Three-dimensional dynamic safety management and control system for power transmission and transformation hoisting operation

Abstract

The invention specifically discloses a three-dimensional dynamic safety management and control system for power transmission and transformation hoisting operation, and relates to the technical field of electric power engineering construction safety. The system comprises a multi-source data sensing layer, a spatio-temporal data processing layer, a digital twin modeling layer, an intelligent decision-making layer and an application output layer. The multi-source data sensing layer collects space-time positioning and visual image data; the spatio-temporal data processing layer realizes data standardization processing and spatio-temporal unification; the digital twinborn modeling layer constructs a static digital twinborn and dynamic target 4D model; the intelligent decision-making layer carries out risk assessment and early warning decision-making; and the application output layer displays information and outputs an alarm. The system overcomes the defects of an existing hoisting safety management and control means, active and intelligent safety management and control are achieved, the safety protection precision and reliability of hoisting operation under complex scenes are improved, and the system can be applied to hoisting operation of high-voltage and extra-high-voltage power transmission and transformation projects.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power engineering construction safety, in particular to a three-dimensional dynamic safety management and control system for hoisting operation of power transmission and transformation. BACKGROUND

[0002] The hoisting operation of power transmission and transformation projects, especially the core hoisting link in the construction of high-voltage and ultra-high-voltage line erection and substation, has long been facing the challenge of extremely complex operating environment. Such scenes are often limited by site space constraints, with densely distributed power equipment and tower components, and are often adjacent to high-voltage live areas, forming a multi-dimensional and three-dimensional operating space, which dramatically increases the risk of interference between the hoisting arm, hoisting object and the surrounding environment during hoisting, and becomes a high-risk link of safety accidents in the construction stage of power transmission and transformation projects, and is also the top priority of safety management and control.

[0003] Currently, the safety guarantee system of hoisting operation still highly depends on traditional technical paths and human intervention, and risk prevention and control is mainly achieved through three types of ways: first, mechanical or electrical limit devices, such as torque limiters and height limiters, which can only trigger an alarm or forced shutdown when a preset threshold is reached in a single physical dimension, and belong to the typical threshold triggered passive protection; second, monitoring systems based on simple sensors such as ultrasonic waves and lasers, which determine the distance between the hoisting arm and obstacles through single-point distance measurement, but lack overall perception of spatial relationships; third, manual lookout, experience judgment and verbal command intervention by on-site command personnel and guardians, which are highly subjective and easily disturbed by the environment.

[0004] Currently, the traditional control mode has cognitive bias on the hoisting operation risk - the essence of hoisting risk is a four-dimensional (three-dimensional space + time) dynamic problem, while the existing technology only reduces the dimension through one-dimensional scalar threshold or two-dimensional local information, resulting in serious loss of key safety information; Specifically, first, it lacks proactive foresight and can only respond passively when the risk is imminent or occurs, and cannot identify potential dangers based on movement trends; second, the environmental perception dimension is insufficient, and it is difficult to restore the three-dimensional spatial relationship between the hoisting arm, hoisting object and complex environment, and cannot distinguish the semantic attributes of obstacles to implement differentiated prevention and control; third, the data support capability is weak, and only low-dimensional information such as distance and on-off quantity can be provided, making it difficult to form structured data to support intelligent decision-making.

[0005] With the advancement of new infrastructure and the large-scale construction of ultra-high-voltage projects, the traditional control mode has been unable to meet the needs of modern power engineering for active and intelligent safety guarantees, and it is urgent to break through the technical bottleneck and build a new safety management and control system that can accurately perceive the environment, dynamically understand the scene and foresee risks in advance. SUMMARY

[0006] The purpose of the present application is to provide a power transmission and transformation hoisting operation three-dimensional dynamic safety management and control system, to realize the paradigm shift from passive response to active early warning, and to improve the safety management and control precision and reliability of hoisting operation in complex scenes.

