A security behavior monitoring and early warning system for an unattended new energy station

Abstract

This invention relates to the field of monitoring and early warning technology for new energy power stations, and particularly to a safety behavior monitoring and early warning system for unmanned new energy power stations. The system includes: a sensing module that collects multi-source data and integrates it to output station safety management data; a risk prediction module that outputs dynamic safety risk prediction values ​​for personnel based on a nonlinear time-series risk prediction model; a behavior recognition module that determines the compliance of behavior and generates tiered early warning signals; a closed-loop optimization module that updates model parameters, recognition weights, and early warning rules based on feedback data; an unmanned management module that executes dynamic risk compensation and tiered automatic takeover strategies for unsupervised operations; and a centralized control and display module that realizes data visualization, targeted early warning push, and command forwarding. In this invention, unified monitoring of multi-source data is achieved through sequential data interaction between the modules, solving the problems of traditional manual inspections, information silos, and data delays that prevent real-time detection of safety risks.

Description

Technical Field

[0001] This invention relates to the field of monitoring and early warning technology for new energy power stations, and in particular to a safety behavior monitoring and early warning system for unattended new energy power stations. Background Technology

[0002] With the rapid expansion of new energy power generation, the number of new energy power plants such as wind farms and photovoltaic power stations continues to increase. In most of these plants, daily operation management and safety monitoring still rely on on-site personnel. Existing technologies mainly rely on on-duty personnel to inspect, record operation logs, and monitor equipment status, working environment, and personnel operation through SCADA systems and environmental sensors. Abnormal equipment operation, environmental changes, and operational violations are manually judged and reported. This on-site handling mode, whether manned or unmanned, can handle emergencies to a certain extent.

[0003] Traditional new energy power plant safety management largely relies on manual inspections and decentralized monitoring. Due to information silos and data delays, potential safety risks cannot be detected in real time. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides a safety behavior monitoring and early warning system for unattended new energy power stations, aiming to improve the problem that traditional new energy power station safety management mostly relies on manual inspections and decentralized monitoring, which easily leads to the inability to detect potential safety risks in real time.

[0005] This invention provides the following technical solution: a safety behavior monitoring and early warning system for unattended new energy power stations includes: The sensing module is used to collect and integrate personnel behavior data, duty status data, work environment data and equipment operation status data, and to obtain personnel safety management data from the existing digital management and control platform of the site, and output site safety management and control data. The risk prediction module has a built-in nonlinear time-series risk prediction model based on the evolution of personnel behavior. It takes the site safety management data output by the sensing module as input and outputs the dynamic safety risk prediction value of personnel. The behavior recognition module is used to determine the compliance of personnel behavior data, generate safety behavior recognition results, and generate graded early warning signals by combining the dynamic safety risk prediction value of personnel. The closed-loop optimization module is used to receive behavior recognition results, graded early warning signals and early warning handling closed-loop feedback data, and update the calculation parameters of the risk prediction module and the feature weights and graded early warning triggering rules of the behavior recognition module in reverse. The unmanned management module is used to determine unattended scenarios and, in conjunction with the dynamic safety risk prediction value and behavior recognition results of personnel, execute dynamic risk compensation strategies and graded automatic takeover strategies for unsupervised operations. The centralized control and display module is used to render and output the site safety management data collected by the sensing module through a graphical interface, send the hierarchical early warning signals generated by the behavior recognition module to the corresponding job terminals according to the job position, and receive the control instructions issued by the remote centralized control system and transmit them to the unmanned control module.

[0006] By adopting the above technical solution, and sequentially establishing a data interaction and communication connection between a perception module, a risk prediction module, a behavior recognition module, a closed-loop optimization module, an unmanned management module, and a centralized control and display module, a dynamic evolution and closed-loop optimization control system for personnel behavior risks based on multi-source data fusion is constructed. This system achieves collaborative modeling and time-series risk prediction of personnel behavior status, working environment, and equipment operating status. It generates hierarchical early warning signals by fusing behavior recognition results with dynamic risk prediction values. At the same time, it adaptively updates the risk prediction model parameters, behavior recognition feature weights, and early warning triggering rules by combining early warning response feedback, thereby achieving continuous correction and dynamic optimization of safety risks. This solves the problems of lagging risk identification, insufficient early warning accuracy, and lack of feedback optimization mechanisms in the safety management of traditional new energy power stations.

