficiency. Finally, the temperature field distribution is mapped to the three-dimensional point cloud model, and the high temperature area is intuitively displayed with color gradients, which provides accurate and efficient technical support for equipment fault diagnosis and operation safety monitoring, and solves the problem of insufficient accuracy caused by the traditional uniform heat source assumption.

[0127] S3, violation determination module, is used to calculate the electric field strength value of the operator's location in real time, capture the operator's action characteristics through the long short-term memory network and use the classification model to calculate the probability of violation behavior;

[0128] Specifically, the real-time calculation of the electric field strength value at the operator's location includes:

[0129] Extract the operator area from the 3D point cloud data generated by the lidar, map the 2D detection results to the 3D point cloud through the image captured by the camera (such as using the YOLO-v8 target detection model to locate the personnel), and use the region growing algorithm to segment the personnel area in the 3D point cloud. The algorithm starts from the target seed point and gradually expands to the neighboring points. Through the segmented personnel point cloud, the PointNet++ model is used to extract the operator skeleton, track the joint points of the skeleton point set frame by frame, calculate the displacement trajectory of the joint points in the time dimension, and generate a skeleton time series. The skeleton time series includes the 3D spatial coordinates of the operator at each time point;

[0130] Convert the three-dimensional space coordinates of the charged body into cylindrical coordinates (r, z y );

[0131] The three-dimensional space coordinate conversion to cylindrical coordinates includes obtaining the three-dimensional space coordinates, length and charge density of the center position of the charged body, taking the three-dimensional space coordinates of the center position of the charged body as the central axis, calculating the shortest radial distance r from the center axis of the charged body in the three-dimensional space coordinates of the operator and obtaining the vertical height z of the three-dimensional space coordinate corresponding to the radial distance y , taking the radial distance and vertical height as cylindrical coordinates;

[0132] According to the current-temperature coupling, the linear charge density is adjusted by the maximum temperature of all nodes in the charged body. The formula is:

[0133] τ(T)=τ0·(1+β·T max ),

[0134] Where τ(T) represents the linear charge density after temperature adjustment, τ0 represents the baseline linear charge density, β represents the temperature influence coefficient, and describes the sensitivity of the linear charge density to temperature changes. max It represents the maximum temperature of all nodes in the charged body;

[0135] When current passes through a charged body, the Joule heat generated will increase the temperature of the charged body. This temperature rise will further affect the conductivity, causing the current distribution and line charge density to change dynamically. The current distribution affects the temperature distribution, and the temperature change in turn affects the conductivity and current distribution, forming a coupling mechanism. If the charged body generates high temperature, it will significantly change the material's conductive properties or charge accumulation ability, causing a significant change in the line charge density.

[0136] Based on the cylindrical coordinates, the equivalent charge method is used to calculate the electric field potential value generated by each charged body. The formula is:

[0137]

[0138] Among them, φ i (r,z y ) represents the i-th charged body in the cylindrical coordinates (r, z y ), L represents the length of the charged body, z s represents the integral variable along the central axis, ε represents the dielectric constant of air, and d represents the integral variable z s Small variables

[0139] The electric field strength value at the operator's location is calculated by the electric field potential value generated by each charged body. The formula is:

[0140]

[0141] Among them, E i represents the electric field strength value of the i-th charged body, A negative value that represents the spatial rate of change of the electric field potential.

[0142] By calculating the electric field strength value of the operator's position in real time, the present invention is based on a combination of lidar point cloud and camera image to accurately extract the operator's skeleton time series to ensure the accuracy of the spatial position. By converting the three-dimensional coordinates of the charged body into cylindrical coordinates and combining the current-temperature coupling effect, the line charge density is dynamically adjusted to fully reflect the dynamic influence of high temperature on the electric field distribution. The electric field strength value is calculated using the equivalent charge method and the spatial rate of change of the electric field potential, which overcomes the problem of traditional methods ignoring the correlation between temperature and electric field. The process realizes high-precision and dynamic electric field strength evaluation, provides a scientific basis for accurately identifying high-risk areas and dangerous electric fields in the working environment, and significantly improves the accuracy and reliability of work safety monitoring.