[0007] To achieve the above-mentioned purpose, the present application provides a power transmission and transformation hoisting operation three-dimensional dynamic safety management and control system, which comprises a multi-source data perception layer, a space-time data processing layer, a digital twin modeling layer, an intelligent decision-making layer and an application output layer. The multi-source data perception layer is used for collecting space-time positioning data and visual image data of the hoisting operation site, and comprises a Beidou real-time dynamic differential positioning module, a multi-view industrial camera module and an inertial measurement module. The Beidou real-time dynamic differential positioning module provides centimeter-level precision position and timestamp information. The industrial camera module collects multi-view video streams on site. The inertial measurement module assists in obtaining device motion attitude data. The space-time data processing layer is used for realizing standardized processing and space-time unification of multi-source data, and comprises a time synchronization unit, a space calibration unit and a data preprocessing unit. The time synchronization unit aligns the timestamps of all sensor data based on Beidou time. The space calibration unit determines the external parameter relationship between sensors through joint calibration and converts it to a unified vehicle coordinate system. The data preprocessing unit completes image denoising, point cloud filtering and abnormal data elimination. The digital twin modeling layer is used for constructing a static digital twin of the operation scene and a dynamic target 4D model, and comprises a static modeling unit and a dynamic reconstruction unit. The static modeling unit fuses ground three-dimensional laser scanning and oblique photography data to generate a centimeter-level three-dimensional model containing geographic reference. The dynamic reconstruction unit synchronously generates the three-dimensional geometric structure and motion trajectory of the hoisted object from the video stream through an end-to-end neural network. The intelligent decision-making layer is used for risk assessment and early warning decision-making, and comprises a collision detection engine, a violation identification engine and a hierarchical alarm unit. The collision detection engine is used for calculating the shortest distance between the predicted trajectory of the hoisted object and the static electronic fence. The violation identification engine identifies the behavior of not wearing a safety helmet and personnel intrusion based on the YOLOv8 model. The hierarchical alarm unit generates comprehensive warning information in combination with geometric, semantic and behavior risk degrees. The application output layer is used for information display and alarm output, and comprises a three-dimensional visualization interface, an audible and visual alarm device and an event log system. The visualization interface presents the digital twin scene and risk information in real time. The audible and visual alarm device triggers differentiated warnings according to the warning level. The log system records all risk events and device states.

[0008] Preferably, the time synchronization unit aligns the timestamps of all sensor data based on Beidou time, and the formula is as follows: ; in, The standard timestamp output by BeiDou This is the sensor's original timestamp. This refers to the time deviation rate. It is a fixed delay.

[0009] Preferably, the spatial calibration unit transforms the extrinsic parameter relationships between sensors to a unified vehicle body coordinate system, as shown in the following formula: ; in, To unify the coordinates of target points in the vehicle body coordinate system, The coordinates of the target point are in the local coordinate system of the sensor. Let be a rotation matrix. It is a translation vector; Rotation matrix Expressed using Euler angles, the formula is as follows: ; in, α For roll angle, β For pitch angle, γ This is the yaw angle.

[0010] Preferably, the static modeling unit also includes a semantic annotation subunit, which annotates entities in the 3D model through AI point cloud segmentation and human interaction, defines the 3D geometry of the entities as a static electronic fence, and assigns security attribute parameters to each entity.

[0011] Preferably, the dynamic reconstruction unit uses the extended Kalman filter algorithm to achieve state estimation based on state prediction and to realize the motion range in the next 100-200 milliseconds through trajectory prediction; The state prediction formula is as follows: ; in, For a moment k Prior state estimation, Here is the state transition matrix. For a moment k- Posterior state estimation of 1, To control the input matrix, For a moment k The control input vector, This is process noise; State transition matrix Based on a uniformly accelerated motion model, the formula is as follows: ; in, I 3 is a 3×3 identity matrix, 03 is a 3×3 zero matrix, Δt The filter period; The trajectory prediction formula is as follows: ; in, To predict the state, , , For a moment k Position state estimation, , , For a moment k Velocity state estimation, , , For at any time k Acceleration state estimation, The time step for prediction.

[0012] Preferably, the collision detection engine uses the GJK algorithm, which calculates the shortest distance between the predicted trajectory of the suspended object and the static electronic fence by solving the support vectors, as shown in the following formula: ; in, As support vectors, convex hull A In direction Support points on the top convex hull B In direction Support points on the top It is the direction vector; The GJK algorithm updates the direction vector iteratively. ,when When the shortest distance is obtained, the formula is as follows: ; in, For the shortest distance, To be perpendicular to A vector pointing to the origin.

