A substation auxiliary monitoring system, device and medium

Abstract

This invention provides a substation auxiliary monitoring system, equipment, and medium, relating to the field of intelligent substation monitoring technology. It includes a data acquisition module for collecting and preprocessing equipment status, environmental parameters, and personnel behavior data, outputting a monitoring dataset that is then merged and transmitted to an adaptive processing module. The adaptive processing module extracts and fuses a set of monitoring features to obtain comprehensive monitoring features, trains and improves a risk assessment model to evaluate these features, outputs the risk assessment result, and transmits it to an early warning and linkage module. The early warning and linkage module determines the result, generates an alarm signal, and executes a response through linkage equipment. A hardware platform supports the operation of all the above modules, enabling task scheduling between edge computing nodes and the cloud platform, monitoring system operating status, and updating the historical monitoring database. This reduces the false alarm rate, balances monitoring effectiveness and privacy compliance, lowers hardware maintenance costs, and improves system scalability and emergency response efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent substation monitoring technology, and in particular to a substation auxiliary monitoring system, equipment and medium. Background Technology

[0002] As a crucial hub in power grid operation, the safe and stable operation of substations directly affects the reliability of power supply. Intelligent auxiliary monitoring is a key technology to ensure the safety of substations.

[0003] Existing substation monitoring systems largely employ an independent and decentralized approach, with equipment monitoring, environmental monitoring, and personnel management operating independently. This lack of data sharing and the creation of information silos hinders comprehensive security control. Furthermore, traditional monitoring algorithms are fixed, hardware configurations are redundant, and high-load computation can sacrifice real-time data performance, making them unsuitable for dynamic substation operation scenarios. In addition, existing personnel monitoring focuses solely on behavioral capture, lacking privacy protection features, which can lead to privacy breaches and violate ethical standards and data security requirements.

[0004] Therefore, it is necessary to provide a substation auxiliary monitoring system, equipment, and medium to solve the above-mentioned technical problems. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a substation auxiliary monitoring system, equipment, and medium, which solves the problems of data silos, high false alarm rate in risk assessment, blind spots in monitoring, lack of privacy protection, high hardware maintenance costs, and poor system scalability in existing substation auxiliary monitoring technologies.

[0006] In a first aspect, the present invention provides a substation auxiliary monitoring system, comprising: The data acquisition module is used to synchronously collect equipment status information, real-time environmental parameters and personnel behavior data through a multimodal sensor network and dynamic tag management unit deployed in multiple areas of the substation, and to perform timestamp calibration and format unification. It is an adaptive processing module that outputs multimodal monitoring datasets and merges them to transmit them to edge computing nodes. The adaptive processing module is used to perform real-time cleaning and feature extraction on the multimodal monitoring data set through the built-in streaming data processing engine to obtain a multimodal monitoring feature set. It also integrates multi-sensor data fusion technology to perform multi-dimensional fusion on the multimodal monitoring feature set to obtain comprehensive monitoring features. Combined with a reinforcement learning model trained on a historical monitoring database, it dynamically optimizes the risk assessment model to obtain an improved risk assessment model. Based on the improved risk assessment model, it performs risk assessment on the comprehensive monitoring features and outputs a comprehensive risk assessment result including regional hazard level coefficient and personnel behavior risk index, which is then transmitted to the early warning and linkage module. The early warning and linkage module is used to classify the comprehensive risk assessment results according to the preset risk level classification rules and generate a graded alarm signal. At the same time, it links the ventilation equipment, access control system and visual interface in the substation to perform corresponding response operations and outputs linkage execution status feedback. The hardware platform is used to support the operation of the data acquisition module, the adaptive processing module, and the early warning and linkage module. It utilizes a cloud-edge collaborative architecture to realize data synchronization and task scheduling between the edge computing nodes and the cloud platform, performs real-time monitoring and fault tolerance processing of the system operation status, outputs system operation status reports, and updates the historical monitoring database.

