Differentiation supervision reminding method and system based on operator safety credit system, and storage medium

Abstract

The invention discloses a differentiated supervision reminding method and system based on an operator safety credit system, and a storage medium, and belongs to the technical field of safety supervision. The method comprises the following steps: collecting basic information of an operator; building a safety credit evaluation model based on a machine learning algorithm, carrying out the multi-dimensional analysis of the qualification validity, violation frequency, score deduction severity and work resume stability of an operator, and generating a quantitative safety credit score; and according to the scoring result, the operators are divided into high, medium and low safety levels, and different supervision reminding rules are configured for different levels, including real-time alarm, dynamic voice prompt, lightweight notification and the like. According to the invention, through linkage of credit scoring and key process risks, accurate identification, dynamic hierarchical management and control and whole-process supervision and reminding of operation risks are realized, and the safety management efficiency and standardization of an operation site are significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of safety supervision, in particular to a differentiated supervision reminding method based on a work personnel safety credit system. BACKGROUND

[0002] In the field of power grid safety supervision, work safety management faces many challenges: In terms of work personnel information management, there is currently a problem that personnel information, qualification data and violation record data are easily covered after the personnel unit information is changed. This leads to loopholes in information management and supervision, and the personnel information management system has a data island problem. When the work personnel's unit, qualification or identity information changes, key information such as historical violation records and deduction data is easily covered or stored in fragments, leading to the fact that cross-regional and cross-unit work personnel can avoid supervision through information tampering, forming a safety hazard.

[0003] At the same time, in terms of work risk control, the work content related to the work plan lacks a unified rule description, making it difficult to identify key procedures from the work plan, and thus it is impossible to accurately control the risks on the work site. Due to the uneven safety awareness and work ability of work personnel, and the complex and variable work site environment, the traditional supervision method is difficult to effectively deal with differentiated work risks, making it difficult to achieve accurate reminding and efficient supervision, and difficult to ensure the safety of power grid work on site.

[0004] In addition, there is a lack of effective linkage mechanism between existing work personnel information management and work risk control, making it difficult to dynamically evaluate and differentially manage work risks according to the safety credit status of work personnel.

[0005] In view of the above problems, an intelligent method that integrates dynamic credit evaluation, key procedure risk matching and differentiated supervision reminding is urgently needed to improve the accuracy and effectiveness of power grid work safety management. SUMMARY

[0006] To solve the above problems, the present application provides a differentiated supervision reminding method, system and storage medium based on a work personnel safety credit system, which calculates the safety credit score in all directions and multiple dimensions by collecting the basic information of work personnel and comprehensively considering work tasks, environment, difficulty and historical records, etc., can accurately reflect the safety work ability and awareness of work personnel, dynamically adapt to work changes, and differentiated reminding methods encourage work personnel to actively pay attention to their own safety behavior, thereby reducing work risks from the source.

[0007] To achieve the above purpose, the technical solution adopted by the present application is: In the first aspect of the present application, a differentiated supervision reminding method based on a work personnel safety credit system is provided, comprising the following steps: S1, collect and preprocess the basic information of the operation personnel, the basic information including personal information, responsibility information and production safety knowledge information; S2, input the preprocessed original data into a deep learning model and generate a personalized operation personnel portrait, and combine an incremental learning strategy to ensure that the operation personnel portrait is continuously optimized according to the original data when the original data changes; S3, constructing a safety credit evaluation model based on a machine learning algorithm, inputting the operation personnel portrait into the safety credit evaluation model, and generating a safety credit score of the operation personnel; S4, setting a corresponding safety level according to the safety credit score of the operation personnel, and setting a differentiated supervision reminding rule for different safety levels.

[0008] As preferred, in S1, the personal information includes age and length of service, the responsibility information includes position, work scope and whether holding relevant professional certificates, and the production safety knowledge information includes the degree of understanding of production safety knowledge.

[0009] As preferred, in S2, a multi-layer perception machine is used as a basic deep learning model to generate a personalized portrait vector of each operation personnel.