[0013] Preferably, the tiered alarm unit generates a comprehensive warning based on the quantification of collision risk level and a multi-dimensional comprehensive risk index; The formula for quantifying collision risk level is as follows: ; in, d For the shortest distance, D 1 is the warning threshold. D 2 represents the danger threshold; The formula for the multi-dimensional risk composite index is as follows: ; wherein, is a comprehensive risk index, is a collision risk coefficient, is a semantic risk coefficient, is a behavior risk coefficient, is a weight coefficient.

[0014] Preferably, the hierarchical alarm unit is built-in with a risk matrix that integrates geometric risk, semantic risk and behavior risk to output a first-level early warning and a second-level alarm; The geometric risk is derived based on the ratio of the warning threshold to the danger threshold of the shortest distance; The semantic risk is based on a weight coefficient set according to the importance of entities; The behavior risk is based on the severity classification of rule-breaking behaviors; The first-level early warning includes visual highlighting and sound prompt; The second-level alarm includes strong visual and sound prompt and collision point indication.

[0015] Preferably, the control system further comprises an edge computing unit integrated in the spatio-temporal data processing layer and the intelligent decision-making layer, which uses lightweight algorithm optimization technology to control the response time of data processing and risk inference within 100 milliseconds.

[0016] Therefore, the present application proposes a three-dimensional dynamic safety control system for power transmission and transformation hoisting operation, which has the following beneficial effects: (1) The present application realizes four-dimensional active early warning, which upgrades the traditional one-dimensional passive protection to four-dimensional active control through spatio-temporal intelligent fusion and 4D reconstruction technology, and predicts collision risks in advance to reserve reaction time for operators; (2) The present application improves the sensing accuracy and coverage, which realizes centimeter-level positioning and full-space non-dead-angle sensing based on Beidou real-time dynamic differential positioning technology and three-dimensional modeling technology, and solves the problems of low accuracy and limited coverage of traditional methods; (3) The present application realizes multi-dimensional risk accurate judgment, which reduces the risk false alarm rate and avoids alarm fatigue through the fusion of geometric, semantic and behavior multi-dimensional risk assessment.

[0017] The technical solutions of the present application will be further described in detail below with the aid of drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 Figure 1 is a schematic diagram of the overall structure of the three-dimensional dynamic safety control system for power transmission and transformation hoisting operation of the present application. DETAILED DESCRIPTION

[0019] To make the technical solutions, advantages, and objectives of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below. The described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0020] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0021] Example like Figure 1 As shown, the present invention provides a three-dimensional dynamic safety management and control system for power transmission and transformation hoisting operations, including a multi-source data perception layer, a spatiotemporal data processing layer, a digital twin modeling layer, an intelligent decision-making layer, and an application output layer; The multi-source data sensing layer is used to collect spatiotemporal positioning data and visual image data at the hoisting operation site. It includes a Beidou real-time dynamic differential positioning module, a multi-view industrial camera module, and an inertial measurement module. The Beidou real-time dynamic differential positioning module provides centimeter-level accuracy position and timestamp information, the industrial camera module collects multi-view video streams at the site, and the inertial measurement module assists in acquiring equipment motion attitude data. The spatiotemporal data processing layer is used to achieve standardized processing and spatiotemporal unification of multi-source data. It includes a time synchronization unit, a spatial calibration unit, and a data preprocessing unit. The time synchronization unit aligns the timestamps of all sensor data with BeiDou time as the reference, as shown in the following formula: ; in, The standard timestamp output by BeiDou This is the sensor's original timestamp. This refers to the time deviation rate. For fixed delay; The spatial calibration unit determines the extrinsic relationships between sensors through joint calibration and transforms them to a unified vehicle coordinate system, as shown in the following formula: ; in, To unify the coordinates of target points in the vehicle body coordinate system, The coordinates of the target point are in the local coordinate system of the sensor. Let be a rotation matrix. It is a translation vector; Rotation matrix Expressed using Euler angles, the formula is as follows: ; in, αFor roll angle, β For pitch angle, γ Yaw angle; The data preprocessing unit performs image denoising, point cloud filtering, and outlier data removal; The digital twin modeling layer is used to construct a static digital twin of the work scene and a dynamic target 4D model, including static modeling units and dynamic reconstruction units; The static modeling unit integrates terrestrial 3D laser scanning and oblique photography data to generate centimeter-level 3D models with geographic references; it also includes a semantic annotation subunit, which annotates entities in the 3D model through AI point cloud segmentation and human interaction, defines the 3D geometry of the entities as static electronic fences, and assigns security attribute parameters to each entity.