[0007] Preferably, the adaptive processing module, which synchronously collects equipment status information, real-time environmental parameters, and personnel behavior data through a multimodal sensor network and dynamic tag management unit deployed in multiple areas of the substation, performs timestamp calibration and format unification, and outputs a multimodal monitoring dataset for merging and transmission to the edge computing node, includes: Based on the substation's topology and equipment distribution, a multimodal sensor network is deployed in key equipment areas, environmentally sensitive areas, and personnel access areas. The multimodal sensor network includes at least visual sensors, infrared sensors, radar, UWB positioning sensors, temperature sensors, humidity sensors, and SF6 gas concentration sensors. The hardware interrupt mechanism controls all sensors in the multimodal sensor network to start data acquisition at the same time. The system synchronously collects equipment status information including equipment aging degree, equipment density and equipment electrical load, real-time environmental parameters including temperature, humidity, SF6 gas concentration and external meteorological data, and personnel behavior data including personnel action images captured by the visual sensor and personnel location information obtained by radio frequency identification tags. The system outputs a time-synchronized set of raw monitoring data. The edge computing node preprocesses the original monitoring data set, including removing abnormal data, filling in missing data, and normalizing data, and outputs a multimodal monitoring data set. The dynamic tag management unit assigns unique encrypted RFID tags to all equipment or personnel in the substation, monitors the equipment or personnel information corresponding to the unique encrypted RFID tags in real time and dynamically updates the unique encrypted RFID tags, outputs tag dynamic management data and integrates it into the multimodal monitoring data set, and transmits the integrated multimodal monitoring data set to the adaptive processing module of the edge computing node.

[0008] Preferably, the step of assigning unique encrypted RFID tags to equipment or personnel within the substation through the dynamic tag management unit, monitoring the equipment or personnel information corresponding to the unique encrypted RFID tags in real time and dynamically updating the unique encrypted RFID tags, outputting tag dynamic management data and integrating it into the multimodal monitoring data set includes: The dynamic tag management unit assigns a unique encrypted RFID tag to the equipment or personnel in the substation, associates and binds the ID of the unique encrypted RFID tag with the equipment information or personnel information, and stores it in the RFID tag database of the cloud platform. The unique encrypted RFID tag is read in real time by RFID readers deployed at the entrance and exit of the substation and in key reading areas. When a change is detected in the equipment information or personnel information associated with the unique encrypted RFID tag, an RFID tag update request is automatically sent to the cloud platform. After receiving the RFID tag update request, the cloud platform performs identity verification and permission review on the RFID tag update request. After the identity verification and permission review are passed, the unique encrypted RFID tag is updated in real time, the tag dynamic management data is output and integrated into the multimodal monitoring data set; When the substation is expanded or upgraded, the unique encrypted RFID tag is automatically assigned to the newly added equipment or personnel, and the reading range of the unique encrypted RFID tag and the deployment location of the RFID reader are adjusted according to the topology of the expanded substation.

[0009] Preferably, the improved risk assessment model is obtained by dynamically optimizing the risk assessment model using a reinforcement learning model trained on a historical monitoring database. Based on the improved risk assessment model, a risk assessment is performed on the comprehensive monitoring features, and a comprehensive risk assessment result including a regional risk level coefficient and a personnel behavior risk index is output and transmitted to the early warning and linkage module, including: Historical operation data, historical fault records, and historical safety event data of the substation within a historical time period are extracted from the historical monitoring database of the cloud platform to construct a risk assessment training dataset. The risk assessment training dataset is then labeled and classified, and the labeled risk assessment training dataset is output. The risk assessment model based on fuzzy logic rules is trained based on the labeled risk assessment training dataset. The membership function and weight coefficient of the fuzzy logic rules are determined. At the same time, the reinforcement learning model is trained as a reinforcement learning optimizer, and the initial risk assessment model and the reinforcement learning optimizer are output. The comprehensive monitoring features are input into the initial risk assessment model to obtain a preliminary risk assessment result. The preliminary risk assessment result is compared with the current security event data to generate a reward signal, which is then input into the reinforcement learning optimizer. The membership function and weight coefficient of the fuzzy logic rule are dynamically adjusted through the reinforcement learning optimizer to output the improved risk assessment model. The comprehensive monitoring features are input into the improved risk assessment model to generate a comprehensive risk assessment result that includes the regional risk level coefficient and the personnel behavior risk index, and then transmitted to the early warning and linkage module.