[0143] Furthermore, the long short-term memory network is used to capture the operator's action characteristics and the classification model is used to calculate the probability of violation behavior, including:

[0144] The generated skeleton time series is input into the long short-term memory network for time series analysis. LSTM can effectively capture the long-term dependencies in the time series and is suitable for the time series modeling of behavioral actions. LSTM captures historical information through memory units, and filters and updates the information through the forget gate, input gate, and output gate to generate action features at each time point.

[0145] The electric field strength value, action characteristics and temperature values ​​of all nodes in each charged body are fused into comprehensive features through dynamic weighting, and the comprehensive features are input into a classification model (such as a fully connected neural network based on deep learning) to determine the violation. The classification model generates the probability of violation through the softmax function.

[0146] The present invention uses a long short-term memory network (LSTM) to perform time series analysis on the skeleton time series of the operators, which can capture the long-term dependencies of behavioral actions, accurately extract dynamic action features, combine the electric field strength value and the temperature value of the charged body, and use dynamic weighted fusion to generate comprehensive features. The multimodal data input classification model that comprehensively considers environmental and behavioral factors is used to determine violations. This method effectively improves the accuracy and reliability of behavior recognition in complex working scenarios, especially in high-voltage electric fields and dynamic temperature environments. It solves the problem of inaccurate identification of complex dynamic behaviors by traditional methods, thereby significantly improving the safety monitoring level of the working area.

[0147] S4, path planning module, used to remind operators in real time and dynamically plan the operation path according to the probability of violation;

[0148] Specifically, real-time reminders to operators based on the probability of violations include:

[0149] A probability threshold is set and the probability of violation generated by the classification model is compared with the probability threshold. If the probability of violation is greater than the probability threshold, all three-dimensional spatial coordinates corresponding to the high-temperature charged body are defined as risk points, and the operator is reminded in real time through the AR device. Otherwise, no operation is performed.

[0150] This step defines the three-dimensional spatial coordinates corresponding to high-temperature live bodies as risk points by setting a probability threshold for violation behavior and combining the judgment results of the classification model. The AR device is used to remind operators in real time, effectively realizing dynamic monitoring and precise warnings for high-risk operations. This method helps operators quickly identify dangerous areas and avoid them in time through intuitive visual reminders, significantly improving the safety of live operations and the efficiency of emergency response, and reducing the possibility of human misjudgment and accidents.

[0151] Furthermore, the dynamic planning operation path includes:

[0152] The live working area is divided into three-dimensional grid units, each of which represents a cubic area. The side length of the three-dimensional grid unit is determined according to the working scene requirements.

[0153] All risk points are integrated into a risk coordinate set, and the three-dimensional spatial coordinates of each risk point in the risk coordinate set are traversed and mapped to each three-dimensional grid unit. If the three-dimensional spatial coordinates of the risk point are in the three-dimensional grid unit, the three-dimensional grid unit is marked as impassable and the path is prohibited from passing through the three-dimensional grid unit. Otherwise, the shortest Euclidean distance from the center point of the three-dimensional grid unit to all risk points is calculated, and the shortest Euclidean distance is used as the risk cost of the three-dimensional grid unit. The smaller the shortest Euclidean distance from the three-dimensional grid unit to the risk point, the higher the risk cost, so that the path planning can avoid these high-risk areas as much as possible;

[0154] According to the target three-dimensional grid unit of the operator, a path cost function of each three-dimensional grid unit is defined;

[0155] The path cost function includes a cumulative cost and an estimated cost;

[0156] The cumulative cost refers to the path cost from the operator's starting point to the current three-dimensional grid unit, including all risk costs of all three-dimensional grid units passed on the path and the Euclidean distance between the center points of all three-dimensional grid units;

[0157] The estimated cost refers to the Euclidean distance from the current grid cell to the target grid cell;

[0158] The D** algorithm is used to expand the three-dimensional grid cell with the minimum total cost, gradually retrieve the optimal path from the starting three-dimensional grid cell to the target three-dimensional grid cell, and generate the optimal path sequence.

