Electrical equipment anti-misoperation alarm method based on image recognition and logic judgment

Abstract

The invention discloses an electrical equipment anti-misoperation alarm method based on image recognition and logic judgment. According to the invention, through deep fusion of image identification and logic judgment, dynamic anti-error control of the electrical equipment operation process is realized. By means of a high-precision image feature extraction technology, the system can capture key information such as an equipment appearance identifier and a mechanical structure alignment state in real time, multi-dimensional verification is carried out in combination with a preset standard template, and accurate matching of an operation object and target equipment is ensured. Meanwhile, a logic judgment algorithm comprehensively considers factors such as operation time sequence rationality and mechanical action coherence, abnormal behaviors are analyzed in real time, misoperation is prevented in time through a grading alarm mechanism, safety accidents caused by the problems of equipment mistaken touch, step reversing and the like are effectively avoided, and an intelligent safety defense line is constructed for operation of the high-voltage power distribution room.

Description

Technical Field

[0001] This invention belongs to the field of electrical equipment maintenance technology, specifically a method for preventing misoperation alarm of electrical equipment based on image recognition and logical judgment. Background Technology

[0002] With the rapid development of the power industry and the continuous advancement of smart grid construction, the safety and reliability of high-voltage substations, as an important component of the power system, are receiving increasing attention. High-voltage substations are key facilities in the power system used to receive and distribute high-voltage electrical energy, and to control, protect, and meter power lines. Their core internal equipment includes high-voltage switchgear, circuit breakers, disconnectors, instrument transformers, busbars, and relay protection devices, collectively forming a complete power supply unit. As a hub connecting the regional power grid and the user side, high-voltage substations, through reliable equipment and strict operating procedures, ensure the safe and stable reduction or distribution of high-voltage electrical energy to the next-level substation or electrical equipment. They are indispensable power infrastructure for ensuring the normal power supply of industrial production, commercial operations, and municipal facilities.

[0003] However, traditional power distribution room operation methods have many safety hazards, such as human error and equipment failure. These problems not only affect the normal operation of the power system, but may also threaten the safety of people's lives and property. Summary of the Invention

[0004] The purpose of this invention is to provide an alarm method for preventing misoperation of electrical equipment based on image recognition and logical judgment in order to solve the problems mentioned above.

[0005] The technical solution adopted in this invention is as follows: an alarm method for preventing misoperation of electrical equipment based on image recognition and logical judgment, the method comprising the following steps: S1: Collect circuit breaker status, equipment temperature and environmental parameters through sensors and cameras to form raw data sequences, which are then synchronously transmitted to the local terminal (providing basic data for S5 status verification and S6 logic verification).

[0006] S2: Workers swipe their cards or register their fingerprints. The system verifies the operator's qualifications. Those who fail the verification are restricted from entering the work area (linked to identity information in S4 behavior trajectory tracking).

[0007] S3: Analyzes digital operation tickets, matches operation tasks with equipment intervals through image recognition, and triggers audio-visual prompts when non-target intervals are close.

[0008] S4: Track personnel movement and hand movements in real time, record operation time and location, and verify the legality of the operation by associating it with S2 authorization information (providing behavioral timing data for S6 logic verification).

[0009] S5: The device status collected by S1 is superimposed on the physical device as a virtual label through the AR terminal. After scanning the key parts, the 3D model is compared with the actual image to verify whether the status is consistent with the operation ticket. If they do not match, a red warning is displayed.

[0010] S6: Extract the operation steps of the S4 record in chronological order, compare them with the five-prevention rule base, and trigger a logic conflict alarm when there is a violation or omission of steps (call the S5 status verification result to confirm compliance).

[0011] S7: Trigger a three-level warning based on the result of S6: Level 1 AR text prompt, Level 2 audible and visual alarm and upload to the monitoring center, Level 3 cut off the operating power and restrict personnel movement (the warning event is synchronized to the S8 record).

[0012] S8: Automatically stores data such as personnel identity, operation sequence, equipment status, and early warning processing, encrypts the storage, and supports traceability query (providing raw data for S9 evaluation).

[0013] S9: Analyze S8 data monthly, calculate operational compliance rate and early warning accuracy rate, and optimize the comparison model and rule base for high-frequency early warning processes. In a preferred embodiment of the present invention, during the system startup phase of step S1, the connection configuration between the AR glasses, Bluetooth computer key, and backend server must be completed, ensuring that the device battery level is not lower than 80% and the network latency is controlled within 50 milliseconds. The AR glasses' cameras, gyroscopes, and other sensors are calibrated, and image acquisition parameters are initialized using grayscale cards and standard color blocks to ensure stable operation of the device in an indoor environment with light intensity of 200-1000 lux. Simultaneously, the 3D model data of the equipment in the current work area is loaded, including equipment room door dimensions, electromechanical equipment appearance features, and lintel marking layout, ensuring that the model's error compared to the actual site does not exceed 2 centimeters.

[0014] In a preferred embodiment of the present invention, in step S2, the near-field sensing module built into the AR glasses reads the RFID chip embedded in the worker's safety helmet, extracts the identity information, compares it with the authorized list in the background, and automatically assigns operation permissions after verification. The system pushes a corresponding operation task list to the AR glasses interface according to the daily work plan. The list includes the equipment number, operation sequence, and safety precautions. Simultaneously, the voice interaction module is activated, and the wake-up word recognition sensitivity is set to 95% to ensure accurate response to commands in industrial environmental noise below 85 decibels.

[0015] In a preferred embodiment of the present invention, in step S3, the AR glasses capture real-time images of the surrounding environment and extract features of the equipment room door, lintel markings, and equipment appearance using a convolutional neural network (CNN) algorithm. These features are then compared with a pre-set standard template in the background. When the accuracy of equipment label text recognition is ≥98% and the door frame outline overlap is ≥95%, the equipment interval is determined to be correct, and a green confirmation icon is displayed on the AR interface. If three consecutive matching attempts fail (matching rate <90%), the system immediately issues a 100-decibel alarm, simultaneously generates a "wrong interval risk" warning record in the background control subsystem, and suspends the current operation process. The operator must then rescan and reposition the equipment or request remote manual review.

[0016] In a preferred embodiment of the present invention, in step S4, the system dynamically generates a virtual operation process in the AR interface based on the content of the digital operation ticket. Operators complete the rehearsal through gesture interaction, and the anti-misoperation subsystem monitors the sequence of operation steps in real time. If a skipped step or reverse operation occurs, a red prohibition icon is displayed on the AR interface, and a voice prompt "Operation sequence error" is triggered, while simultaneously locking the next operation permission. During the rehearsal, the Bluetooth computer key reads the current status of the equipment in real time (such as the open / closed position of the disconnector switch and the air pressure value of the circuit breaker). When the air pressure deviation of the pneumatic operating mechanism exceeds 0.1 MPa, the rehearsal is automatically paused, and the AR interface displays the text prompt "Air pressure abnormal, please check equipment status." The rehearsal can only continue after the parameters return to normal.

[0017] In a preferred embodiment of the present invention, in step S5, the operator scans key parts of the high-voltage equipment using an AR terminal. The terminal's built-in high-definition camera captures the physical features of the equipment surface in real time and transmits the image data to the local edge computing unit. First, the image is preprocessed, and then a lightweight convolutional neural network model is used to extract the equipment feature vector. This vector contains key information such as equipment type, current status identifier, and relative position of the mechanical structure. Simultaneously, the system retrieves the standard status model corresponding to the current operation task from the equipment's 3D model database, compares the actual extracted feature vector with the standard model vector dimension by dimension, and calculates the Euclidean distance and structural similarity between the two. If the similarity is higher than a preset threshold, the AR terminal overlays a green virtual label "Status Verification Passed" on the equipment surface; if the similarity is lower than the threshold, a red warning box is immediately displayed, and a voice prompt is given: "The actual equipment status does not match the operation ticket; please check the location of the trip indicator." The system also freezes the next operation permission until manual confirmation or system re-verification. The entire process requires the integration of real-time operating data collected in step S1 as auxiliary verification data to ensure the accuracy of the verification of the consistency between the virtual and real states. The formula for verifying the creativity of AR virtual-real overlay state is: ; In the formula: S represents the comprehensive verification score of the device status (judgment threshold ≥95, red warning is triggered if it is below the threshold), and the value range is 0~100; w_1w1 represents the image feature weight coefficient (reflecting the importance of visual features such as device identification and color, with an empirical value of 0.45), and its value ranges from 0 to 1; F represents the image feature matching degree (the percentage of the AR scanned image that matches the grayscale, outline, and text logo of the standard template), with a value range of 0 to 100; k represents the fluctuation attenuation coefficient (controlling the degree of influence of fluctuations in continuous frame matching, with a value of 0.12 in industrial scenarios), and its value ranges from 0 to 1; σ F The standard deviation of feature fluctuation (the degree of dispersion of matching degree of 10 consecutive frames of images, reflecting the recognition stability) ranges from 0 to 20. w2 represents the mechanical structure weighting coefficient (reflecting the importance of the physical alignment accuracy of the equipment, with an empirical value of 0.35), and its value ranges from 0 to 1; C represents the alignment accuracy of the mechanical structure (the degree of coincidence between the AR scan 3D coordinates and the standard model coordinates), with a value range of 0 to 100. ΔP represents the current air pressure deviation value (the difference between the actual air pressure and the rated air pressure of the circuit breaker pneumatic operating mechanism, in MPa), with a value range of 0~0.2; P max This indicates the maximum permissible air pressure deviation (industry standard threshold 0.2 MPa), with a value of 0.2. w3 represents the time series consistency weighting coefficient (reflecting the rationality and importance of the time series of the operation process, with an empirical value of 0.2), and its value ranges from 0 to 1. T real Indicates the actual operation time (time from the start of scanning to feature stabilization, in seconds), with a value range of 0~30; T std This represents the standard operation duration (the average time taken in historical compliant operation statistics, in seconds), with a value range of 5 to 20.

[0018] In a preferred embodiment of the present invention, in step S6, the AR glasses capture first-person perspective operation video at a rate of 25 frames per second and transmit it to the backend management subsystem via Wi-Fi. The video stream uses H.265 encoding to reduce bandwidth usage. Simultaneously, a two-way voice call function is enabled, allowing remote monitoring personnel to view the operation screen in real time and issue guidance commands through the backend system. The voice transmission latency must be controlled within 300 milliseconds. The system performs AI recognition on key actions during the operation process (such as the rotation angle of the operating handle and the opening status of the equipment cabinet door). When an action not performed according to the standard procedure is detected, the AR interface flashes a yellow warning border and overlays text prompts.

[0019] In a preferred embodiment of the present invention, in step S7, after each operation, the AR glasses automatically take a photo confirming the device status (resolution not less than 1920×1080 pixels), focusing on recording the position of the opening / closing indicator, the lock status, and the instrument readings. The photo is associated with the corresponding steps on the digital operation ticket and stored. The Bluetooth computer key reads the device operation counter value and verifies it against the operation ticket completion status to ensure that all steps are 100% executed. The system automatically calculates the duration of this operation and compares it with the historical average duration, generating an operation efficiency evaluation report, which is simultaneously uploaded to the work behavior statistical analysis module.

[0020] As a preferred embodiment of the present invention, in step S8, based on AI image recognition and logical judgment algorithms, the system monitors illegal operations during the operation process in real time (such as unauthorized opening of the equipment back cover or crossing the safety fence). When a risky behavior is identified, a three-level alarm is immediately triggered: the first-level alarm displays a flashing red border on the AR interface and issues a voice warning; the second-level alarm triggers the on-site sound and light alarm device (the red warning light flashes at a frequency of 1Hz); the third-level alarm automatically pushes fault information to the background management subsystem, including the type of abnormality, the time of occurrence, and the real-time video link, and freezes all operation permissions in the work area, which must be manually lifted by the administrator.

[0021] As a preferred embodiment of the present invention, in step S9, after the operation is completed, all process data (including audio and video recordings, operation step logs, and device status parameters) are encrypted and stored through a unified data management platform. Data access permissions follow the "principle of least privilege," limiting access to system administrators and authorized auditors only. The system performs monthly statistical analysis of historical data, focusing on evaluating image matching success rate, operational compliance rate, and alarm response timeliness. When the feature matching failure rate of a certain type of device exceeds 5% for three consecutive months, the model optimization process is automatically triggered, and image samples of the device area are re-collected to update the feature template library, ensuring recognition stability.

[0022] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention achieves dynamic anti-misoperation control of electrical equipment operation through the deep integration of image recognition and logical judgment. Utilizing high-precision image feature extraction technology, the system can capture key information such as equipment appearance markings and mechanical structure alignment in real time. Combined with preset standard templates, multi-dimensional verification is performed to ensure accurate matching between the operated object and the target equipment. Simultaneously, the logical judgment algorithm comprehensively considers factors such as the rationality of the operation sequence and the continuity of mechanical actions, performing immediate analysis of abnormal behavior. A tiered alarm mechanism promptly prevents misoperation, effectively avoiding safety accidents caused by accidental equipment contact or reversed procedures, thus constructing an intelligent safety defense line for high-voltage power distribution room operations.

[0023] 2. This invention deeply integrates virtual verification with physical operation through the synergy of AR (Augmented Reality) and AI intelligent decision-making. The system not only provides real-time feedback on potential risks during operation but also continuously optimizes the recognition model through historical data accumulation, improving adaptability and accuracy in complex environments. This dynamic error prevention mechanism significantly reduces the pressure on manual monitoring, standardizes work processes, and provides traceable digital records for equipment operation. It strongly supports the intelligent upgrade of power system safety management and promotes the transformation of work modes from experience-driven to data-driven. Attached Figure Description

[0024] 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.

[0025] Figure 1 This is a schematic diagram illustrating the process principle of the present invention. Detailed Implementation

[0026] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0027] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0028] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0029] Example: Reference Figure 1 , An alarm method for preventing misoperation of electrical equipment based on image recognition and logical judgment, the method includes the following steps: S1: Real-time acquisition of equipment status and environmental parameters Infrared sensors, high-definition cameras, and temperature and humidity monitoring devices deployed in high-voltage power distribution rooms continuously collect data on the opening and closing status, surface temperature, and operating mechanism positions of key equipment such as circuit breakers and disconnect switches, as well as parameters such as ambient light and dust concentration, forming a raw monitoring data sequence. The collected data is synchronously transmitted to a local processing terminal, providing a basis for subsequent equipment identification and status judgment (data exchange is required with S5 equipment status verification and S6 logic verification).

[0030] S2: Dual verification of operator identity and qualifications Before entering the power distribution room, personnel must register their identity by swiping an IC card or using fingerprint recognition. The system automatically retrieves the personnel qualification database to verify whether they possess the authorization qualifications for the current operation (such as the validity period of the high-voltage equipment operation qualification certificate and the level of job authority). The verification results are fed back to the on-site display terminal in real time. Personnel who fail the verification will be restricted from entering the work area (personnel identity information must be linked to the S4 behavior trajectory tracking).

[0031] S3: Intelligent matching of operation tasks and equipment intervals Based on the digital operation ticket issued by the shift dispatcher, the system automatically parses the operation object (such as the #1 main transformer high-voltage side circuit breaker) and the sequence of steps. Through equipment identification plate image recognition (such as equipment name, number, and bay number), the operation task is matched with the actual equipment bay on site. If personnel are detected approaching a non-target bay, an audible and visual prompt "Current bay does not match the task" is triggered (the equipment location data collected by S1 needs to be called to verify and lock the target equipment for S5 status verification).

[0032] S4: Real-time monitoring of dynamic operation behavior trajectory By using cameras deployed along the critical path, the system tracks the movement trajectory and hand movements of workers in real time, and determines the intention of the operation through limb key point localization technology (such as wrist and finger movement characteristics). For example, when a person reaches out to touch the circuit breaker operating handle, the system automatically records the time and location of the action, and associates it with the authorization information in S2 to confirm whether it is an operation action permitted for the current task (providing behavioral timing data for the logical rule verification in S6).

[0033] S5: Augmented Reality Overlay Device Status Verification Using a portable augmented reality terminal (such as AR glasses), real-time equipment status data collected by S1 (such as circuit breaker tripping indication and energy storage motor power status) is overlaid onto the physical equipment surface as virtual tags. Operators need to scan key parts of the equipment (such as circuit breaker mechanical indicator plates and cabinet door locking devices) using the terminal. The system simultaneously calls up the equipment's 3D model and compares its features with the actual image to verify whether the equipment's current status matches the operation ticket requirements (e.g., "It should be in the tripping state before operation"). If the virtual and real statuses do not match, the terminal immediately displays a red warning box and provides a voice prompt: "Equipment status is abnormal, please verify" (this requires integrating the real-time data from S1 with the task matching results from S3 to provide status basis for the logical verification of S6).

[0034] S6: Comparison of operation sequence with preset procedure logic The system extracts the operation steps (such as "open the disconnect switch → check the trip position → close the grounding switch") in chronological order based on the operation behavior trajectory recorded by S4, and compares them one by one with the preset five-prevention logic rule library (such as "It is strictly forbidden to open the disconnect switch under load" and "The grounding switch must be tested for voltage before operation"). If an error in the operation sequence is found (such as closing the grounding switch without testing for voltage) or a key step is omitted, a logic conflict alarm is immediately triggered (it is necessary to call the equipment status verification result of S5 to confirm whether the status before operation meets the requirements of the procedure).

[0035] S7: Multi-level risk warning and emergency intervention linkage Based on the logical comparison results of S6, the system triggers three levels of early warning according to risk level: Level 1 warning (minor deviation) displays a yellow warning message on the AR terminal; Level 2 warning (operational violation) activates the on-site audible and visual alarm and simultaneously sends a warning message to the monitoring center; Level 3 warning (emergency danger, such as accidental contact with live equipment) automatically cuts off the power supply to the operating mechanism and links the power distribution room access control system to restrict personnel movement. After the warning is triggered, it is necessary to wait for the monitoring center or a supervisor to unlock the system using an authorization code (the warning event must be synchronized with the operation record of S8 in real time).

[0036] S8: Closed-loop recording of full-process operation data From personnel verification in S2 to warning cancellation in S7, the system automatically records key data during the operation, including: personnel identity information, operation timestamps, equipment status change curves, AR terminal interaction logs, warning triggering and processing results, etc. The data is stored in encrypted format on the local server and supports traceability queries by time period, equipment name, or personnel name (providing raw data support for the effectiveness evaluation of S9).

[0037] S9: Operational Effectiveness Evaluation and Dynamic Optimization of the Rule Base Monthly statistical analysis is performed on the operational data recorded by S8 to calculate key indicators such as: operation standardization rate (percentage of steps conforming to procedures), early warning accuracy rate (actual misoperation / total number of early warnings), and average operation time. For operational steps that frequently trigger early warnings (such as misjudgments in the status verification of specific equipment), technical personnel review the virtual-to-real comparison model parameters of S5 and the logical rule base of S6. The feature comparison algorithm is optimized by adding typical case samples (such as equipment identification images under different lighting conditions) to continuously improve system reliability (requiring iterative optimization by calling historical operation data and current system parameters).

[0038] In step S1, during the system startup phase, the connection configuration between the AR glasses, Bluetooth computer key, and backend server must be completed, ensuring that the device battery level is not lower than 80% and the network latency is controlled within 50 milliseconds. The AR glasses' cameras, gyroscopes, and other sensors are calibrated, and image acquisition parameters are initialized using grayscale cards and standard color blocks to ensure stable operation of the device in an indoor environment with light intensity of 200-1000 lux. Simultaneously, the 3D model data of the equipment in the current work area is loaded, including equipment room door dimensions, electromechanical equipment appearance features, and lintel marking layout, ensuring that the model's error compared to the actual site does not exceed 2 centimeters.

[0039] In step S2, the near-field sensing module built into the AR glasses reads the RFID chip embedded in the worker's safety helmet, extracts the identity information, and compares it with the authorized list in the background. Once verification is successful, operation permissions are automatically assigned. The system pushes a corresponding operation task list to the AR glasses interface based on the day's work plan. The list includes the equipment number, operation sequence, and safety precautions. Simultaneously, the voice interaction module is activated, with the wake-up word recognition sensitivity set to 95% to ensure accurate response to commands in industrial environmental noise levels below 85 decibels.

[0040] In step S3, the AR glasses capture real-time images of the surrounding environment and extract features of the equipment room door, lintel markings, and equipment appearance using a convolutional neural network (CNN) algorithm. These features are then compared with a pre-set standard template in the background. When the accuracy of equipment label text recognition is ≥98% and the door frame outline overlap is ≥95%, the equipment interval is determined to be correct, and a green confirmation icon is displayed on the AR interface. If three consecutive matching attempts fail (match rate <90%), the system immediately issues a 100-decibel alarm, simultaneously generates a "wrong interval risk" warning record in the background control subsystem, and suspends the current operation process. The operator must then rescan and reposition the equipment or request remote manual verification.

[0041] In step S4, the system dynamically generates a virtual operation process on the AR interface based on the digital operation ticket content. Operators complete the rehearsal through gesture interaction, and the anti-misoperation subsystem monitors the sequence of operation steps in real time. If a step is skipped or reversed, a red prohibition icon is displayed on the AR interface, triggering a voice prompt "Operation sequence error," and simultaneously locking access to the next operation. During the rehearsal, the Bluetooth computer key reads the current status of the equipment in real time (such as the open / closed position of the disconnector switch and the air pressure value of the circuit breaker). When the air pressure deviation of the pneumatic operating mechanism exceeds 0.1 MPa, the rehearsal is automatically paused, and the AR interface displays the text prompt "Air pressure abnormal, please check equipment status." The rehearsal can only continue after the parameters return to normal.

[0042] In step S5, the operator scans key parts of the high-voltage equipment using an AR terminal. The terminal's built-in high-definition camera captures the physical features of the equipment surface in real time and transmits the image data to the local edge computing unit. First, the image is preprocessed, and then a lightweight convolutional neural network model is used to extract the equipment feature vector. This vector contains key information such as equipment type, current status identifier, and relative position of the mechanical structure. Simultaneously, the system retrieves the standard status model corresponding to the current operation task from the equipment's 3D model database. The extracted feature vector is compared dimension-by-dimensionally with the standard model vector, calculating the Euclidean distance and structural similarity between the two. If the similarity is higher than a preset threshold, the AR terminal overlays a green virtual label "Status Verification Passed" on the equipment surface; if the similarity is lower than the threshold, a red warning box is immediately displayed, and a voice prompt is given: "The actual equipment status does not match the operation ticket; please check the location of the trip indicator." The system also freezes the next operation permission until manual confirmation or system re-verification. The entire process requires the integration of real-time operational data collected in step S1 as auxiliary verification to ensure the accuracy of the verification of the consistency between the virtual and real statuses. The formula for verifying the creativity of AR virtual-real overlay state is: ; In the formula: S represents the comprehensive verification score of the device status (judgment threshold ≥95, red warning is triggered if it is below the threshold), and the value range is 0~100; w_1w1 represents the image feature weight coefficient (reflecting the importance of visual features such as device identification and color, with an empirical value of 0.45), and its value ranges from 0 to 1; F represents the image feature matching degree (the percentage of the AR scanned image that matches the grayscale, outline, and text logo of the standard template), with a value range of 0 to 100; k represents the fluctuation attenuation coefficient (controlling the degree of influence of fluctuations in continuous frame matching, with a value of 0.12 in industrial scenarios), and its value ranges from 0 to 1; σ FThe standard deviation of feature fluctuation (the degree of dispersion of matching degree of 10 consecutive frames of images, reflecting the recognition stability) ranges from 0 to 20. w2 represents the mechanical structure weighting coefficient (reflecting the importance of the physical alignment accuracy of the equipment, with an empirical value of 0.35), and its value ranges from 0 to 1; C represents the alignment accuracy of the mechanical structure (the degree of coincidence between the AR scan 3D coordinates and the standard model coordinates), with a value range of 0 to 100. ΔP represents the current air pressure deviation value (the difference between the actual air pressure and the rated air pressure of the circuit breaker pneumatic operating mechanism, in MPa), with a value range of 0~0.2; P max This indicates the maximum permissible air pressure deviation (industry standard threshold 0.2 MPa), with a value of 0.2. w3 represents the time series consistency weighting coefficient (reflecting the rationality and importance of the time series of the operation process, with an empirical value of 0.2), and its value ranges from 0 to 1. T real Indicates the actual operation time (time from the start of scanning to feature stabilization, in seconds), with a value range of 0~30; T std This represents the standard operation duration (the average time taken in historical compliant operation statistics, in seconds), with a value range of 5 to 20.

[0043] In step S6, the AR glasses capture first-person perspective operation video at a rate of 25 frames per second and transmit it to the backend management subsystem via Wi-Fi. The video stream uses H.265 encoding to reduce bandwidth consumption. Simultaneously, two-way voice communication is enabled, allowing remote monitoring personnel to view the operation screen in real time and issue guidance commands through the backend system. The voice transmission latency must be controlled within 300 milliseconds. The system uses AI to recognize key actions during the operation (such as the rotation angle of the operating handle and the opening status of the equipment cabinet door). When an action not performed according to the standard procedure is detected, the AR interface flashes a yellow warning border and overlays the text "Please confirm the operation is in compliance with regulations."

[0044] In step S7, after each operation, the AR glasses automatically take a photo confirming the device status (resolution no less than 1920×1080 pixels), focusing on recording the position of the opening / closing indicator, the lock status, and the instrument readings. The photo is associated with the corresponding steps on the digital operation ticket and stored. The Bluetooth computer key reads the device operation counter value and verifies it against the operation ticket completion status to ensure that all steps are 100% executed. The system automatically calculates the duration of this operation and compares it with the historical average duration, generating an operation efficiency evaluation report, which is simultaneously uploaded to the work behavior statistical analysis module.

[0045] In step S8, based on AI image recognition and logical judgment algorithms, the system monitors illegal operations during the operation process in real time (such as unauthorized opening of the equipment back cover or crossing the safety fence). When a risky behavior is identified, a three-level alarm is immediately triggered: the first-level alarm displays a flashing red border on the AR interface and issues a voice warning; the second-level alarm triggers the on-site sound and light alarm device (the red warning light flashes at a frequency of 1Hz); the third-level alarm automatically pushes fault information to the background management subsystem, including the type of abnormality, the time of occurrence, and the real-time video link, and freezes all operation permissions in the work area, which must be manually lifted by the administrator.

[0046] In step S9, after the operation is completed, all process data (including audio and video recordings, operation step logs, and device status parameters) are encrypted and stored through a unified data management platform. Data access permissions follow the "principle of least privilege," limiting access to system administrators and authorized auditors only. The system performs monthly statistical analysis of historical data, focusing on evaluating image matching success rate, operational compliance rate, and alarm response timeliness. When the feature matching failure rate of a certain type of device exceeds 5% for three consecutive months, the model optimization process is automatically triggered, re-collecting image samples of the device area to update the feature template library and ensure recognition stability.

[0047] From the above, we can conclude that: (1) Deep integration and application of AR+AI human-machine recognition technology This application deeply integrates AR (Augmented Reality) and AI (Artificial Intelligence) technologies to construct a high-precision human-machine recognition system. This system can identify the identity, location, and actions of operators in real time, ensuring that only authorized personnel can perform specific operations at the correct time and place. The combination of AR and AI technologies represents a cutting-edge trend in current technological development, particularly in the field of industrial safety, where its application can significantly improve operational efficiency and safety. This application keeps pace with technological trends, applying this advanced technology to the safety management of high-voltage power distribution rooms, demonstrating significant timeliness.

[0048] (2) Innovation of AI-based image recognition and logical judgment algorithms This application develops an AI-based image recognition and logical judgment algorithm for logical verification and illegal operation alarms on equipment back covers. The algorithm can monitor equipment status in real time, determine the legality of operations based on operating rules, and issue alarm signals when necessary. Image recognition and logical judgment algorithms are important application areas of AI technology, particularly in industrial equipment monitoring and safety management. This application, through innovative algorithm design, achieves precise monitoring and intelligent early warning of high-voltage power distribution room equipment, demonstrating significant cutting-edge technology and timeliness.

[0049] (3) Integration and optimization of the five-prevention system for high-voltage switchgear in thermal power plants based on AI bidirectional recognition This application integrates the aforementioned innovative technologies to construct a five-prevention system for high-voltage switchgear in thermal power plants based on AI-powered bidirectional recognition. This system enables real-time monitoring, intelligent early warning, and behavioral correction of the operational process, significantly improving operational safety and efficiency. Intelligent safety management systems are a hot research area in the field of industrial safety, especially in high-risk operating environments such as high-voltage distribution rooms. Through system integration and optimization, this application achieves comprehensive monitoring and intelligent management of the operational process, demonstrating significant cutting-edge technology and timeliness.

[0050] This application is expected to achieve significant innovative results in AR+AI human-machine recognition technology, AI-based image recognition and logical judgment algorithms, and the integration and optimization of the five-prevention system for high-voltage switches in thermal power plants based on AI bidirectional recognition. These results are not only cutting-edge and timely, but also possess methodological, theoretical, and intellectual property characteristics, providing strong support for the high-quality development of the power industry.

[0051] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended 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. An electrical equipment anti-misoperation alarm method based on image recognition and logical judgment, characterized in that: The method comprises the following steps: S1: Collect the circuit breaker state, equipment temperature and environmental parameters through sensors and cameras to form a raw data sequence, and synchronously transmit to a local terminal; S2: The operator registers by swiping a card or fingerprint, and the system checks the operation qualification. If it fails, the operator is restricted from entering the operation area; S3: Analyze the digital operation ticket, match the operation task and equipment interval through image recognition, and trigger an audible and light prompt when approaching a non-target interval; S4: Real-time track the personnel movement trajectory and hand movements, record the operation time and position, and associate the S2 authorized information to confirm the operation legality; S5: Superimpose the equipment state collected in S1 on the physical equipment as a virtual label through the AR terminal, compare the three-dimensional model and actual image after scanning the key parts, and verify whether the state is consistent with the operation ticket. If not, a red warning is displayed; S6: Extract the operation steps recorded in S4 in chronological order, compare them with the five-prevention rule library, and trigger a logical conflict alarm when there are rule violations or missing steps; S7: Trigger a three-level early warning according to the results of S6: level one AR text prompt, level two audible and light alarm and upload to the monitoring center, and level three cut off the operation power and restrict the personnel movement; S8: Automatically store personnel identity, operation time sequence, equipment state, early warning processing and other data, and support traceability query after encryption storage; S9: Analyze S8 data every month, calculate the operation specification rate and early warning accuracy rate, and optimize the comparison model and rule library for high-frequency early warning links.

2. The image recognition and logic judgment based electrical equipment misoperation prevention warning method of claim 1, wherein: In the step S1, the system startup stage needs to complete the connection configuration of AR glasses, Bluetooth computer key and background server, ensure that the device power is not less than 80% and the network delay is controlled within 50 milliseconds; calibrate the sensors such as camera and gyroscope of AR glasses, complete image acquisition parameter initialization through gray card and standard color block, and make the device keep stable operation in indoor environment with light intensity of 200-1000 lux; synchronously load the three-dimensional model data of the equipment in the current operation area, including equipment room door size, mechanical and electrical equipment appearance characteristics and disc ledge identification layout, to ensure that the model and the actual error is not more than 2 cm.

3. The image recognition and logic judgment based electrical equipment misoperation prevention warning method of claim 1, wherein: In the step S2, the near-field induction module built-in AR glasses reads the RFID chip built-in safety helmet of the operator, extracts the identity information and compares it with the background authorized list, and verifies it automatically after passing the verification. The system pushes the corresponding operation task list to the AR glasses interface according to the daily operation plan, which includes equipment number, operation sequence and safety precautions; at the same time, the voice interaction module is started, the wake-up word recognition sensitivity is set to 95%, and the system ensures accurate response in an industrial environment noise below 85 decibels.

4. The image recognition and logic judgment based electrical equipment misoperation prevention warning method of claim 1, wherein: In the step S3, the AR glasses real-time capture the on-site environment image, extract the equipment room door, the disc lintel identification and the equipment appearance features through the convolution neural network algorithm, and match and compare with the standard template preset in the background; when the equipment identification text recognition accuracy is ≥98% and the door frame contour coincidence degree is ≥95%, it is determined that the equipment interval is correct, and the AR interface displays a green confirmation mark; if the matching fails for 3 times in a row, the system immediately issues a 100 decibel buzzing alarm, generates a "wrong interval risk" early warning record in the background management and control subsystem, and suspends the current operation process, which needs to be re-scanned and positioned by the operator or applied for remote manual review.

5. The image recognition and logic judgment based electrical equipment misoperation prevention warning method of claim 1, wherein: In the step S4, the system dynamically generates a virtual operation process in the AR interface according to the digital operation ticket content, and the operator completes the rehearsal through gesture interaction, and the error prevention operation subsystem monitors the operation step sequence in real time; if the steps are skipped or operated in reverse, the AR interface displays a red prohibited icon and triggers a voice prompt "operation sequence error", and locks the next operation permission; during the rehearsal process, the Bluetooth computer key reads the current state of the equipment in real time, and when the pneumatic operation mechanism pressure deviation exceeds 0.1 MPa, the rehearsal is automatically paused and the AR interface displays a text prompt "abnormal pressure, please check the equipment state", which needs to be continued after the parameters return to normal.

6. The image recognition and logic judgment based electrical equipment misoperation prevention warning method of claim 1, wherein: In the step S5, the operator scans the key parts of the high-voltage equipment through the AR terminal, the high-definition camera built-in the terminal captures the equipment surface physical features in real time, and the image data is transmitted to the local edge computing unit; first, the image is preprocessed, and then the lightweight convolutional neural network model is used to extract the equipment feature vector, which contains the equipment type, current state identification, mechanical structure relative position and other key information; at the same time, the system retrieves the standard state model corresponding to the current operation task from the equipment three-dimensional model database, compares the actual extracted feature vector with the standard model vector dimension by dimension, calculates the Euclidean distance and structural similarity of the two; if the similarity is higher than the preset threshold, the AR terminal superimposes a green virtual label "state verification passed" on the equipment surface; if the similarity is lower than the threshold, a red warning box is immediately displayed and a synchronous voice prompt "the actual state of the equipment does not match the operation ticket, please review the position of the brake instruction board" is given, and the next operation permission is frozen until manual confirmation or system re-verification is passed; the real-time running data collected in the S1 step needs to be fused as an auxiliary verification basis to ensure the accuracy of the virtual and real state consistency verification; The AR virtual and real superposition state verification formula is: ； In the formula: S represents the equipment state comprehensive verification score, with a value range of 0~100; w1 represents the image feature weight coefficient, with a value range of 0~1; F represents the image feature matching degree, with a value range of 0~100; k represents the fluctuation attenuation coefficient, with a value range of 0~1; σ F σ represents the standard deviation of characteristic fluctuations, and has a value ranging from 0 to 20; w2 represents the mechanical structure weight coefficient, with a value range of 0~1; C represents the mechanical structure alignment accuracy, with a value range of 0~100; ΔP represents the current pressure deviation value, with a value range of 0~0.2; P max represents the maximum allowable pressure deviation, and takes the value 0.2; w3 represents the time sequence consistency weight coefficient, with a value range of 0~1. T real represents the actual operation duration, and takes a value in the range of 0-30; T std represents the standard operation duration, and takes a value in the range of 5-20.

7. The image recognition and logic judgment based electrical equipment misoperation prevention warning method of claim 1, wherein: In the step S6, the AR glasses collect the first perspective operation video at a rate of 25 frames per second, and transmit the video to the background control subsystem through the Wi-Fi network. The video stream adopts the H.265 encoding format to reduce the bandwidth occupation, and the two-way voice call function is started at the same time. The remote monitor can view the operation picture in real time through the background system and issue guidance instructions. The voice transmission delay needs to be controlled within 300 milliseconds.

8. The image recognition and logic judgment based electrical equipment misoperation prevention warning method of claim 1, wherein: In the step S7, after each operation is completed, the AR glasses automatically take a device state confirmation photo, which focuses on recording the position of the split and combination indicator, the status of the lock and the instrument reading. The photo is stored in association with the corresponding step of the digital operation ticket. The Bluetooth computer key reads the device operation counter value, which is checked with the operation ticket completion condition to ensure that all steps are 100% executed in place. The system automatically calculates the operation duration of this time and compares it with the historical average duration to generate an operation efficiency evaluation report, which is uploaded to the job behavior statistical analysis module at the same time.

9. The image recognition and logic judgment based electrical equipment misoperation prevention warning method of claim 1, wherein: In the step S8, based on the AI image recognition and logical judgment algorithm, the system monitors illegal operations in real time during the operation. When a risk behavior is identified, a three-level alarm is triggered immediately. The first-level alarm displays a red flashing frame through the AR interface and issues a voice warning. The second-level alarm triggers the on-site sound and light alarm device. The third-level alarm automatically pushes the fault information to the background control subsystem, including the abnormal type, occurrence time and real-time video link. At the same time, all operation permissions of the operation area are frozen, which need to be manually released by the administrator.

10. The image recognition and logic judgment based electrical equipment misoperation prevention warning method of claim 1, wherein: In the step S9, after the operation is completed, all process data are stored through the unified data management platform. The data access authority follows the "minimum authorization principle" and is limited to system administrators and authorized auditors. The system performs statistical analysis on historical data every month, focusing on evaluating the image matching success rate, operation specification compliance rate and alarm response time. When the feature matching failure rate of a certain type of device exceeds 5% for three consecutive months, the model optimization process is automatically triggered, and the image samples of the device area are re-collected to update the feature template library, ensuring the stability of identification.

Images