Relay protection skill intelligent training evaluation device and method based on state perception and visual identification fusion

Abstract

The invention belongs to the technical field of electric power, and particularly relates to a relay protection skill intelligent training evaluation device and method based on state perception and visual identification fusion. Comprising a state sensing unit which is used for constructing a distributed state sensing network so as to obtain operation details of a student on a protection pressing plate, a secondary circuit and a test instrument; the visual identification unit is used for building an image analysis system so as to capture action details of trainees in safety measure arrangement, tool use and device operation; the man-machine interaction unit is integrated with an intelligent evaluation system and is used for realizing man-machine interaction operation, providing operation guidance and feedback and storing operation data; and the mode selection unit is used for providing scene selection of a training mode, an assessment mode and a debugging mode. The system not only can capture the real-time state of the operation, but also can record the subtle process of the operation in the whole process, converts the operation process of the trainee into quantifiable and high-precision time sequence data, and lays a solid data foundation for objective evaluation.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of electric power, and particularly relates to a relay protection skill intelligent training and evaluation device and method based on state perception and visual recognition fusion. BACKGROUND

[0002] Relay protection is a key link to ensure the safe and stable operation of the power grid, and its correct debugging and troubleshooting ability is a core skill that power system professionals must have. At present, the training and evaluation of relay protection skills mainly rely on traditional training devices and manual observation and evaluation methods, which have the following obvious problems:

[0003] First, the existing training devices have single functions and lack real-time and accurate perception of the operation process of students. Especially at key operation nodes such as protection pressure plate switching, secondary circuit wiring and instrument use, the operation details cannot be automatically identified and recorded, which makes it difficult to monitor and backtrack the operation process.

[0004] Second, the traditional devices have weak human-computer interaction capabilities and cannot provide real-time voice prompts or risk warnings according to the operation process of students, lack intelligent guidance functions for operation errors, and are difficult to realize individualized training and process correction.

[0005] Third, the evaluation process relies on the subjective experience of evaluators and lacks unified and quantitative evaluation standards. The operation data records are incomplete, which makes it difficult to form traceable scoring basis and is not conducive to post-analysis and skill improvement.

[0006] In order to overcome the above-mentioned defects, in recent years, some training systems have tried to introduce sensors or image recognition technology, but there are still problems such as single perception dimension, insufficient data fusion and imperfect evaluation algorithm, which have not realized the whole-process closed loop from "operation perception" to "intelligent evaluation". SUMMARY

[0007] The purpose of the present application is to provide a relay protection skill intelligent training and evaluation device and method based on state perception and visual recognition fusion to solve the problems in the background art.

[0008] In order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a relay protection skill intelligent training and evaluation device based on state perception and visual recognition fusion, the device comprising:

[0009] a state perception unit for constructing a distributed state perception network to obtain the operation details of students on the protection pressure plate, secondary circuit and test instrument;

[0010] A visual recognition unit is configured to build an image analysis system to capture details of the student's actions in the arrangement of safety measures, use of tools and operation of devices, and to supplement and verify data of the test wiring operation;

[0011] A human-computer interaction unit is integrated with an intelligent evaluation system to realize human-computer interaction operation, provide operation guidance and feedback, store operation data, and quantitatively score the operation process based on a fuzzy comprehensive evaluation method;

[0012] A mode selection unit is configured to provide scene selection of training mode, examination mode and debugging mode.

[0013] Preferably, the state sensing unit comprises:

[0014] A protection pressure plate operation sensing module is composed of a high-precision micro switch installed at the pressure plate shaft to detect the input / exit mechanical position state of the pressure plate;

[0015] A device instrument test operation sensing module is integrated in the test instrument and secondary circuit terminal row, and is composed of a micro current sensor arranged in the measurement hole and wiring terminal area to monitor the branch current value and measurement reading in real time to determine the correctness of the wiring form and the loop conduction state.

[0016] Preferably, the micro current sensor is an IC packaged Hall current sensor based on Hall effect, which adopts a ring structure to non-contact wrap the wire, and uses closed loop compensation technology to realize current measurement.

[0017] Preferably, the visual recognition unit comprises a high-definition wide-angle camera deployed above the operation station to record the whole process of the student's operation action and conduct behavior analysis based on machine vision technology, and to supplement and review the signals of the state sensing unit.

[0018] Preferably, the human-computer interaction unit comprises:

[0019] A capacitive touch screen is used as a human-computer interaction interface to support multi-point touch control and glove operation;

[0020] A high-fidelity voice module is configured to provide Chinese and English speech synthesis to broadcast operation guidance and incorrect operation feedback;

[0021] A data storage module is configured to compress, encrypt and store process data of training and examination;

[0022] An interactive module is configured to provide work mode determination, question selection and voice guidance request functions;

[0023] The intelligent evaluation system is based on expert knowledge base and machine learning algorithm, and has a built-in scoring rule base containing operation specification, safety requirement and task completion quality dimension, and adopts fuzzy comprehensive evaluation method to quantitatively score the operation process in all directions.

[0024] Preferably, the mode selection unit comprises:

[0025] The training mode is used for skill practice, and provides voice guidance, operation record and post-practice analysis;

[0026] The examination mode is used for skill examination and grade evaluation, and automatically records operation steps, analyzes and scores, and generates an evaluation report;

[0027] The debugging mode is used for system maintenance, and allows the administrator to add training projects, scoring rules and import and modify external question banks.

[0028] Preferably, the fuzzy comprehensive evaluation method executed by the intelligent evaluation system comprises the following steps:

[0029] (1) establishing an index set U = {u1, u2, …, u n} corresponding to all scoring rules;

[0030] (2) establishing a comment set V = {v1, v2, …, v n} corresponding to all evaluation grades;

[0031] (3) constructing an evaluation matrix R, wherein the element r ij in the evaluation matrix R represents the membership value of the i-th element in the index set on the j-th element in the comment set;

[0032] (4) setting a weight set W = {w1, w2, …, w n} to determine the weight of each scoring rule by the analytic hierarchy process;

[0033] (5) using a fuzzy synthesis operator to synthesize the weight set W and the evaluation matrix R into a fuzzy vector H on the comment set V, i.e.

[0034] H = WR = {h1, h2, …, h n}

[0035] (6) determining the final comment according to the fuzzy vector H, and calculating the final score according to a preset grade set S as:

[0036] SCORE = |HS T |.

[0037] Preferably, the step of determining the weight set W by the analytic hierarchy process comprises:

[0038] Construct a judgment matrix P, where the elements p ij Let represent the importance scale of the i-th indicator relative to the j-th indicator, and satisfy .

[0039] Calculate the geometric mean of the elements in each row of the judgment matrix P.

[0040]

[0041] For vectors Normalization is performed to obtain the weights of each indicator.

[0042]

[0043] Form a weight set W;

[0044] A consistency test is performed on the judgment matrix P. When the consistency ratio CR < 0.1, the weight allocation is considered reasonable. The consistency index... m is the number of indicators, and the random consistency index. λ max To determine the largest eigenvalue of matrix P.

[0045] Preferably, the intelligent evaluation system supports the continuous optimization of the network parameters of the scoring model based on the difference between expert scores and system scores through deep reinforcement learning algorithms, so as to dynamically adjust the evaluation matrix R or the weight set W and improve the scoring accuracy.

[0046] This invention also discloses a smart training and assessment method for relay protection skills based on the aforementioned device, characterized in that the method includes the following steps:

[0047] The system uses a state perception unit and a visual recognition unit to collect multi-source state information and visual information of trainees during the operation process in real time.

[0048] The system receives mode selection instructions through the human-computer interaction unit and enters the corresponding workflow according to the selected mode.

[0049] During the training or assessment process, the operation steps are recorded in real time based on the collected information, and voice guidance is provided in the training mode. Risk warnings are issued in both modes.

[0050] After the operation is completed, the operation process is quantitatively scored by the intelligent evaluation system and the fuzzy comprehensive evaluation method according to claim 7 or 8, and an evaluation report is generated.

[0051] Compared with existing related technologies, the beneficial effects of the present invention are as follows:

[0052] 1. Multi-dimensional state perception capability is provided to realize accurate digital reproduction of the operation process. The application constructs a distributed state perception network by deeply integrating multiple types of sensors such as micro-moment and electrical sensors at key operation nodes such as secondary circuit wiring terminals, protection pressure plates, and instrument measurement holes. Combined with visual recognition auxiliary verification, it can capture the real-time operation state of trainees in the whole process and from multiple angles, such as the input and output of the pressure plate, the connection of the test wiring, etc. Combined with visual recognition behavior analysis, it accurately records the complete operation process of the trainee, laying a data foundation for auxiliary practice and objective evaluation.

[0053] 2. An intelligent scoring system based on expert knowledge base and fuzzy comprehensive evaluation is constructed to realize the objectivity and standardization of evaluation. The application improves the traditional training and evaluation form which relies on the subjective experience of teachers, constructs a relay protection skill operation expert knowledge base, and covers multi-dimensional evaluation standards such as operation specifications, safety requirements, and task completion quality. The evaluation system uses fuzzy comprehensive evaluation method to form a preliminary evaluation of the accurately collected quantitative data and operation process through machine learning, combines with the rule weight determined by the analytic hierarchy process, maps it to quality comments, and finally obtains a quantifiable score result, effectively eliminating the subjectivity of human evaluation.

[0054] 3. Various modes such as training, examination, and debugging are developed to provide instant prompts and risk warnings according to the operation steps of the trainees, automatically record incorrect operations and generate evaluation reports, and carry out review analysis after the operation is completed. Not only can it provide personalized intensive training programs for trainees, but also can effectively reduce the burden of evaluators and improve the fairness of examination. DETAILED DESCRIPTION

[0055] In the description of the present disclosure, it should be understood that the orientations or positional relationships indicated by the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like are based on the orientations or positional relationships shown in the drawings, and are only for the convenience of describing the present disclosure and simplifying the description, and therefore cannot be understood as indicating or implying that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present disclosure.

[0056] The terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present disclosure, unless otherwise stated, the meaning of "multiple" is two or more.

[0057] In the description of this disclosure, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," "linking," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection, an electrical connection, or a communication connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this disclosure according to the specific circumstances.

[0058] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings and preferred embodiments.

[0059] Appendix Figure 1 This is a smart training and assessment device for relay protection skills based on the fusion of state perception and visual recognition. The device includes:

[0060] The state awareness unit constructs a distributed high-precision state awareness network to acquire a series of operational details performed by trainees in various parts such as the protection pressure plate, secondary circuit, and testing instruments, thereby providing a basis for assessment and behavior analysis.

[0061] The visual recognition unit establishes a full-dimensional image analysis system to capture operational details in locations where electrical information cannot be collected, such as the arrangement of safety measures for trainees, device operation, and use of tools. It also further supplements the data on electrical operations such as test wiring, refining the monitoring dimensions.

[0062] The human-computer interaction unit has a series of convenient, multi-level operation functions and operating logic, and has a built-in intelligent evaluation system based on an expert knowledge base. As a multi-functional integrated software platform, it is used to realize human-computer interaction operation.

[0063] The mode selection unit is used to provide students, teachers and operation and management personnel with a mode scenario selection that is suitable for the current task.

[0064] Furthermore, the state perception unit adopts a distributed sensor network architecture, consisting of a protection pressure plate operation perception module and a device instrument test operation perception module.

[0065] The protective pressure plate operation sensing module consists of high-precision micro switches. These high-precision micro switches are installed at each pressure plate shaft and directly connected to the shaft via a coupling. They accurately detect the mechanical position (engaged / disengaged) of the pressure plate and possess excellent corrosion and oxidation resistance.

[0066] The device instrument test operation perception module is integrated in the test instrument and secondary circuit terminal row, and a micro current sensor is arranged in the instrument and meter measurement hole and secondary circuit wiring terminal area. The micro current sensor is embedded in the measurement hole and secondary circuit wiring terminal, and adopts a ring structure to wrap the wire. The sensor is designed based on the Hall effect, adopts an IC packaged Hall current sensor to isolate interference, uses a closed loop compensation technology to realize high precision and non-contact current measurement, and realizes real-time monitoring of the branch current value, so as to judge whether the wiring form is correct and the circuit is turned on as expected. When the student performs the relay protection tester and secondary circuit wiring, the system automatically perceives and records the behavior, and real-time monitors the current wiring form and electrical measurement value. When the multimeter and other measuring instruments are used, when the meter pen contacts the measured point, the sensor collects the actual reading of this measurement, and records the operation process in real time.

[0067] Further, the visual recognition unit is composed of a global operation visual auxiliary verification module, and the state perception unit forms a dual configuration. The global operation visual auxiliary verification module is arranged on the front and back of the operation station, and a high-definition wide-angle camera is arranged above the station. Based on machine vision technology, the student's safety measure arrangement, tool use, device operation and test wiring are recorded and analyzed, which is an effective supplement and verification of the electrical signal perception, and is used for reviewing the key operation action.

[0068] Further, the man-machine interaction unit adopts an integrated design and is installed on the side of the operation station, and is composed of a capacitive touch screen, a high-fidelity voice module, a data storage module, an interactive module and an intelligent evaluation system. The capacitive touch screen is the main body of the man-machine interaction unit, supports multi-point touch and glove operation, and the surface is treated with anti-glare. The high-fidelity voice module is integrated in the man-machine interaction unit, provides Chinese and English speech synthesis function, and supports providing clear operation guidance and real-time feedback of incorrect operation to the student. The data storage module is configured with large capacity memory, has data compression, encryption and export functions, and supports complete recording of data in long time training and examination process. The interactive module is integrated in the man-machine interaction unit as the core system, provides interactive functions such as determining the working mode, extracting or selecting the question, and seeking voice guidance. The intelligent evaluation system is based on expert knowledge base and machine learning algorithm, and has a built-in scoring rule base including operation specification, safety requirement and task completion quality. The fuzzy comprehensive evaluation method is used to quantitatively score the operation process from all aspects, and an evaluation report of the operation process is generated.

[0069] Further, the mode selection unit is composed of a training mode, an examination mode and a debugging mode. The training mode is used for the scene of daily skill training, the trainee selects the skill point to be practiced and starts operation, and if difficulty is encountered, voice guidance can be played along with the operation steps, the operation steps of the trainee are recorded in real time so as to carry out review and analysis after practice, and an evaluation report is formed. The examination mode is used for training examination and skill level evaluation, the teacher extracts questions in the system, and the operation steps are recorded in real time along with the operation process of the examinee, the system automatically analyzes the operation process and completes scoring, generates an evaluation report, and quickly completes objective evaluation of the skill operation process. The debugging mode is used for daily maintenance, only allows the administrator to enter, and can be used for the administrator to add training projects and scoring rules by himself, or to import and modify an external question bank.

[0070] Based on the above technical scheme, the application uses the advanced technology in the field of Internet of Things and electric power, and proposes a relay protection skill intelligent evaluation device based on state sensing technology, realizes “non-inductive and full-dimensional” state sensing and operation reconstruction. A distributed state sensing network is constructed by deeply integrating multiple types of sensors such as micro torque and electricity at key operation nodes such as secondary circuit wiring terminals, protection pressure plates and instrument measuring holes, and combined with visual identification auxiliary verification, the complete operation process of the trainee is accurately recorded, which lays a data foundation for auxiliary practice and objective evaluation; an expert knowledge base of relay protection skill operation is constructed, which comprehensively covers multi-dimensional evaluation standards such as operation specifications, safety requirements and task completion quality, uses machine learning and fuzzy comprehensive evaluation method to map the accurately collected quantitative data and operation process to quality comments, and finally obtains a quantifiable score result, which effectively eliminates the subjectivity of human evaluation; and various modes such as training, examination and debugging are developed, instant prompts and risk warnings are provided according to the operation steps of the trainee, error operations are automatically recorded and evaluation reports are generated, which not only provides personalized intensive training schemes for the trainee, but also effectively reduces the burden of the evaluators and improves the fairness of the examination.

[0071] Attached Figure 2 The control flowchart for the operation of the relay protection skill intelligent training and evaluation device based on state sensing and visual identification fusion is shown in the following figure, and the specific steps are as follows:

[0072] The relay protection skill intelligent training and evaluation device based on state sensing and visual identification fusion has three working modes, which are a training mode, an examination mode and a debugging mode.

[0073] (1) Training mode

[0074] The trainee arrives at the work station, selects a training mode, and starts practicing. The training mode supports overall training and module practice. When the trainee selects overall training, the trainee needs to perform complete operation on a skill point; when the trainee selects module practice, the trainee can perform targeted practice on part of the operation steps of a skill point, and needs to select the practice content. The trainee selects the skill point to be practiced in the system, and the system broadcasts “start practicing” through voice. The state sensing unit and the visual recognition unit dynamically capture the real-time running state of the device instrument and the operation steps of the trainee, and the training and evaluation device records each operation step in time sequence. When encountering operation difficulties, the trainee clicks “voice guidance”, the system broadcasts the subsequent operation points according to the current real-time operation steps, and records the operation steps unfamiliar to the trainee. If the trainee performs dangerous operation, the system automatically identifies the safety risk and issues a risk warning. After the operation of the skill point is completed, the system calls the expert knowledge base and uses the fuzzy comprehensive evaluation method to rate and score the trainee according to the trainee's performance, comprehensively analyzes the entire operation process, displays the error operation and unfamiliar steps of the trainee on the screen, forms an evaluation report for export, and facilitates the trainee to check and supplement the missing parts.

[0075] (2) Examination mode

[0076] The examinee arrives at the work station, and the examiner selects the examination mode. The examination mode supports two ways of free selection and random selection of questions. If the questions are selected freely, the examiner can select the skill point to be examined in the system; if the questions are selected randomly, the examiner needs to select the skill level and examination direction of the examinee, and the system randomly selects the test questions. Start the examination, and the system broadcasts “start the examination” through voice. The state sensing unit and the visual recognition unit dynamically capture the operation steps of the trainee and the real-time state of the device instrument, and record each operation step in time sequence with each operation of the trainee. If the trainee performs dangerous operation, the system automatically identifies the safety risk and issues a risk warning. After the operation of the question is completed, the system calls the expert knowledge base and uses the fuzzy comprehensive evaluation method to rate and score the trainee according to the trainee's performance, comprehensively analyzes the entire operation process, generates an evaluation report for the error operation of the trainee on the skill point, and facilitates the examiner to export from the system.

[0077] (3) Debugging mode

[0078] The operation manager arrives at the work station and selects the debugging mode. The debugging mode supports modification of system settings and import and arrangement of external question bank. Entering this mode needs to verify the identity of the administrator, and the administrator can add training, examination projects and scoring rules by himself / herself.

[0079] Appendix Figure 3 It is a control flow chart of an intelligent evaluation system in a relay protection skill intelligent evaluation device based on state sensing technology, and the specific steps are:

[0080] The intelligent evaluation system is based on an expert knowledge base and is divided into three dimensions, including several scoring rules:

[0081] (1) Operation specification dimension: including operation sequence, process standard degree, etc.

[0082] (2) Safety requirement dimension: including safety measure arrangement, use of instruments and tools, dangerous operation, etc.

[0083] (3) Task completion quality dimension: including correct wiring, task completion degree, etc.

[0084] The scoring algorithm uses a fuzzy comprehensive evaluation method:

[0085] (1) Establish an index set U = {u1, u2, …, u n} corresponding to all scoring rules;

[0086] (2) Establish a comment set V = {v1, v2, …, v n} corresponding to all evaluation levels, which can be set as V = {Excellent, Good, Qualified, Unqualified};

[0087] (3) Based on the machine learning algorithm, construct the evaluation matrix R, and convert the evaluation levels of each index in the index set into the evaluation matrix. The element r ij of the matrix R represents the membership value of the i-th element in the index set on the j-th element in the comment set. The evaluation matrix R satisfies

[0088] (4) Set the weight set W = {w1, w2, …, w n} and determine the weight of each scoring rule by the analytic hierarchy process;

[0089] (5) Use fuzzy operators to map the weight vector W on the index set U to the fuzzy vector H on the comment set V, i.e.

[0090] H = WR = {h1, h2, …, h n}

[0091] (6) Take the maximum value H k in the fuzzy vector H, and the final comment for this student is V k .

[0092] (7) Determine the grade S of each element in the comment set V. According to the difficulty of skill point operation and the overall level of the examinee, it can be set as S = {100, 75, 60, 45}. Calculate the comprehensive score, and the final score of this student is

[0093] SCORE = |HS T |

[0094] The method for determining the weight of each scoring rule by analytic hierarchy process is as follows:

[0095] The 1-9 scale method is used to represent the relative importance between the row index and the column index. The scale 1 represents that the row index and the column index are equally important, the scale 3 represents that the row index is slightly more important than the column index, the scale 5 represents that the row index is more important than the column index, the scale 7 represents that the row index is very important than the column index, and the scale 9 represents that the row index is absolutely important than the column index. The scales 2, 4, 6 and 8 represent the evaluation scales between the above and below scales. Based on the above rules, a judgment matrix P can be constructed, and the element p ij in the jth column and the ith row of the matrix P represents the importance of the ith row index compared with the jth column index. Since the comparison between the row index and the column index is relative, the element p ji in the ith row and the jth column of the judgment matrix P is the reciprocal of p ij , that is,

[0096]

[0097] For an index set with m indexes, the square root method can be used to calculate the weight of each index, and the formula is as follows:

[0098]

[0099] Through the above formula, the weight w i of the ith index can be obtained, and a weight vector W = {w1, w2, …, w n} is formed.

[0100] In order to ensure the rationality of the judgment matrix P, consistency check is needed. First, the eigenvalue λ of the judgment matrix P is solved, that is,

[0101] PX = λX

[0102] X is the eigenvector of the judgment matrix P. The maximum value λ max in the eigenvalue λ is extracted, and the consistency index CI is calculated, that is,

[0103]

[0104] The larger the CI value is, the greater the inconsistency of the judgment matrix P is. The random consistency index RI can be obtained by looking up the table, and the ratio CR of CI and RI is calculated, that is,

[0105]

[0106] If CR < 0.1, it is considered that the consistency check is passed; if CR ≥ 0.1, the consistency check is not passed, and the evaluation matrix P needs to be adjusted again.

[0107] The system supports online learning function, continuously optimizes the scoring model through deep reinforcement learning algorithm, continuously updates network parameters based on the difference between expert scoring and system scoring, and continuously improves scoring accuracy to provide objective and fair skill evaluation results.

[0108] It should be noted that, for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered within the scope of protection of the present application.

Claims

1. A smart training and assessment device for relay protection skills based on the fusion of state perception and visual recognition, characterized in that, The device includes: The state awareness unit is used to build a distributed state awareness network to obtain the operational details of trainees on the protection pressure plate, secondary circuit and test instruments. The visual recognition unit is used to build an image analysis system to capture the details of trainees' actions in setting up safety measures, using tools and operating devices, and to supplement and verify the data of the test wiring operation. The human-computer interaction unit integrates an intelligent evaluation system to realize human-computer interaction operations, provide operation guidance and feedback, store operation data, and quantify and score the operation process based on the fuzzy comprehensive evaluation method. The mode selection unit is used to provide scenario selection for training mode, assessment mode and debugging mode.

2. The apparatus according to claim 1, characterized in that, The state sensing unit includes: The pressure plate operation sensing module consists of a high-precision micro switch installed on the pressure plate's rotating shaft, used to detect the pressure plate's engaged / disengaged mechanical position. The device instrument testing operation sensing module is integrated inside the testing instrument and secondary circuit terminal block. It consists of miniature current sensors configured in the measurement hole and wiring terminal area, used to monitor the branch current value and measurement reading in real time to determine the correctness of the wiring and the circuit continuity status.

3. The apparatus according to claim 2, characterized in that, The miniature current sensor is an IC-packaged Hall current sensor based on the Hall effect. It uses a ring structure to non-contactly wrap the wire and utilizes closed-loop compensation technology to achieve current measurement.

4. The apparatus according to claim 1, characterized in that, The visual recognition unit includes a high-definition wide-angle camera deployed above the workstation. Based on machine vision technology, it records and analyzes the trainee's actions throughout the process, and is used to supplement and verify the signals of the state perception unit.

5. The apparatus according to claim 1, characterized in that, The human-computer interaction unit includes: Capacitive touchscreens serve as human-computer interaction interfaces, supporting multi-touch and operation while wearing gloves. A high-fidelity voice module is used to provide Chinese and English voice synthesis to broadcast operation instructions and feedback on erroneous operations; The data storage module is used to compress, encrypt, and store training and assessment process data; The interactive module provides functions for determining the work mode, selecting questions, and requesting voice guidance. The intelligent evaluation system, based on an expert knowledge base and machine learning algorithms, has a built-in scoring rule library that includes operational procedures, safety requirements, and task completion quality dimensions. It also uses a fuzzy comprehensive evaluation method to quantitatively score the operation process in all aspects.

6. The apparatus according to claim 1, characterized in that, The mode selection unit includes: The training model is used for skills practice and provides voice guidance, operation records, and post-training review and analysis. The assessment mode is used for skills assessment and level evaluation, automatically recording operation steps, analyzing scores and generating evaluation reports; Debug mode is used for system maintenance, allowing administrators to add training programs, scoring rules, and import and modify external question banks.

7. The apparatus according to claim 5, characterized in that, The fuzzy comprehensive evaluation method executed by the intelligent evaluation system includes the following steps: (1) Establish an index set U = {u1, u2, ..., u} n The set of indicators corresponds to all scoring rules; (2) Establish the comment set V = {v1, v2, ..., v} n The set of comments corresponds to all rating levels; (3) Construct the evaluation matrix R, wherein the elements r in the evaluation matrix R are... ij This represents the membership value of the i-th element in the index set to the j-th element in the comment set; (4) Define the weight set W = {w1, w2, ..., w n The weight set is determined by the analytic hierarchy process (AHP) to determine the weights of each scoring rule. (5) Use the fuzzy synthesis operator to synthesize the weight set W and the evaluation matrix R into a fuzzy vector H on the comment set V, i.e. H=WR={h1,h2,…,h n } (6) Determine the final comments based on the fuzzy vector H, and calculate the final score based on the preset level set S: SCORE=|HS T |。 8. The apparatus according to claim 7, characterized in that, The steps for determining the weight set W using the analytic hierarchy process include: Construct a judgment matrix P, where the elements p ij Let represent the importance scale of the i-th indicator relative to the j-th indicator, and satisfy . Calculate the geometric mean of the elements in each row of the judgment matrix P. For vectors Normalization is performed to obtain the weights of each indicator. Form a weight set W; A consistency test is performed on the judgment matrix P. When the consistency ratio CR < 0.1, the weight allocation is considered reasonable. The consistency index... m is the number of indicators, and the random consistency index. λ max To determine the largest eigenvalue of matrix P.

9. The apparatus according to claim 7 or 8, characterized in that, The intelligent evaluation system supports the use of deep reinforcement learning algorithms to continuously optimize the network parameters of the scoring model based on the difference between expert scores and system scores, so as to dynamically adjust the evaluation matrix R or the weight set W and improve the scoring accuracy.

10. A method for intelligent training and assessment of relay protection skills based on the device described in any one of claims 1 to 9, characterized in that, The method includes the following steps: The system uses a state perception unit and a visual recognition unit to collect multi-source state information and visual information of trainees during the operation process in real time. The system receives mode selection instructions through the human-computer interaction unit and enters the corresponding workflow according to the selected mode. During the training or assessment process, the operation steps are recorded in real time based on the collected information, and voice guidance is provided in the training mode. Risk warnings are issued in both modes. After the operation is completed, the operation process is quantitatively scored by the intelligent evaluation system and the fuzzy comprehensive evaluation method according to claim 7 or 8, and an evaluation report is generated.

Images