[0007] The present invention has the following beneficial effects: By sequentially establishing a data interaction and communication connection between a perception module, a risk prediction module, a behavior recognition module, a closed-loop optimization module, an unmanned management module, and a centralized control and display module, a dynamic evolution and closed-loop optimization control system for personnel behavior risks based on multi-source data fusion was constructed. This system enables collaborative modeling and time-series risk prediction of personnel behavior status, working environment, and equipment operating status. It generates tiered early warning signals by fusing behavior recognition results with dynamic risk prediction values. Simultaneously, it adaptively updates the risk prediction model parameters, behavior recognition feature weights, and early warning triggering rules based on early warning response feedback, thereby achieving continuous correction and dynamic optimization of safety risks. This solves the problems of lagging risk identification, insufficient early warning accuracy, and lack of feedback optimization mechanisms in the safety management of traditional new energy power plants.

[0008] 2. In this invention, compliance judgment is made on the collected personnel behavior data and a graded early warning signal is generated by combining the dynamic safety risk prediction value of personnel. This enables the accurate identification of abnormal behavior and potential risks, thereby improving the problem that traditional safety early warnings mostly rely on fixed thresholds and experience judgments, which ignore the evolution of personnel behavior and changes in the environment, resulting in high false alarm rates and delayed early warnings.

[0009] 3. In this invention, by sending graded early warning signals to the corresponding job terminals according to job positions and linking them with the unmanned control module to execute strategies, rapid response and automatic handling of risk events can be achieved. This improves the problem that traditional site safety control mostly relies on manual alarm reception and manual intervention, which results in slow response speed and dependence on manual judgment, leading to delays in accident handling and accumulation of safety hazards. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the architecture of a safety behavior monitoring and early warning system for an unattended new energy power station proposed in this invention; Figure 2 This is a flowchart illustrating a safety behavior monitoring and early warning method for unattended new energy power stations proposed in an embodiment of the present invention. Detailed Implementation

[0011] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] Example 1: In the first embodiment of the present invention, the present invention provides a safety behavior monitoring and early warning system for unattended new energy power stations, such as... Figure 1 As shown, it includes a perception module, a risk prediction module, a behavior recognition module, a closed-loop optimization module, an unmanned management and control module, and a centralized control and display module that establish data interaction and communication connections in sequence. The sensing module is used to collect and integrate personnel behavior data, duty status data, work environment data and equipment operation status data, and to obtain personnel safety management data from the existing digital management and control platform of the site, and output site safety management and control data. Furthermore, in the sensing module, the steps for collecting and integrating personnel behavior data, on-duty status data, work environment data, and equipment operating status data include: Data on human behavior is collected through the deployment of video cameras, infrared sensors, access control systems, and positioning sensors. Data on duty status is obtained by collecting duty records, operation logs, and work plans; Environmental sensors are used to collect temperature, humidity, wind speed, and dust concentration to obtain operational environment data. Real-time operating status data of equipment is collected through industrial bus, PLC interface and IoT interface; Personnel behavior data, duty status data, work environment data, and equipment operation status data are synchronized and integrated according to timestamps and spatial locations to output site safety management data.

[0013] Specifically, let the personnel behavior data be... ;in Indicates the first Personnel movement status and location coordinates within a specific time period; duty status data is as follows: , This represents duty logs, operation logs, and work plan information; the work environment data is... This represents environmental parameters at a specific time and location, including temperature, humidity, wind speed, and dust concentration; equipment operating status data is... , This represents the real-time operating indicators of the equipment, such as current, voltage, and speed. During data integration, a unified safety management data matrix can be formed through time alignment and spatial mapping. The calculation formula is as follows: ;in Represents functions for data synchronization, alignment, and integration. It includes the behavioral status of each person at each time point, their duty status, work environment, equipment operating status, and safety management information, and the output is... It includes information on personnel behavior status, duty conditions, working environment, and equipment operation status at each time point and spatial location, forming input data that can be used by the risk prediction module and behavior recognition module. This data is then used to calculate the dynamic safety risk prediction value and safety behavior recognition results of personnel, providing a foundation for the continuity of safety management and data-driven analysis.

[0014] Furthermore, within the sensing module, the steps for acquiring personnel safety management data from the existing digital control platform at the site include: Establish an interface communication connection with the existing digital management and control platform of the site, and verify data access permissions; Retrieve personnel safety management data, including training records, qualification certificates, historical violation records, and job safety scores; The acquired personnel safety management data is converted into a standardized data structure consistent with the collected on-site data; Personnel safety management data is integrated into the site safety control data and output to the risk prediction module and behavior recognition module.

[0015] Specifically, in the perception module, by establishing an interface communication connection with the existing digital management and control platform of the site and verifying data access permissions, personnel safety management data is acquired. This personnel safety management data includes training records, qualification certificates, historical violation records, and job safety scores. Let this data set be... ;in Indicates the first Safety management information for personnel. , This indicates training records. Indicates qualification certificate information, This indicates historical violation records. This indicates the job safety score, with data retrieved via platform API and mapped according to personnel ID. Subsequently... Converted to data collected on site A unified data structure and standardized mapping functions are provided. ;in This represents the process of uniformly formatting and encoding different data types, and the output is... This creates a standardized management data vector for each individual. Finally, The on-site data collected and integrated by the sensing module are fused according to timestamps and personnel IDs to form a complete site safety management and control data matrix. Output As input data for the risk prediction module and behavior recognition module, it is used to generate dynamic safety risk prediction values ​​and safety behavior recognition results for personnel, so as to realize safety management and early warning based on full data.

