Power operation supervision method and system based on ai behavior recognition and semantic analysis

Abstract

The application discloses a power operation supervision method and system based on AI behavior recognition and semantic analysis, belongs to the technical field of power system safety, and comprises a control framework and a collection device arranged in a power system, which continuously collect voiceprint data and video stream data; the identity of an operator is verified through voiceprint recognition, the semantic analysis of voice instructions is carried out based on a power standard term library, the consistency of instruction content and a task issued by the system is judged, the position of the operator is recognized through visual positioning and matched with the interval number of a target device in space, meanwhile, the standardization of operation behavior is detected through video analysis; voice interaction prompts are generated according to instruction ambiguity, and hierarchical alarms are triggered through multi-modal data fusion, and illegal operation data is recorded and stored synchronously. The application improves the supervision efficiency of the power system, ensures authorized operation and reduces human errors, guarantees safety through visual monitoring, triggers alarms through multi-modal fusion, responds to illegal operations in time and records evidence, and supports traceable safety management.

Description

Technical Field

[0001] This invention relates to the field of power system safety technology, specifically to a power operation supervision method and system based on AI behavior recognition and semantic analysis. Background Technology

[0002] The 6kV switch room in a thermal power plant is a core control area of ​​the power system, where operators frequently perform high-risk operations such as equipment start-up, shutdown, and switching. Traditional operation supervision relies on manual monitoring and a two-person confirmation system, which has the following problems: human fatigue can easily lead to misjudgments or omissions; unclear or misspoken voice commands are difficult to correct in real time; it is difficult to detect when operators go to the wrong interval (accidentally entering non-operation areas); there is a lack of full-process digital records, making accident backtracking difficult.

[0003] Currently, in this field, especially in power plants, the relevant technologies rely solely on manual monitoring to prevent incorrect intervals and misoperations. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problems solved by this invention are: how to solve the risks of misoperation, including the accidents easily caused by ambiguous voice commands (such as "disconnect segment A" and "close segment A"); how to solve the problem of going to the wrong interval, such as operators accidentally entering non-target equipment areas; and how to solve the problem of judging violations, as there is a lack of automated recording and judgment of the entire operation process (voice + behavior), and the inability to monitor and warn violations in real time.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a power operation supervision method based on AI behavior recognition and semantic analysis, comprising the following steps:

[0007] Deploy the control architecture and data acquisition equipment of the power system to continuously collect voiceprint data and video stream data; verify operator identity through voiceprint recognition and perform semantic analysis on voice commands based on the power standard terminology library to determine the consistency between command content and system-issued tasks; identify operator location through visual positioning and spatially match it with target equipment interval numbers, while simultaneously detecting the standardization of operational behavior through video analysis; generate voice interaction prompts based on command ambiguity, trigger hierarchical alarms through multimodal data fusion, and simultaneously record and store data on violations.

[0008] As a preferred embodiment of the power operation supervision method based on AI behavior recognition and semantic analysis described in this invention, the continuous collection of voiceprint data and video stream data includes an edge computing device deployed in the central control room, which integrates voiceprint recognition, semantic analysis and behavior recognition models.

[0009] A distributed high-precision microphone array and infrared camera group form a spatial acoustic-visual collaborative acquisition network.

[0010] As a preferred embodiment of the power operation supervision method based on AI behavior recognition and semantic analysis described in this invention, the voiceprint recognition includes matching operators with a pre-authorized task list through voiceprint features;

[0011] By combining the power standard terminology database, multi-dimensional semantic analysis is performed on the instructions, actions, equipment numbers, and operating parameters to identify logical conflicts between instructions and tasks.

[0012] As a preferred embodiment of the power operation supervision method based on AI behavior recognition and semantic analysis described in this invention, the semantic analysis includes dynamically updating the power standard terminology database and supporting the conversion between dialect terms and standard terms.

[0013] Generate interactive voice prompts for non-standard instructions to guide operators in repeating the standard instructions.

[0014] As a preferred embodiment of the power operation supervision method based on AI behavior recognition and semantic analysis described in this invention, the visual positioning includes establishing a spatial mapping relationship between personnel positions and equipment intervals through infrared visual positioning.

[0015] Real-time analysis of video stream data is used to verify the compliance of electrical testing procedures and the wearing of safety protective equipment.

[0016] As a preferred embodiment of the power operation supervision method based on AI behavior recognition and semantic analysis described in this invention, the hierarchical alarm includes a hierarchical response that triggers voice alarms, light warnings, and remote notifications based on multimodal fusion of voiceprint, semantic, spatial location, and behavioral data.

