Employee ability mapping system and method based on AI and digital twinning

By using an AI- and digital twin-based employee competency mapping system, the problems of data errors, assessment delays, and training disconnects in the power industry's employee competency assessment have been solved. This system enables efficient integration and real-time assessment of multi-source data, improving the accuracy of assessments and the practicality of training.

CN120931160APending Publication Date: 2025-11-11STATE GRID BEIJING ELECTRIC POWER CO +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202511230119.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-08-29
Publication Date
2025-11-11

AI Technical Summary

Technical Problem

The existing employee competency assessment in the power industry suffers from problems such as large data errors, assessment lag, high costs of cross-system integration, lack of objective data support for soft skills, disconnect between assessment and training, and slow assessment feedback, which cannot meet the needs of real-time assessment and training.

Method used

An AI- and digital twin-based employee competency mapping system is adopted, including a data acquisition and preprocessing module, a multimodal AI analysis module, a digital twin model construction and application module, a competency assessment and feedback module, and a user interface module. Through multi-source data acquisition, multimodal AI analysis, digital twin model construction and application, and edge-cloud collaborative computing, a closed loop of real-time assessment and training is achieved.

Benefits of technology

It enables efficient fusion and querying of multi-source data, reduces data error in assessment, dynamically adapts assessment strategies, improves the matching degree between training and actual business, and enhances the real-time performance and accuracy of employee skills assessment and training.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120931160A_ABST
    Figure CN120931160A_ABST
Patent Text Reader

Abstract

The invention discloses an employee ability mapping system based on AI and digital twinning, which relates to the technical field of employee ability mapping and comprises a data acquisition and preprocessing module, a multi-modal AI analysis module, a digital twinning model construction and application module, an ability evaluation and feedback module and a user interaction interface module. The data acquisition and preprocessing module comprises a multi-source data acquisition unit which is used for acquiring data in real time from a plurality of heterogeneous business systems and field sensors of a power enterprise; the data quality management and control unit is internally provided with an integrity verification rule base, a rationality verification rule base and a consistency verification rule base, and adopts a machine learning model to predict and correct a data quality problem in advance; the multi-source data fusion engine realizes unified acquisition and association of data of a plurality of business systems, so that the data integration efficiency is improved, and the cross-department data query response time is shortened; the data quality management and control mechanism improves the qualified rate of data, reduces the error rate of evaluation data, and avoids evaluation deviation caused by data problems.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of employee competency mapping technology, specifically an employee competency mapping system and method based on AI and digital twins. Background Technology

[0002] With the accelerated construction of new power systems, the power industry's demand for young employees' "cross-business collaboration ability, new technology adaptability, and emergency response ability" has significantly increased. However, existing employee competency assessment and management techniques have the following core shortcomings: 1. The electricity metering data collected by outdated monitoring equipment has an error of ≥10%, which cannot accurately reflect the employees' electricity metering skills; the data update is lagging, resulting in a deviation rate of over 20% in the assessment of employees' ability to cope with new working conditions. 2. High cost and low efficiency of data integration: Different systems use different databases such as Oracle and MySQL, and communication protocols include Modbus, MQTT, etc. Cross-system data integration requires customized development of interfaces, with a single interface development cost exceeding 50,000 yuan. Moreover, the data query response time after integration is ≥10 seconds, which cannot support real-time evaluation. 3. The training scenarios generated by the digital twin model are inconsistent with the safety standards in actual work, resulting in differences in actual operation after employee training; the model analysis results cannot be synchronized to the human resources system, leading to a disconnect between assessment and talent development decisions. 4. Soft skills such as innovation and communication skills rely on subjective evaluations from superiors without objective data support. For example, the evaluation of employees' communication skills is based solely on qualitative descriptions of whether they actively communicate, without quantifying key indicators such as the efficiency of cross-departmental communication in solving problems and the accuracy of information transmission. 5. Traditional assessments rely on quarterly manual evaluations, which take at least one month from data collection to feedback, making it difficult to identify employees' skill gaps in a timely manner. Summary of the Invention