[0159] The present invention divides the live working area into three-dimensional grid units, combines the dynamic marking of risk areas and the design of cost functions, accurately evaluates the passability of each grid unit, and uses the D** algorithm to plan the optimal working path in real time based on the principle of minimizing the total cost. This path planning method can dynamically avoid high-risk areas and ensure that workers always move on a safe path in a complex and constantly changing working environment, effectively improving the safety of live working, the flexibility of path planning and the working efficiency, and solving the problem of insufficient adaptability of traditional path planning to dynamic high-risk scenarios.

[0160] S5, data storage module, used to store and manage all data in a relational database;

[0161] Specifically, storing and managing all data in a relational database includes:

[0162] Choose a relational database to store and manage data and analysis results, design the database table structure to store different types of data, set up regular backup tasks, back up all data in the database, manage permissions for database users and encrypt and store data.

[0163] By selecting a relational database to store and manage data and analysis results, designing a reasonable table structure to store multiple types of data, and combining a regular backup mechanism to ensure data security and integrity, while ensuring data confidentiality and access control through user authority management and data encryption storage, the standardization of data storage, the efficiency of management, and the security of data use are effectively improved, providing a solid foundation for the long-term and stable operation of the system.

[0164] Embodiment 2, if the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0165] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0166] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0167] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0168] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A live working safety monitoring system based on artificial intelligence, characterized by: include: The data acquisition module is used to collect the operation data in the live operation area in real time and perform preprocessing, and extract the point cloud data of the live body based on the operation data; A high temperature determination module is used to divide all charged bodies into finite element grid units and determine high temperature charged bodies; The violation determination module is used to calculate the electric field strength value of the operator's location in real time, capture the operator's action characteristics through the long short-term memory network, and use the classification model to calculate the probability of violation behavior; The path planning module is used to remind operators in real time and dynamically plan the operation path based on the probability of violation; The data storage module is used to store and manage all data in a relational database.

2. The artificial intelligence-based live working safety monitoring system according to claim 1, characterized in that: The real-time collection of operation data in the live working area includes: Install Hall effect current sensors and temperature sensors on each live body in the live working area and set the sampling frequency of the sensors. Use a wireless transmission module to transmit the collected current density and temperature values ​​to the central control system in real time. Start the laser radar and set the scanning frequency of the laser radar. Set the acquisition range to the spherical space of the live working area. The laser radar generates point cloud data in real time through multi-line laser emission and reflection measurement; The point cloud data refers to the three-dimensional space coordinates of each reflection point.

3. The artificial intelligence-based live working safety monitoring system according to claim 1, characterized in that: The pre-processing comprises: Check the integrity of the point cloud data, remove erroneous data during the collection process, calculate the average distance of the field points and remove abnormal points. For the edge area of ​​the live working area, use the density calculation formula, define the point density as the number of field points within the unit spherical radius, detect isolated points through local density changes and remove them. The current density and temperature values ​​were cleaned and the data were smoothed using the sliding smoothing filter method.

4. The artificial intelligence-based live working safety monitoring system according to claim 3, characterized in that: The extracting of charged body point cloud data based on operation data comprises: The point cloud data is aligned and converted into a global coordinate system using a coarse registration method based on feature points and a fine registration method using an iterative closest point algorithm. Based on the center position of the live working area, the point cloud data in the global coordinate system is spatially cropped, and the point cloud data centered on the live body is retained; Voxel grid downsampling is used to compress point cloud data, and the height distribution of point cloud data is used to identify ground points. The background objects around the charged body are removed through density and curvature analysis to extract the point cloud data of the charged body.