[0016] The risk prediction module has a built-in nonlinear time-series risk prediction model based on the evolution of personnel behavior. It takes the site safety management data output by the sensing module as input and outputs the dynamic safety risk prediction value of personnel. Furthermore, in the risk prediction module, the steps for outputting dynamic safety risk prediction values ​​for personnel, using the site safety management data output by the perception module as input, include: For the personnel behavior data, duty status data, work environment data, equipment operation status data and personnel safety management data output by the sensing module, time alignment, missing value filling and standardization are performed in sequence. Standardization includes unit unification and numerical normalization. Extract personnel behavioral characteristics, job condition characteristics, work environment characteristics, and equipment operation characteristics; The extracted features are input into a nonlinear time-series risk prediction model based on the evolution of human behavior to generate dynamic safety risk prediction values ​​for the corresponding personnel. The dynamic safety risk prediction values ​​are linked and organized with personnel ID, timestamp, and work scenario information, and then output to the behavior recognition module and the unmanned control module. A nonlinear time-series risk prediction model based on the evolution of human behavior is built with time-series decay calculation logic, behavior evolution inertia calculation logic, and risk propagation coefficient calculation logic.

[0017] Specifically, in the risk prediction module, the site safety management data output by the perception module is used as input, and this data matrix is ​​denoted as follows: ;in Represents personnel behavior data, This indicates the data on duty status. This represents the data related to the working environment. This indicates the equipment's operating status data. To represent personnel safety management data, firstly... Time alignment, missing value imputation, and standardization are performed sequentially. Standardization includes unit unification and numerical normalization. The processed data is denoted as... Then from Extracting feature vectors For personnel behavioral characteristics, Based on the characteristics of the job conditions, Due to the characteristics of the working environment, Based on the equipment's operational characteristics, the feature vector is then input into a nonlinear time-series risk prediction model for calculation. This model generates dynamic safety risk prediction values ​​according to the following relationship: ;in This represents the risk value at the previous time point. The time-series decay coefficient, The inertia coefficient for behavioral evolution. This represents the change in current behavioral characteristics. For risk propagation coefficient, Weighting of risk transmission among personnel Indicates with personnel The relevant personnel will be gathered, and ultimately... The output set is formed by associating personnel ID, timestamp, and work scenario information, and is used for subsequent processing by the behavior recognition module and the unmanned control module.

[0018] The behavior recognition module is used to determine the compliance of personnel behavior data, generate safety behavior recognition results, and generate graded early warning signals by combining the dynamic safety risk prediction value of personnel. Furthermore, in the behavior recognition module, the steps for determining the compliance of personnel behavior data and generating safe behavior recognition results include: The personnel behavior data output by the sensing module is organized and labeled according to personnel ID and time series; Extract behavioral features, including action type, duration, frequency, and degree of matching with work specifications; The behavioral characteristics are compared with a pre-established behavioral compliance rule base, which is a set of standardized operation compliance judgment rules pre-stored in the behavior recognition module. Determine whether each action complies with the work specifications; Generate security behavior identification results, including behavior ID, compliance judgment result, timestamp, and relevant personnel information.

[0019] Specifically, in the behavior recognition module, the personnel behavior data from the site safety management data output by the perception module is used as input. Let the input data be... ;in Indicates the first The behavioral sequences of individuals were collected at various time points, first categorized by individual ID and time sequence. Organize and annotate to generate Then from Extracting behavioral feature vectors ;in Encode the action category, Duration of the behavior For behavior frequency, To determine the degree of matching with the work specifications, the degree of matching By standardizing the comparison function Calculation, where For behavioral compliance rule base, This represents the rule matching scoring function, and its output is... Then, based on the judgment logic To determine whether the behavior conforms to the work specifications, This indicates compliance. This indicates non-compliance, among which This is a rule-based comparison function that compares action category, duration, behavior frequency, and matching degree with rule thresholds to generate a judgment result, ultimately producing a set of safe behavior recognition results. ;in For behavior timestamps, personnel Output the corresponding personnel identifier. This is used to generate tiered early warning signals by combining dynamic personnel safety risk prediction values. These signals serve as inputs for the unmanned control module to execute risk compensation and tiered automatic takeover strategies. (Feature parameters) The rule base is calculated from continuously collected behavioral data. It is generated by standardized work specifications, and the output results can reflect whether each behavior meets the work safety requirements, and can be used for subsequent safety management and early warning strategy triggering.