[0017] Automatically extract audio and video clips within a set time window before and after the violation to generate a traceable chain of evidence.

[0018] As a preferred embodiment of the power operation supervision method based on AI behavior recognition and semantic analysis described in this invention, the hierarchical alarm further includes establishing a task-driven closed-loop supervision logic, and associating the verification results of operator identity, instruction content, spatial location, and behavioral norms with the same task number.

[0019] A behavioral profile of the personnel involved in repeated violations is generated and fed back to the security management platform.

[0020] Another objective of this invention is to provide a power operation monitoring system based on AI behavior recognition and semantic analysis.

[0021] To address the aforementioned technical problems, this invention provides the following technical solution: a power operation monitoring system based on AI behavior recognition and semantic analysis, comprising: a data acquisition module, a voiceprint recognition module, a semantic analysis module, a behavior recognition module, a decision-making and alarm module, and a data storage module;

[0022] The data acquisition module is equipped with the control architecture and acquisition equipment of the power system, and continuously acquires voiceprint data and video stream data.

[0023] The voiceprint recognition module verifies the operator's identity through voiceprint recognition;

[0024] The semantic analysis module performs semantic analysis on voice commands based on the power standard terminology library to determine the consistency between the command content and the task issued by the system.

[0025] The behavior recognition module identifies the operator's position through visual positioning and spatially matches it with the target equipment interval number. At the same time, it detects the standardization of the operation behavior through video analysis.

[0026] The decision-making and alarm module generates voice interaction prompts based on the ambiguity of the instructions and triggers hierarchical alarms through multimodal data fusion.

[0027] The data storage module synchronously records and stores data on violations.

[0028] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the power operation supervision method based on AI behavior recognition and semantic analysis.

[0029] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the power operation supervision method based on AI behavior recognition and semantic analysis.

[0030] The beneficial effects of this invention are as follows: By deploying the control architecture and acquisition equipment of the power system, continuous acquisition of voiceprint and video stream data is achieved, establishing an efficient monitoring network and improving the efficiency and accuracy of power operation supervision. Voiceprint recognition ensures that only authorized personnel perform operations, and semantic analysis determines the consistency between instructions and tasks, reducing the risk of human error. Visual positioning and behavioral norm detection monitor the operator's position and actions in real time, ensuring safety and compliance and reducing accidents caused by insufficient personal safety protection. The hierarchical alarm mechanism triggered by multimodal data fusion can respond promptly to violations and automatically record audio and video data, providing a traceable chain of evidence for safety management, thereby achieving safe and efficient management in the power industry. Attached Figure Description

[0031] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 The above is a flowchart of an overall power operation supervision method based on AI behavior recognition and semantic analysis, provided as an embodiment of the present invention. Detailed Implementation

[0033] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0034] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a power operation supervision method based on AI behavior recognition and semantic analysis, including:

[0035] S1. Deploy the control architecture and acquisition equipment of the power system to continuously collect voiceprint data and video stream data.

[0036] It should be noted that the edge computing equipment deployed in the central control room integrates voiceprint recognition, semantic analysis, and behavior recognition models;

[0037] A distributed high-precision microphone array and infrared camera group form a spatial acoustic-visual collaborative acquisition network.

[0038] Among them, the edge computing device (host system) is deployed in the central control room. The shift leader uses it to assign tasks, and it also serves as a server for real-time data storage and processing. The sound and video collected on site are all processed centrally by it.

[0039] A high-precision microphone array collects operation commands and ambient sounds; and a microphone is used to collect the voice of the operator, with one microphone for each switch cabinet.

[0040] Infrared cameras support low-light environments and capture personnel positions and movements; multiple cameras provide comprehensive on-site monitoring to determine the accurate position of operators (mainly whether the personnel's position corresponds to the position of the corresponding switch cabinet) and the standardization of personnel behavior.

[0041] S2. Verify the operator's identity through voiceprint recognition, and perform semantic analysis on the voice commands based on the power standard terminology library to determine the consistency between the command content and the tasks issued by the system.

[0042] It should be noted that voiceprint features are used to match operators with the pre-authorized task list;

[0043] Specifically, the system verifies the operator's identity by extracting voiceprint features; all operators' voiceprints are pre-recorded in the system for identification. For example, if an operation is scheduled to be performed by personnel A and B, but personnel C actually goes, the system will detect the mismatch in personnel identity based on C's voice when C is announcing the votes and issue an alarm.

