Non-stop operation data intelligent acquisition and management method and system

Abstract

The application discloses a kind of uninterruptible operation data intelligent acquisition and management method and system, the method includes the following steps: through the multi-modal operation data of the smart safety helmet worn by operating personnel in the collection of operation site;The smart safety helmet passes through first communication link, and the multi-modal operation data is transmitted to handheld terminal in real time;The multi-modal operation data received by the handheld terminal is time-synchronized and encapsulated, and digital fingerprint based on block chain technology is added to form trusted data package, and is uploaded to cloud platform by second communication link in real time;The cloud platform and / or the handheld terminal are based on the trusted data package, and real-time fusion analysis and intelligent decision are carried out, and corresponding operation instruction or risk early warning is generated.The application can realize the active early warning of data acquisition and management, process controllable, so that early warning and prediction result are more accurate, more reliable, more credible.

Description

Technical Field

[0001] This invention relates to the field of power operation safety and intelligent management technology, specifically to a method and system for intelligent data acquisition and management of power outage operation scenarios in power distribution networks. Background Technology

[0002] Live-line work on power distribution networks is a key means of improving power supply reliability, but the process is always accompanied by extremely high risks such as high-voltage electric shock and falls from heights. Traditional safety management models rely heavily on the experience of workers, visual supervision by safety monitors, and paper-based operation tickets and records, which have the following prominent drawbacks: (1) There are blind spots and delays in process supervision. The perspective and energy of the supervisors are limited, making it difficult to detect minor violations in real time (such as insufficient safety distance at the millimeter level or incomplete insulation shielding). All process information is recorded after the fact, which cannot provide real-time decision support for ongoing operations, and the early warning is seriously delayed.

[0003] (2) Data such as video surveillance, environmental monitoring, personnel positioning, and work permits are scattered across different systems, forming "data silos." The data is fragmented and isolated, making it impossible to cross-verify and reconstruct the complete work chain. Paper or simple electronic records are susceptible to being added or tampered with after the fact, resulting in low data credibility and insufficient legal validity in accident tracing.

[0004] (3) Existing safety warnings are mostly based on fixed thresholds (such as fixed distance alarms), which cannot combine dynamic factors such as real-time wind speed causing conductor swaying and personnel's specific working postures to conduct accurate and predictive risk assessments. Summary of the Invention

[0005] To address the technical problems in existing live-line work safety management, such as blind spots and delays in process supervision, scattered and isolated data with low reliability, and the inability to accurately predict safety warnings by combining dynamic parameters, this invention provides an intelligent data acquisition and management method for live-line work. This method enables reliable perception of multimodal data throughout the entire work process, real-time fusion analysis, dynamic risk warning, and automatic auditing of procedures. It transforms the safety management model from passive response and post-event tracing to proactive warning and process control, making the warning and prediction results more accurate, reliable, and trustworthy.

[0006] On one hand, this invention discloses a method for intelligent data acquisition and management during live-line work, comprising the following steps: S1: Collect multimodal work data at the work site through the smart safety helmet worn by the workers. The multimodal work data includes at least the work view video stream, the work environment audio stream, the orientation and attitude data, and the environmental sensor data. S2: The smart safety helmet transmits the multimodal operation data to the handheld terminal in real time through the first communication link; S3: The handheld terminal performs time synchronization and encapsulation on the received multimodal operation data, and adds a digital fingerprint based on blockchain technology to form a trusted data packet, which is then uploaded to the cloud platform in real time through the second communication link; S4: The cloud platform and / or the handheld terminal perform real-time fusion analysis and intelligent decision-making based on the trusted data packet, and generate corresponding operation instructions or risk warnings.

[0007] The aforementioned audio stream of the work environment mainly refers to sound signals directly related to the work process, collected through the microphone array integrated into the smart safety helmet. These include, but are not limited to, voice commands and responses between workers, mechanical sounds generated by tool operation, and abnormal sounds generated by equipment discharge or arcing. Its core function is to capture auditory events (mechanical sounds, abnormal sounds) and semantic information (voice commands and responses) during the work process for compliance auditing (such as password verification) and abnormal state identification (such as discharge sound warnings).

[0008] In contrast, the environmental sensing data mainly refers to quantitative parameter data describing the objective physical state of the work site, collected by dedicated physical sensors (such as anemometers, temperature and humidity sensors, and gas sensors), such as wind speed, temperature, humidity, and concentration of harmful gases. Therefore, the two types of data, work environment audio stream and environmental sensing data, are fundamentally different in modality, content, and application purpose.

[0009] The orientation and attitude data refers to the real-time measurement and output of the inertial measurement unit (IMU) and / or global navigation satellite system (GNSS) module integrated into the smart helmet. This data reflects the position information (such as latitude, longitude, and elevation) and body attitude information (such as pitch angle, roll angle, and yaw angle of the head) of the actual worker (i.e., the smart helmet wearer) in three-dimensional space. This data is used to drive the corresponding virtual worker model in the digital twin model, achieving synchronization between the virtual and real worlds.

[0010] Furthermore, in step S3, the handheld terminal performs time synchronization and encapsulation of the multimodal operation data, specifically including: assigning unified time and space labels to the operation perspective video stream, operation environment audio stream, orientation and attitude data and environmental sensor data based on the timing signal, and logically binding data with spatiotemporal correlation.

[0011] Furthermore, in step S4, the real-time fusion analysis and intelligent decision-making includes dynamic security boundary early warning based on digital twins, specifically including: S41: Within the cloud platform or the handheld terminal, a digital twin model corresponding to the current work scenario is pre-built or generated in real time; S42: Combine the orientation and attitude data in the trusted data packet with the real-time analysis of the video stream from the work perspective to drive the virtual workers and their virtual tools in the digital twin model in real time. S43: In the digital twin model, multi-source data is integrated to simulate and calculate the estimated motion trajectory of the virtual worker and the virtual tools they hold in real time, as well as the minimum electrical distance between them and surrounding charged bodies; the multi-source data includes real-time pose and motion trend data of the virtual worker and tools, environmental sensing data, and scene model data; S44: Compare the estimated motion trajectory with the safety boundary defined by the dynamic minimum electrical distance. When it is determined that the estimated motion trajectory will intrude into the safety boundary within a preset time, generate a predictive warning instruction.

[0012] Furthermore, in step S4, the real-time fusion analysis and intelligent decision-making includes automatic auditing of work procedure compliance, specifically including: The standard operating procedure is digitally represented as a procedure template that includes multiple key action nodes and the logical relationships between nodes; Real-time analysis of the operational perspective video stream and operational environment audio stream in the trusted data packet to identify the currently executed action node and key passwords; The identified action nodes and key passwords are compared with the procedure template in real time to determine the sequence of steps, the completeness of actions, and the compliance of the confirmed passwords. When skipping steps, out-of-order sequences, or missing key passwords are detected, a procedure deviation warning is generated and recorded in the structured audit log.