[0022] The dynamic reconstruction unit synchronously generates the three-dimensional geometric structure and motion trajectory of the suspended object from the video stream through an end-to-end neural network. It adopts the extended Kalman filter algorithm, realizes state estimation based on state prediction, and realizes the motion range in the next 100-200 milliseconds through trajectory prediction. The state prediction formula is as follows: ; in, For a moment k Prior state estimation, Here is the state transition matrix. For a moment k- Posterior state estimation of 1, To control the input matrix, For a moment k The control input vector, This is process noise; State transition matrix Based on a uniformly accelerated motion model, the formula is as follows: ; in, I 3 is a 3×3 identity matrix, 03 is a 3×3 zero matrix, Δ t The filter period; The trajectory prediction formula is as follows: ; in, To predict the state, , , For a moment k Position state estimation, , , For a moment k Velocity state estimation, , , For at any time k Acceleration state estimation, The time step for prediction.

[0023] The intelligent decision-making layer is used for risk assessment and early warning decisions, including a collision detection engine, a violation recognition engine, and a hierarchical alarm unit. The collision detection engine is used to calculate the shortest distance between the predicted trajectory of the suspended object and the static electronic fence. The violation recognition engine is based on the YOLOv8 model to identify behaviors such as not wearing a safety helmet and personnel intrusion. The hierarchical alarm unit combines geometric, semantic, and behavioral hazard levels to generate comprehensive early warning information. The collision detection engine uses the GJK algorithm, which calculates the shortest distance between the predicted trajectory of the suspended object and the static electronic fence by solving the support vectors. The formula is as follows: ; in, As support vectors, convex hull A In direction Support points on the top convex hull B In direction Support points on the top It is the direction vector; The GJK algorithm updates the direction vector iteratively. ,when When the shortest distance is obtained, the formula is as follows: ; in, For the shortest distance, To be perpendicular to A vector pointing to the origin.

[0024] The tiered alarm unit generates comprehensive warnings based on collision risk level quantification and multi-dimensional risk comprehensive index. The formula for quantifying collision risk level is as follows: ; in, d For the shortest distance, D 1 is the warning threshold. D 2 represents the danger threshold; The formula for the multi-dimensional risk composite index is as follows: ; in, As a comprehensive risk index, For collision risk factor, For semantic risk coefficient, For behavioral risk coefficient, These are the weighting coefficients.

[0025] The tiered alarm unit also has a built-in risk matrix, which integrates geometric risk, semantic risk, and behavioral risk to output a first-level warning and a second-level alarm. Geometric hazard is calculated based on the ratio of the shortest distance between the warning and hazard thresholds; Semantic risk is determined by weighting coefficients based on entity importance; Behavioral hazard level is graded based on the severity of the violation; Level 1 warnings include visual highlighting and audible alerts; Level 2 alarms include strong audible and visual warnings and collision point indications.

[0026] The application output layer is used for information display and alarm output, including a 3D visualization interface, audible and visual alarm devices, and an event log system. The visualization interface presents digital twin scenes and risk information in real time, the audible and visual alarm devices trigger differentiated warnings according to the warning level, and the log system records all risk events and equipment status.

[0027] The control system also includes an edge computing unit, which is integrated into the spatiotemporal data processing layer and the intelligent decision-making layer. It uses lightweight algorithm optimization technology to control the response time of data processing and risk reasoning within 100 milliseconds.

[0028] The invention will be further illustrated below through specific implementation examples.

[0029] At the construction site of an ultra-high voltage substation, the BeiDou real-time dynamic differential positioning module used was the Huace T10 Pro (positioning accuracy: horizontal ±8mm+1ppm, elevation ±15mm+1ppm), the industrial camera used was the Hikvision MV-CA050-10GC (5 megapixels, 30fps), the IMU used was the ADI ADIS16460 (angular rate accuracy 0.005° / s), and the edge computing unit used was the NVIDIA Jetson AGX Xavier. The Beidou real-time dynamic differential positioning module is installed at the center of the top of the crane. Industrial cameras are deployed at the top of the boom, in front of the cab, and on both sides of the crane. The IMU is integrated at the root of the boom. A 200mm×200mm checkerboard calibration plate is used to complete the camera intrinsic parameter calibration. By slowly rotating the boom of the crane, sensor data under multiple postures are collected to complete the extrinsic parameter calibration.