[0044] It should also be noted that, by combining the power standard terminology database, multi-dimensional semantic analysis of command actions, equipment numbers, and operating parameters is performed to identify logical conflicts between commands and tasks.

[0045] The power standard terminology database is dynamically updated, and the conversion between dialect terms and standard terms is supported.

[0046] Generate interactive voice prompts for non-standard instructions to guide operators in repeating the standard instructions.

[0047] Specifically, in one embodiment of the present invention, power operation terms are pre-entered into the system. For example, A and B are arranged to jointly perform "closing the 6KV-A switch". However, when B outputs "disconnecting the 6KV-A switch" or "closing the 6KV-B switch" during the vote counting, the system detects through semantic analysis that it is inconsistent with the task being performed and issues an alarm.

[0048] An alarm may also be triggered if the operator uses non-standard terminology during the vote counting process.

[0049] In an optional embodiment, multi-dimensional semantic parsing can involve collecting and organizing a keyword library for power operation instructions, including common operation commands and equipment names. Natural Language Processing (NLP) technology is used to extract keywords from the operator's voice commands, identifying keywords related to task parameters. Contextual analysis is then used to determine the relevance of the keywords to the current power system status. For example, when an operator issues a "close" command, it is necessary to identify whether any equipment is currently disconnected and to determine the rationality of the operation. If the identified keywords do not match the instructions in the task list, the system will prompt the operator through a visual interface or audio feedback, such as "Please confirm whether you want to close the 6kV-A switch?".

[0050] In another optional embodiment, multi-dimensional semantic parsing can also integrate the operator's historical operation records into the power standard terminology library, update instructions and operation specifications in real time, form a dynamic terminology library, and update the non-standard expressions in the terminology library by analyzing non-standard instructions generated in previous operations, establish a mapping relationship between non-standard expressions and standard expressions, and when the operator issues an instruction, perform similarity analysis with historical operation records to determine whether the current instruction conforms to the previous operation mode; for example, if the operator frequently uses "switching 6KV-A switch" instead of the standard "closing", the system can automatically identify and prompt that it is a non-standard instruction, and guide the operator to repeat the standard terminology.

[0051] S3. The operator's position is identified through visual positioning and spatially matched with the target equipment interval number. At the same time, the standardization of the operation behavior is detected through video analysis.

[0052] It should be noted that the spatial mapping relationship between personnel positions and equipment spacing is established through infrared visual positioning;

[0053] Real-time analysis of video stream data is used to verify the compliance of electrical testing procedures and the wearing of safety protective equipment.

[0054] Specifically, in one embodiment of the present invention, electrical operations have standardized actions. These standardized actions are pre-entered into the system. If the physical actions performed by on-site personnel do not match the standard actions in the action library, the operation is deemed non-compliant. For the operation of high-voltage electrical switches, before each action is performed, the monitoring personnel will repeat the action content. This allows the system to determine whether the action is compliant based on voice and video.

[0055] Video analysis is used to determine the operator's location (whether they are within the target equipment bay) and the compliance of their actions (e.g., whether the voltage testing operation is standardized and whether safety protective equipment is worn). Each switchgear is pre-numbered and located using on-site cameras. For example, if A and B are assigned to "disconnect the 6KV-A switch," but before the operation, A and B are not standing in front of the 6KV-A switchgear but in front of the 6KV-B switchgear, the on-site camera will capture their actions, detect the incorrect location, issue an alarm, and verbally inform them "Beware of entering the wrong bay." Simultaneously, a microphone can also detect incorrect bay entry; if an operator standing in front of the 6KV-A switchgear gives the instruction to operate the 6KV-B switchgear, an alarm will be triggered. This achieves dual protection through video and audio.

[0056] Additionally, depending on the task, if certain personal protective equipment (PPE) is required, for example, if operators A and B are not wearing safety helmets or insulated gloves when operating the switch to turn off or on power, an alarm will be triggered, and a voice message will be sent informing them to wear PPE correctly.

[0057] In one optional embodiment, video analytics can be used to detect the standardization of operational behavior by deploying accelerometers, pressure sensors, and temperature sensors around various electrical devices to monitor the operator's environment and behavior. The accelerometers detect the operator's movement (such as walking, stopping, bending, etc.), and the pressure sensors can sense the pressure applied by the operator on the equipment to determine whether an operation has been performed. Data characteristics of standard electrical operation actions are predefined to form a feature library. The real-time sensor data is compared with the standard action library to determine whether the operation conforms to the standard. If the detected non-compliance with the standard is detected, the system will issue an alarm through voice and visual prompts to remind the operator to pay attention.

[0058] In another optional embodiment, the standardization of video analysis detection of operational behavior can also be achieved by labeling standard actions in different operation processes based on video stream data, using the collected labeled video data to train a deep learning model, enabling it to automatically identify and classify different power operation actions. During actual operation, the trained model is used to analyze the video stream in real time to identify the operator's actions. The system compares the identification results with standard operation actions, and if non-standard behavior is found, an alarm is immediately triggered and the operator is prompted via voice.

[0059] S4. Generate voice interaction prompts based on the ambiguity of the instructions, trigger hierarchical alarms through multimodal data fusion, and simultaneously record and store data on violations.

[0060] In one embodiment of the present invention, a hierarchical response of voice alarm, light warning and remote notification is triggered based on multimodal fusion of voiceprint, semantic, spatial location and behavioral data.

[0061] Automatically extract audio and video clips within a set time window before and after the violation, and generate a traceable chain of evidence;

[0062] Establish a task-driven closed-loop supervision logic, and associate the verification results of operator identity, instruction content, spatial location, and behavioral norms with the same task number;

[0063] A behavioral profile of the personnel involved in repeated violations is generated and fed back to the security management platform.

[0064] In an optional embodiment, tiered alarms can be implemented by integrating voiceprints, video streams, and environmental data to establish a context-aware model. This model captures the operator's voice, location, actions, and changes in the surrounding environment, such as temperature and humidity, in real time. Based on ordinary operation records, a behavioral model is built to perform standardized analysis of power operations, setting thresholds for normal and abnormal behaviors. During operation, artificial intelligence is used to analyze the current operational context, fusing environmental factors with operational behavior. For example, if it is detected that an operator is not wearing protective equipment in a high-temperature environment, a corresponding alarm is triggered. Based on the context analysis results, multi-level alarm triggering strategies are set. Minor violations (such as improper clothing) are indicated by audio prompts, while more serious violations (such as not wearing necessary safety equipment) trigger strong warning signals and notify relevant management personnel. After all alarms are generated, the system automatically records audio and video clips and synchronizes the corresponding alarm information to the safety management platform, forming a traceable chain of evidence.

[0065] In another optional embodiment, the tiered alarm system can also collect historical operational data, including voiceprints, video streams, and corresponding ambient sounds such as machine operation sounds, to create a multimodal operational database. This database extracts behavioral features from the audio and video data, such as the frequency and amplitude variations of the sound and the movement trajectories of personnel in the video. These features are then preprocessed and standardized. Machine learning techniques are applied, and supervised learning algorithms (such as decision trees and random forests) are used to train the model to identify normal and abnormal operational behaviors. Normal operational data can be used as positive samples, and non-standard behaviors as negative samples for learning. In actual operation, the behavior of operators is monitored in real time. Newly collected data is input into the trained model, and the model's prediction results are used to determine whether the behavior is compliant. If the model determines that the behavior is abnormal, a tiered alarm is set according to the degree of abnormality. Minor violations are notified via audio prompts, while serious violations trigger a high-intensity alarm and notify safety supervisors.

[0066] Example 2 is the second embodiment of the present invention, which differs from the previous embodiment in that:

[0067] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0068] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0069] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0070] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any of the following techniques known in the art and combinations thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0071] Example 3 is the third embodiment of the present invention. This embodiment provides a power operation monitoring system based on AI behavior recognition and semantic analysis, including: a data acquisition module, a voiceprint recognition module, a semantic analysis module, a behavior recognition module, a decision-making and alarm module, and a data storage module;

[0072] The data acquisition module is used to deploy the control architecture and acquisition equipment of the power system and continuously collect voiceprint data and video stream data.

[0073] The voiceprint recognition module verifies the operator's identity through voiceprint recognition;

[0074] The semantic analysis module performs semantic analysis on voice commands based on the power standard terminology library to determine the consistency between the command content and the tasks issued by the system.

[0075] The behavior recognition module identifies the operator's position through visual positioning and spatially matches it with the target equipment interval number. At the same time, it detects the standardization of the operation behavior through video analysis.

[0076] The decision-making and alarm module generates voice interaction prompts based on the ambiguity of the instructions and triggers hierarchical alarms through multimodal data fusion;

[0077] The data storage module synchronously records and stores data on violations.