[0013] In step S4, the real-time fusion analysis and intelligent decision-making are decomposed into multiple analysis tasks, and each analysis task is dynamically allocated to the cloud platform or the handheld terminal as its computing power execution node according to the preset allocation strategy. The allocation strategy is determined based on at least the real-time requirements and computational complexity requirements of each analysis task.

[0014] Furthermore, the first type of analysis tasks with real-time requirements exceeding the first threshold are assigned to the handheld terminal for execution. These tasks include at least dynamic security boundary early warning calculation and action node identification in automatic auditing of work procedure compliance. The second type of analysis tasks, whose computational complexity requirements exceed the second threshold, are assigned to the cloud platform for execution. These tasks include at least mining hidden risk patterns based on historical job data and training and updating AI models for the real-time fusion analysis based on the data mining results.

[0015] Furthermore, in step S4, the real-time fusion analysis and intelligent decision-making includes group collaborative risk perception, specifically including: When it is determined through the first communication link or the second communication link that there are multiple working terminals, the system constructs a group working situation diagram; If any worker is detected performing a predefined high-risk operation, the alert level for the activity status of other workers within a preset range centered on that worker will be raised, and a collaborative early warning will be generated.

[0016] The construction of the group operation situation map is based on: orientation and attitude data (especially GNSS / IMU fusion positioning data) with unified spatiotemporal tags uploaded by each operator's handheld terminal, as well as operator identification. The system uses this as the core data source to draw and update the spatial location, movement trajectory, orientation, and status information (such as whether they are in a high-risk operation state) of all operators in real time on a two-dimensional or three-dimensional electronic map.

[0017] The main uses of this group operation situation map include: 1) Visual monitoring, providing remote monitoring personnel with a global overview of the work site. 2) Spatial relationship analysis, calculating the spatial distance between any workers in real time, and automatically identifying other related workers within a preset spatial range around any worker performing a high-risk operation based on predefined rules (such as the radius of the safety isolation zone at different work stages). 3) Collaborative early warning triggering, serving as the basis for generating collaborative early warnings when related personnel are identified entering a risk area, or when the group operation status meets a specific risk pattern (such as multiple people entering a small, high-risk area simultaneously).

[0018] Furthermore, the method also includes step S5: the cloud platform uses machine learning algorithms to mine implicit correlation patterns between near-error events and multi-dimensional features based on historical trusted data packets; when the matching degree between the feature data of the real-time operation scenario and a high-risk historical pattern exceeds a preset threshold, a predictive risk warning is generated.

[0019] On the other hand, the present invention also discloses an intelligent data acquisition and management system for uninterrupted power supply operations, used to implement the method described above, including: The smart safety helmet collects multimodal work data from the work site. The multimodal work data includes at least the work perspective video stream, the work environment audio stream, orientation and attitude data, and environmental sensor data. The handheld terminal receives multimodal work data transmitted by the smart safety helmet through the first communication link. It is equipped with an edge AI analysis module and a trusted evidence storage module, which are used to synchronize and encapsulate the received multimodal work data in time, and add a digital fingerprint based on blockchain technology to form a trusted data packet. The cloud platform is connected to the handheld terminal via a second communication link and contains a digital twin engine and a big data mining module. The cloud platform's digital twin engine and big data mining module and / or the handheld terminal's edge AI analysis module perform real-time fusion analysis and intelligent decision-making based on the trusted data packets, generating corresponding work instructions or risk warnings.

[0020] Furthermore, the smart helmet integrates a miniature camera, an array microphone, an inertial measurement unit, a global navigation satellite system receiving module, and an environmental sensing unit that includes at least wind speed and temperature and humidity sensors; the handheld terminal integrates a first wireless module for communicating with the smart helmet, a second wireless module for accessing a second communication link, and a processor for running a lightweight AI model.

[0021] Furthermore, the cloud platform includes: The data access and service layer is used to receive and parse data streams from multiple work sites; The capability platform layer integrates the aforementioned digital twin engine, real-time AI analysis engine, and rule engine; The data asset layer is used to store raw data, analysis results, and blockchain-based digital archives of operations. The application service layer provides interfaces for risk warning, remote collaboration, and digital archive traceability services.

[0022] Furthermore, the system is also applicable to low-voltage DC live-line work scenarios; the smart safety helmet or handheld terminal is equipped with a dedicated sensor for detecting specific spectral or magnetic field characteristics of DC arcs, and the analysis model of the cloud platform or handheld terminal is adapted accordingly to DC polarity judgment and arc risk identification.

[0023] Compared with the prior art, the present invention has at least the following beneficial effects: (1) Through the two core functions of dynamic safety boundary early warning and automatic procedure audit, safety management is transformed from passive monitoring that relies on people to proactive control driven by the system, realizing millisecond-level real-time intervention in operational risks and compliance.

[0024] (2) From the synchronous collection and intelligent processing of multimodal data at the collection end to the blockchain evidence storage during the transmission process, a complete and reliable data value chain has been constructed, which ultimately forms an unalterable digital archive of operations. This completely solves the industry problems of scattered, isolated and low credibility of existing operation data, and provides a reliable data foundation and reliable evidence for accident tracing, insurance claims, training and assessment.

[0025] (3) The collaborative architecture and dynamic task scheduling mechanism of “cloud (cloud platform)-edge (edge ​​AI analysis module)-end (handheld terminal)” ensures that the core dynamic security early warning function can still be maintained by relying on edge computing power (on the handheld terminal) in extreme situations such as network interruption, ensuring the system robustness under extreme conditions and greatly enhancing the system reliability.

[0026] (4) It not only solves real-time risks, but also improves overall operational safety through group collaborative perception, and realizes the prediction of hidden risks through data mining, thereby enhancing the depth and breadth of intelligent decision-making and demonstrating the system's ability to continuously evolve.

[0027] (5) The modular design of the system of the present invention enables the system to quickly adapt to various new scenarios such as low-voltage DC operation and cable operation by adding dedicated sensors and updating AI models, and has higher scalability and adaptability, and a longer technology life cycle. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a flowchart illustrating the intelligent data acquisition and management method for uninterrupted power supply operations according to the present invention.

[0030] Figure 2 This is a system architecture diagram of the intelligent data acquisition and management system for uninterrupted power supply operations according to the present invention. Detailed Implementation

[0031] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0032] It should be understood that the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention. Example 1

[0033] Taking the replacement of intermediate pole insulators on a 10kV distribution line as an example, the implementation process of the method of this invention is illustrated. (Reference) Figure 1 The present invention provides an intelligent data acquisition and management method for uninterrupted power supply operations, the method comprising the following steps: S1: The smart safety helmet worn by the worker collects multimodal work data on site. The multimodal work data includes at least the video stream from the work perspective, the audio stream of the work environment, the orientation and attitude data, and the environmental sensor data. S2: The smart safety helmet transmits multimodal operation data to the handheld terminal in real time through the first communication link; S3: The handheld terminal performs time synchronization and encapsulation on the received multimodal operation data, adds a digital fingerprint based on blockchain technology to form a trusted data packet, and then uploads it to the cloud platform in real time through the second communication link; S4: The cloud platform and / or handheld terminal perform real-time fusion analysis and intelligent decision-making based on trusted data packets, and generate corresponding work instructions or risk warnings.