[0030] The static modeling unit uses a Faro Focus S70 laser scanner to collect point cloud data and combines it with DJI drones for oblique photography to obtain 200 images with 80% overlap. A 3D model is reconstructed using Context Capture software, and georeferencing is achieved using BeiDou real-time dynamic differential positioning control points. The AI ​​point cloud uses the PointNet++ algorithm to automatically segment the model, identifying 12 types of entities such as the main transformer and GIS equipment. After manual correction, a static electronic fence is generated. During the main transformer hoisting operation, the dynamic reconstruction unit generates a 4D model of the hoisted object in real time, accurately predicting the swing trajectory of the object. When the distance between the hoisted object and the GIS equipment approaches the warning threshold (1.5m), the system triggers a level one warning. When the distance drops to the danger threshold (0.8m), a level two alarm is triggered, indicating the collision point location. The operator adjusts the boom in time to avoid a collision accident.

[0031] Construction results show that the static model reconstruction accuracy reaches ±2cm, the positioning error of the suspended object is less than 5cm, and the trajectory prediction error is less than 10cm; the collision detection response time is less than 50ms, and the violation identification accuracy rate reaches 96%; under complex conditions such as light intensity of 500-10000 lux and dusty weather, the system operates stably, with a warning accuracy rate of 98% and a false alarm rate of less than 2%.

[0032] It is worth noting that all contents not described in detail in this invention are existing technologies and are well known to those skilled in the art.

[0033] Therefore, this invention provides a three-dimensional dynamic safety management and control system for power transmission and transformation hoisting operations, realizing the transformation from manual observation to proactive early warning. It can accurately depict the four-dimensional risk status, predict the trajectory of hoisted objects in advance, quantify collision risks in real time, and integrate multi-dimensional hazards to generate accurate graded early warnings. It achieves full-space safety management without blind spots, ensuring the safety of personnel and equipment, and providing an intelligent safety solution for power transmission and transformation hoisting operations.

[0034] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A three-dimensional dynamic safety control system for power transmission and transformation hoisting operations, characterized in that, It includes a multi-source data perception layer, a spatiotemporal data processing layer, a digital twin modeling layer, an intelligent decision-making layer, and an application output layer; The multi-source data sensing layer is used to collect spatiotemporal positioning data and visual image data at the hoisting operation site. It includes a Beidou real-time dynamic differential positioning module, a multi-view industrial camera module, and an inertial measurement module. The Beidou real-time dynamic differential positioning module provides centimeter-level accuracy position and timestamp information, the industrial camera module collects multi-view video streams at the site, and the inertial measurement module assists in acquiring equipment motion attitude data. The spatiotemporal data processing layer is used to achieve standardized processing and spatiotemporal unification of multi-source data. It includes a time synchronization unit, a spatial calibration unit, and a data preprocessing unit. The time synchronization unit aligns the timestamps of all sensor data with BeiDou time as the reference. The spatial calibration unit determines the external parameter relationship between sensors through joint calibration and transforms it to a unified vehicle coordinate system. The data preprocessing unit completes image denoising, point cloud filtering, and abnormal data removal. The digital twin modeling layer is used to construct a static digital twin and a dynamic target 4D model of the operation scene. It includes a static modeling unit and a dynamic reconstruction unit. The static modeling unit integrates ground 3D laser scanning and oblique photography data to generate a centimeter-level 3D model with geographic reference. The dynamic reconstruction unit synchronously generates the 3D geometry and motion trajectory of the suspended object from the video stream through an end-to-end neural network. The intelligent decision-making layer is used for risk assessment and early warning decisions, including a collision detection engine, a violation recognition engine, and a hierarchical alarm unit. The collision detection engine is used to calculate the shortest distance between the predicted trajectory of the suspended object and the static electronic fence. The violation recognition engine identifies the behavior of not wearing a safety helmet and personnel intrusion based on the YOLOv8 model. The hierarchical alarm unit generates comprehensive early warning information by combining geometric, semantic, and behavioral hazard levels. The application output layer is used for information display and alarm output, including a 3D visualization interface, an audible and visual alarm device, and an event log system. The visualization interface presents digital twin scenes and risk information in real time. The audible and visual alarm device triggers differentiated warnings according to the warning level. The log system records all risk events and equipment status.