[0010] As preferred, in S3, the safety credit evaluation formula of the safety credit evaluation model for the operation personnel is as follows: wherein, represents the safety credit score of the operation personnel i, represents the preliminary safety score of the operation personnel i calculated by the multi-layer perception machine model, represents the personalized portrait vector feature matrix of the operation personnel i, represents the decay weight coefficient of the operation personnel i, used to adjust the influence of the operation personnel on the decay of the operation danger degree, represents a high-risk operation decay function, e is a constant, represents a violation operation influence coefficient, represents a comprehensive score deduction value of the violation operation.

[0011] As preferred, in S4, the operation personnel is divided into high, medium and low safety levels according to the safety credit score; High-risk level personnel: automatically triggering real-time alarm when the operation plan is released, pushing the historical violation records and control points to the supervisors and operation responsible persons, and forcibly reminding through the whistle mechanism before the start of the key process; Medium-risk level personnel: pushing a standardized risk prompt to the operation responsible person when the operation starts, and triggering dynamic voice or SMS reminders at key process steps; Low-risk level personnel: send lightweight notifications to supervisors only during critical process execution, for record supervision.

[0012] More preferably, the whistle-blowing mechanism specifically comprises: before the execution of a critical process, the system automatically matches risk control points according to the safety level of the worker, and sends early warning information containing risk description, control standard and real-time monitoring link to supervisors, supervisors and support personnel through multi-terminal linkage.

[0013] As a preferred, in the S4, the reminding mode includes SMS, voice, APP push, and on-site broadcast, which delivers prompt information to workers and managers.

[0014] As a preferred, the method further comprises a dynamic updating mechanism of credit score: according to the real-time behavior data of the worker during execution, the completion degree of rectification and the newly added violation records, the safety credit score and the corresponding level are updated periodically.

[0015] In the second aspect of the present application, a differential supervision reminding system based on a worker safety credit system is provided, comprising sequentially connected: A data acquisition and preprocessing module for acquiring and preprocessing basic information of workers; A worker portrait generation module for generating personalized worker portraits based on a deep learning model; A safety credit evaluation module for generating a safety credit score of a worker based on a safety credit evaluation model; A supervision reminding module for setting corresponding safety levels and conducting differential supervision reminding for workers of different safety levels.

[0016] In the third aspect of the present application, a storage medium having a computer program stored thereon is also provided, wherein the program is executed by a processor to implement the differential supervision reminding method based on the worker safety credit system as described above.

[0017] Compared with the prior art, the present application has the following advantages: Accurate assessment of worker safety quality: by collecting various basic information of workers and comprehensively considering work tasks, environment, difficulty and historical records, etc., the safety credit score is calculated in a comprehensive and multi-dimensional manner, which can accurately reflect the safety operation ability and consciousness of workers, and provide a scientific basis for personnel safety management.

[0018] Differential supervision management is achieved: according to the safety credit score, safety levels are divided, personalized prompt methods are set for different levels, and the traditional "one-size-fits-all" supervision mode is changed. The supervision intensity on high-risk workers is strengthened, resources are concentrated to control key risk points, supervision resource allocation is optimized, and management efficiency is improved.

[0019] Dynamic adaptation to job changes: The safety credit score can be dynamically adjusted according to the real-time job data of the job personnel, so that the supervision reminders closely match the actual job situation and timely respond to the risk fluctuations brought by changes in job environment and tasks, ensuring the effectiveness and timeliness of the reminder information.

[0020] Strengthening safety awareness: Differentiated reminders encourage job personnel to actively focus on their own safety behavior, and high-risk personnel enhance their safety awareness and standardize their operation due to frequent and detailed reminders, which helps to cultivate good safety habits and reduce job risks from the source.

[0021] Perfect personnel information management: Collect all kinds of information of job personnel in data collection link, build unified and comprehensive personnel information database, plug the information change loophole, realize the transparency and uniqueness of job personnel information, provide solid data foundation for cross-regional and cross-project operation supervision, prevent job personnel from escaping supervision by changing information. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 It is a flow chart of the differential supervision and reminding method based on the safety credit system of the job personnel of the present application.