[0034] In step S1, the multimodal operation data acquisition requires workers to wear smart safety helmets integrated with multiple types of sensors. These integrated sensors include at least a miniature camera, a microphone, and an environmental sensor array. The miniature camera and microphone are used to collect real-time video streams from the work perspective and audio streams from the work environment, respectively. The environmental sensor array is used to collect real-time orientation and attitude data as well as environmental sensing data, including at least temperature, humidity, wind speed, and harmful gas concentrations. Furthermore, when integrating the miniature camera, microphone, and environmental sensor array into the smart safety helmet, it is essential to ensure that the sensors can operate stably in complex electromagnetic environments (such as near high-voltage electric fields), and that the electromagnetic compatibility level of the sensors meets electrical work standards (such as the GB / T 17626 series).

[0035] For example, in a scenario where worker A is replacing an insulator on a 10kV power distribution line, worker A wears an integrated smart safety helmet and carries an explosion-proof handheld terminal to the site. The miniature camera on the smart safety helmet continuously records video from worker A's working perspective, the microphone collects on-site sounds (including conversations with colleagues, tool operation sounds, and possible discharge sounds), and the environmental sensor group monitors wind speed in real time (ensuring it is below the operating threshold of the boom truck), temperature, and humidity.

[0036] In step S2, a high-bandwidth, low-latency, and highly interference-resistant first wireless link needs to be established between the smart helmet and the handheld terminal. This first wireless link uses a local wireless link (such as a self-organizing network Wi-Fi). For example, multimodal operational data such as collected video streams, audio streams, and environmental sensor data are stably transmitted to the handheld terminal via the self-organizing network Wi-Fi between the smart helmet and the handheld terminal.

[0037] In an embodiment of the present invention, the packaged multimodal data is transmitted in real time and stably to the handheld terminal held by the person in charge of the work, C, via a first (local) communication link based on the Wi-Fi 6 protocol, in a low-latency and high-bandwidth manner. This handheld terminal can serve as an edge gateway and data aggregation point on site.

[0038] In step S3, the handheld terminal processes the received multimodal operational data. After processing, these data streams are uploaded to the cloud platform in real time via a second wireless link (wide area network, such as the on-site 5G network). Simultaneously, the app running on the handheld terminal displays current environmental parameters and its own device status. This handheld terminal processes the received multimodal operational data, transforming the raw, heterogeneous sensor data into standardized, reliable, spatiotemporally aligned data packets that can be used for advanced intelligent analysis.

[0039] Preferably, in step S3, the handheld terminal processes the received multimodal job data, specifically including: first, performing time synchronization and encapsulation (which includes spatiotemporal alignment and logical binding), and then adding a digital fingerprint based on blockchain technology to form a trusted data packet. This processing includes the following three key sub-steps: (3.1) Spatiotemporal Alignment. The goal is to establish a unified spatiotemporal reference system for data from all sources, enabling subsequent fusion. To achieve high-precision timestamp synchronization of data from multiple sources, the handheld terminal acts as the master clock, obtaining high-precision standard time reference UTC (Coordinated Universal Time, a standard time reference with microsecond-level accuracy) from the Global Navigation Satellite System (GNSS) module. If the GNSS signal is lost, the terminal relies on its built-in high-stability crystal oscillator for timekeeping and resynchronizes after signal recovery. All received video frames, audio packets, and sensor data packets are marked with this unified timestamp. This ensures that regardless of any delays in data transmission, it is known that, for example, an action in a video, an attitude recorded by the IMU (Inertial Measurement Unit), and a command recorded by the microphone occur at the same absolute moment. When unifying the spatial coordinates of data from multiple sources, various types of data are mapped to the same three-dimensional coordinate system. This can be achieved through technologies such as SLAM (Simultaneous Localization and Mapping, independent of external signals like GPS) or GNSS+IMU (Global Navigation Satellite System + Inertial Measurement Unit) fusion positioning. This determines the precise position and attitude of the handheld terminal (i.e., the data aggregation point) within a global coordinate system (such as the WGS-84 geodetic coordinate system) or a pre-established local coordinate system at the work site. Based on a pre-calibrated sensor extrinsic parameter matrix (i.e., the position and angular relationship between the camera and IMU relative to the terminal center), the pixel information of the video from the work perspective and the head orientation measured by the IMU are uniformly transformed into the global coordinate system of the handheld terminal. This provides clear three-dimensional spatial location information for objects captured by the camera (video from the work perspective) and the orientation of the worker's head (position and attitude), enabling the computer to understand the spatial relationships between objects.

[0040] (3.2) Logical binding, i.e., building associations. This sub-step, based on spatiotemporal alignment, associates data of different types but describing the same event or state according to business logic. This includes time-based frame-level binding and event-based feature-level binding. Time-based frame-level binding means that the system uses time as an index to package all data within the same time window (e.g., ±10 milliseconds) into a single data frame. For example, the video keyframe at T=14:30:05.123, the corresponding 100-millisecond audio clip, the IMU attitude quaternion at that moment, and the current wind speed value are bound into a single logical unit. This provides subsequent models with simultaneous, multi-angle information input, enabling cross-validation and fusion judgment. Event-based feature-level binding refers to identifying key event points in the operation through preset rules or lightweight real-time analysis (such as a simple motion detection model running on the terminal). For example, when the system detects the action characteristic of "insulated rod contacting the conductor," it strongly correlates all relevant data within a certain period before and after this moment (video, audio such as the "verify no power" command, force sensor data on the tool, ambient temperature and humidity) and tags it with "event label: power verification." This structures the massive data stream, forming event-centric, semantically rich data fragments, greatly improving the efficiency of subsequent retrieval and analysis.

[0041] (3.3) Generate a digital fingerprint and store it on the blockchain. This sub-step utilizes blockchain technology to provide an immutable digital fingerprint (also known as an identity fingerprint) for key data to form a trusted data packet, ensuring its legal validity for subsequent tracing. For the key data packets bound in step two (especially those marked with key events), the handheld terminal uses a cryptographic hash algorithm (such as SHA-256) to calculate its unique data digest (i.e., hash value). Any minor modification to the original data will result in a significant change in the hash value, thus generating a unique digital fingerprint for the data packet. The handheld terminal instantly sends a small amount of metadata (rather than large volumes of data such as the original video) such as the hash value, timestamp, and event tag of the key data packet to a pre-set blockchain node (usually an enterprise or consortium blockchain) via a mobile network (such as a 5G network). The blockchain network packages it into a new block, forming a permanent and publicly verifiable transaction record. Thus, by utilizing the decentralized, immutable, and timestamp service characteristics of blockchain, a legally valid third-party time authentication is provided for the critical operational moments in on-site operations. If a dispute arises later regarding the authenticity of the data, it can be verified simply by recalculating the data hash and comparing it with the on-chain record.