[0020] Furthermore, in the behavior recognition module, the steps for generating tiered early warning signals by combining dynamic safety risk prediction values ​​of personnel include: Receive the safety behavior recognition results generated by the behavior recognition module and the dynamic safety risk prediction values ​​of personnel output by the risk prediction module; The behavioral identification results are correlated and matched with the risk prediction values ​​according to the personnel ID and time period; The compliance results of behavior recognition and the risk prediction values ​​are input into the graded early warning judgment rules, which are a set of early warning level classification rules pre-stored in the behavior recognition module. Generate corresponding graded early warning signals based on the judgment rules; The graded early warning signals are associated with personnel ID, behavior ID, and timestamp, and the results are output for the closed-loop optimization module and the unmanned control module.

[0021] Specifically, in the behavior recognition module, the set of security behavior recognition results generated in the previous step is... The set of personnel dynamic safety risk prediction values ​​output by the risk prediction module As input, where For the results of the behavior compliance assessment, For behavior timestamps, personnel For personnel identification, For the corresponding personnel in time The risk prediction value, For work scenario information, first sort by personnel ID and time period. and Perform association matching to generate associated data Then associate the data Input the tiered early warning judgment rules ,in The rules represent the tiered warning levels, generating tiered warning signals by mapping risk thresholds and behavioral compliance. The graded early warning signal is combined with behavior ID, personnel ID, timestamp, and work scenario information to generate an output set. Output The data is transmitted to the closed-loop optimization module and the unmanned control module for closed-loop feedback updates of risk prediction model parameters, behavior recognition feature weights, and triggering rules, as well as for triggering the unmanned control module to execute dynamic risk compensation and graded automatic takeover strategies for unsupervised operations. and The timestamps are calculated by the behavior recognition module and the risk prediction module, respectively. With personnel Inherited from the original collected data, including work scenario information. Generated from site safety management data. It can reflect the risk status of personnel behavior in real time and guide subsequent safety management and early warning measures. In the behavior recognition module, personnel behavior data is input into a rule matching function to calculate the degree of matching between behavior and norms. The degree of matching is expressed as: ;in This represents a scoring function based on the degree of matching between behavioral data and the rule base. It is used to quantify the degree of consistency between behavior and work standards, and the output is used for subsequent behavior compliance determination.

[0022] The closed-loop optimization module is used to receive behavior recognition results, graded early warning signals and early warning handling closed-loop feedback data, and update the calculation parameters of the risk prediction module and the feature weights and graded early warning triggering rules of the behavior recognition module in reverse. Furthermore, in the closed-loop optimization module, the steps for receiving behavior recognition results, tiered early warning signals, and closed-loop feedback data for early warning handling include: The closed-loop optimization module establishes a reverse data interaction communication connection with the risk prediction module and the behavior recognition module, and receives the safety behavior recognition results through the data interface of the behavior recognition module. Receive graded early warning signals through the data interface of the behavior recognition module; Obtain closed-loop feedback data for early warning and response through the data interface of the centralized control display module or the unmanned control module; Time alignment and integration are performed on the received safety behavior identification results, graded early warning signals, and closed-loop feedback data of early warning response. The integrated data is stored in the local cache of the closed-loop optimization module for use in subsequent updates.

[0023] Specifically, in the closed-loop optimization module, a reverse data interaction communication connection is established with the risk prediction module and the behavior recognition module to receive the set of safety behavior recognition results output by the behavior recognition module. With graded early warning signal set Simultaneously, it receives a set of early warning and response closed-loop feedback data through the centralized control display module or the unmanned management module interface. The data is then aligned and integrated based on personnel ID and timestamp to form a unified dataset. The data is stored in the local cache of the closed-loop optimization module, and the data in the local cache is used for subsequent closed-loop optimization parameter update processing.

[0024] Furthermore, in the closed-loop optimization module, the steps for reversing the updating of the operational parameters of the risk prediction module and the feature weights and triggering rules of the behavior recognition module include: The integrated security behavior identification results, graded early warning signals, and early warning handling closed-loop feedback data are obtained from the local cache of the closed-loop optimization module. Deviation calculation is performed on the safety behavior identification results, graded early warning signals, and closed-loop feedback data; Based on the deviation calculation results, the operational parameters of the nonlinear time series risk prediction model are updated through the gradient descent iterative algorithm, including the risk propagation coefficient, the behavior evolution inertia coefficient, and the time series decay parameter. Adjust the feature weights of the behavior recognition module, including action category weight, behavior duration weight, behavior frequency weight, and work specification matching weight; Update the triggering rules for the tiered early warning module, including the early warning level thresholds corresponding to different risk values, the triggering conditions for early warning actions, and the job-specific tiered push strategy.