[0010] Preferably, the step of classifying the comprehensive risk assessment results according to preset risk level classification rules to generate graded alarm signals, and simultaneously linking the ventilation equipment, access control system, and visual interface within the substation to perform corresponding response operations, and outputting linkage execution status feedback, includes: Based on the range of values ​​for the regional risk level coefficient and the personnel behavior risk index in the comprehensive risk assessment results, the comprehensive risk is divided into four levels: Level 1, Level 2, Level 3, and Level 4. Alarm methods, response procedures, and a list of linked devices are formulated for each risk level, and a hierarchical early warning rule library is output. The comprehensive risk assessment results are matched with the hierarchical early warning rule base to determine the current risk level and generate corresponding alarm signals, including at least audible and visual alarms, SMS alarms and platform pop-up alarms. At the same time, the time and location of the alarm are recorded and the alarm event record is output. Based on the alarm event records and the hierarchical early warning rule base, control commands are sent to the corresponding linked devices. When an SF6 gas leak is detected, the ventilation equipment is activated to start air exchange. When personnel are detected entering a dangerous area, the access control system is activated to close the passage. When an equipment malfunction is detected, the visualization interface is activated to display equipment information, and the linkage execution status feedback is output.

[0011] Preferably, the visualization interface in the early warning and linkage module provides contextual analysis functionality for the personnel behavior data, specifically including: Collect contextual information of the personnel behavior data, including the time and location of the personnel behavior, personnel identity, current task type and surrounding equipment operating status, extract personnel behavior features corresponding to the personnel behavior data and associate them with the contextual information, and output a personnel behavior dataset with context labels; The context-labeled personnel behavior dataset is input into a pre-trained AI behavior classification model. The AI ​​behavior classification model identifies the differences in contextual features between normal behavior data and abnormal behavior data, thus distinguishing between the normal behavior data and the abnormal inspection data. The abnormal behavior data is verified, and the duration, movement trajectory and impact range of the abnormal behavior data are analyzed in combination with historical personnel behavior records to eliminate misjudgments caused by environmental interference or normal operation, and output true abnormal inspection data. An abnormal behavior analysis report is generated based on the actual abnormal behavior data and pushed to the visualization interface in real time.

[0012] Preferably, the hardware platform is used to support the operation of the data acquisition module, the adaptive processing module, and the early warning and linkage module. It utilizes a cloud-edge collaborative architecture to achieve data synchronization and task scheduling between the edge computing nodes and the cloud platform, performs real-time monitoring and fault tolerance handling of the system's operating status, outputs a system operating status report, and updates the historical monitoring database, including: The hardware platform is built using a plug-and-play modular design, with the data acquisition module, the adaptive processing module, and the early warning and linkage module designed as independent functional modules. The number and type of the functional modules are flexibly configured according to the scale and requirements of the substation. The rotatable monitor in the hardware platform is equipped with a non-contact ultrasonic self-cleaning device. The non-contact ultrasonic self-cleaning device generates high-frequency vibrations through an ultrasonic generator, causing dust, moisture and dirt to fall off the surface of the monitor lens. It automatically adjusts the frequency and working duration of the ultrasonic waves according to the ambient dust concentration and provides real-time feedback of cleaning control signals and lens cleaning status data. The system adopts a hybrid power supply method that combines solar power and grid power. The hardware platform is equipped with solar panels and energy storage batteries. The system energy consumption is optimized through low-power chips and dynamic power management technology. The power supply mode is automatically switched according to the light intensity and system load, and energy consumption management data and power status information are fed back in real time. Multiple edge computing nodes are deployed in the hardware platform. Through a cloud-edge collaborative architecture, the edge computing nodes and the cloud platform achieve bidirectional data synchronization. When an edge computing node fails, its tasks are automatically migrated to other normally operating edge computing nodes, and the fault tolerance processing results and system reliability data are fed back in real time.