[0042] For example, after receiving data, the handheld terminal held by the person in charge, C, performs the following core operations: The handheld terminal uses its built-in high-precision GNSS module to provide high-precision timing signals, assigning a unified millisecond-level timestamp and spatial coordinate label to each video frame, each audio segment, each IMU group, and each environmental data packet. Subsequently, data with strong logical correlation generated at the same time (such as a video frame at a certain moment, the corresponding head posture, and environmental wind speed) are logically bound together to form a spatiotemporally aligned multimodal data frame. For key operational nodes (such as "starting power testing" or "installing shielding"), the handheld terminal calls its built-in trusted evidence storage module to calculate the hash value of the multimodal data frames bound within a certain period before and after that moment, generating a digital fingerprint. This digital fingerprint is instantly uploaded to the blockchain node of the cloud platform via the 5G network for evidence storage, ensuring the immutability of key process data. Subsequently, the complete spatiotemporally aligned data is continuously uploaded to the cloud platform via the 5G network.

[0043] The highly structured data packets with built-in trust anchors obtained through the above data processing steps lay a solid and reliable data foundation for subsequent high-value and credible integrated analysis (such as dynamic security assessment and procedural auditing) on ​​the cloud platform. This processing flow is also an important technical step in the method of this invention, moving from simple data collection to intelligent and trustworthy analysis.

[0044] In step S4, the cloud platform and the handheld terminal perform real-time fusion analysis and intelligent decision-making based on real-time network conditions and computing power requirements.

[0045] The real-time fusion analysis and intelligent decision-making, as well as the generation of work instructions or risk warnings, have execution nodes that support multiple configuration modes to adapt to different on-site network conditions, computing power requirements, and security level requirements. Specifically, this includes at least one of the following: Centralized cloud processing mode: suitable for scenarios with stable network conditions and reliable connection to the cloud platform; in this mode, the cloud platform executes all analysis model calculations and decision-making logic, and generates the final operation instructions or risk warnings, which are then sent to the handheld terminal via the second communication link; Edge Autonomy Mode: Suitable for scenarios with poor or interrupted network signals, or extreme real-time requirements (such as millisecond-level security alerts); In this mode, the handheld terminal uses local computing power to independently execute all necessary real-time analysis and decision-making, and immediately generate and trigger operation instructions or risk alerts. Cloud-edge collaborative processing mode: As a preferred implementation of the present invention, it is applicable to normal operation scenarios that need to balance real-time response and complex calculation. In this mode, the complete analysis and decision-making task can be intelligently decomposed into multiple analysis tasks according to the preset dynamic scheduling strategy, and dynamically allocated and collaboratively executed between the cloud platform and the handheld terminal to jointly complete the generation of analysis decisions and instruction warnings.

[0046] In one optional embodiment, the fusion analysis and intelligent decision-making is based on dynamic security boundary early warning using digital twins. Preferably, in step S4, the real-time fusion analysis and intelligent decision-making includes dynamic security boundary early warning based on digital twins, specifically comprising the following sub-steps: S41: Pre-build or generate in real time a digital twin model corresponding to the current work scenario within a cloud platform or handheld terminal; S42: Combine the orientation and posture data (including personnel head posture) in the trusted data packet with real-time analysis of the video stream from the work perspective (such as identifying the tool type and approximate holding position through a lightweight model) to jointly drive the virtual worker and the virtual tool model held in the digital twin model, so that the twin keeps the movement synchronized with the real work scene. S43: In digital twin models, dynamic secure computation is performed by fusing multi-source data. The core input multi-source data includes: Driving data: Real-time pose and motion trends of virtual workers and tools (derived from orientation and posture data and video analysis); Environmental dynamic data: real-time wind speed and direction data (used to simulate the swaying of conductors and tools due to wind). Environmental static / auxiliary data: real-time temperature and humidity data (which can be used to correct electrical parameter models such as air insulation strength); Scene model data: Preset locations of live parts, voltage levels, and minimum electrical distance reference values ​​required by safety regulations.

[0047] Based on the aforementioned fused data, the digital twin engine uses built-in physical simulation models (such as rigid body dynamics models that consider wind loads and electrical clearance calculation models based on electric field distribution) to simulate and calculate in real time the predicted movement trajectory of virtual workers and their tools in the future (such as the next 0.5-2 seconds), as well as the dynamically changing minimum electrical distance (i.e., dynamic safety boundary) between them and surrounding charged bodies, which is corrected based on real-time environmental conditions.

[0048] S44: The system continuously compares the estimated motion trajectory with the safety boundary defined by the dynamic minimum electrical distance. When an intrusion is predicted to occur within a preset warning time (e.g., 0.5 seconds in the future), a predictive warning command is generated. This method elevates the fixed safety distance threshold to a dynamic safety boundary linked to attitude and environment, enabling more precise (e.g., millimeter-level) warnings.

[0049] In an embodiment of the invention, for example, the cloud platform preloads a digital twin model of the current tower and line based on the work task ticket. The model receives orientation and attitude data from worker A and drives the virtual worker in the model in real time. Simultaneously, real-time wind speed data is integrated to simulate conductor swaying in the digital twin model. The system continuously calculates the dynamic minimum electrical distance between the virtual operating tool head and the simulated swaying live conductor (e.g., a current dynamic safety threshold of 0.82 meters is calculated at a wind speed of 4 m / s). When it is predicted that worker A's next action will cause the distance between the tool head and the conductor to be less than this threshold in 0.5 seconds, a warning command is generated and issued within 200 milliseconds. Upon receiving the command, worker A's smart safety helmet immediately triggers an audible, visual, and vibration alarm.

[0050] Example 2

[0051] In another optional embodiment, the fusion analysis and intelligent decision-making includes automatic auditing of work procedure compliance. Preferably, in step S4, the automatic auditing of work procedure compliance specifically includes: S4a: Digitally represent standard operating procedures as a procedure template containing multiple key action nodes and the logical relationships between nodes; S4b: Real-time analysis of video stream from the work perspective and audio stream from the work environment to identify currently executed action nodes and key commands; S4c: The identified action nodes and key passwords are compared with the procedure template in real time to determine the sequence of steps, the completeness of actions, and the compliance of the confirmation passwords. S4d: When skipping steps, out-of-order sequences, or missing critical passwords are detected, a procedure deviation warning is generated and recorded in the structured audit log.

