Metering device inspector training system based on artificial intelligence

Abstract

The invention discloses a metering device inspector training system based on artificial intelligence, and relates to the technical field of practical training devices, a personal center module generates a trainee portrait based on the reply of a pre-stored questionnaire and a deep learning algorithm, and inputs the trainee portrait into a knowledge graph module through a main control module; the knowledge graph module carries out online learning and testing based on the trainee portrait; questions generated by the students are sent to the AI teaching assisting module through the main control module, and questions are answered based on resources pre-stored by the knowledge graph module; and the virtual simulation module is in line connection with the main control module, and performs student operation scoring based on a pre-stored scene in combination with an operation sequence, operation time and scene key points in actual operation of the student. The invention provides a set of efficient and intelligent training system, online training of related personnel is realized, and the urgent task of the industry for the requirements of professional metering device inspectors is met.

Description

Technical Field

[0001] The present invention relates to the technical field of training devices, and more particularly to a training system for metering device inspectors based on artificial intelligence. Background Art

[0002] In the fields of modern industrial production and scientific research, the accuracy and reliability of metering devices are crucial, which are directly related to product quality control, the accuracy of experimental data, and the stability of production processes. As the key role in ensuring the normal operation of metering devices, the professional skills and operation levels of metering device inspectors play a decisive role in the quality of metering work.

[0003] Traditional training methods for metering device inspectors have many limitations. On the one hand, relying on the combination of theoretical lectures and actual equipment operations, theoretical teaching is often abstract and difficult for trainees to intuitively understand the internal principles and complex operation processes of metering devices; actual operation training is limited by resources such as the number of devices, venues, and training teachers, and trainees have limited opportunities to operate and are difficult to fully practice. On the other hand, this training method lacks a precise evaluation system, is difficult to comprehensively capture and quantitatively analyze the detailed problems in the trainees' operation process, and cannot provide trainees with targeted improvement suggestions, resulting in difficult-to-guarantee training effects. The trained inspectors may affect the accuracy and efficiency of metering work due to problems such as unskilled operations and improper handling of emergencies in actual work.

[0004] With the rapid development of artificial intelligence technology, it has shown great potential in the field of education and training. Knowledge graph technology can integrate and correlate a large amount of knowledge information, provide trainees with structured and systematic learning resources, and help trainees better understand and master the knowledge context; virtual simulation technology can create highly realistic virtual scenarios, allowing trainees to practice repeatedly in a risk-free environment, breaking through the limitations of traditional training in terms of equipment and venues; AI teaching assistants can, based on big data and intelligent algorithms, answer trainees' questions in real time and provide personalized learning guidance. However, currently in the field of training metering device inspectors, the application of these artificial intelligence technologies is still in the exploratory stage, and a mature and perfect training system has not yet been formed. Deeply integrating artificial intelligence technology into the training of metering device inspectors and constructing an efficient and intelligent training system have become an urgent task to improve training quality and meet the industry's demand for professional metering device inspectors, and it is also a problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a training system for metering device inspectors based on artificial intelligence to solve the problems in the above background art.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A training system for metering device inspectors based on artificial intelligence, comprising: a knowledge graph module, a virtual simulation module, a main control module, an AI teaching assistant module and a personal center module. The personal center module generates a student portrait based on the responses to the pre-stored questionnaire and deep learning algorithms, and inputs the student portrait into the knowledge graph module through the main control module; the knowledge graph module pre-stores teaching and test resources in the form of a knowledge graph, and conducts online learning and tests based on the student portrait; the questions generated by the student in the knowledge graph module are sent to the AI teaching assistant module through the main control module, and answers are provided based on the resources pre-stored in the knowledge graph module; the virtual simulation module is connected to the main control module by a circuit, and based on the pre-stored scenarios, combined with the operation sequence, operation time and scenario key points in the actual operation of the student, the student's operation is scored.

[0008] Preferably, the virtual simulation module includes:

[0009] A key point evaluation module, which selects key points for the pre-stored scenarios, evaluates the importance of the key points, and obtains importance data;

[0010] A time step evaluation module, which determines the completion time and operation sequence of each step of the student during the simulation according to the sensing devices set on the student's body, and conducts time evaluation based on the completion time and operation sequence to obtain time evaluation data;

[0011] A comprehensive evaluation module, which conducts a comprehensive evaluation based on the importance data and time evaluation data, combined with the error and omission evaluation data and step evaluation data in the simulation, to obtain the student's operation score.

[0012] Preferably, the key point evaluation module includes:

[0013] A significance calculation module, which collects the component nodes of the metering device as key points, marks the connections between every two adjacent components, calculates the effective performance parameters of each connection, calculates the connectivity of the key points based on the effective performance parameters, and further obtains the significance key degree;

[0014] A destructiveness calculation module, which establishes the initial parameters of the operation of the metering device, simulates key point failures, and calculates the destructiveness key degree of the key points;

[0015] A key point comprehensive calculation module, which calculates the importance data according to the significance key degree and destructiveness key degree of node k, and repeats the calculation to obtain the importance data of each node:

[0016]

[0017] where λ is an adjustable parameter, Kk is the importance data of node k, is the significance criticality, is the destructive criticality.

[0018] Preferably, the significance calculation module includes: collecting the component nodes of the metering device as key points, numbering the N component nodes, marking the connections between every two adjacent components, and calculating the effective performance parameters of each connection in the metering device:

[0019]

[0020] where D(a ij ) represents the degree of connection a ij , l ij represents the performance parameter corresponding to connection a ij , and β is an adjustable parameter;

[0021] Count the number of paths with the optimal effective performance parameters between component i and component j under parameter β Set the key point of the metering device to be evaluated as node k, and count the number of paths with the optimal effective performance parameters between component i and component j passing through node k Calculate the effective connectivity of node k, and the formula is as follows:

[0022]

[0023] Calculate the effective performance parameters of each node in the metering device, find the maximum value B max , and calculate the significance criticality of node k:

[0024]

[0025] Preferably, the destructive calculation module includes: establishing the initial parameters of the operation of the metering device, and the initial measurement error of connection a ij is denoted as x ij (0), and the allowable error range is denoted as U ij , simulate the failure of node k, make the performance of the connections and components related to node k fail, and redistribute the measurement errors on the failed components to the remaining components of the device. For the errors of the measured results, they are distributed proportionally according to the error-bearing capacity of the adjacent components; for the errors that have not produced measured results, they are redistributed according to the measurement principle and error transfer model of the device;

[0026] For each time step t after the error redistribution, judge whether the measurement error of each component satisfies x ij (t) > p * U ij; If satisfied, the component fails, is deleted from the device (marked as the failure state), the device model is updated, and the error is reallocated; if not satisfied, the measurement error of the component is updated; when no component fails, record the state at this time to obtain the destruction ability G of node k k , the formula is: where |A′| represents the number of remaining effective connections in the device model, and |A| represents the number of initial connections;

[0027] Calculate the destruction ability of each node in the device, and select the maximum value G max , and use the formula to calculate the destructive criticality of node k:

[0028]

[0029] The key point comprehensive calculation module calculates the importance data according to the significance criticality and destructive criticality of node k, and further obtains the importance data of each node.

[0030] Preferably, the time step evaluation module specifically includes:

[0031] Obtain the operation time data and each operation step of each step in the standard operation process, and obtain the covariance matrix according to the correlation coefficient between the operation time data and each operation step;

[0032] Based on the actual operation time data and the covariance matrix, generate operation time random numbers, and use the operation time random numbers to replace the actual distribution of the operation time data;

[0033] In the order of each operation step, add up the operation time of each step in turn to obtain N possible total operation times of this operation process;

[0034] Sort the obtained N total operation times in ascending order, set the standard total operation time T, count the number of total operation times less than or equal to T, set it as n, and the formula for calculating the time evaluation data is:

[0035] Preferably, the comprehensive evaluation module specifically includes:

[0036] The error and omission evaluation calculation module calculates the error and omission evaluation data based on the error operation deduction and the importance data;

[0037]

[0038] where, K i represents the importance weight of the i-th key point of the metering device, and α i ={0,1}, which measures the severity of the operation error of the i-th key point of the metering device. When α i =0, deduct Ki Q1 points; when α i = 1, deduct K i Q2 points. Q1 is the deduction for serious errors, and Q2 is the deduction for minor errors;

[0039] The step evaluation calculation module determines that the key operations of the standard measuring device are [K1, K2....K r , and the key operations of the trainee's actual measuring device are [K1′, K2′....K r ′];

[0040] For each operation step K j , if K j = K j ′, then the score for this step sequence is full marks. If K j ≠K j ′, the correct position of K j in the standard sequence is k, and |j - k| is the number of deviation steps. The score calculation formula for the key operation of this measuring device is: S order,j = 100 - |j - k|×Q3, where Q3 is the deviation deduction;

[0041] The total score of the step sequence is the weighted average of the scores of the key operations of each measuring device, specifically:

[0042] The total evaluation module calculates the operation score S of the trainee, specifically S = S time - S error + S order .

[0043] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a training system for metering device inspectors based on artificial intelligence, which generates a trainee portrait based on pre-stored questionnaire responses and deep learning algorithms, enabling the knowledge graph module to accurately push online learning and test resources for trainees accordingly. Different from the traditional "one-size-fits-all" training mode, this system can provide customized content according to the knowledge base, learning habits and skill levels of each trainee, greatly improving learning efficiency; for trainees with a better foundation, it provides more challenging case analysis and complex fault diagnosis learning materials; the virtual simulation module scores from multiple dimensions such as operation sequence, operation time and scene key points, and the evaluation method is comprehensive and objective. Among them, the key point evaluation module determines the importance of each node by calculating the significance key degree and the destructive key degree, ensuring that the key parts involved in the trainee's operation are considered key; the time step evaluation module generates random numbers and statistically analyzes them by combining the standard operation process and the actual operation time, making the time evaluation more scientific; the comprehensive evaluation module integrates multi-faceted data to calculate the operation score of the trainee, which can accurately reflect the operation level of the trainee and provide a strong basis for teaching improvement and trainee improvement. Brief Description of the Drawings

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0045] Figure 1 It is the system structure diagram provided by the present invention;

[0046] Figure 2 It is the structure diagram of the training system provided by the present invention;

[0047] Figure 3 It is the virtual simulation module diagram provided by the present invention;

[0048] Figure 4 It is the structure diagram of the personal center module provided by the present invention. Detailed Embodiments

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0050] An inspection personnel training system for metering devices based on artificial intelligence is disclosed in an embodiment of the present invention, as Figure 1 shown, including: a knowledge graph module, a virtual simulation module, a main control module, an AI teaching assistant module, and a personal center module. The personal center module generates a student portrait based on the responses to the pre-stored questionnaires and deep learning algorithms, and inputs the student portrait into the knowledge graph module through the main control module; the knowledge graph module pre-stores teaching and test resources in the form of a knowledge graph, and conducts online learning and tests based on the student portrait; the questions generated by the students in the knowledge graph module are sent to the AI teaching assistant module through the main control module for answering based on the resources pre-stored in the knowledge graph module; the virtual simulation module is connected to the main control module by a line, and scores the students' operations based on the pre-stored scenarios, combined with the operation sequence, operation time, and scenario key points in the students' actual operations.

[0051] In a specific embodiment, the virtual simulation module includes:

[0052] A key point evaluation module that selects key points for the pre-stored scenarios, evaluates the importance of the key points, and obtains importance data;

[0053] A time step evaluation module, based on a sensing device set on the trainee's body, determines the completion time and operation sequence of each step of the trainee during the simulation, conducts a time evaluation based on the completion time and operation sequence, and obtains time evaluation data;

[0054] A comprehensive evaluation module, based on the importance data and time evaluation data, combines the error and omission evaluation data and step evaluation data in the simulation to conduct a comprehensive evaluation and obtain the trainee's operation score.

[0055] In a specific embodiment, the sensing device set on the trainee's body is specifically:

[0056] Arms and hands: Wear an intelligent bracelet integrated with an IMU and a pressure sensor on the trainee's wrist. The IMU is used to monitor the movement posture and acceleration of the arm, judge actions such as the extension, bending, and rotation of the arm, so as to determine the start and end times of the operation. The pressure sensor can sense the grasping force and changes of the hand on the tool, further refining the judgment of the operation actions. Install small pressure sensors at the finger joints to capture the subtle movements of the fingers, such as operations like pressing buttons and rotating knobs, to ensure the accurate judgment of the operation sequence.

[0057] Key parts of the body: Install IMUs on the trainee's back and waist to monitor the overall posture changes of the body. For example, when performing some operations that require body movement or turning, the data from the IMUs on the back and waist can accurately judge the trainee's action sequence and completion time. At the same time, install pressure sensors at the bottom of the trainee's shoes to monitor the trainee's foot movements and moving directions, which is very crucial for judging the walking path and the time to reach a specific position of the trainee in the simulation scenario.

[0058] Operating tools and equipment: Install pressure sensors and micro IMUs at key operating parts such as the operating handle, buttons, and knobs of the measuring device. When the trainee operates these components, the pressure sensor can detect the operating force and contact time, and the IMU can monitor the rotation angle and acceleration changes of the components, accurately recording the details and time of the operation.

[0059] The data collected by the sensing device is sent to a local receiving terminal, such as a tablet computer or a laptop, through Bluetooth wireless transmission technology. Bluetooth transmission has the characteristics of low power consumption and stable short-distance transmission, and is suitable for use in simulation scenarios. In order to ensure the stability and real-time performance of data transmission, Bluetooth multi-connection technology is adopted to connect multiple sensing devices simultaneously, and the data is transmitted and verified in packets to ensure the integrity of the data. A special data processing algorithm is run on the receiving terminal. First, the collected IMU data is filtered to remove noise interference and improve the accuracy of the data. Then, through preset action models and threshold judgments, the IMU data is converted into specific operation actions, such as raising, lowering, and rotating the arm. For the data of the pressure sensor, according to the pressure change curve and the preset pressure threshold, the start and end of the operation are judged. Finally, combined with the Bluetooth positioning data, according to the position change of the trainee and the setting of the operation area, the operation sequence is determined, and the completion time of each operation step is calculated; the above settings of the sensing device on the trainee's body are all prior art.

[0060] In a specific embodiment, the key point evaluation module includes:

[0061] A significance calculation module, which collects the component nodes of the metering device as key points, marks the connections between every two adjacent components, calculates the effective performance parameters of each connection, calculates the connectivity of the key points based on the effective performance parameters, and further obtains the significance criticality;

[0062] A destructiveness calculation module, which establishes the initial parameters of the operation of the metering device, simulates the key point failures, and calculates the destructive criticality of the key points;

[0063] A key point comprehensive calculation module, which calculates the importance data according to the significance criticality and the destructive criticality of node k, and repeats the calculation to obtain the importance data of each node:

[0064]

[0065] where λ is an adjustable parameter, K k is the importance data of node k, is the significance criticality, is the destructive criticality.

[0066] In a specific embodiment, the significance calculation module includes: collecting the component nodes of the metering device as key points, numbering the N component nodes, marking the connections between every two adjacent components, and calculating the effective performance parameters of each connection in the metering device:

[0067]

[0068] where D(a ij ) represents connection aij Degree, l ij Indicates the connection a ij The corresponding performance parameter, where β is an adjustable parameter (which can be adjusted according to the device characteristics, generally taken as 1);

[0069] Count the number of paths with the optimal effective performance parameters between component i and component j under parameter β Set the key point of the metering device to be evaluated as node k, and count the number of paths with the optimal effective performance parameters between component i and component j passing through node k Calculate the effective connectivity of node k, and the formula is as follows:

[0070]

[0071] Calculate the effective performance parameters of each node in the metering device and find the maximum value B max , calculate the significance criticality of node k:

[0072]

[0073] In a specific embodiment, the destructive calculation module includes: establishing the initial parameters of the operation of the metering device, connecting a ij The initial measurement error of is denoted as x ij (0), and the allowable error range is denoted as U ij , simulate the failure of node k, make the connections and component performances related to node k fail, and redistribute the measurement errors on the failed components to the remaining components of the device. For the errors of the components that have generated measurement results, they are distributed proportionally according to the error-bearing capacity of adjacent components; for the errors that have not generated measurement results, they are redistributed according to the measurement principle and error transfer model of the device;

[0074] For each time step t after the error redistribution, determine whether the measurement error of each component satisfies x ij (t) > p * U ij ; if it is satisfied, the component fails, is deleted from the device, and the device model is updated (construct the device model with the connection points of the metering device components as nodes and the connection relationships between components as edges), and the error is redistributed; if it is not satisfied, update the component measurement error; when no component fails, record the state at this time to obtain the destruction ability G of node k k , and the formula is: Where |A′| represents the number of remaining effective connections in the device model, and |A| represents the number of initial connections;

[0075] Calculate the destruction ability of each node in the device and select the maximum value G max , use the formula to calculate the destructive criticality of node k:

[0076]

[0077] The key point comprehensive calculation module calculates importance data based on the significance key degree and destructiveness key degree of node k, and further obtains the importance data of each node.

[0078] In a specific embodiment, the time step evaluation module specifically includes:

[0079] Obtain the operation time data and each operation step of each step in the standard operation process, and obtain the covariance matrix according to the correlation coefficient between the operation time data and each operation step;

[0080] Based on the actual operation time data and the covariance matrix, generate operation time random numbers, and use the operation time random numbers to replace the actual distribution of the operation time data;

[0081] In the order of each operation step, add up the operation time of each step in turn to obtain N possible total operation times of this operation process;

[0082] Sort the obtained N total operation times in ascending order, set the standard total operation time T, count the number of total operation times less than or equal to T, set it as n, and the formula for calculating the time evaluation data is:

[0083] In a specific embodiment, the covariance matrix includes:

[0084]

[0085] Among them, Cov(·,·) represents the covariance of the operation time between operation steps; σ i represents the standard deviation of the operation time of operation step i; ρ represents the correlation coefficient of the operation time between operation steps.

[0086] In a specific embodiment, the comprehensive evaluation module specifically includes:

[0087] The error and omission evaluation calculation module calculates error and omission evaluation data based on the error operation deduction points and importance data;

[0088]

[0089] Among them, K i represents the importance degree weight of the key point of the i-th metering device, α i ={0,1}, which measures the severity of the operation error of the key point of the i-th metering device. When α i =0, deduct K i Q1 points; when α i =1, deduct K i Q2 points, Q1 is the deduction for serious errors, and Q2 is the deduction for minor errors;

[0090] Step evaluation calculation module, determining that the key operations of the standard measuring device are [K1, K2....K r , and the key operations of the trainee's actual measuring device are [K1′, K2′....K r ′];

[0091] For each operation step K j , if K j = K j ′, then the score for this step sequence is full marks. If K j ≠ K j ′, the correct position of K j in the standard sequence is k, |j - k| is the number of deviation steps, and the score calculation formula for the key operation of this measuring device is: S order,j = 100 - |j - k| × Q3, where Q3 is the deduction for deviation;

[0092] The total score of the step sequence is the weighted average of the scores of the key operations of each measuring device, specifically:

[0093] Total evaluation module, calculating the operation score S of the trainee, specifically S = S time - S error + S order .

[0094] In a specific embodiment, the present invention includes two major scenarios: online and offline. As Figure 2 shown, the online scenario includes a knowledge graph module, a virtual simulation module, an AI teaching assistant module, and a personal center module. Among them, the knowledge graph module and the virtual simulation module are associated. The knowledge graph module and the virtual simulation module are respectively embedded with the design concept of gamification, and trainees can select roles and conduct level-passing exercises. The offline scenario includes four major modules: on-site verification of electric energy meters, on-site inspection of metering voltage transformers, on-site inspection of metering current transformers, and detection of metering secondary circuits.

[0095] Trainees first conduct online scenario learning. When the online comprehensive score reaches 90 points or above, they can obtain the permission for offline practical operation. The calculation of the comprehensive score comes from the personal center module. The score composition includes the assessment score of theoretical learning in the knowledge graph module, the learning completion rate, the level-passing points in the virtual simulation module, and the situation of after-class extended learning. The scores of each part are set by the teacher with conversion algorithms and proportion weights, and the comprehensive score is automatically calculated using artificial intelligence.

[0096] Online scenario: Based on the metrology inspection textbooks and teaching syllabi, an ontology is constructed to extract knowledge point entities, entity attributes, and the relationships between entities. The extracted information is classified and imported into the Neo4j graph database for storage to achieve the visual display of knowledge. The knowledge entities in the metering device inspection training are associated with multi-modal teaching resource entities in a standardized and formalized manner. The associated resources include: (1) Linking teaching resources, such as linking the courseware, videos and other teaching resources developed by teachers to the knowledge graph; (2) Linking test and examination resources, classifying and summarizing theoretical and practical exercises according to the knowledge points they examine and linking them to the relevant knowledge entities for the convenience of students to search and practice; (3) Linking online teaching resources to expand the scope of knowledge search, improve the knowledge update speed, meet the learning needs of students in a wider range and at a deeper level, and link relevant resources such as books, textbooks, papers, standards, MOOCs, open courses, etc. to facilitate students' guided learning after class; (4) Linking safety warning education resources and typical scenario cases.

[0097] Based on the WebLearn virtual simulation teaching experiment platform and using the modeling tool AlgDesignerV3.0, the entire on-site test process of the transformer is abstracted into common key technologies such as unified virtual test components, underlying mathematical models, and logical mechanisms, forming an extensible virtual experiment construction library to provide background logical support operations for the virtual experiment platform. In the foreground, virtual reality technology is used to build a visual experimental scenario, experimental items, and experimental logic to achieve an integrated experimental environment that supports demonstration, interaction, calculation, and design. Restore the on-site real environment and equipment status 1:1, provide a safe learning space, and reduce safety risks.

[0098] In all aspects of the knowledge graph module and the virtual simulation module, gamified teaching thinking is incorporated. The knowledge points in the knowledge graph module are interrelated with the operation key points in the virtual simulation module. Requirements for the completion degree of knowledge points are set. Students can choose to perform virtual simulation operations after completing the corresponding knowledge points. Simulation training is not allowed if the corresponding knowledge points in the knowledge graph module have not been learned.

[0099] There are two modes in the virtual simulation module, namely the practice mode and the level-passing mode, such as Figure 3As shown, in the practice mode, there are hint and error correction functions. Students can practice according to the operation hints by themselves. The operation steps are as follows: The first step is role selection. Students can independently select the roles set in the system, such as the person in charge of the work, members of the work team, etc. The system can automatically match other work roles to form a work group that meets the requirements. Students can also freely form work groups by themselves; The second step is to select the virtual scene of on-site operation. There are four virtual scenes available for selection: on-site verification of power meters, on-site inspection of metering voltage transformers, on-site inspection of metering current transformers, and detection of metering secondary circuits; The third step is to perform virtual simulation operations. During the operation, students can view the hint window as needed, and the system will give real-time feedback and guidance according to the students' operations to help students correct errors in a timely manner. In the clearance mode, the system supports the function of independently generating test papers, conducts assessments and evaluations of knowledge in each link, generates test papers of different difficulties, conducts comprehensive knowledge assessments, assigns roles by teachers or the system automatically, and students independently complete operations in specific scenarios according to their roles. The standardized operation steps are decomposed into operable game elements, and each operation is designed as a level. Each level corresponds to a specific learning module. Students need to complete the learning and operation of the current module to unlock the next module and complete the clearance. The progressive learning method helps students master knowledge and skills more systematically.

[0100] The personal center module provides students' personal portraits and personalized learning path recommendations. For example, Figure 4 as shown, a questionnaire is designed on the login interface. Through the students' selections, the initial portraits of the students are obtained. Based on the students' learning situations, usage records, and behavioral habits, the students' portraits are adjusted on the basis of the initial portraits. The BERT pre-trained model is introduced into the Bi-LSTM+CRF model. Based on the model output, the students' interest entities are obtained, mapped into the vector space to obtain the interest entity vectors, and added to the students' portraits to generate dynamic portraits for the students. The personal portraits can be displayed in the form of graphs. Students can view their weak knowledge points and the mastery of previous and subsequent knowledge through the personal portraits, and have a clear understanding of their own learning situations. According to the students' dynamic portraits, the students' learning situations and interest entities are obtained. After inputting the learning situations, interest entity system relationship weight matrices, and knowledge point connection graphs into the KgRank algorithm, the algorithm gives the sorted paths. According to the sequential paths, the relevant learning resources of the corresponding knowledge points in the knowledge graph can be retrieved for precise push, providing students with personalized learning paths. After students carry out personalized learning, the system can dynamically adjust the students' portraits according to the students' learning behaviors, truly realizing individualized teaching according to students' aptitudes.

[0101] AI intelligent Q&A module based on knowledge graph. Students can ask questions at any time. Teachers have imported a large number of questions and standard answers in advance. The AI teaching assistant can also search for standard answers from the uploaded documents and the Internet and give the sources of the answers, which is convenient for students to conduct in-depth learning. The AI teaching assistant is continuously trained according to the questions of students, and a powerful question knowledge base is built through deep learning algorithms to answer various personalized questions of students.

[0102] In the offline scenario, there are on-site verification module for electric energy meters, on-site inspection module for metering voltage transformers, on-site inspection module for metering current transformers, and detection module for metering secondary circuits. Each module in the offline scenario corresponds one-to-one with the virtual scenario of the online virtual simulation module and is the real scene of the virtual scenario in the online virtual simulation module. Students can perform actual operations and truly experience the operation process.

[0103] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description in the method section.

[0104] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An artificial intelligence-based training system for metering device inspectors, characterized in that, It includes: A knowledge graph module, a virtual simulation module, a main control module, an AI teaching assistant module, and a personal center module. The personal center module generates a student portrait based on the responses to a pre-stored questionnaire and a deep learning algorithm, and inputs the student portrait into the knowledge graph module through the main control module. The knowledge graph module pre-stores teaching and test resources in the form of a knowledge graph, and conducts online learning and tests based on the student portrait. The questions generated by the student in the knowledge graph module are sent to the AI teaching assistant module through the main control module, and answers are provided based on the resources pre-stored in the knowledge graph module. The virtual simulation module is connected to the main control module by a circuit, and based on the pre-stored scenarios, the operation scores of the student are evaluated in combination with the operation sequence, operation time, and scenario key points in the student's actual operation.

2. The training system for metering device inspectors based on artificial intelligence according to claim 1, characterized in that The virtual simulation module includes: A key point evaluation module that selects key points for the pre-stored scenarios, evaluates the importance of the key points, and obtains importance data; A time step evaluation module that, according to the sensing devices set on the student's body, judges the completion time and operation sequence of each step during the simulation process, and conducts time evaluation based on the completion time and operation sequence to obtain time evaluation data; A comprehensive evaluation module that conducts a comprehensive evaluation based on the importance data and time evaluation data, in combination with the error and omission evaluation data and step evaluation data in the simulation, to obtain the operation scores of the student.

3. The training system for metering device inspectors based on artificial intelligence according to claim 2, wherein The key point evaluation module includes: A significance calculation module that collects the component nodes of the measurement device as key points, marks the connections between every two adjacent components, calculates the effective performance parameters of each connection, calculates the key point connectivity based on the effective performance parameters, and further obtains the significance key degree; A destructiveness calculation module that establishes the initial parameters of the operation of the measurement device, simulates key point failures, and calculates the destructiveness key degree of the key points; A key point comprehensive calculation module that calculates the importance data based on the significance key degree and destructiveness key degree of node k, and repeats the calculation to obtain the importance data of each node: where λ is an adjustable parameter, and K k is the importance data of node k, is the significance criticality, is the destructive criticality.

4. An inspection personnel training system for a metering device based on artificial intelligence according to claim 3, characterized in that, The significance calculation module includes: collecting the component nodes of the measurement device as key points, numbering the N component nodes, marking the connections between every two adjacent components, and calculating the effective performance parameters of each connection in the measurement device: Among them, D(a ij ) represents the degree of connection of a ij , l ij represents the performance parameter corresponding to the connection of a ij , and β is an adjustable parameter; Count the number of paths with the optimal effective performance parameters between component i and component j under parameter β Set the key points of the metering device to be evaluated as node k, and count the number of paths with the optimal effective performance parameters between component i and component j passing through node k Calculate the effective connectivity of node k, and the formula is as follows: Calculate the effective performance parameters of each node in the computing and metering device, and find the maximum value B max , calculate the significance criticality of node k:

5. The training system for metering device inspectors based on artificial intelligence according to claim 3, wherein, The destructive computing module includes: establishing the initial parameters of the metering device operation, connecting a ij The initial measurement error is denoted as x ij (0), and the allowable error range is denoted as U ij , simulating the failure of node k, causing the connections and component performances related to node k to fail, and redistributing the measurement errors on the failed components to the remaining components of the device. For the errors of the measurement results that have been generated, they are distributed proportionally according to the error-bearing capacity of the adjacent components; for the errors of the measurement results that have not been generated, they are redistributed according to the device measurement principle and the error transfer model; For each time step t after the reallocation of each error, determine whether the measurement error of each component satisfies x ij (t) > p * U ij ; if it is satisfied, the component fails, is deleted from the device, the device model is updated, and the error is reallocated; if it is not satisfied, the component measurement error is updated; when no component fails, record the state at this time to obtain the damage capacity G of node k k , and the formula is: where |A′| represents the number of remaining effective connections in the device model, and |A| represents the number of initial connections; Calculate the destruction ability of each node in the computing device and select the maximum value G max , and calculate the destructive criticality of node k using the formula: A key point comprehensive calculation module that calculates the importance data based on the significance key degree and destructiveness key degree of node k, and further obtains the importance data of each node.

6. The training system for metering device inspectors based on artificial intelligence according to claim 3, wherein, The time step evaluation module specifically includes: Obtaining the operation time data and each operation step of each step in the standard operation process, and obtaining the covariance matrix according to the correlation coefficient between the operation time data and each operation step; Generating operation time random numbers based on the actual operation time data and the covariance matrix, and using the operation time random numbers to replace the actual distribution of the operation time data; Adding the operation time of each step in sequence according to the order of each operation step to obtain N possible total operation times of this operation process; Sort the obtained N total operation times in ascending order, set the standard total operation time T, count the number of total operation times less than or equal to T, denoted as n, and the formula for calculating the time evaluation data is:

7. An inspection personnel training system for a metering device based on artificial intelligence according to claim 6, characterized in that, The comprehensive evaluation module specifically includes: An error and omission evaluation calculation module that calculates the error and omission evaluation data based on the deduction of points for incorrect operations and the importance data; Among them, K i represents the importance weight of the key point of the i-th metering device, and α i = {0, 1}, which measures the severity of the operation error of the key point of the i-th metering device. When α i = 0, K i Q1 points are deducted; when α i = 1, K i Q2 points are deducted. Q1 is the deduction for serious errors, and Q2 is the deduction for minor errors; The step evaluation calculation module determines that the key operations of the standard measuring equipment are [K1, K2....K r , and the key operations of the trainee's actual measuring equipment are [K1′, K2′....K r ′]; For each operation step K j , if K j = K j ', then the score for this step sequence is full marks. If K j ≠ K j ', and the correct position of K j in the standard sequence is k, |j - k| is the number of deviation steps. The calculation formula for the score of the key point operation of this measuring device is: S order,j = 100 - |j - k| × Q3, where Q3 is the deduction for deviation; The total score of the step sequence is the weighted average of the operation scores of the key points of each metering device, specifically: Total evaluation module, calculates the operation score S of the trainee, specifically S = S time - S error + S order .

Images