[0025] Specifically, in the closed-loop optimization module, the integrated set of security behavior recognition results is obtained from the local cache. Tiered early warning signal set Data set for closed-loop feedback of early warning and response ;in For the results of the behavior compliance assessment, For timestamps, personnel For personnel identification, The warning levels are divided into tiers. For work scenario information, To indicate the results of the early warning response, firstly... and Align and integrate data based on personnel ID and timestamp to form a unified dataset. Then the deviation was calculated. ;in This represents a level quantization function that maps warning levels to response results as continuous numerical values, used to unify the quantization representation of different types of data. Furthermore, the following optimization objective function is constructed: The parameter optimization objective is constructed by minimizing the sum of squared deviations. Based on the objective function, the deviations are backpropagated using a gradient descent iterative algorithm to update the operational parameters of the nonlinear time-series risk prediction model. Feature weights of the behavior recognition module and tiered early warning triggering rules ,in Includes risk propagation coefficient, behavioral evolution inertia coefficient, and time-series decay parameter. This includes weights for action category, duration of behavior, frequency of behavior, and matching degree to work specifications. It includes early warning level thresholds, early warning triggering conditions, and job posting strategies. The updated parameters are used for subsequent model calculations and safety management strategy adjustments to improve the accuracy of dynamic safety risk prediction and the reliability of behavior compliance judgment. It also provides an adjustment basis for the unmanned management module, enabling the optimization of dynamic risk compensation strategies and graded automatic takeover strategies for unsupervised operations.

[0026] The unmanned management module is used to determine unattended scenarios and, in conjunction with the dynamic safety risk prediction value and behavior recognition results of personnel, execute dynamic risk compensation strategies and graded automatic takeover strategies for unsupervised operations. Furthermore, in the unmanned control module, the steps for implementing the dynamic risk compensation strategy and the graded automatic takeover strategy for unsupervised operations include: Receive the dynamic safety risk prediction values ​​of personnel output by the risk prediction module and the safety behavior recognition results generated by the behavior recognition module; The predicted value of personnel dynamic safety risks is matched with the results of safety behavior identification and then matched with the corresponding policy level in the built-in policy library. The policy library is a set of execution rules corresponding to different risk levels that are pre-stored in the unmanned control module. Based on the matched strategy level, control instructions containing dynamic risk compensation actions or automatic takeover actions for unsupervised operations are generated and sent to the corresponding equipment or system for execution through the site equipment control communication interface; Dynamic risk compensation strategies include compensation for monitoring accuracy, compensation for early warning thresholds, compensation for monitoring redundancy, and compensation for access control. The graded automatic takeover strategy for unsupervised operations includes alert-level takeover, intervention-level takeover, control-level takeover, and emergency-level takeover.

[0027] Specifically, the unmanned control module receives a set of dynamic safety risk prediction values ​​for personnel from the risk prediction module. Set of security behavior recognition results from the behavior recognition module ;in This represents the predicted risk value. For the results of the behavior compliance assessment, For timestamps, For work scene information, personnel Personnel are identified by a policy library pre-stored within the unmanned control module. The matching process maps each individual action to its corresponding risk value and then to a policy level. ,when And determine the corresponding execution rules. Subsequently, according to Generate control commands The instructions include dynamic risk compensation actions and automatic takeover actions for unsupervised operations. Dynamic risk compensation actions cover monitoring accuracy compensation, early warning threshold compensation, monitoring redundancy compensation, and access control compensation. Automatic takeover actions for unsupervised operations are categorized into alert, intervention, control, and emergency levels, and are communicated through the station equipment control communication interface. The command is sent to the corresponding device or system for execution, and the execution result is as follows. Feedback can be fed back to the closed-loop optimization module to update the risk prediction model parameters and behavior recognition feature weights, so as to achieve dynamic control and safety assurance of personnel behavior and operational risks, and ensure that operational risks in unattended or minimally staffed scenarios are compensated in a timely manner and automatically intervened.

[0028] The centralized control and display module is used to render and output the site safety management data collected by the sensing module through a graphical interface, send the hierarchical early warning signals generated by the behavior recognition module to the corresponding terminal according to the job position, and receive the control instructions issued by the remote centralized control system and transmit them to the unmanned control module. Furthermore, in the centralized control and display module, the steps for sending the hierarchical early warning signals generated by the behavior recognition module to the corresponding job terminals according to job positions include: Acquire the graded early warning signals generated by the behavior recognition module and their corresponding job information; Obtain the personnel position information corresponding to the graded early warning signal. The personnel position information is the personnel position and terminal mapping data pre-stored in the centralized control and display module. Match the graded early warning signal with the position information to determine the target position terminal corresponding to each signal. Convert the graded early warning signals into a transmittable data format; Data is sent to the corresponding terminal via the network communication interface of the centralized control and display module; The terminal receives data and completes the display or notification processing.

[0029] Specifically, the centralized control and display module receives a set of tiered early warning signals from the behavior recognition module. With pre-stored personnel job and terminal mapping data Each tiered early warning signal is associated with its corresponding personnel position and terminal through matching rules. The matching formula is: terminal Among them, the risk level Indicates the level of security risk corresponding to the behavior. For timestamps, actions For behavioral identification, personnel For personnel identification, the matched signal set Convert to a transmittable data format The data is then sent to the corresponding terminal via the network communication interface of the centralized control and display module. After receiving the data, the terminal completes the display or notification processing and generates a visual output of the job position. Meanwhile, the centralized control and display module receives control commands issued by the remote centralized control system. The information is then transmitted to the unmanned control module to trigger corresponding equipment control or safety control actions. The entire process ensures that the hierarchical early warning information can be accurately transmitted and visualized according to the job position to support operational safety monitoring and risk response.

[0030] Example 2: In the second embodiment of the present invention, the present invention provides a method for monitoring and early warning of safety behaviors in unattended new energy power stations, such as... Figure 2 As shown, it includes the following steps: S1. Collect and integrate personnel behavior data, duty status data, work environment data and equipment operation status data, and obtain personnel safety management data from the existing digital management and control platform of the site, and output site safety management and control data; S2 has a built-in nonlinear time-series risk prediction model based on the evolution of personnel behavior. It takes the site safety management data output by S1 as input and outputs the predicted value of dynamic safety risks of personnel. S3. Perform compliance assessment on personnel behavior data to generate safety behavior identification results, and generate graded early warning signals by combining the dynamic safety risk prediction values ​​of personnel. S4 receives behavior recognition results, graded early warning signals and early warning handling closed-loop feedback data, and updates the operation parameters of S2 and the feature weights and graded early warning triggering rules of S3 in reverse. S5. Determine unattended scenarios and, in conjunction with the predicted values ​​of dynamic safety risks to personnel and the results of behavior recognition, execute dynamic risk compensation strategies and graded automatic takeover strategies for unsupervised operations. S6. Render and output the site safety control data collected by S1 through the graphical interface, send the graded early warning signals generated by S3 to the corresponding job terminals according to the job position, and at the same time receive the control instructions issued by the remote centralized control system and transmit them to S5.

[0031] In the unmanned operation of large-scale wind farms or photovoltaic power stations, the dispersed distribution of personnel, the unpredictable inspection routes, and the significant impact of weather and equipment operating conditions on the site environment easily lead to problems such as undetected violations, inability to dynamically assess accumulated risks, and delayed early warning information transmission. This results in difficulties in timely intervention of safety hazards, and technical issues such as insufficient precision and delayed response in operational safety management. To address these problems, this invention provides a safety behavior monitoring and early warning method for unmanned new energy power stations, the process of which is as follows: Figure 2 As shown. The specific implementation process of this method is as follows: S1. Collect and integrate personnel behavior data, duty status data, work environment data and equipment operation status data, and obtain personnel safety management data from the existing digital management and control platform of the site, output site safety management and control data. Through the unified integration of multi-source heterogeneous data, personnel behavior status, environmental changes and equipment operation status can be expressed under the same data system, thereby achieving a comprehensive perception of the on-site operation status and improving the integrity and consistency of the data foundation. S2 has a built-in nonlinear time-series risk prediction model based on the evolution of personnel behavior. It takes the site safety management data output by S1 as input and outputs the dynamic safety risk prediction value of personnel. By introducing time-series evolution and risk propagation mechanisms, it realizes the dynamic characterization of the future risk trend of personnel, thereby improving the foresight and continuity of risk identification. S3. Perform compliance assessment on personnel behavior data to generate safety behavior identification results, and combine them with the dynamic safety risk prediction value of personnel to generate graded early warning signals. By integrating the assessment of behavior compliance and risk value, the assessment can be transformed from a single rule judgment to a comprehensive risk assessment, thereby improving the accuracy and pertinence of early warning classification. S4 receives behavior recognition results, graded early warning signals, and closed-loop feedback data for early warning handling. It then updates the computational parameters of S2 and the feature weights and triggering rules of graded early warnings in reverse. Through closed-loop feedback, it drives the adaptive adjustment of model parameters and rules, enabling the system to continuously optimize according to actual operation, thereby improving the adaptability and reliability of the risk prediction model and behavior recognition results. S5. Determine the unattended scenario and combine the dynamic safety risk prediction value and behavior recognition results of personnel to implement dynamic risk compensation strategy and graded automatic takeover strategy for unsupervised operation. By matching corresponding compensation and takeover measures to different risk levels, a graded response from prompting intervention to automatic control is achieved, thereby ensuring operation safety and reducing reliance on human intervention under unattended or minimally staffed conditions. S6 renders and outputs the site safety management data collected by S1 through a graphical interface, sends the graded early warning signals generated by S3 to the corresponding job terminals according to job positions, and receives management instructions issued by the remote centralized control system and transmits them to S5. By realizing data visualization and accurate distribution of early warning information, different positions can obtain safety information related to their responsibilities in a timely manner, and support the linkage execution of remote centralized control instructions, thereby improving the overall safety management efficiency and response capability.

[0032] Finally, it should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A safety behavior monitoring and early warning system for unattended new energy power stations, characterized in that, include: The sensing module is used to collect and integrate personnel behavior data, duty status data, work environment data and equipment operation status data, and to obtain personnel safety management data from the existing digital management and control platform of the site, and output site safety management and control data. The risk prediction module has a built-in nonlinear time-series risk prediction model based on the evolution of personnel behavior. It takes the site safety management data output by the sensing module as input and outputs the dynamic safety risk prediction value of personnel. The behavior recognition module is used to determine the compliance of personnel behavior data, generate safety behavior recognition results, and generate graded early warning signals by combining the dynamic safety risk prediction value of personnel. The closed-loop optimization module is used to receive behavior recognition results, graded early warning signals and early warning handling closed-loop feedback data, and update the calculation parameters of the risk prediction module and the feature weights and graded early warning triggering rules of the behavior recognition module in reverse. The unmanned management module is used to determine unattended scenarios and, in conjunction with the dynamic safety risk prediction value and behavior recognition results of personnel, execute dynamic risk compensation strategies and graded automatic takeover strategies for unsupervised operations. The centralized control and display module is used to render and output the site safety management data collected by the sensing module through a graphical interface, send the hierarchical early warning signals generated by the behavior recognition module to the corresponding job terminals according to the job position, and receive the control instructions issued by the remote centralized control system and transmit them to the unmanned control module.

2. The safety behavior monitoring and early warning system for unmanned new energy power stations according to claim 1, characterized in that, In the sensing module, the steps of collecting and integrating personnel behavior data, duty status data, work environment data, and equipment operating status data include: Data on human behavior is collected through the deployment of video cameras, infrared sensors, access control systems, and positioning sensors. Data on duty status is obtained by collecting duty records, operation logs, and work plans; Environmental sensors are used to collect temperature, humidity, wind speed, and dust concentration to obtain operational environment data. Real-time operating status data of equipment is collected through industrial bus, PLC interface and IoT interface; Personnel behavior data, duty status data, work environment data, and equipment operation status data are synchronized and integrated according to timestamps and spatial locations to output site safety management data.

3. The safety behavior monitoring and early warning system for unattended new energy power stations according to claim 1, characterized in that, In the sensing module, the step of acquiring personnel safety management data from the existing digital management and control platform of the site includes: Establish an interface communication connection with the existing digital management and control platform of the site, and verify data access permissions; Retrieve personnel safety management data, including training records, qualification certificates, historical violation records, and job safety scores; The acquired personnel safety management data is converted into a standardized data structure consistent with the collected on-site data; Personnel safety management data is integrated into the site safety control data and output to the risk prediction module and behavior recognition module.

4. The safety behavior monitoring and early warning system for unattended new energy power stations according to claim 1, characterized in that, In the risk prediction module, the step of taking the site safety management data output by the sensing module as input and outputting the predicted value of dynamic personnel safety risks includes: The personnel behavior data, duty status data, work environment data, equipment operation status data and personnel safety management data output by the sensing module are sequentially aligned with time, filled with missing values ​​and standardized. The standardization process includes unit unification and numerical normalization. Extract personnel behavioral characteristics, job condition characteristics, work environment characteristics, and equipment operation characteristics; The extracted features are input into a nonlinear time-series risk prediction model based on the evolution of human behavior to generate dynamic safety risk prediction values ​​for the corresponding personnel. The dynamic safety risk prediction values ​​are linked and organized with personnel ID, timestamp, and work scenario information, and then output to the behavior recognition module and the unmanned control module. The nonlinear time-series risk prediction model based on the evolution of human behavior incorporates time-series decay calculation logic, behavior evolution inertia calculation logic, and risk propagation coefficient calculation logic.

5. The safety behavior monitoring and early warning system for unattended new energy power stations according to claim 1, characterized in that, In the behavior recognition module, the steps of performing compliance assessments on personnel behavior data to generate security behavior recognition results include: The personnel behavior data output by the sensing module is organized and labeled according to personnel ID and time series; Extract behavioral features, including action type, duration, frequency, and degree of matching with work specifications; The behavioral characteristics are compared with a pre-established behavioral compliance rule library, which is a set of standardized operation compliance judgment rules pre-stored in the behavior recognition module; Determine whether each action complies with the work specifications; Generate security behavior identification results, including behavior ID, compliance judgment result, timestamp, and relevant personnel information.

6. The safety behavior monitoring and early warning system for unattended new energy power stations according to claim 1, characterized in that, In the behavior recognition module, the step of generating a graded early warning signal by combining the predicted value of dynamic personnel safety risks includes: Receive the safety behavior recognition results generated by the behavior recognition module and the dynamic safety risk prediction values ​​of personnel output by the risk prediction module; The behavioral identification results are correlated and matched with the risk prediction values ​​according to the personnel ID and time period; The compliance results of behavior recognition and the risk prediction value are input into the graded early warning judgment rules, which are a set of early warning level classification rules pre-stored in the behavior recognition module; Generate corresponding graded early warning signals based on the judgment rules; The graded early warning signals are associated with personnel ID, behavior ID, and timestamp, and the results are output for the closed-loop optimization module and the unmanned control module.

7. The safety behavior monitoring and early warning system for unattended new energy power stations according to claim 1, characterized in that, In the closed-loop optimization module, the steps of receiving behavior recognition results, graded early warning signals, and closed-loop feedback data for early warning processing include: The closed-loop optimization module establishes a reverse data interaction communication connection with the risk prediction module and the behavior recognition module, and receives the safety behavior recognition results through the data interface of the behavior recognition module. Receive graded early warning signals through the data interface of the behavior recognition module; Obtain closed-loop feedback data for early warning and response through the data interface of the centralized control display module or the unmanned control module; Time alignment and integration are performed on the received safety behavior identification results, graded early warning signals, and closed-loop feedback data of early warning response. The integrated data is stored in the local cache of the closed-loop optimization module for use in subsequent updates.

8. The safety behavior monitoring and early warning system for unattended new energy power stations according to claim 1, characterized in that, In the closed-loop optimization module, the steps of reversing the updating of the operational parameters of the risk prediction module and the feature weights and hierarchical warning triggering rules of the behavior recognition module include: The integrated security behavior identification results, graded early warning signals, and early warning handling closed-loop feedback data are obtained from the local cache of the closed-loop optimization module. Deviation calculation is performed on the safety behavior identification results, graded early warning signals, and closed-loop feedback data; Based on the deviation calculation results, the operational parameters of the nonlinear time series risk prediction model are updated through the gradient descent iterative algorithm, including the risk propagation coefficient, the behavior evolution inertia coefficient, and the time series decay parameter. Adjust the feature weights of the behavior recognition module, including action category weight, behavior duration weight, behavior frequency weight, and work specification matching weight; Update the triggering rules for the tiered early warning module, including the early warning level thresholds corresponding to different risk values, the triggering conditions for early warning actions, and the job-specific tiered push strategy.

9. The safety behavior monitoring and early warning system for unattended new energy power stations according to claim 1, characterized in that, In the unmanned management module, the steps for implementing the dynamic risk compensation strategy and the graded automatic takeover strategy for unsupervised operations include: Receive the dynamic safety risk prediction values ​​of personnel output by the risk prediction module and the safety behavior recognition results generated by the behavior recognition module; The predicted value of dynamic safety risks of personnel is matched with the results of safety behavior identification and then matched with the corresponding policy level in the built-in policy library. The policy library is a set of execution rules corresponding to different risk levels that are pre-stored in the unmanned control module. Based on the matched strategy level, control instructions containing dynamic risk compensation actions or automatic takeover actions for unsupervised operations are generated and sent to the corresponding equipment or system for execution through the site equipment control communication interface; The dynamic risk compensation strategy includes monitoring accuracy compensation, early warning threshold compensation, monitoring redundancy compensation, and access control compensation. The hierarchical automatic takeover strategy for unsupervised operations includes alert-level takeover, intervention-level takeover, control-level takeover, and emergency-level takeover.

10. A safety behavior monitoring and early warning system for an unattended new energy power station according to claim 1, characterized in that, In the centralized control and display module, the step of sending the hierarchical early warning signal generated by the behavior recognition module to the corresponding job terminal according to the job position includes: Acquire the graded early warning signals generated by the behavior recognition module and their corresponding job information; Obtain the personnel position information corresponding to the graded early warning signal. The personnel position information is the personnel position and terminal mapping data pre-stored in the centralized control and display module. Match the graded early warning signal with the position information to determine the target position terminal corresponding to each signal. Convert the graded early warning signals into a transmittable data format; Data is sent to the corresponding terminal via the network communication interface of the centralized control and display module; The terminal receives data and completes the display or notification processing.

Images