5. The artificial intelligence-based live working safety monitoring system according to claim 1, characterized in that: The method of dividing all charged bodies into finite element grid units and determining high-temperature charged bodies includes: The surface of the geometric shape of the charged body is reconstructed based on the point cloud data to generate the corresponding three-dimensional equipment model. According to the number of charged bodies, the global coordinate system is divided into multiple finite element mesh units using the three-dimensional meshing technology. The type of each finite element mesh unit is set, and physical properties are assigned to all nodes in each finite element mesh unit. The current density value is assigned to the central node of each finite element mesh unit using the inverse distance weighted interpolation method to obtain the current density value of each node in the finite element mesh unit, and the temperature value of the node in each finite element mesh unit is calculated; The finite element mesh unit types include hexahedral units and tetrahedral units; The physical data includes node coordinates and material parameters; The material parameters include thermal conductivity, specific heat capacity and density of the charged body; The calculating the temperature value of the node in each finite element mesh unit includes establishing an initial temperature field control equation; Dynamic non-uniform heat source distribution modeling is introduced to express the volume heat source intensity of each node as a dynamic distribution related to the current density value, initial temperature value and environmental conditions. The formula is: in, represents the volume heat source intensity of the g-th node after adjustment, ω0 represents the initial volume heat source intensity, f() represents the exponential function of the nonlinear function, I g represents the current density value of the gth node, T g represents the initial temperature value of the g-th node in the finite element mesh unit, and t represents the current time point; The expression of the exponential function of the nonlinear function is: Among them, b1 represents the adjustment parameter, b2 represents the temperature sensitivity coefficient, b3 represents the amplitude coefficient, and w represents the frequency; All temperature values ​​in the charged body obtained by the temperature sensor are converted into absolute temperature, and the thermal radiation term is introduced to describe the heat loss on the surface of the charged body. The improved temperature field control equation is established in combination with the heat source intensity. The formula is: Where ρ represents density, cc represents specific heat capacity, represents the partial derivative, λ represents the thermal conductivity, is the Laplace operator of temperature, T g represents the initial temperature of the gth node in the finite element mesh unit, σ represents the Stefan-Bohr slow constant, ∈ represents the surface emissivity, T a Represents the absolute mean temperature of the charged body, T e Indicates the absolute temperature of the live working area; Set boundary conditions and initial conditions, and obtain the temperature value of each node through numerical iteration; The initial conditions include setting the initial temperature value and initial volume heat source intensity of each node in the finite element grid unit at the time point t=0; The numerical iteration includes using an explicit difference method to perform discrete calculations based on boundary conditions and using Newton iteration to solve the nonlinear term σ∈(T a 4 -T e 4 ); Set an error threshold. If the temperature change of each node is less than the error threshold in two adjacent iterations, the node stops iterating. Otherwise, it continues iterating. When all nodes stop iterating, the iteration process is completed. Set a high temperature safety threshold. If the temperature of a node is higher than the high temperature safety threshold, the node is marked as a high temperature node. Otherwise, no operation is performed. A ratio threshold is set. When the ratio of high-temperature nodes in the charged body is higher than the ratio threshold, the charged body is judged to be a high-temperature charged body. Otherwise, no operation is performed.

6. The artificial intelligence-based live working safety monitoring system according to claim 1, characterized in that: The real-time calculation of the electric field strength value at the location of the operator includes: The operator area is extracted from the 3D point cloud data generated by the lidar. The 2D detection results are mapped to the 3D point cloud through the image captured by the camera. The region growing algorithm is used to segment the operator area in the 3D point cloud. The algorithm starts from the target seed point and gradually expands to the neighboring points. The operator skeleton is extracted through the segmented point cloud using the PointNet++ model. The joint points of the skeleton point set are tracked frame by frame, and the displacement trajectory of the joint points in the time dimension is calculated to generate a skeleton time series. The skeleton time series includes the 3D spatial coordinates of the operator at each time point. Convert the three-dimensional space coordinates of the charged body into cylindrical coordinates (r, z y ); The three-dimensional space coordinate conversion to cylindrical coordinates includes obtaining the three-dimensional space coordinates, length and charge density of the center position of the charged body, taking the three-dimensional space coordinates of the center position of the charged body as the central axis, calculating the shortest radial distance r from the center axis of the charged body in the three-dimensional space coordinates of the operator and obtaining the vertical height z of the three-dimensional space coordinate corresponding to the radial distance y , taking the radial distance and vertical height as cylindrical coordinates; According to the current-temperature coupling, the linear charge density is adjusted by the maximum temperature of all nodes in the charged body. The formula is: τ(T)=τ0·(1+β·T max ), Where τ(T) represents the linear charge density after temperature adjustment, τ0 represents the baseline linear charge density, β represents the temperature influence coefficient, and T max It represents the maximum temperature of all nodes in the charged body; Based on the cylindrical coordinates, the equivalent charge method is used to calculate the electric field potential value generated by each charged body. The formula is: Among them, φ i (r,z y ) represents the i-th charged body in the cylindrical coordinates (r, z y ), L represents the length of the charged body, z s represents the integral variable along the central axis, ε represents the dielectric constant of air, and d represents the integral variable z s Small variables The electric field strength value at the operator's location is calculated by the electric field potential value generated by each charged body. The formula is: Among them, E i represents the electric field strength value of the i-th charged body, A negative value that represents the spatial rate of change of the electric field potential.

7. The artificial intelligence-based live working safety monitoring system according to claim 1, characterized in that: The method of capturing the action features of operators through the long short-term memory network and using the classification model to calculate the probability of violation behavior includes: The generated skeleton time series is input into the long short-term memory network for time series analysis. LSTM captures historical information through memory units, and filters and updates the information through the forget gate, input gate, and output gate to generate action features at each time point. The electric field strength value, action characteristics and the temperature values ​​of all nodes in each charged body are fused into comprehensive features through dynamic weighting, and the comprehensive features are input into the classification model to determine the violation behavior. The classification model generates the violation probability through the softmax function.

8. The artificial intelligence-based live working safety monitoring system according to claim 1, characterized in that: The real-time reminder to the operator according to the probability of violation includes: A probability threshold is set and the probability of violation generated by the classification model is compared with the probability threshold. If the probability of violation is greater than the probability threshold, all three-dimensional spatial coordinates corresponding to the high-temperature charged body are defined as risk points, and the operator is reminded in real time through the AR device. Otherwise, no operation is performed.

9. The artificial intelligence-based live working safety monitoring system according to claim 1, characterized in that: The dynamic planning operation path includes: Divide the live working area into three-dimensional grid units, each of which represents a cubic area; All risk points are integrated into a risk coordinate set, and the three-dimensional spatial coordinates of each risk point in the risk coordinate set are traversed and mapped to each three-dimensional grid unit. If the three-dimensional spatial coordinates of the risk point are in the three-dimensional grid unit, the three-dimensional grid unit is marked as inaccessible, and the path is prohibited from passing through the three-dimensional grid unit. Otherwise, the shortest Euclidean distance from the center point of the three-dimensional grid unit to all risk points is calculated, and the shortest Euclidean distance is used as the risk cost of the three-dimensional grid unit. The smaller the shortest Euclidean distance between the three-dimensional grid unit and the risk point, the higher the risk cost. According to the target three-dimensional grid unit of the operator, a path cost function of each three-dimensional grid unit is defined; The path cost function includes a cumulative cost and an estimated cost; The cumulative cost refers to the path cost from the operator's starting point to the current three-dimensional grid unit, including all risk costs of all three-dimensional grid units passed on the path and the Euclidean distance between the center points of all three-dimensional grid units; The estimated cost refers to the Euclidean distance from the current grid cell to the target grid cell; The D** algorithm is used to expand the three-dimensional grid cell with the minimum total cost, gradually retrieve the optimal path from the starting three-dimensional grid cell to the target three-dimensional grid cell and generate the optimal path sequence.

10. The artificial intelligence-based live working safety monitoring system according to claim 9, characterized in that: The storing and managing of all data in a relational database includes: Choose a relational database to store and manage data and analysis results, design the database table structure to store different types of data, set up regular backup tasks, back up all data in the database, manage permissions for database users and encrypt and store data.

Images