[0052] In embodiments of this invention, the cloud platform's real-time AI analysis engine operates in parallel. This AI analysis engine has a pre-built digital procedure template for "replacing intermediate pole insulators," including key action nodes and sequential logic such as "voltage testing," "shielding," and "removing old insulators." The AI ​​analysis engine analyzes the video from the work perspective of worker A in real time, recognizing their hand movements (e.g., "picking up the voltage detector"); simultaneously, it analyzes the audio, recognizing key commands (e.g., "verify no power"). The system automatically compares the recognition results with the procedure template. If worker A attempts to remove the insulator without issuing the "shielding complete" command, the rules engine immediately determines it as a "skipping step violation," generates a warning, and sends it to worker C's (the person in charge of the work) handheld terminal. Simultaneously, this violation, along with the evidenced video footage, is recorded in the structured audit log. Therefore, this method is equivalent to introducing a tireless AI virtual monitor, achieving rigid constraints on the execution of standard operating procedures (SOPs).

[0053] Example 3 Optionally, this fusion analysis and intelligent decision-making includes group collaborative risk perception. Specifically, this group collaborative risk perception includes: constructing a group operational situation map when the system determines that multiple work terminals exist; if the system detects that the first worker is performing a high-risk operation, it automatically raises the alert level for the activities of other workers within a preset range centered on that worker, and generates a collaborative early warning.

[0054] For example, the system uses location information from handheld terminals to construct a group operation situation map including the locations of worker A (in-bucket electrician), worker B (ground electrician), and worker C (work supervisor). When the early warning module of the digital twin model determines that worker A is in a high-risk operation phase, the system automatically marks the ground area with a 3-meter radius below worker A as a "high-risk fall zone" and upgrades the electronic fence alert level for that area. If worker B's location information shows that he has entered this area ("high-risk fall zone"), the system will simultaneously issue a coordinated early warning of "ground personnel entering high-risk area" to the terminals of workers A, B, and C, thus achieving an upgrade from individual protection to group collaborative intelligent protection.

[0055] Example 4 Optionally, in step S4, the real-time fusion analysis and intelligent decision-making are decomposed into multiple analysis tasks, and each analysis task is dynamically allocated to a cloud platform or handheld terminal as its computing power execution node according to a preset allocation strategy. The allocation strategy is determined at least based on the real-time requirements and computing power complexity requirements of each analysis task.

[0056] Specifically, the allocation strategy is determined based on the real-time requirements of each analysis task, the computational complexity requirements, and the current network communication and node load status. The allocation strategy includes: (1) For analysis tasks with real-time requirements higher than the first threshold (such as dynamic security boundary calculation and key action recognition), regardless of their computational complexity, they are preferentially assigned to handheld terminals for execution. (2) For analysis tasks with computational complexity requirements higher than the second threshold (such as hidden risk mining and model training), if they also meet the real-time requirements higher than the first threshold, a simplified version is executed on the handheld terminal according to strategy (1); if not, it is assigned to the cloud platform for execution. (3) For analysis tasks with real-time requirements not exceeding the first threshold and computational complexity requirements not exceeding the second threshold, the task is dynamically allocated to a node with better resources on the cloud platform or handheld terminal based on the real-time network status and node load.

[0057] To achieve optimal performance and reliability in real-time fusion analysis and intelligent decision-making, this embodiment employs an intelligent computing power scheduling mechanism based on multi-dimensional constraints in step S4. The system decomposes the overall goal of real-time fusion analysis and intelligent decision-making into multiple independently schedulable analysis tasks, such as: real-time action recognition based on video streams, key password recognition based on audio streams, safe distance prediction based on digital twins, and offline risk pattern mining based on historical data.

[0058] Scheduling decisions are executed by a built-in dynamic scheduler (which can be deployed on handheld terminals or cloud platforms), based on a pre-defined, hierarchical scheduling strategy. The strategy's input can include two static attributes and two dynamic states: Static task attributes: Each analysis task is predefined with a real-time requirement level (bounded by a first threshold) and a computational complexity level (bounded by a second threshold).

[0059] Dynamic system status: including current network communication quality (such as latency and bandwidth) and computing node load (such as CPU / memory utilization of handheld terminals and cloud platforms).

[0060] The first threshold primarily focuses on the maximum permissible delay from data generation to the generation of an alert / instruction, typically ranging from 50 to 200 milliseconds. This first threshold is a dynamically adjusted threshold based on the task's criticality and the current operational risk. Its core requirement is that any analytical task that could directly lead to personal injury or serious equipment accidents must have an end-to-end delay less than human reaction time (approximately 250 ms) and a safety margin, usually within 100 ms. When the task type has extremely high real-time requirements, such as dynamic safety boundary warnings (e.g., predicting tool collisions with wires) and personnel fall / electric shock emergency stop judgments, the first threshold is required to be ≤100 ms. Furthermore, during high-risk operations such as "equipotential work" or "proximity to live conductors (<0.7 meters)," the system will automatically tighten the first threshold (e.g., set to 80 ms), incorporating more tasks into edge processing to ensure absolute safety. For task-driven tasks such as identifying key actions for compliance with regulations (e.g., "voltage detector contacting the wire"), key audio commands (e.g., "no power"), and simple violations (e.g., "not wearing insulated gloves"), the first threshold is set to a range of 100ms to 200ms. However, for low-risk tasks such as "ground potential work" or "preparation phase," the first threshold can be appropriately relaxed (e.g., set to 150ms) to allow more tasks to be performed on the cloud platform for more accurate analysis.

[0061] The second threshold is a dynamically adjusted threshold value based on the computational complexity of the task and the real-time status of the edge terminal. It represents the efficiency boundary of resources, ensuring optimal system load under limited resources. Its core requirement is that complex analysis tasks should not overload or cripple the already resource-constrained handheld terminal; the handheld terminal must have sufficient capacity to guarantee the absolute priority execution of the first category of tasks. For example, tasks requiring more than 80% of the handheld terminal's available computing power, such as pixel-level semantic segmentation of a 1080p image frame (e.g., identifying the precise contours of all wires, hardware, and human limbs), multimodal deep fusion inference (simultaneously understanding the correlation between video, audio, and text commands), and high-fidelity digital twin physical simulation; these tasks have many model parameters and complex computational graphs, and are forcibly classified as second-category tasks, executed by the cloud. Tasks requiring 20% ​​to 80% of the terminal's computing power, such as lightweight object detection (outlining tools and hands), audio event classification (identifying tool knocking sounds and discharge sounds), and simplified model calculations for dynamic gaps. These tasks can be dynamically determined to be executed on the handheld terminal or the cloud platform based on network conditions and the load on the handheld terminal.