[0078] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. 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 be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A power operation supervision method based on AI behavior recognition and semantic analysis, characterized in that: include, Deploy the control architecture and acquisition equipment of the power system to continuously collect voiceprint data and video stream data; Operator identity is verified through voiceprint recognition, and semantic analysis of voice commands is performed based on the power standard terminology database to determine the consistency between the command content and the tasks issued by the system. The system identifies the operator's location using visual positioning and spatially matches it with the target equipment's interval number. Simultaneously, it uses video analysis to detect the standardization of the operator's behavior. Based on the ambiguity of the instructions, voice interaction prompts are generated, and hierarchical alarms are triggered through multimodal data fusion, while simultaneously recording and storing data on violations. The voiceprint recognition includes matching operators with a pre-authorized task list based on voiceprint features; Verify the operator's identity and extract voiceprint features; All operators' voiceprints were pre-recorded into the system for identification purposes; By combining the power standard terminology database, multi-dimensional semantic analysis is performed on the command actions, equipment numbers and operating parameters to identify logical conflicts between commands and tasks; The semantic analysis includes dynamically updating the power standard terminology database and supporting the conversion between dialect terms and standard terms; Generate interactive voice prompts for non-standard instructions to guide operators in repeating the standard instructions; The visual positioning includes establishing a spatial mapping relationship between personnel positions and equipment intervals through infrared visual positioning. Based on real-time analysis of video stream data, the compliance of the voltage testing operation process and the wearing of safety protective equipment is verified. Standardized actions are pre-entered into the system. If the physical actions performed by on-site personnel do not match the standard actions in the action library, they are judged to be non-compliant. For the operation of high-voltage electrical switches, before each action is performed, the supervisor repeats the action content and judges whether the action is compliant based on the voice and video. The tiered alarm system includes a tiered response that triggers voice alarms, light warnings, and remote notifications based on multimodal fusion of voiceprint, semantic, spatial location, and behavioral data. Automatically extract audio and video clips within a set time window before and after the violation to generate a traceable chain of evidence.

2. The power operation supervision method based on AI behavior recognition and semantic analysis as described in claim 1, characterized in that: The continuous collection of voiceprint data and video stream data includes edge computing devices deployed in the central control room, which integrate voiceprint recognition, semantic analysis and behavior recognition models; A distributed high-precision microphone array and infrared camera group form a spatial acoustic-visual collaborative acquisition network. Among them, the edge computing device is deployed in the central control room, through which the shift leader dispatches tasks. It also serves as a server for real-time data storage and processing. The sound and video collected on site are all centrally processed by the edge computing device. A high-precision microphone array collects operation commands and ambient sounds; a microphone is used for each switch cabinet to collect the operator's voice. Infrared cameras support low-light environments and capture personnel positions and movements; two or more cameras can provide all-around monitoring of the site to determine the accurate location of operators and the standardization of their behavior.

3. The power operation supervision method based on AI behavior recognition and semantic analysis as described in claim 2, characterized in that: The tiered alarm also includes establishing a task-driven closed-loop monitoring logic, which associates the verification results of operator identity, instruction content, spatial location, and behavioral norms with the same task number. A behavioral profile of the personnel involved in repeated violations is generated and fed back to the security management platform.

4. A power operation monitoring system based on AI behavior recognition and semantic analysis, employing the power operation monitoring method based on AI behavior recognition and semantic analysis as described in any one of claims 1 to 3, characterized in that, include: The system includes a data acquisition module, a voiceprint recognition module, a semantic analysis module, a behavior recognition module, a decision-making and alarm module, and a data storage module. The data acquisition module is equipped with the control architecture and acquisition equipment of the power system, and continuously acquires voiceprint data and video stream data. The voiceprint recognition module verifies the operator's identity through voiceprint recognition; The semantic analysis module performs semantic analysis on voice commands based on the power standard terminology library to determine the consistency between the command content and the task issued by the system. The behavior recognition module identifies the operator's position through visual positioning and spatially matches it with the target equipment interval number. At the same time, it detects the standardization of the operation behavior through video analysis. The decision-making and alarm module generates voice interaction prompts based on the ambiguity of the instructions and triggers hierarchical alarms through multimodal data fusion. The data storage module synchronously records and stores data on violations.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the power operation supervision method based on AI behavior recognition and semantic analysis as described in any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the power operation supervision method based on AI behavior recognition and semantic analysis as described in any one of claims 1 to 3.

Images