[0003] The purpose of this invention is to provide an employee capability mapping system and method based on AI and digital twins to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: An employee competency mapping system based on AI and digital twins includes: a data acquisition and preprocessing module, a multimodal AI analysis module, a digital twin model construction and application module, a competency assessment and feedback module, and a user interface module. Data acquisition and preprocessing module: Multi-source data acquisition unit: used to acquire data in real time from multiple heterogeneous business systems and field sensors of power companies; Data quality control unit: It has a built-in rule base for integrity verification, rationality verification, and consistency verification, and uses machine learning models to predict and correct data quality problems in advance; Data storage unit: performs data cleaning, transformation, and standardized storage; Multimodal AI analysis module: Multimodal fusion framework: Based on the multimodal Transformer model, it integrates voice, text, image, and device status data, and learns cross-modal relationships through a self-attention mechanism; Industry-adapted pre-trained model: Based on ERNIE 3.0, incorporating industry-standard knowledge; Dynamic evaluation model; based on reinforcement learning, it uses employees' real-time work performance as state input, evaluation strategy adjustment as action output, and ability improvement effect as reward feedback to achieve dynamic iteration of evaluation strategy; Digital twin model construction and application module: Includes a high-precision equipment modeling unit: using a multi-physics coupling method, a digital twin model integrating electrochemical, thermal, and electrical characteristics is constructed for power equipment, and parameters are dynamically calibrated through real-time sensor data and historical operating data; Edge-cloud collaborative computing unit: Decentralizes data preprocessing and local state analysis tasks to the substation edge terminal, and only transmits key feature data to the cloud; Virtual-Real Collaboration Platform: Seamlessly integrates digital twin models with production management and human resource systems to generate personalized virtual training scenarios, provide real-time feedback on employee operational effectiveness, and synchronize training data to employee skill profiles; Capability assessment and feedback module: includes a comprehensive indicator system, quantitative assessment unit, and real-time feedback unit; User interface module: including employee and management sides.

[0005] As a further aspect of the present invention, the data acquisition and preprocessing module also includes a knowledge graph driving unit: constructing a knowledge graph for the power industry and realizing cross-system data association queries based on the Neo4j graph database.

[0006] As a further aspect of the present invention, the dynamic evaluation model of the multimodal AI analysis module also includes a key event triggering mechanism: when an employee participates in a new training program, obtains skills certification, or experiences a performance fluctuation of ≥20%, the evaluation strategy adjustment is triggered immediately without waiting for periodic iterations.

[0007] As a further aspect of the present invention: the high-precision equipment modeling unit of the digital twin model construction and application module can simulate the charging and discharging efficiency and life decay curve for the digital twin model of the new energy storage device, and support the evaluation of employee operation effects under different working conditions.

[0008] As a further aspect of the present invention: the full-dimensional indicator system includes primary indicators, secondary indicators, and tertiary indicators; The quantitative evaluation unit uses the entropy method to calculate the index weights and conducts quantitative evaluation through multi-source data. The real-time feedback unit generates an assessment report, displays employees' strengths and weaknesses using a skills radar chart, and pushes personalized training suggestions.

[0009] As a further aspect of the present invention: the employee terminal is used to display competency assessment results, conduct comparative analysis with other employees in the same position, plan career development paths, and support training course registration and online consultation; The management interface provides a visual representation of the team's capability distribution and capability gap analysis, as well as decision support functions for talent selection and development.

[0010] An employee capability mapping method based on AI and digital twins, the method comprising: S1. Real-time acquisition of data from 12 business systems and on-site sensor data through Apache Flink processing stream; S2. Clean the data using rules of integrity, rationality, and consistency; correct outliers using machine learning models; and standardize the format. S3. Store standardized data in a Greenplum data warehouse by partitioning it according to "employee ID + time" and create indexes to optimize query efficiency; S4. Construct a multimodal Transformer fusion framework to convert employee voice commands into text, input them into the model along with device image features and real-time status data, and generate fused feature vectors through a self-attention mechanism. S5. Employ industry-adaptive pre-trained models to analyze fusion features and identify the accuracy of employee operational intentions and professional expressions. S6. Based on a reinforcement learning dynamic evaluation model, the evaluation weights are adjusted using the intent recognition results and the employee's historical performance as inputs. S7. For novel energy storage devices, a multi-physics coupled model is constructed by integrating electrochemical, thermal, and electrical equations, and parameters are calibrated using real-time sensor data to ensure that the deviation between the model and the actual state is ≤5%. S8. Through the edge-cloud collaborative architecture, the edge terminal preprocesses the on-site data, and the cloud simulates the employee operation effect based on the digital twin model; S9. A virtual training scenario is generated on the virtual-real collaboration platform, and employee operation data is fed back to the digital twin model in real time to evaluate the standardization and effectiveness of operations. S10. Based on the three-level indicator system, the weight of each indicator is calculated using the entropy method, and the score is quantified through multi-source data. S11 generates hourly assessment reports, displays employee skill distribution using radar charts, and pushes personalized training suggestions. S12. The management side provides a visual display of team capability distribution, supporting talent selection and training decisions. Employees can query capability trends and career development paths. S13. Collect new business data monthly, incrementally train multimodal AI models and digital twin models, and update evaluation indicator weights; S14. Collect feedback from employees and managers quarterly to optimize the user interface and evaluation rules, ensuring that the system adapts to technological changes in the power industry and changes in the company's talent needs.

[0011] As a further aspect of the present invention: the fine-tuning data of the industry-adaptive pre-trained model in step S5 includes the company's equipment operation manuals, fault handling reports, and technical specification documents from the past 5 years. The training process uses the Adam optimizer with a learning rate set to 0.001 and iterates until the loss function converges.

[0012] As a further aspect of the present invention: in step S7, the parameter calibration of the digital twin model of the novel energy storage device adopts the particle swarm optimization algorithm, with the "mean square error between the model prediction value and the sensor measured value" as the objective function, and the calibration is stopped after the error is ≤5%.

[0013] As a further aspect of the present invention: In step S10, the quantitative formula for the third-level indicator "number of technical improvement suggestions" is: score Σ = suggestion innovation score * feasibility score, where innovation is evaluated by technical experts on a scale of 1-5, and feasibility is adjusted based on the project implementation rate.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. The multi-source data fusion engine enables unified collection and association of data from multiple business systems, improving data integration efficiency and shortening cross-departmental data query response time; the data quality control mechanism improves the data qualification rate, reduces the evaluation data error rate, and avoids evaluation bias caused by data problems. 2. The three-level indicator system covers professional skills, safety awareness, and innovation capabilities, which is an improvement over the traditional single-skill assessment dimension and comprehensively identifies employees' potential and shortcomings. 3. The enhanced learning dynamic assessment model shortens the assessment strategy update cycle, adapting to power technology changes and employee growth; 4. The edge-cloud collaborative architecture shortens the time for single-round strategy iterations and meets the needs of real-time evaluation; the virtual-physical collaborative platform improves the matching degree between training and actual business, and reduces the error rate of actual operation after employee training. Attached Figure Description

[0015] Figure 1This is an architecture diagram of an employee capability mapping system based on AI and digital twins according to the present invention; Figure 2 This is a flowchart of an employee capability mapping method based on AI and digital twins according to the present invention. Detailed Implementation

[0016] Please see Figure 1 An employee capability mapping system based on AI and digital twins includes: a data acquisition and preprocessing module, a multimodal AI analysis module, a digital twin model construction and application module, a capability assessment and feedback module, and a user interaction interface module. Data acquisition and preprocessing module: We used Apache Flink to build a real-time processing stream and developed a custom connector to adapt to 12 power business systems to achieve unified collection of "equipment operation and maintenance data + employee performance data + training records + field sensor data". Multi-source data acquisition unit: used to acquire data in real time from multiple heterogeneous business systems and field sensors of power companies; Data quality control unit: It has a built-in rule base for integrity verification, rationality verification, and consistency verification, and uses machine learning models to predict and correct data quality problems in advance; Establish a three-level validation rule base: Integrity verification: Automatically detects missing values ​​in key fields such as performance scores and skill ratings, and fills in the missing data from related systems with a success rate of ≥90%; Reasonableness verification: Abnormal data such as performance scores > 100 points and equipment operation time < 1 minute are removed using the 3σ criterion, with an anomaly identification accuracy rate ≥ 95%; Consistency verification: Data such as device model and date format are standardized to achieve 100% data consistency; Meanwhile, LSTM models are used to train historical data to predict future data quality risks, triggering calibration in advance and improving the data quality pass rate. Data storage unit: Cleans, transforms, and standardizes data storage; constructs a power industry knowledge graph, defining employee-skill-equipment-project entities and relationships, such as employee A possessing 10kV operation and maintenance skills and associated with substation B maintenance projects; enables cross-system data association queries based on the Neo4j graph database, shortening query response time; data storage uses a Greenplum data warehouse, organizing data by employee ID + time, and creating composite indexes to support multi-dimensional queries based on employee, time, and skills. Multimodal AI analysis module: Multimodal fusion framework: Based on the multimodal Transformer model, it integrates voice, text, image, and device status data, and learns cross-modal relationships through a self-attention mechanism; Industry-adapted pre-trained model: Based on ERNIE 3.0, incorporating industry-standard knowledge; based on ERNIE 3.0, adapted for the power industry: Corpus expansion: Incorporate power industry terminology and standards, such as the "Power Safety Work Regulations" and fault handling cases, to build a dedicated power industry corpus; Fine-tuning training: Supervised fine-tuning was adopted, using power text-intent labels as training data, and the AdamW optimizer was used for 50 iterations until the loss function converged, which improved the model's accuracy in understanding power-related professional expressions. A dynamic evaluation model, built upon reinforcement learning, uses real-time employee performance as input, adjustments to the evaluation strategy as output, and improved skills as reward feedback, enabling dynamic iteration of the evaluation strategy. A reinforcement learning framework is constructed to achieve adaptive evaluation strategies. State space: employee's current ability score, real-time work performance (such as troubleshooting time), and external environment; Action space: Adjust the weight of evaluation indicators and switch evaluation scenarios; Reward function: Skills enhancement and job matching assessment bias, where skills enhancement is the increase in employee skill scores, job matching is the degree of matching between ability and job requirements, and assessment bias is the error between model prediction and actual performance; Strategy optimization: The PPO algorithm is used to iteratively optimize the strategy network, which is updated quarterly to ensure that the evaluation strategy is adapted to employee growth and technological change; A multimodal Transformer fusion framework for the power industry, designed with a multimodal fusion process specifically for power scenarios: Voice processing: Employee operation commands are collected through noise-canceling microphones, features are extracted using MFCC, and then converted into text by CTC speech recognition; Image processing: YOLOv8 is used to recognize employee gestures and equipment status in inspection images, and target detection boxes and category labels are output; Equipment status processing: Extracting trend characteristics of data such as voltage and current through a timing encoder; Cross-modal fusion: Input speech text embedding, image feature vectors, and device status temporal features into a multimodal Transformer, and learn the association between command, gesture and device status through a self-attention mechanism. For example, the collaborative analysis of bus inspection command + pointing to 10kV bus + bus voltage anomaly significantly improves the accuracy of intent recognition by fusing feature vectors. Digital twin model construction and application module: Includes a high-precision equipment modeling unit: employing a multiphysics coupling method, a digital twin model integrating electrochemical, thermal, and electrical characteristics is constructed for power equipment, dynamically calibrating parameters through real-time sensor data and historical operating data; constructing the digital twin model: Electrochemical model: Based on the PNGV equation, the lithium-ion migration process is described, and the charge-discharge capacity change is simulated; Thermal model: Fourier's law is used to calculate the internal temperature distribution of the battery, taking into account the influence of the heat dissipation structure; Electrical model: Voltage and current characteristics are simulated based on Thevenin equivalent circuit; Coupling equations: The three-physics model is coupled using the finite element method to establish a unified mathematical model; Parameter calibration: Data input: Real-time acquisition of voltage, current, and temperature data, as well as historical operating data of the energy storage device; Optimized calibration: The particle swarm optimization algorithm is adopted, with the mean square error between the model prediction and the measured value as the objective function, to ensure that the model is highly consistent with the actual equipment state; Edge-Cloud Collaborative Computing Unit: Deploys data preprocessing and local state analysis tasks to the substation edge terminal, transmitting only key feature data to the cloud; Edge Node Deployment: Deploys edge terminals in the substation, responsible for: Data Preprocessing: Filtering and dimensionality reduction of sensor data, reducing data volume by 70%; Local Computation: Running lightweight models to quickly determine if there are any urgent anomalies; Cloud computing: Cloud servers run complete digital twin models and multimodal AI models, responsible for: Global analysis: simulating the overall operating status of equipment based on key feature data transmitted from the edge; Capability assessment: generating assessment results by combining employee operation data; A virtual-physical collaboration platform seamlessly integrates digital twin models with production management and human resources systems, generating personalized virtual training scenarios, providing real-time feedback on employee operational effectiveness, and synchronizing training data to employee skill profiles; achieving deep integration of digital twins and actual business operations. Virtual training scenario generation: Based on the enterprise's business needs, virtual scenarios are generated using a digital twin model, with scenario parameters consistent with the actual power plant; Operational feedback and evaluation: Employees operate in a virtual scenario, and the model provides real-time feedback on the operation effect, synchronously recording the operation steps and duration; Data synchronization and application: Training data is automatically synchronized to the human resources system to update employees' "new energy operation and maintenance skills" files; evaluation results are used to adjust actual work arrangements; Capability assessment and feedback module: includes a comprehensive indicator system, quantitative assessment units, and real-time feedback units; achieving a closed loop of data collection, analysis, assessment, and feedback. Real-time data acquisition: The edge terminal uploads employee operation data and equipment status data every 10 minutes; Rapid analysis and computation: Cloud-based multimodal AI and digital twin models perform parallel computation, updating capability scores within 30 minutes; Visualized feedback: Employees can view real-time scores through a skills radar chart, identify weaknesses, and receive innovative thinking training courses; management can generate a team capability heatmap to assist in developing training plans. User interface module: including employee and management sides.

[0017] The data acquisition and preprocessing module also includes a knowledge graph-driven unit: constructing a knowledge graph for the power industry and enabling cross-system data association queries based on the Neo4j graph database.

[0018] The dynamic evaluation model of the multimodal AI analysis module also includes a key event triggering mechanism: when an employee participates in a new training program, obtains a skills certification, or experiences a performance fluctuation of ≥20%, the evaluation strategy adjustment is triggered immediately without waiting for periodic iterations.

[0019] The high-precision equipment modeling unit of the digital twin model construction and application module can simulate charging and discharging efficiency and life decay curves for digital twin models of new energy storage devices, and support the evaluation of employee operation effects under different working conditions.

[0020] The comprehensive indicator system includes primary, secondary, and tertiary indicators; The quantitative evaluation unit uses the entropy method to calculate the weight of indicators and conducts quantitative evaluation through multi-source data. The real-time feedback unit generates assessment reports, displays employees' strengths and weaknesses using a skills radar chart, and pushes personalized training suggestions.

[0021] The employee side is used to display competency assessment results, conduct comparative analysis with similar positions, plan career development paths, and support training course registration and online consultation. The management interface provides a visual representation of the team's capability distribution and analysis of capability gaps, offering decision support functions for talent selection and development.

[0022] Please see Figure 2 A method for mapping employee capabilities based on AI and digital twins, the method includes: S1. Real-time acquisition of data from 12 business systems and on-site sensor data through Apache Flink processing stream; S2. Clean the data using rules of integrity, rationality, and consistency; correct outliers using machine learning models; and standardize the format. S3. Store standardized data in a Greenplum data warehouse by partitioning it according to "employee ID + time" and create indexes to optimize query efficiency; S4. Construct a multimodal Transformer fusion framework to convert employee voice commands into text, input them into the model along with device image features and real-time status data, and generate fused feature vectors through a self-attention mechanism. S5. Industry-adaptive pre-trained models are used to analyze and integrate features to identify the accuracy of employees' operational intentions and professional expressions. The fine-tuning data for the industry-adaptive pre-trained models includes equipment operation manuals, fault handling reports, and technical specification documents from the past 5 years. The training process uses the Adam optimizer with a learning rate of 0.001 and iterates until the loss function converges. S6. Based on a reinforcement learning dynamic evaluation model, the evaluation weights are adjusted using the intent recognition results and the employee's historical performance as inputs. S7. For novel energy storage devices, a multi-physics coupled model is constructed by integrating electrochemical, thermal, and electrical equations. Parameters are calibrated using real-time sensor data to ensure that the deviation between the model and the actual state is ≤5%. The parameter calibration of the digital twin model of the novel energy storage device adopts the particle swarm optimization algorithm, with the "mean square error between the model prediction value and the sensor measurement value" as the objective function. The calibration is stopped after the error is ≤5%. S8. Through the edge-cloud collaborative architecture, the edge terminal preprocesses the on-site data, and the cloud simulates the employee operation effect based on the digital twin model; S9. A virtual training scenario is generated on the virtual-real collaboration platform, and employee operation data is fed back to the digital twin model in real time to evaluate the standardization and effectiveness of operations. S10. Based on the three-level indicator system, the weight of each indicator is calculated using the entropy method, and the score is quantified through multi-source data. The quantification formula for the three-level indicator "number of technical improvement suggestions" is: Score Σ = suggestion innovation score * feasibility score, where innovation is evaluated by technical experts on a scale of 1-5, and feasibility is adjusted according to the project implementation rate. S11 generates hourly assessment reports, displays employee skill distribution using radar charts, and pushes personalized training suggestions. S12. The management side provides a visual display of team capability distribution, supporting talent selection and training decisions. Employees can query capability trends and career development paths. S13. Collect new business data monthly, incrementally train multimodal AI models and digital twin models, and update evaluation indicator weights; S14. Collect feedback from employees and managers quarterly to optimize the user interface and evaluation rules, ensuring that the system adapts to technological changes in the power industry and changes in the company's talent needs.

[0023] This invention's multi-source data fusion engine enables unified collection and association of data from multiple business systems, improving data integration efficiency and shortening cross-departmental data query response time. A data quality control mechanism increases data qualification rates and reduces assessment error rates, avoiding assessment biases caused by data issues. A three-level indicator system covers professional skills, safety awareness, and innovation capabilities, surpassing traditional single-skill assessment dimensions and comprehensively identifying employee potential and weaknesses. A reinforcement learning dynamic assessment model shortens the assessment strategy update cycle, adapting to power technology changes and employee growth. An edge-cloud collaborative architecture shortens single-round strategy iteration time, meeting real-time assessment needs. A virtual-physical collaboration platform improves the alignment between training and actual business operations, reducing the error rate in actual operations after employee training.

[0024] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An employee capability mapping system based on AI and digital twins, characterized in that, include: The module includes data acquisition and preprocessing, multimodal AI analysis, digital twin model construction and application, capability assessment and feedback, and user interface. Data acquisition and preprocessing module: Multi-source data acquisition unit: used to acquire data in real time from multiple heterogeneous business systems and field sensors of power companies; Data quality control unit: It has a built-in rule base for integrity verification, rationality verification, and consistency verification, and uses machine learning models to predict and correct data quality problems in advance; Data storage unit: performs data cleaning, transformation, and standardized storage; Multimodal AI analysis module: Multimodal fusion framework: Based on the multimodal Transformer model, it integrates voice, text, image, and device status data, and learns cross-modal relationships through a self-attention mechanism; Industry-adapted pre-trained model: Based on ERNIE 3.0, incorporating industry-standard knowledge; Dynamic evaluation model; based on reinforcement learning, it uses employees' real-time work performance as state input, evaluation strategy adjustment as action output, and ability improvement effect as reward feedback to achieve dynamic iteration of evaluation strategy; Digital twin model construction and application module: Includes a high-precision equipment modeling unit: using a multi-physics coupling method, a digital twin model integrating electrochemical, thermal, and electrical characteristics is constructed for power equipment, and parameters are dynamically calibrated through real-time sensor data and historical operating data; Edge-cloud collaborative computing unit: Decentralizes data preprocessing and local state analysis tasks to the substation edge terminal, and only transmits key feature data to the cloud; Virtual-Real Collaboration Platform: Seamlessly integrates digital twin models with production management and human resource systems to generate personalized virtual training scenarios, provide real-time feedback on employee operational effectiveness, and synchronize training data to employee skill profiles; Capability assessment and feedback module: includes a comprehensive indicator system, quantitative assessment unit, and real-time feedback unit; User interface module: including employee and management sides.

2. The employee capability mapping system based on AI and digital twins according to claim 1, characterized in that, The data acquisition and preprocessing module also includes a knowledge graph-driven unit: constructing a knowledge graph for the power industry and enabling cross-system data association queries based on the Neo4j graph database.

3. The employee capability mapping system based on AI and digital twins according to claim 1, characterized in that, The dynamic evaluation model of the multimodal AI analysis module also includes a key event triggering mechanism: when an employee participates in a new training program, obtains a skills certification, or experiences a performance fluctuation of ≥20%, the evaluation strategy adjustment is triggered immediately without waiting for periodic iterations.

4. The employee capability mapping system based on AI and digital twins according to claim 1, characterized in that, The high-precision equipment modeling unit of the digital twin model construction and application module can simulate charging and discharging efficiency and life decay curves for digital twin models of new energy storage devices, and support the evaluation of employee operation effects under different working conditions.

5. The employee capability mapping system based on AI and digital twins according to claim 1, characterized in that, The comprehensive indicator system includes primary indicators, secondary indicators, and tertiary indicators; The quantitative evaluation unit uses the entropy method to calculate the index weights and conducts quantitative evaluation through multi-source data. The real-time feedback unit generates an assessment report, displays employees' strengths and weaknesses using a skills radar chart, and pushes personalized training suggestions.

6. The employee capability mapping system based on AI and digital twins according to claim 1, characterized in that, The employee interface is used to display competency assessment results, conduct comparative analysis with similar positions, plan career development paths, and support training course registration and online consultation. The management interface provides a visual representation of the team's capability distribution and capability gap analysis, as well as decision support functions for talent selection and development.

7. An employee capability mapping method based on AI and digital twins, applied to the system described in any one of claims 1-6, the method comprising: S1. Real-time acquisition of data from 12 business systems and on-site sensor data through Apache Flink processing stream; S2. Clean the data using rules of integrity, rationality, and consistency; correct outliers using machine learning models; and standardize the format. S3. Store standardized data in a Greenplum data warehouse by partitioning it according to "employee ID + time" and create indexes to optimize query efficiency; S4. Construct a multimodal Transformer fusion framework to convert employee voice commands into text, input them into the model along with device image features and real-time status data, and generate fused feature vectors through a self-attention mechanism. S5. Employ industry-adaptive pre-trained models to analyze fusion features and identify the accuracy of employee operational intentions and professional expressions. S6. Based on a reinforcement learning dynamic evaluation model, the evaluation weights are adjusted using the intent recognition results and the employee's historical performance as inputs. S7. For novel energy storage devices, a multi-physics coupled model is constructed by integrating electrochemical, thermal, and electrical equations, and parameters are calibrated using real-time sensor data to ensure that the deviation between the model and the actual state is ≤5%. S8. Through the edge-cloud collaborative architecture, the edge terminal preprocesses the on-site data, and the cloud simulates the employee operation effect based on the digital twin model; S9. A virtual training scenario is generated on the virtual-real collaboration platform, and employee operation data is fed back to the digital twin model in real time to evaluate the standardization and effectiveness of operations. S10. Based on the three-level indicator system, the weight of each indicator is calculated using the entropy method, and the score is quantified through multi-source data. S11 generates hourly assessment reports, displays employee skill distribution using radar charts, and pushes personalized training suggestions. S12. The management side provides a visual display of team capability distribution, supporting talent selection and training decisions. Employees can query capability trends and career development paths. S13. Collect new business data monthly, incrementally train multimodal AI models and digital twin models, and update evaluation indicator weights; S14. Collect feedback from employees and managers quarterly to optimize the user interface and evaluation rules, ensuring that the system adapts to technological changes in the power industry and changes in the company's talent needs.

8. The employee capability mapping method based on AI and digital twins according to claim 7, characterized in that, In step S5, the fine-tuning data for the industry-adapted pre-trained model includes the company's equipment operation manuals, fault handling reports, and technical specification documents from the past 5 years. The training process uses the Adam optimizer with a learning rate of 0.001, iterating until the loss function converges.

9. The employee capability mapping method based on AI and digital twins according to claim 7, characterized in that, In step S7, the parameter calibration of the digital twin model of the novel energy storage device adopts the particle swarm optimization algorithm, with the "mean square error between the model prediction value and the sensor measurement value" as the objective function, and the calibration stops after the error is ≤5%.

10. The employee capability mapping method based on AI and digital twins according to claim 7, characterized in that, In step S10, the quantitative formula for the third-level indicator "number of technical improvement suggestions" is: Score Σ = suggestion innovation score * feasibility score, where innovation is evaluated by technical experts on a scale of 1-5, and feasibility is adjusted based on the project implementation rate.