2. The three-dimensional dynamic safety control system for power transmission and transformation hoisting operations according to claim 1, characterized in that, The time synchronization unit aligns the timestamps of all sensor data using BeiDou time as the reference, as shown in the following formula: ； in, The standard timestamp output by BeiDou This is the sensor's original timestamp. This refers to the time deviation rate. It is a fixed delay.

3. The three-dimensional dynamic safety control system for power transmission and transformation hoisting operations according to claim 1, characterized in that, The spatial calibration unit transforms the extrinsic relationships between sensors to a unified vehicle coordinate system, using the following formula: ； in, To unify the coordinates of target points in the vehicle body coordinate system, The coordinates of the target point are in the local coordinate system of the sensor. Let be a rotation matrix. It is a translation vector; Rotation matrix Expressed using Euler angles, the formula is as follows: ； in, α For roll angle, β For pitch angle, γ This is the yaw angle.

4. The three-dimensional dynamic safety control system for power transmission and transformation hoisting operations according to claim 1, characterized in that, The static modeling unit also includes a semantic annotation subunit, which uses AI point cloud segmentation and human interaction to annotate entities in the 3D model, and defines the 3D geometry of the entities as static electronic fences, assigning security attribute parameters to each entity.

5. The three-dimensional dynamic safety control system for power transmission and transformation hoisting operations according to claim 1, characterized in that, The dynamic reconstruction unit uses the extended Kalman filter algorithm to achieve state estimation based on state prediction and to realize the motion range in the next 100-200 milliseconds through trajectory prediction. The state prediction formula is as follows: ； in, For a moment k Prior state estimation, Here is the state transition matrix. For a moment k- Posterior state estimation of 1, To control the input matrix, For a moment k The control input vector, This is process noise; State transition matrix Based on a uniformly accelerated motion model, the formula is as follows: ； in, I 3 is a 3×3 identity matrix, 03 is a 3×3 zero matrix, Δ t The filter period; The trajectory prediction formula is as follows: ； in, To predict the state, , , For a moment k Position state estimation, , , For a moment k Velocity state estimation, , , For at any time k Acceleration state estimation, The time step for prediction.

6. The three-dimensional dynamic safety control system for power transmission and transformation hoisting operations according to claim 1, characterized in that, The collision detection engine uses the GJK algorithm, which calculates the shortest distance between the predicted trajectory of the suspended object and the static electronic fence by solving the support vectors. The formula is as follows: ； in, As support vectors, convex hull A In direction Support points on the top convex hull B In direction Support points on the top It is the direction vector; The GJK algorithm updates the direction vector iteratively. ,when When the shortest distance is obtained, the formula is as follows: ； in, For the shortest distance, To be perpendicular to A vector pointing to the origin.

7. The three-dimensional dynamic safety control system for power transmission and transformation hoisting operations according to claim 1, characterized in that, The tiered alarm unit generates comprehensive warnings based on collision risk level quantification and multi-dimensional risk comprehensive index. The formula for quantifying collision risk level is as follows: ； in, d For the shortest distance, D 1 is the warning threshold. D 2 represents the danger threshold; The formula for the multi-dimensional risk composite index is as follows: ； in, As a comprehensive risk index, For collision risk factor, For semantic risk coefficient, For behavioral risk coefficient, These are the weighting coefficients.

8. The three-dimensional dynamic safety control system for power transmission and transformation hoisting operations according to claim 1, characterized in that, The tiered alarm unit has a built-in risk matrix, which integrates geometric risk, semantic risk, and behavioral risk to output a first-level warning and a second-level alarm. Geometric hazard is calculated based on the ratio of the shortest distance between the warning and hazard thresholds; Semantic risk is determined by weighting coefficients based on entity importance; Behavioral hazard level is graded based on the severity of the violation; Level 1 warnings include visual highlighting and audible alerts; Level 2 alarms include strong audible and visual warnings and collision point indications.

9. A three-dimensional dynamic safety control system for power transmission and transformation hoisting operations according to claim 1, characterized in that, The control system also includes an edge computing unit, which is integrated into the spatiotemporal data processing layer and the intelligent decision-making layer. It uses lightweight algorithm optimization technology to control the response time of data processing and risk reasoning within 100 milliseconds.

Images