[0023] Figure 2 It is a block diagram of the differential supervision and reminding system based on the safety credit system of the job personnel of the present application. DETAILED DESCRIPTION

[0024] Please refer to Figure 1 In the first aspect of the present application, a differential supervision and reminding method based on the safety credit system of the job personnel is provided, comprising the following steps: S1, collecting and preprocessing the basic information of the job personnel, including personal information, responsibility information and production safety knowledge information; The basic information of the job personnel is obtained from the power grid safety supervision system, the job personnel management platform, the third-party qualification certification agency or the company system, including personal information, responsibility information and production safety knowledge information; wherein the name, ID number, age, length of service, position and responsibility, work scope, and examination results are collected through the online safety knowledge examination system.

[0025] Integrate historical data: including the validity period of the qualification certificate of the job personnel (such as electrician certificate, high-altitude operation certificate), the violation record in the past 5 years (such as violation type, occurrence time, rectification result), deduction data (such as deduction value corresponding to each violation) and work history (such as the type of operation participated, working hours).

[0026] In the data collection step, the basic information of the workers should also include the workers' past accident records and safety production reward information, etc., in order to more comprehensively evaluate the safety credit status of the workers.

[0027] The above data is encrypted and stored through blockchain technology, ensuring that the data cannot be tampered with, and establishing a real-name information library to achieve real-time data synchronization.

[0028] When the personnel information changes, including unit transfer, qualification expiration, etc., the system automatically triggers the update process, synchronizes the change content to the associated work risk type library and key process library, and generates a timestamped change log to avoid historical data being overwritten.

[0029] Example: A worker moves from A to B due to unit transfer, and the system automatically associates his historical violation records (such as 2022 electric shock risk violation) to the new work unit account, ensuring continuity of credit evaluation.

[0030] The output of this step S1 is a structured behavior sample set which is obtained by preprocessing and merging, and the sample set is a matrix, where is the number of samples (i.e. the number of historical work times), and each row represents the feature data of a work behavior. This output will be used as the input for the worker modeling in the subsequent steps. Specifically, each row in contains personal information, job information, and production safety knowledge information. This output will be used as input for the worker behavior modeling in the subsequent steps.

[0031] Through this step S1, multi-source data from personal information, job information, and production safety knowledge information is effectively aligned, normalized, and missing value imputed to generate a unified format input feature vector . In the processing process, the consistency and computational rationality of the data are ensured, and the unstable factors caused by time misalignment, feature scale difference, or missing data are avoided. Through these processes, high-quality and reliable input data are provided for the subsequent worker behavior modeling, ensuring the stability and operability of the entire system.

[0032] S2, input the preprocessed original data into the deep learning model and generate a personalized worker portrait, and use the incremental learning strategy to ensure that the worker portrait is continuously optimized according to the original data when the original data changes; Each row represents a work behavior sample, containing the following three features: personal information, job responsibility information, and production safety knowledge information. The size of the input matrix is where is the number of worker samples, and 3 is the feature dimension of each sample, which is input into the deep learning model.

[0033] In the process of generating worker portraits, a multi-layer perceptron (MLP) is used as the basic deep learning model. The model structure includes an input layer, multiple hidden layers, and an output layer. Through training, the model can extract deep personalized features from worker behavior data to generate a personalized portrait vector for each worker , which represents the personalized features of the worker. This vector will be the input of the subsequent recommendation system.

[0034] A multi-layer perceptron is used as the basic deep learning model to generate a personalized portrait vector for each worker.

[0035] With the input of worker behavior data, the incremental learning strategy fine-tunes the model, ensuring that the model can adapt to the latest changes in worker behavior. Each time the model is updated, it only adjusts the parameters based on new data, rather than starting from scratch. This greatly improves the training efficiency of the model.

[0036] Through this step S2, the entire system can accurately generate personalized worker portraits while protecting worker privacy, and ensure that the model continuously optimizes and adapts to changes in worker behavior as new data is added.