[0013] Preferably, it also includes a privacy protection unit, which processes the personnel behavior data, specifically including: The edge computing node performs differential privacy desensitization processing on the collected personnel behavior data, which masks sensitive behavioral details by adding noisy data and outputs desensitized personnel behavior data. Based on the desensitized personnel behavior data, the personnel identity information, personnel facial features and personnel location information are anonymized in the visualization interface, and anonymized display data is output. Establish a role-based data access permission hierarchical mechanism, classify personnel into operation and maintenance level, management level and security level, assign different data access permissions to personnel of different levels, record all personnel's data access operations and generate data access logs, and output permission management results and access audit reports; The personnel behavior data is encrypted using national cryptographic algorithms during processing, and is stored on the cloud platform using block encryption and distributed storage methods.

[0014] Secondly, the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor performs the function of a substation auxiliary monitoring system.

[0015] Thirdly, the present invention provides a readable storage medium storing a computer program, which, when executed by a processor, is used to implement the functions of a substation auxiliary monitoring system.

[0016] Compared with related technologies, the substation auxiliary monitoring system, equipment, and medium provided by the present invention have the following beneficial effects: This invention includes a data acquisition module for synchronously collecting equipment status information, real-time environmental parameters, and personnel behavior data through a multimodal sensor network and dynamic tag management unit deployed in multiple areas of a substation. The module performs timestamp calibration and format unification, outputting a multimodal monitoring dataset which is then merged and transmitted to an edge computing node. The adaptive processing module uses a built-in streaming data processing engine to perform real-time cleaning and feature extraction on the multimodal monitoring data set to obtain a multimodal monitoring feature set. It then integrates multi-sensor data fusion technology to perform multi-dimensional fusion of the multimodal monitoring feature set to obtain comprehensive monitoring features. Finally, it combines a reinforcement learning model trained on a historical monitoring database to dynamically optimize a risk assessment model, resulting in an improved risk assessment model. Based on this improved model, it performs a risk assessment on the comprehensive monitoring features, outputting data including regional hazard coefficients and personnel behavior data. The system generates a comprehensive risk assessment result based on the risk index and transmits it to the early warning and linkage module. This module, according to preset risk level classification rules, categorizes the comprehensive risk assessment result into different levels, generating tiered alarm signals. Simultaneously, it links the ventilation equipment, access control system, and visual interface within the substation to execute corresponding response operations, outputting linkage execution status feedback. The hardware platform supports the operation of the data acquisition module, adaptive processing module, and early warning and linkage module. Utilizing a cloud-edge collaborative architecture, it achieves data synchronization and task scheduling between edge computing nodes and the cloud platform, enabling real-time monitoring and fault tolerance of the system's operating status. It outputs system operating status reports and updates the historical monitoring database, thereby breaking down data silos, reducing false alarm rates, eliminating monitoring blind spots, balancing monitoring effectiveness and privacy compliance, reducing hardware maintenance costs, and improving system scalability and emergency response efficiency.

[0017] This invention adapts to environmental changes, reduces subjective bias, and improves the accuracy and reliability of risk assessment. Compared to existing technologies that rely on fixed formulas, this invention achieves continuous optimization through training with historical data, reducing false alarm rates. This invention eliminates monitoring blind spots (such as corners or low-light areas), enhancing the continuous tracking capability of personnel and equipment. Compared to existing technologies that are susceptible to tag damage or visual conditions, this invention ensures full coverage through data complementarity, enhancing system robustness. This invention's system reduces wear on moving parts, extends hardware lifespan, and lowers maintenance frequency and costs. Compared to existing technologies that rely on gravity drives and gear transmissions, this invention's non-contact cleaning method is more efficient. While ensuring monitoring effectiveness, this invention prevents data leakage and malicious tampering, enhancing user trust and compliance. It also supports real-time response to emergencies, and its modular design facilitates the expansion of new sensors or devices. Compared to the high integration risks of existing technologies, this invention's redundant design prevents single-point failures, reduces false alarms (such as normal inspections being misjudged as abnormal), and improves the overall operation and maintenance efficiency of substations through integration with the power grid system. Attached Figure Description

[0018] