Electric power skill training and scoring evaluation method based on large model algorithm

Abstract

The invention discloses an electric power skill training and scoring evaluation method based on a large model algorithm, and belongs to the technical field of electric power skill training and scoring evaluation, and the method comprises the steps: collecting and processing student information data and historical operation behavior data, and integrating the processed data into a unified data warehouse for storage; analyzing the information data and the power skill post demand of each student by using a natural language processing technology to generate a demand analysis result; generating a training content scheme according to the demand analysis result based on a pre-selected large model; by analyzing learning behavior data, constructing a student learning style model, and further optimizing a training content scheme to obtain an optimal training scheme; a three-dimensional electric power working scene model is constructed by using computer graphics and VR / AR technologies; through establishing an operation evaluation model, real-time scoring evaluation is carried out on electric power skill operation of students, and an evaluation result of each student is obtained.

Description

Technical Field

[0001] This invention relates to the field of power skills training and assessment technology, specifically to a power skills training and assessment method based on a large model algorithm. Background Technology

[0002] In the power industry, power skills training and assessment are crucial for ensuring the safe and stable operation of the power system. Traditional power skills training and assessment methods typically employ fixed training courses and standardized assessment methods, which are insufficient to meet the personalized learning needs of trainees.

[0003] On the one hand, traditional training courses lack specificity and cannot be flexibly adjusted according to the trainees' knowledge base, learning ability, and the specific needs of different power skills positions. For example, trainees with weak foundations may not be able to keep up with the pace of learning according to a uniform course, while for trainees with strong abilities, their potential cannot be fully explored.

[0004] On the other hand, in the scoring and evaluation process, traditional methods rely heavily on manual observation and subjective evaluation, which has problems such as inconsistent scoring standards and the assessment results being greatly affected by human factors, making it difficult to objectively, comprehensively, and in real time evaluate trainees' electrical skills.

[0005] With the rapid development of the power industry and the increasing demands for intelligentization, existing power skills training and assessment technologies can no longer meet the industry's needs for cultivating power professionals. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a power skills training and assessment method based on large model algorithms to solve the problems mentioned in the background art.

[0007] The objective of this invention can be achieved through the following technical solution: a power skills training and assessment method based on a large model algorithm, comprising the following steps:

[0008] Step 1: Collect and process student information data and historical operation behavior data, and integrate the processed data into a unified data warehouse for storage;

[0009] Step 2: Analyze the information data of each trainee and the requirements of the power skills position using natural language processing technology to generate a requirements analysis result; based on the pre-selected large model, generate a training content plan according to the requirements analysis result;

[0010] Step 3: By analyzing learning behavior data, construct a learning style model for each learner to identify their learning style category, further optimize the training content plan, and obtain the optimal training plan.

[0011] Step 4: Using computer graphics and VR / AR technology, construct a 3D power work scene model for each trainee to practice power work.

[0012] Step 5: By establishing an operation evaluation model, the trainees' electrical skills operation is scored and evaluated in real time to obtain the evaluation results for each trainee.

[0013] Preferably, in step one, the student information data includes the student's age, education level, professional background, and learning behavior data;

[0014] Historical operational behavior data: By deploying sensors and video monitoring equipment during training practice, action data, equipment operation data, and video data of trainees operating electrical equipment are collected; action data includes operation sequence, operation force, and time interval; equipment operation data includes voltage, current, and power.

[0015] Preferably, in step one, processing the collected data includes:

[0016] The student information data is cleaned and normalized to remove outliers.

[0017] Calibrate historical operational data to eliminate data deviations caused by equipment errors;

[0018] Feature extraction is performed on historical operational behavior data to obtain key features, including:

[0019] Extracting operation step sequence features from operation sequence data;

[0020] Extract fluctuation and amplitude characteristics from the voltage and current of the equipment during operation;

[0021] For video and motion data, computer vision technology is used to extract the motion features of trainees, including using human pose estimation algorithms to obtain the position information of body joints during operation;

[0022] Based on the characteristics of the operation step sequence, fluctuation characteristics, amplitude characteristics, and action characteristics, the key features of historical operation behavior data are determined.

[0023] Preferably, step two includes:

[0024] Keyword extraction was performed on the professional background text of each student's information data using the TextRank algorithm;

[0025] A graph model of professional background text is constructed, with words in the professional background text as nodes in the graph. The graph model is constructed based on the co-occurrence relationship between words, and the weight of each node is calculated iteratively until the number of iterations reaches a preset standard.

[0026] Obtain the final calculated weights of all nodes, sort all nodes from largest to smallest according to their weights, and select the nodes that meet the preset rules as keywords; the preset rule is to select the top 20% of nodes sorted from largest to smallest as keywords.

[0027] Based on the knowledge system and skill requirements of the power industry, keywords are classified and mapped to obtain their corresponding knowledge domains;

[0028] Each student's score in each knowledge area is calculated by weighted summation and then summed to obtain each student's overall knowledge base score.

[0029] Preferably, step two further includes:

[0030] For the requirements of power skills positions, obtain the job description text, and calculate the importance score of each word in the job description text using the TF-IDF algorithm;

[0031] The calculation formula is:

[0032]

[0033]

[0034] In the formula, TF(t,D) represents the word frequency of each word t in the job description text D, and n t t is the number of times each word t appears in the job description text D, N is the sum of the number of occurrences of all words in the job description text D; IDF(t,D) represents the inverse document frequency, |D| is the size of the job description text set, |{d∈D:t∈d}| is the number of job description text sets containing word t; log is the logarithmic function;

[0035] Multiply the calculated term frequency and inverse document frequency to obtain the importance score of each word in the job description text;

[0036] Set a keyword threshold, compare the importance score of each word with the keyword threshold, and obtain words with an importance score greater than or equal to the keyword threshold to determine the demand information, including key knowledge and skills requirements for power skills positions;

[0037] The results of the needs analysis consist of each student's comprehensive knowledge base score and needs information.

[0038] Preferably, the process of generating a training content plan based on a pre-selected large model and the results of the needs analysis includes:

[0039] The process involves acquiring a pre-selected large model, encoding the demand information from the demand analysis results, determining input prompts based on the encoded demand information and each trainee's comprehensive knowledge base score, and inputting these prompts into the large model. The process also involves searching for corresponding related knowledge in the pre-trained knowledge space of the large model, using a greedy search algorithm or a bundle search algorithm to select the optimal word sequence, and generating a training content plan.

[0040] Preferably, the process of constructing the learner learning style model in step three includes:

[0041] Based on each student's learning behavior data, clustering analysis is performed using a clustering algorithm; the learning behavior data of all students are denoted as a feature vector set {x1, x2, ..., x...}. M}, M is the total number of data points; using the K-Means clustering algorithm, k cluster centers are iteratively updated to minimize the sum of the distances from each data point to its respective cluster center, thus obtaining the learner's learning style model;

[0042] The distance calculation formula is as follows:

[0043]

[0044] In the formula, d(x) i ,c j ) represents data point x i and cluster center c j The distance; i = 1, 2, ..., M, where M is the total number of data points; j = 1, 2, ..., k, where k is the total number of cluster centers; and These are data points x i and cluster center c j The m-th feature value; m = 1, 2, ..., p, where p is the feature dimension.

[0045] Preferably, step five includes:

[0046] The system uses sensors and computer vision technology to monitor each student's actions in the virtual environment and identifies each student's real-time action data.

[0047] Perform the same feature extraction operation as in step one on the real-time operation behavior data to obtain real-time key features, and denot them as action sequence vector A;

[0048] Based on the optimal training plan and knowledge data in the power sector, obtain the corresponding preset standard operating actions and convert the standard operating actions into an action sequence vector S.

[0049] The action sequence vector A and the action sequence vector S are compared and analyzed, and the similarity between the two is calculated using the dynamic time warping algorithm.

[0050] Define a distance matrix, calculate the distance between each feature vector in action sequence vector A and action sequence vector S using the Euclidean distance formula, and calculate the cumulative distance of the regularized path W using dynamic programming to minimize the sum of distances on the path. Normalize the cumulative distance of the regularized path W to obtain the similarity between action sequence vector A and action sequence vector S.

[0051] Preferably, step five also includes: obtaining the knowledge level test scores of trainees after training, the operation time of the power skills simulation test, and the similarity between the extracted real-time key features and standard operation actions, calculating the comprehensive score of each trainee, and realizing the scoring and evaluation of each trainee;

[0052] The calculation formula is as follows:

[0053] DF=α1·CJ+α2·T+α3·SD;

[0054] In the formula, DF is the trainee's overall score, CJ is the knowledge level test score after training, T is the operation time of the power skills simulation test, SD is the similarity between the extracted real-time key features and the standard operation actions, and α1, α2, and α3 are the corresponding weight coefficients, and the sum of the three values ​​is 1.

[0055] Compared to existing solutions, the beneficial effects achieved by this invention are:

[0056] This invention collects and processes student information data and historical operational behavior data, and uses natural language processing technology to analyze student situations and job requirements. Based on a large model, it generates training content plans and optimizes them by constructing a student learning style model based on learning behavior data, thereby achieving a high degree of personalization of training content.

[0057] This invention utilizes computer graphics and VR / AR technology to construct a three-dimensional power work scene model, creating an immersive learning environment for trainees. Trainees can conduct skills training in a near-realistic scenario, enhancing their operational capabilities in actual work scenarios and significantly improving training effectiveness.

[0058] This invention establishes an operational evaluation model to conduct real-time scoring and assessment of trainees' electrical skills operations, changing the traditional method that relies on subjective human evaluation. Based on objective data and algorithm rules, this model can achieve the standardization of scoring criteria and the automation of the evaluation process, ensuring that the evaluation results are objective, fair, and accurate, and providing a reliable basis for assessing trainees' skill levels. Attached Figure Description

[0059] The invention will now be further described with reference to the accompanying drawings.

[0060] Figure 1This is a flowchart of the power skills training and assessment method based on large model algorithm proposed in this invention. Detailed Implementation

[0061] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0062] like Figure 1 As shown, this invention is a power skills training and assessment method based on a large model algorithm, including:

[0063] Step 1: Collect and process student information data and historical operation behavior data, and integrate the processed data into a unified data warehouse for storage;

[0064] The student information data includes the student's age, education, professional background, and learning behavior data, including but not limited to the number of training sessions, online learning time, and frequency of questions asked in the past learning process.

[0065] Historical operational behavior data: This data is collected by deploying sensors and video monitoring equipment during training practice to gather action data, equipment operation data, and video data of the operation process from trainees operating electrical equipment. Action data includes operation sequence, force, and time interval; equipment operation data includes voltage, current, and power.

[0066] It should be further explained that the process of processing the collected data includes:

[0067] The student information data is cleaned and normalized to remove outliers; outliers are data whose values ​​exceed a reasonable range.

[0068] Calibrate historical operational data to eliminate data deviations caused by equipment errors;

[0069] For example, the operating force data is corrected according to the calibration parameters of the equipment; wherein, the standard parameters of the equipment can explain the accuracy characteristics of the equipment in different measurement ranges and the deviation relationship between the equipment and the ideal measurement value. The operating force data is corrected by the standard parameters of the equipment to obtain the correct operating force data.

[0070] Feature extraction is performed on historical operational behavior data to obtain key features, including:

[0071] Extracting operation step sequence features from operation sequence data;

[0072] Extract fluctuation and amplitude characteristics from the voltage and current of the equipment during operation;

[0073] For video and motion data, computer vision technology is used to extract the motion features of trainees, including using human pose estimation algorithms to obtain the position information of body joints during operation;

[0074] Based on the characteristics of the operation step sequence, fluctuation characteristics, amplitude characteristics, and action characteristics, the key characteristics of historical operation behavior data are determined.

[0075] In the implementation steps of this invention, by collecting trainee information data, operational behavior data, and fault case data, and processing and storing the collected data, a precise data foundation is provided for subsequently developing the best training plan for each trainee.

[0076] Step 2: Analyze the information data of each trainee and the requirements of the power skills position using natural language processing technology to generate a requirements analysis result; based on the pre-selected large model, generate a training content plan according to the requirements analysis result;

[0077] It should be further explained that the process of analyzing each trainee's information data and the requirements of the power skills position using natural language processing technology includes:

[0078] Keyword extraction was performed on the professional background text of each student's information data using the TextRank algorithm;

[0079] A graph model of the professional background text is constructed, with words in the professional background text as nodes in the graph. The graph model is constructed based on the co-occurrence relationship between words, and the weight of each node is calculated iteratively until the number of iterations reaches a preset standard. Among them, the TextRank algorithm is a link analysis algorithm used to evaluate the importance of nodes. Its calculation formula will not be elaborated here.

[0080] Obtain the final calculated weights of all nodes, sort all nodes from largest to smallest according to their weights, and select the nodes that meet the preset rules as keywords; the preset rule is to select the top 20% of nodes sorted from largest to smallest as keywords.

[0081] Based on the knowledge system and skill requirements of the power industry, keywords are classified and mapped to obtain their corresponding knowledge domains;

[0082] The scores of each student in each knowledge area are calculated by weighted summation and then summed to obtain the comprehensive knowledge base score of each student.

[0083] It should be noted that a higher overall knowledge base score indicates a better knowledge base for the learner.

[0084] Furthermore, for the requirements of power skills positions, the job description text is obtained, and the importance score of each word in the job description text is calculated using the TF-IDF (Term Frequency-Inverse Document Frequency) algorithm;

[0085] The calculation formula is:

[0086]

[0087] In the formula, TF(t,D) represents the word frequency of each word t in the job description text D, and n t t is the number of times each word t appears in the job description text D, N is the sum of the number of occurrences of all words in the job description text D; IDF(t,D) represents the inverse document frequency, |D| is the size of the job description text set, |{d∈D:t∈d}| is the number of job description text sets containing word t; log is the logarithmic function;

[0088] Multiply the calculated term frequency and inverse document frequency to obtain the importance score of each word in the job description text;

[0089] Set keyword thresholds, where the specific values ​​of the keyword thresholds are determined by those skilled in the art based on big data experience;

[0090] The importance score of each word is compared with the keyword threshold, and words with an importance score greater than or equal to the keyword threshold are obtained to determine the demand information, including key knowledge and skills requirements for power skills positions.

[0091] The comprehensive knowledge base score and needs information of each student constitute the needs analysis results.

[0092] It should be further explained that the process of generating training content solutions based on the pre-selected large model and the results of the needs analysis includes:

[0093] The process involves acquiring a pre-selected large model, encoding the demand information from the demand analysis results, determining input prompts based on the encoded demand information and each trainee's comprehensive knowledge base score, and inputting them into the large model; searching for corresponding related knowledge in the pre-trained knowledge space of the large model, and using a greedy search algorithm or a bundle search algorithm to select the optimal word sequence, and generating a training content plan.

[0094] The training content includes, but is not limited to, theoretical knowledge explanations and watching videos explaining relevant theoretical knowledge; virtual operation exercises, allowing trainees to practice in a simulated environment and familiarize themselves with the operation process; and practical operation guidance, providing specific operation suggestions and guidance for actual work scenarios to improve trainees' practical skills.

[0095] It should be explained that the pre-selected large model can be GPT-4, which is a Decoder-only large model developed by OpenAI. The specific pre-training process will not be elaborated on here.

[0096] Step 3: By analyzing learning behavior data, construct a learning style model for each learner to identify their learning style category, further optimize the training content plan, and obtain the optimal training plan.

[0097] It should be further explained that the process of constructing the learner learning style model includes:

[0098] Based on each student's learning behavior data, clustering analysis is performed using a clustering algorithm; the learning behavior data of all students are denoted as a feature vector set {x1, x2, ..., x...}. M}, M is the total number of data points; using the K-Means clustering algorithm, k cluster centers are iteratively updated to minimize the sum of the distances from each data point to its respective cluster center, thus obtaining the learner's learning style model;

[0099] The distance calculation formula is as follows:

[0100]

[0101] In the formula, d(x) i ,c j ) represents data point x i and cluster center c j The distance; i = 1, 2, ..., M, where M is the total number of data points; j = 1, 2, ..., k, where k is the total number of cluster centers; and These are data points x i and cluster center c j The m-th feature value; m = 1, 2, ..., p, where p is the feature dimension;

[0102] It should be noted that each data point represents the learning behavior data of each student. The feature value of the data point is the learning behavior data, including but not limited to the number of training sessions, online learning time, and question frequency in the student's past learning process; the cluster center represents different learning style categories.

[0103] The learning style categories include, but are not limited to, visual learning style, auditory learning style, and hands-on learning style; for example, if a student spends a lot of time learning online, it may be a visual learning style.

[0104] Furthermore, the training content plan is adjusted according to the learning style of the trainees; for example, for visual learning style, priority is given to learning theoretical knowledge and watching relevant theoretical knowledge explanation videos.

[0105] Step 4: Using computer graphics and VR / AR technology, construct a 3D power work scene model for each trainee to practice power work.

[0106] It should be further explained that the process of constructing a three-dimensional power working scene model using computer graphics and VR / AR technology includes:

[0107] The physical characteristics of power equipment operation are simulated based on a physics engine, including but not limited to equipment movement, collisions, and electrical connections, all of which are modeled and simulated according to actual physical laws and electrical principles. In the VR scene, computer vision technology is used to extract the real-time motion features of the learner, enabling natural interaction between the user and the virtual environment. VR / AR technology refers to virtual reality / augmented reality technology.

[0108] Step 5: By establishing an operational evaluation model, the trainees' electrical skills operation is scored and evaluated in real time to obtain the evaluation results for each trainee;

[0109] It should be further explained that the process of establishing an operational evaluation model to conduct real-time scoring and assessment of trainees' electrical skills operations includes:

[0110] The system uses sensors and computer vision technology to monitor each student's actions in the virtual environment and identifies each student's real-time action data.

[0111] Perform the same feature extraction operation as in step one on the real-time operation behavior data to obtain real-time key features, and denot them as action sequence vector A;

[0112] Based on the optimal training plan and knowledge data in the power sector, obtain the corresponding preset standard operating actions and convert the standard operating actions into an action sequence vector S.

[0113] The action sequence vector A and the action sequence vector S are compared and analyzed, and the similarity between the two is calculated using the dynamic time warping algorithm.

[0114] Define a distance matrix, calculate the distance between each feature vector in action sequence vector A and action sequence vector S using the Euclidean distance formula, and calculate the cumulative distance of the regularized path W using dynamic programming to minimize the sum of distances on the path. Normalize the cumulative distance of the regularized path W to obtain the similarity between action sequence vector A and action sequence vector S. The cumulative distance of the regularized path W can be calculated programmatically using the dynamic time warping algorithm, which is an existing technology and will not be elaborated on here.

[0115] The system obtains the knowledge level test scores of trainees after training, the operation time of the power skills simulation test, and the similarity between the extracted real-time key features and standard operation actions. It then calculates the comprehensive score of each trainee to achieve the evaluation and assessment of each trainee.

[0116] The calculation formula is as follows:

[0117] DF=α1·CJ+α2·T+α3·SD;

[0118] In the formula, DF is the trainee's comprehensive score, CJ is the knowledge level test score after training, T is the operation time of the power skills simulation test, SD is the similarity between the extracted real-time key features and the standard operation actions, α1, α2, and α3 are the corresponding weight coefficients, and the sum of the three values ​​is 1. The specific values ​​are obtained by those skilled in the art through big data calculation.

[0119] The overall score of the trainees is the evaluation result.

[0120] In the several embodiments provided by this invention, it should be understood that the disclosed system can be implemented in other ways. For example, the embodiments of the invention described above are merely illustrative; for example, the division of modules is only a logical functional division, and there may be other division methods in actual implementation.

[0121] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0122] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0123] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0124] Finally, 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.

Claims

1. A power skills training and assessment method based on a large model algorithm, characterized in that, include: Step 1: Collect and process student information data and historical operation behavior data, and integrate the processed data into a unified data warehouse for storage; Step 2: Analyze the information data of each trainee and the requirements of the power skills position using natural language processing technology to generate a requirements analysis result; based on the pre-selected large model, generate a training content plan according to the requirements analysis result; Step 3: By analyzing learning behavior data, construct a learning style model for each learner to identify their learning style category, further optimize the training content plan, and obtain the optimal training plan. Step 4: Using computer graphics and VR / AR technology, construct a 3D power work scene model for each trainee to practice power work. Step 5: By establishing an operation evaluation model, the trainees' electrical skills operation is scored and evaluated in real time to obtain the evaluation results for each trainee.

2. The power skills training and assessment method based on large model algorithm according to claim 1, characterized in that, In step one, the student information data includes the student's age, education level, professional background, and learning behavior data; Historical operational behavior data: Data on trainees' actions, equipment operation, and video recordings of the operation process are collected by deploying sensors and video monitoring equipment during training practice. Action data includes operation sequence, operation force, and time interval; equipment operation data includes voltage, current, and power.

3. The power skills training and assessment method based on large model algorithm according to claim 2, characterized in that, Step one involves processing the collected data, including: The student information data is cleaned and normalized to remove outliers. Calibrate historical operational data to eliminate data deviations caused by equipment errors; Feature extraction is performed on historical operational behavior data to obtain key features, including: Extracting operation step sequence features from operation sequence data; Extract fluctuation and amplitude characteristics from the voltage and current of the equipment during operation; For video and motion data, computer vision technology is used to extract the motion features of trainees, including using human pose estimation algorithms to obtain the position information of body joints during operation; Based on the characteristics of the operation step sequence, fluctuation characteristics, amplitude characteristics, and action characteristics, the key features of historical operation behavior data are determined.

4. The power skills training and assessment method based on large model algorithm according to claim 3, characterized in that, Step two includes: Keyword extraction was performed on the professional background text of each student's information data using the TextRank algorithm; A graph model of professional background text is constructed, with words in the professional background text as nodes in the graph. The graph model is constructed based on the co-occurrence relationship between words, and the weight of each node is calculated iteratively until the number of iterations reaches a preset standard. Obtain the final calculated weights of all nodes, sort all nodes from largest to smallest according to their weights, and select the nodes that meet the preset rules as keywords; the preset rule is to select the top 20% of nodes sorted from largest to smallest as keywords. Based on the knowledge system and skill requirements of the power industry, keywords are classified and mapped to obtain their corresponding knowledge domains; Each student's score in each knowledge area is calculated by weighted summation and then summed to obtain each student's overall knowledge base score.

5. The power skills training and assessment method based on large model algorithm according to claim 4, characterized in that, Step two also includes: For the requirements of power skills positions, obtain the job description text and calculate the importance score of each word in the job description text using the TF-IDF algorithm; The calculation formula is: In the formula, TF(t,D) represents the word frequency of each word t in the job description text D, and n t t is the number of times each word t appears in the job description text D, N is the sum of the number of occurrences of all words in the job description text D; IDF(t,D) represents the inverse document frequency, |D| is the size of the job description text set, |{d∈D:t∈d}| is the number of job description text sets containing word t; log is the logarithmic function; Multiply the calculated term frequency and inverse document frequency to obtain the importance score of each word in the job description text; Set a keyword threshold, compare the importance score of each word with the keyword threshold, and obtain words with an importance score greater than or equal to the keyword threshold to determine the demand information, including key knowledge and skills requirements for power skills positions; The results of the needs analysis consist of each student's comprehensive knowledge base score and needs information.

6. The power skills training and assessment method based on large model algorithm according to claim 5, characterized in that, The process of generating training content plans based on pre-selected large models and the results of needs analysis includes: The process involves acquiring a pre-selected large model, encoding the demand information from the demand analysis results, determining input prompts based on the encoded demand information and each trainee's comprehensive knowledge base score, and inputting these prompts into the large model. The process also involves searching for corresponding related knowledge in the pre-trained knowledge space of the large model, using a greedy search algorithm or a bundle search algorithm to select the optimal word sequence, and generating a training content plan.

7. The power skills training and assessment method based on large model algorithm according to claim 6, characterized in that, Step three involves constructing a learner learning style model, which includes: Based on each student's learning behavior data, clustering analysis is performed using a clustering algorithm; the learning behavior data of all students are denoted as a feature vector set {x1, x2, ..., x...}. M }, M is the total number of data points; using the K-Means clustering algorithm, k cluster centers are iteratively updated to minimize the sum of the distances from each data point to its respective cluster center, thus obtaining the learner's learning style model; The distance calculation formula is as follows: In the formula, d(x) i ,c j ) represents data point x i and cluster center c j The distance; i = 1, 2, ..., M, where M is the total number of data points; j = 1, 2, ..., k, where k is the total number of cluster centers; and These are data points x i and cluster center c j The m-th feature value; m = 1, 2, ..., p, where p is the feature dimension.

8. The power skills training and assessment method based on large model algorithm according to claim 7, characterized in that, Step five includes: The system uses sensors and computer vision technology to monitor each student's actions in the virtual environment and identifies each student's real-time action data. Perform the same feature extraction operation as in step one on the real-time operation behavior data to obtain real-time key features, and denot them as action sequence vector A; Based on the optimal training plan and knowledge data in the power sector, obtain the corresponding preset standard operating actions and convert the standard operating actions into an action sequence vector S. The action sequence vector A and the action sequence vector S are compared and analyzed, and the similarity between the two is calculated using the dynamic time warping algorithm. Define a distance matrix, calculate the distance between each feature vector in action sequence vector A and action sequence vector S using the Euclidean distance formula, and calculate the cumulative distance of the regularized path W using dynamic programming to minimize the sum of distances on the path. Normalize the cumulative distance of the regularized path W to obtain the similarity between action sequence vector A and action sequence vector S.

9. The power skills training and assessment method based on large model algorithm according to claim 8, characterized in that, Step five also includes: obtaining the knowledge level test scores of trainees after training, the operation time of the power skills simulation test, and the similarity between the extracted real-time key features and standard operation actions, calculating the comprehensive score of each trainee, and realizing the scoring and evaluation of each trainee; The calculation formula is as follows: DF=α1·CJ+α2·T+α3·SD; In the formula, DF is the trainee's overall score, CJ is the knowledge level test score after training, T is the operation time of the power skills simulation test, SD is the similarity between the extracted real-time key features and the standard operation actions, and α1, α2, and α3 are the corresponding weight coefficients, and the sum of the three values ​​is 1.

Images