[0037] S3, a safety credit evaluation model is constructed based on machine learning algorithms, and the worker portrait is input into the safety credit evaluation model to generate a safety credit score for the worker; The goal of this step S3 is to model the device capabilities based on the worker portrait through a deep learning model. A multi-layer perceptron (MLP) network will be used for training. Since the data has been normalized in step S1, the standardized feature matrix is directly used as input to avoid repeated processing.

[0038] First, the worker portrait vector generated in step S2 and the device state data in step S1 are combined into a new feature matrix as input to the safety credit evaluation model. The dimension of the combined feature matrix is , where is the dimension of the worker's work data (e.g., job responsibilities and online safety knowledge test scores).

[0039] ​After the data is prepared, the input feature matrix is next input into a multi-layer perceptron (MLP) model, i.e., a safety credit assessment model. The MLP model is composed of multiple fully connected layers (FC layers), and the output of each layer represents the ability score of the device. The specific network structure is as follows: Input layer: The dimension of the input layer is , i.e., all the job personnel portrait features and device state features; Hidden layer: Two hidden layers are designed, the first layer has 64 neurons, and the second layer has 32 neurons, and the activation function uses ReLU (Rectified Linear Unit). The calculation formula of the ReLU activation function is: ReLU activation function can increase the non-linear expression ability of the model, thereby effectively capturing the complex relationship between the safety credit of the job personnel and the work behavior; Output layer: The output layer is a single neuron, which outputs the safety credit score of the job personnel i , and the score range is [0, 100], indicating the work ability of the job personnel i.

[0040] The input features include: Static features: age, length of service, validity of qualifications (such as the proportion of the remaining validity period of the certificate); Dynamic features: frequency of violations in the past 3 years, number of repeated violations of the same type (such as electric shock-related violations), severity of deductions (quantified according to the ratio of the deduction value to the standard threshold), stability of work history (such as the frequency of job changes); Behavioral features: timeliness of rectification feedback (average length of time from discovery of violation to completion of rectification), score of safety knowledge assessment.

[0041] The model is trained through historical data (containing 10,000 job personnel records), and the output is a credit score of 0-100, with a higher score representing better safety credit.

[0042] For personnel who have committed the same type of violation 2 or more times within the past 3 months, increase the deduction weight by 20%; Risk weight superposition: combine the risk levels in the job type matching key process library (such as the risk weight of high-altitude work is 0.8, and the risk weight of device installation is 0.5) to dynamically adjust the final score.

[0043] The safety credit assessment formula of the safety credit assessment model for the job personnel is as follows: wherein, represents the safety credit score of the job personnel i, represents the preliminary safety score of the job personnel i calculated by the multi-layer perceptron model, a personalized portrait vector feature matrix representing the worker i, a decay weight coefficient representing worker i, used to adjust the influence of worker on the degree of work danger decay, a high-risk work decay function, e is a constant, a violation work influence coefficient, a comprehensive score deduction value of the violation work.

[0044] Safety Credit Score is a comprehensive evaluation result based on the work behavior, personal information, responsibility information and production safety knowledge information of the worker, reflecting the performance of the worker i under specific working conditions.

[0045] S4, according to the safety credit score of the worker, set the corresponding safety level, and set different supervision reminding rules for different safety levels.

[0046] The safety level is divided into high, medium and low three levels, each level corresponds to a corresponding range of safety credit score, and the prompt method corresponding to each level is different. For example, workers with high safety level only need to receive general prompts; workers with medium safety level receive general prompts and some detailed prompt content; workers with low safety level receive comprehensive detailed prompts, real-time prompts and multi-channel prompts and other prompt methods.

[0047] Safety level division: High risk (0-40 points): about 10% of the total, including habitual violators or high-risk workers; Medium risk (41-70 points): about 60% of the total, including occasional violators but timely rectification personnel; Low risk (71-100 points): about 30% of the total, including long-term non-violation and valid qualification personnel.

[0048] High-risk personnel: When the work plan is released, the system automatically pushes a pop-up window alarm to the supervisor, including the worker's last three violation records and the associated control points (such as "high-altitude work needs to check the safety belt locking device"); Before the key process starts, the "whistle mechanism" is used to force the on-site broadcast reminder, and at the same time, a short message is sent to the work supervisor and the supervisor.

[0049] The whistle mechanism is as follows: before the execution of the key process, the system automatically matches the risk control points according to the safety level of the worker, and sends warning information containing risk description, control standard and real-time monitoring link to the supervisor, supervisor and support personnel through multi-terminal linkage.

[0050] Medium-risk personnel: At the beginning of the work, the standardization prompt is pushed to the person in charge (such as "this work contains 3 electric shock risks, and the insulating tool needs to be checked"). During the execution of the key process, the smart bracelet vibrates or the voice prompts (such as "work step 5: wearing insulating gloves for electric test operation").

[0051] Low-risk personnel: Only after the completion of the key process, the light record notice is sent to the supervision system (such as "Zhang San has completed the tower installation, and the risk control meets the standard").

[0052] The on-site supervision personnel upload the work behavior data (such as step completion time, control measure implementation situation) in real time through the mobile terminal; if new violations are found, the system immediately deducts the score and recalculates the credit score, triggers the level adjustment; periodically (such as every month) generate a credit report and push it to the management personnel for decision optimization.

[0053] As an option, in S4, the prompting method includes SMS, voice, APP push, and on-site broadcast to deliver prompt information to the work personnel and the management personnel.

[0054] Each level corresponds to a corresponding range of safety credit scores, and the prompt methods corresponding to each level are different. For example, low-risk personnel only need to receive general prompts; medium-risk personnel receive general prompts and some detailed prompt content; high-risk personnel receive comprehensive detailed prompts, real-time prompts, and multiple-channel prompts, etc. to improve the reliability and effectiveness of information transmission.

[0055] As an option, the method further includes a dynamic updating mechanism for the credit score: periodically updating the safety credit score and the corresponding level according to the real-time behavior data of the work personnel during the execution process, the completion degree of rectification, and the newly added violation records.

[0056] Through this embodiment, the power grid operation and maintenance unit can effectively master the safety credit status of the work personnel, implement differentiated supervision and prompting strategies according to different safety levels, improve the work safety supervision efficiency, and reduce the operation and maintenance safety risks. In practical applications, as the work personnel's work behavior is continuously monitored and the data is updated, the safety credit score is dynamically adjusted, the safety level is correspondingly changed, and the prompting method is also optimized, thereby ensuring the accurate adaptation of safety supervision to the safety quality changes of the work personnel.

[0057] By this step, the job worker portrait in step S2 is successfully combined with the job worker safety credit score in step S3. By collecting the job worker's basic information in multiple aspects, and comprehensively considering the job task, environment, difficulty and historical record, etc., the safety credit score is calculated in all directions and multiple dimensions, which can accurately reflect the safety operation ability and consciousness of the job worker, and provide a scientific basis for personnel safety management. The incremental learning strategy ensures that the system can continuously adapt to changing work requirements and maximize system efficiency.

[0058] The innovation of this step is that by combining the job worker's work behavior data, safety credit score and strategy execution, an adaptive and real-time optimized execution system is created, which can effectively adapt to changing work requirements and maximize system efficiency.

[0059] As shown in Figure 2 In the second aspect of the present application, a differentiated supervision and reminding system based on a job worker safety credit system is provided, which comprises, which are connected in sequence: A data acquisition and preprocessing module for acquiring and preprocessing the basic information of the job worker; A job worker portrait generation module for generating personalized job worker portraits based on a deep learning model; A safety credit evaluation module for generating a safety credit score of the job worker based on a safety credit evaluation model; A supervision and reminding module for setting corresponding safety levels and conducting differentiated supervision and reminding for job workers with different safety levels.

[0060] In the third aspect of the present application, a storage medium having a computer program stored thereon is also provided, wherein the program is executed by a processor to realize the differentiated supervision and reminding method based on the job worker safety credit system as described above.

[0061] The above embodiments only describe the preferred embodiments of the present application, and do not limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements to the technical solutions of the present application made by ordinary engineering and technical personnel in the art shall fall within the protection scope determined by the claims of the present application.

Claims

1. A differentiated supervision and reminder method based on the safety credit system of operators, characterized in that, Includes the following steps: S1. Collect and preprocess the basic information of the workers. The basic information includes personal information, job information, and production safety knowledge information. S2. Input the preprocessed raw data into the deep learning model and generate personalized worker profiles. Combine with the incremental learning strategy to ensure that the worker profiles are continuously optimized based on the raw data as the raw data changes. S3. Construct a safety credit assessment model based on machine learning algorithms, input the operator profile into the safety credit assessment model, and generate the operator's safety credit score; S4. Set corresponding safety levels based on the safety credit scores of the workers, and set differentiated supervision and reminder rules for different safety levels.

2. The differentiated supervision and reminder method based on the safety credit system of operators as described in claim 1, characterized in that, In S1, the personal information includes age and length of service, the job information includes position, scope of work and whether relevant professional certificates are held, and the production safety knowledge information includes the level of understanding of production safety knowledge.

3. The differentiated supervision and reminder method based on the safety credit system of operators as described in claim 1, characterized in that, In S2, a multilayer perceptron is used as the basic deep learning model to generate a personalized profile vector for each worker.

4. The differentiated supervision and reminder method based on the safety credit system of operators as described in claim 1, characterized in that, In S3, the safety credit assessment model uses the following formula for assessing the safety credit of operators: in, This represents the safety credit score of worker i. This indicates that the multilayer perceptron model calculates the preliminary safety score for worker i. This represents the personalized profile vector feature matrix of worker i. This represents the decay weighting coefficient for worker i, used to adjust for the worker's impact on the decay of job hazard level. This represents the decay function for high-risk operations, where e is a constant. Indicates the impact coefficient of illegal operations. This represents the total score deduction for violations.

5. A differentiated supervision and reminder method based on the safety credit system of operators as described in claim 1, characterized in that, In S4, workers are divided into three safety levels: high, medium, and low, based on their safety credit scores. High-risk personnel: When the work plan is released, a real-time alarm will be automatically triggered, and their historical violation records and key control points will be pushed to the supervisors and work supervisors. A whistleblowing mechanism will be used to forcefully remind them before the start of key processes. For personnel at medium risk level: Standardized risk warnings will be sent to the person in charge of the work when the work starts, and dynamic voice or SMS reminders will be triggered at key process steps; Low-risk personnel: Lightweight notifications are sent to inspectors only when critical processes are being performed, for record-keeping and supervision purposes.

6. A differentiated supervision and reminder method based on the safety credit system of operators, as described in claim 5, is characterized in that... The aforementioned whistleblowing mechanism is as follows: before the execution of key processes, the system automatically matches risk control points based on the safety level of the operators, and sends early warning information containing risk descriptions, control standards, and real-time monitoring links to supervisors, managers, and support personnel through multi-terminal linkage.

7. A differentiated supervision and reminder method based on the safety credit system of operators as described in claim 1, characterized in that, In S4, the reminder methods include SMS, voice, APP push, and on-site broadcast, to convey reminder information to operators and managers.

8. A differentiated supervision and reminder method based on the safety credit system of operators according to claim 1, characterized in that, The method also includes a dynamic update mechanism for credit scores: based on the real-time behavioral data of operators during the execution process, the degree of rectification completion, and new violation records, the safety credit score and corresponding level are updated periodically.

9. A differentiated supervision and reminder system based on the safety credit system of operators, characterized in that, Including those connected sequentially: The data acquisition and preprocessing module is used to collect basic information about the working fish and perform preprocessing. The worker profile generation module is used to generate personalized worker profiles based on deep learning models. The safety credit assessment module is used to generate safety credit scores for workers based on the safety credit assessment model. The supervision and reminder module is used to set corresponding safety levels and provide differentiated supervision and reminders to workers at different safety levels.

10. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the differentiated supervision and reminder method based on the safety credit system of operators as described in any one of claims 1 to 8.

Images