[0062] Take, for example, the application scenario in underground cable well operations where public network signals are interrupted. In this scenario, the handheld terminal cannot maintain a stable connection with the cloud platform. The handheld terminal automatically detects the network interruption with the cloud platform and switches to edge autonomous mode. All real-time analysis tasks originally performed by the cloud platform (such as dynamic safety boundary calculation and procedure action recognition) are now handled by the edge AI analysis module on the handheld terminal's local machine. All raw data and analysis results are encrypted and cached within the handheld terminal. When the workers return to the surface and the network is restored, the handheld terminal automatically resumes the transmission of cached trusted data packets and local analysis logs to the cloud platform, completing data synchronization and archiving, and ensuring the integrity of the work chain. This method ensures that the most critical personal safety early warning function remains uninterrupted under harsh network conditions, achieving high system robustness. Once the network is restored, the cloud platform synchronizes the data again for complete in-depth analysis.

[0063] Example 5

[0064] Preferably, the method of the present invention further includes step S5: The cloud platform, based on historical reliable data packets, uses machine learning algorithms to mine implicit correlation patterns between near-error events (violations that do not cause consequences) and multi-dimensional features (personnel, environment, task type). When the matching degree between the feature data of the real-time operation scenario and a certain high-risk historical pattern exceeds a preset threshold, a predictive risk warning is generated. For example, the big data mining module of the cloud platform analyzes the digital archives of operations over the past year and discovers an implicit pattern: when the "complex multi-wire shielding" task is performed by "operators with less than 3 years of experience" during the "high-temperature afternoon in summer", the probability of "near-error events" (such as system warnings for inadequate shielding) is 3 times the average. When a new round of operations begins, if the system detects that the current scenario of operator A matches this high-risk pattern, even if the operation has just started and no real-time alarm has been triggered, the cloud platform will push a predictive risk warning to the handheld terminal of the work supervisor C: "The current scenario matches the historical high-risk pattern XXX. It is recommended to strengthen supervision or have experienced personnel take the lead." This achieves true prevention before problems arise. Therefore, this method achieves a leap from post-event analysis to pre-event predictive prevention.

[0065] Example 6

[0066] Taking the live-line replacement of a pole-mounted intelligent switch (vacuum circuit breaker) on a 10kV distribution line as an example, the roles of the operators are: Electrician A inside the hopper (responsible for performing the main operations inside the insulated hopper), Ground Electrician B (responsible for passing and coordinating ground tools), and Work Supervisor C (responsible for overall on-site monitoring and command). In this embodiment, the present invention provides an intelligent data acquisition and management system for live-line work, used to implement the aforementioned method for intelligent data acquisition and management of live-line work. Figure 2 As shown, the system of the present invention mainly includes: a smart safety helmet, a handheld terminal, and a cloud platform.

[0067] The smart safety helmet is used to collect multimodal work data. It integrates a miniature camera, array microphone, inertial measurement unit (IMU), Global Navigation Satellite System (GNSS) receiver module, and an environmental sensing unit that includes at least wind speed and temperature / humidity sensors.

[0068] The handheld terminal communicates with the smart safety helmet. It integrates a first wireless module (such as Wi-Fi 6) for communication with the helmet, a second wireless module (such as 5G) for accessing a second communication link, and a dedicated processor (such as an NPU) for running lightweight AI models. Additionally, the handheld terminal includes an edge AI analysis module and a trusted evidence storage module. The handheld terminal is used to receive, process, and upload multimodal operational data.

[0069] The cloud platform communicates with handheld terminals for data storage, fusion analysis, and centralized decision-making. It includes a digital twin engine and a big data mining module.

[0070] For example, in an embodiment of the present invention, the system hardware deployment includes: both electrician A inside the bucket and electrician B on the ground wear smart safety helmets (e.g., model IH-800). The work supervisor C holds an industrial-grade explosion-proof handheld terminal (e.g., model HT-10). The backend is deployed on the "Smart Electricity Safety Supervision" cloud platform of a provincial State Grid company's cloud data center. A micro-weather station (wind speed, temperature, and humidity) and a vehicle-mounted gateway are installed on the outside of the bucket of the on-site insulated bucket truck, and connected to the work supervisor C's handheld terminal via Bluetooth. During multimodal data acquisition, A's smart safety helmet has a built-in 120° wide-angle 8-megapixel camera with electronic image stabilization (EIS) to capture A's first-person perspective video at 25fps, covering his hands, tools, and the overhead wires and switching equipment in front. A dual-microphone array is used; the main microphone picks up A's voice commands and reports, while the noise-canceling microphone collects ambient sounds (tool operation sounds, possible discharge sounds). The system integrates a nine-axis IMU (three-axis accelerometer + three-axis gyroscope + three-axis magnetometer) to output helmet attitude (pitch, roll, yaw) data at a frequency of 100Hz. It also features a built-in high-precision GNSS / BeiDou dual-mode positioning module, outputting the worker's latitude, longitude, and elevation. A micro-weather station collects real-time data on wind speed (4.5m / s), wind direction, temperature (32℃), relative humidity (65%), and atmospheric pressure at the work site. All sensor data is transmitted via a LoRa wireless module to the handheld terminal (model HT-10) of the person in charge, C.

[0071] The handheld terminal (HT-10) of device C acts as an edge gateway, receiving real-time multimodal data streams from two smart helmets (A and B) via built-in dual-band Wi-Fi (2.4GHz / 5GHz); it also receives data from a micro-weather station via Bluetooth. The handheld terminal (HT-10) uses its GNSS module to obtain the UTC time published by the National Time Service Center as its master clock, achieving millisecond-level accuracy. Utilizing GNSS and the built-in inertial navigation system (INS) fusion positioning, it determines the precise location of device C (i.e., the data convergence point) in the WGS-84 coordinate system (e.g., longitude: E 118.123456°, latitude: N 31.987654°, elevation: 45.2m).

[0072] The handheld terminal (HT-10) marks each received video frame, audio packet, and sensor data with a unified "timestamp-spatial stamp." Using time as an index, the system packages video frames, audio segments, IMU attitude data (A / B), and ambient wind speed data from the same moment (time window ±20ms) into a multimodal data frame. When the system (through a lightweight edge AI model) initially identifies a key operational event (such as "starting power testing" or "connecting the lead wire"), the handheld terminal (HT-10) immediately triggers the evidence preservation process. The security chip (SE) within the handheld terminal performs a SHA-256 hash operation on the key multimodal data frame (compressed) containing 10 seconds before and after the event, generating a unique digital fingerprint. The handheld terminal (HT-10) sends the fingerprint, timestamp, event type, and operation ID, among other metadata, to the evidence preservation smart contract deployed on the "State Grid Chain" (a blockchain platform of the State Grid Corporation of China) via a 5G network. This contract returns an evidence preservation receipt containing information such as block height and transaction hash. After completing the above processing, the handheld terminal (HT-10) continuously streams the complete trusted data packets with spatiotemporal tags and evidence references to the "Smart Electricity Safety Supervision" cloud platform through 5G SA network slicing (ensuring high priority and low latency).

[0073] Optionally, the cloud platform adopts a microservice architecture, specifically including: The data access and service layer is used to receive and parse data streams from multiple work sites at high concurrency. The capability platform layer integrates a digital twin engine, a real-time AI analysis engine, and a configurable rule engine; The data asset layer adopts a cold and hot data tiered storage strategy and uses blockchain technology to store key operational data, forming an immutable digital archive of operations. The application service layer provides standardized interfaces for business applications such as risk warning, remote two-way audio and video collaboration, and digital archive traceability.

[0074] Example 7 In a preferred embodiment, the system of the present invention further includes a collaborative intelligent analysis module. This module can intelligently distribute analysis tasks between the handheld terminal (HT-10) and the cloud platform based on the operation stage and network conditions.

[0075] The edge AI analysis module within the handheld terminal (HT-10) performs the first type of analysis task, which demands the highest real-time performance. Using its built-in action detection model (lightweight convolutional neural network CNN), the edge AI analysis module analyzes A's work-view video in real time, quickly detecting seven basic actions such as "raising hand," "holding a tool," and "tool contacting the conductor." Based on a dynamic gap estimation model (lightweight computational model) using physical kinematics, and taking into account A's head posture (IMU), the known length of the insulating rod (preset), and real-time wind speed, it quickly estimates the minimum dynamic distance between the tool head and the live conductor. For example, when A uses the insulating rod to clamp the guide wire in preparation for connection, the edge AI analysis module's action detection model identifies the tool contacting the conductor. Simultaneously, the dynamic gap estimation model calculates the current dynamic safe distance threshold as 0.85 meters, while the real-time estimated distance is 0.92 meters. The edge AI analysis module determines this to be safe but remains in a state of alert; all analysis results are cached locally and uploaded with the data stream.

[0076] The cloud platform's real-time AI analytics engine performs the second type of analytical task, which requires powerful computing capabilities, complex models, or global data analysis. The real-time AI analytics engine further includes: 1) High-precision visual recognition sub-engine (core is a large Mask R-CNN model) performs pixel-level segmentation on uploaded videos and accurately identifies dozens of targets and their precise outlines, such as "A-phase conductor", "B-phase conductor", "switch terminal", "insulation shield", "electrician's right index finger" and so on.

[0077] 2) The multimodal procedure audit sub-engine incorporates a digital procedure template for "Live Replacement of Pole-Mounted Switches." This template is a directed graph that defines the sequence of steps: "Verification → Covering → Removal of Old Switch → Installation of New Switch → Restoration," along with compliance conditions for each step (e.g., "Verification must be performed before covering, and the result must be no power"). This sub-engine simultaneously receives high-precision visual recognition results and audio text translated by the cloud platform's voice recognition service, performing real-time comparisons. For example, in the "Covering" step, the sub-engine checks if the visual recognition results show the state "Insulation cover covers the upper pole head," and checks if the audio text contains the standard command "A-phase covering completed." If the action is recognized but the command is missing, the built-in rules engine immediately triggers a "Procedure Deviation Warning."

[0078] 3) The digital twin and dynamic safety assessment sub-engine loads a high-precision 3D point cloud digital twin model of the operation route and equipment. It receives tool and limb coordinates, as well as wind speed data, from high-precision visual recognition by the cloud platform. High-fidelity physical simulation is performed within the twin to calculate the dynamic electrical clearance, accurate to the millimeter level, between the tool and each phase conductor under the most unfavorable wind deflection. A predictive warning is generated when it is predicted that the clearance may fall below the dynamic threshold within the next second.

[0079] 4) Group Collaborative Sensing Module. Receives location and status information from all terminals of A, B, and C. Establishes a group situational awareness map. When the cloud platform's real-time AI analysis engine determines that A is performing the high-risk operation of "attaching the drain line," the group collaborative sensing module automatically marks a 5-meter radius area directly below the insulated bucket as a "falling object risk zone." It also monitors B's location in real time; once B enters this risk zone, a "collaborative warning" is immediately generated and simultaneously sent to the terminals of A, B, and C.

[0080] This collaborative intelligent analysis module implements the following three working modes: When the network is good, the edge AI analysis module performs quick detection and preliminary estimation, while the cloud platform performs high-precision and complex model analysis. Early warning instructions issued by the cloud platform (such as procedural deviation warnings) and real-time warnings from the handheld terminal (such as dynamic gap alarms) are displayed together on the handheld terminal (HT-10) interface.

[0081] When network latency increases, for example, if the handheld terminal (HT-10) detects a communication latency exceeding 200ms with the cloud platform, the handheld terminal automatically reclaims all judgment authority for core security analyses such as dynamic gap estimation locally, only uploading the results and summary logs. At this time, the system operates in a degraded manner, but the core security early warning function (Category 1 task) remains intact.

[0082] When the network is interrupted, the handheld terminal (HT-10) enters "edge autonomous mode". All Category I analysis tasks are fully localized, and warning information is only displayed locally and stored in encrypted form. All data (including evidence storage requests) is cached in a local queue and automatically resumed and synchronized after the network is restored.

[0083] When the system (whether cloud platform or handheld terminal) generates a high-risk warning (e.g., gap less than 0.8 meters), A's smart safety helmet will trigger a triple visual warning: a voice broadcast in the helmet's earpiece, flashing red LED lights on the side, and activation of a micro-vibration motor. Meanwhile, in the background, all operational data, analysis results, warning logs, and blockchain-based evidence are linked together in the cloud platform's data asset layer to form a complete digital archive of the operation. This archive supports one-click backtracking, allowing video playback along a timeline and linking the viewing of sensor data, AI analysis results, and warning information at any given moment. This archive serves as an attachment to the electronic work order for this operation, is automatically archived, and cannot be tampered with. After the operation is completed, the big data mining module of the "Smart Electricity Safety Supervision" cloud platform will perform the following operations: i) Analyze all warning events and operation sequences during the operation. ii) Compare the actual operational path (e.g., the actual movement trajectory of the insulated bucket) with the ideal safety path planned by the cloud platform to optimize the path planning algorithm. iii) Use the new data generated during the operation (especially complex scenarios that are difficult for the edge AI analysis module to determine) to incrementally train the cloud platform's real-time AI analysis engine. A portion of the trained and optimized lightweight model (such as the improved "action fast detection model") is distributed to each handheld terminal (HT-10) through the cloud platform to complete the iterative upgrade of the edge AI analysis module, making the system as a whole more intelligent with use.

[0084] As an extension, the system of the present invention is also particularly suitable for new scenarios such as low-voltage DC live-line work. To this end, the smart safety helmet or handheld terminal is equipped with a dedicated sensor for detecting specific spectral or magnetic field characteristics of DC arcs, and the analysis model of the cloud platform or handheld terminal needs to be adapted accordingly for DC polarity determination and arc risk identification.

[0085] In addition, for easier distinction, the following explanations are provided regarding AI-related names and relationships: The lightweight AI model is the core algorithm deployed in the edge AI analysis module; while the edge AI analysis module (located on the handheld terminal) and the real-time AI analysis engine (located on the cloud platform) are two parallel and collaborative AI computing power carriers in the system. They each run AI models suitable for their own positioning, together forming a complete analysis capability.

[0086] It should be noted that in this paper, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply these relationships. There is no such actual relationship or order between entities or operations. Furthermore, the terms "including" and "package" do not apply. The word "comprise" or any other variation thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0087] In this embodiment of the invention, the term "and / or" describes the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The character " / " generally indicates that the preceding and following associated objects have an "or" relationship.

[0088] The various embodiments in this specification are described in a related manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0089] In particular, the device embodiments are basically similar to the method embodiments, so they are described in a simpler way. For relevant details, please refer to the description of the method embodiments.

[0090] For ease of description, the above apparatus is described by dividing it into various functional units / modules. Of course, in implementing this invention, the functions of each unit / module can be implemented in one or more software and / or hardware.

[0091] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0092] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for intelligent data acquisition and management of live-line working operations, characterized in that, Includes the following steps: S1: The smart safety helmet worn by the worker collects multimodal work data at the work site. The multimodal work data includes at least the work perspective video stream, the work environment audio stream, the orientation and attitude data, and the environmental sensor data. S2: The smart safety helmet transmits the multimodal operation data to the handheld terminal in real time through the first communication link; S3: The handheld terminal performs time synchronization and encapsulation on the received multimodal operation data, and adds a digital fingerprint based on blockchain technology to form a trusted data packet, which is then uploaded to the cloud platform in real time through the second communication link; S4: The cloud platform and / or the handheld terminal perform real-time fusion analysis and intelligent decision-making based on the trusted data packet, and generate corresponding operation instructions or risk warnings.

2. The intelligent data acquisition and management method for live-line work as described in claim 1, characterized in that, In step S3, the handheld terminal performs time synchronization and encapsulation of the multimodal operation data, specifically including: assigning unified time and space labels to the operation view video stream, operation environment audio stream, orientation and attitude data and environmental sensor data based on the timing signal, and logically binding data with spatiotemporal correlation.

3. The intelligent data acquisition and management method for live-line work as described in claim 1, characterized in that, In step S4, the real-time fusion analysis and intelligent decision-making includes dynamic security boundary early warning based on digital twins, specifically including: S41: Within the cloud platform or the handheld terminal, a digital twin model corresponding to the current work scenario is pre-built or generated in real time; S42: Combine the orientation and attitude data in the trusted data packet with the real-time analysis of the video stream from the work perspective to drive the virtual workers and their virtual tools in the digital twin model in real time. S43: In the digital twin model, multi-source data is integrated to simulate and calculate the estimated movement trajectory of the virtual worker and the virtual tools they hold in the future, as well as the minimum electrical distance between them and the surrounding charged bodies; the multi-source data includes real-time pose and movement trend data of the virtual worker and tools, environmental sensing data, and scene model data; S44: Compare the estimated motion trajectory with the safety boundary defined by the dynamic minimum electrical distance. When it is determined that the estimated motion trajectory will intrude into the safety boundary within a preset time, generate a predictive warning instruction.

4. The intelligent data acquisition and management method for live-line work as described in claim 1, characterized in that, In step S4, the real-time fusion analysis and intelligent decision-making includes automatic auditing of work procedure compliance, specifically including: The standard operating procedure is digitally represented as a procedure template that includes multiple key action nodes and the logical relationships between nodes; Real-time analysis of the operational perspective video stream and operational environment audio stream in the trusted data packet to identify the currently executed action node and key passwords; The identified action nodes and key passwords are compared with the procedure template in real time to determine the sequence of steps, the completeness of actions, and the compliance of the confirmed passwords. When skipping steps, out-of-order sequences, or missing key passwords are detected, a procedure deviation warning is generated and recorded in the structured audit log.

5. The intelligent data acquisition and management method for live-line work as described in claim 1, characterized in that, In step S4, the real-time fusion analysis and intelligent decision-making are decomposed into multiple analysis tasks, and each analysis task is dynamically allocated to the cloud platform or the handheld terminal as its computing power execution node according to the preset allocation strategy. The allocation strategy is determined based on at least the real-time requirements and computational complexity requirements of each analysis task.

6. The intelligent data acquisition and management method for live-line work as described in claim 5, characterized in that, The first type of analysis tasks with real-time requirements exceeding the first threshold are assigned to the handheld terminal for execution. These tasks include at least dynamic security boundary early warning calculation and action node identification in automatic audit of work procedure compliance. The second type of analysis tasks, whose computational complexity requirements exceed the second threshold, are assigned to the cloud platform for execution. These tasks include at least mining hidden risk patterns based on historical job data and training and updating AI models for the real-time fusion analysis based on the data mining results.

7. The intelligent data acquisition and management method for live-line work as described in claim 1, characterized in that, In step S4, the real-time fusion analysis and intelligent decision-making includes group collaborative risk perception, specifically including: When it is determined through the first communication link or the second communication link that there are multiple working terminals, a group working situation map is constructed; If any worker is detected performing a predefined high-risk operation, the alert level for the activity status of other workers within a preset range centered on that worker will be raised, and a collaborative early warning will be generated.

8. The intelligent data acquisition and management method for live-line work as described in claim 1, characterized in that, The method further includes step S5: the cloud platform uses machine learning algorithms to mine implicit correlation patterns between near-error events and multi-dimensional features based on historical trusted data packets; when the matching degree between the feature data of the real-time operation scenario and the high-risk historical pattern exceeds a preset threshold, a predictive risk warning is generated.

9. A live-line work data intelligent acquisition and management system, used to implement the live-line work data intelligent acquisition and management method according to any one of claims 1-8, characterized in that, include: The smart safety helmet collects multimodal work data from the work site. The multimodal work data includes at least the work perspective video stream, the work environment audio stream, orientation and attitude data, and environmental sensor data. The handheld terminal receives multimodal work data transmitted by the smart safety helmet through the first communication link. It is equipped with an edge AI analysis module and a trusted evidence storage module, which are used to synchronize and encapsulate the received multimodal work data in time, and add a digital fingerprint based on blockchain technology to form a trusted data packet. The cloud platform is connected to the handheld terminal via a second communication link and contains a digital twin engine and a big data mining module. The cloud platform's digital twin engine and big data mining module and / or the handheld terminal's edge AI analysis module perform real-time fusion analysis and intelligent decision-making based on the trusted data packets, generating corresponding work instructions or risk warnings.

10. The intelligent data acquisition and management system for uninterrupted power supply operations according to claim 9, characterized in that, The smart safety helmet integrates a miniature camera, an array microphone, an inertial measurement unit, a global navigation satellite system receiving module, and an environmental sensing unit that includes at least wind speed and temperature and humidity sensors; the handheld terminal integrates a first wireless module for communicating with the smart safety helmet, a second wireless module for accessing a second communication link, and a processor for running a lightweight AI model.

Images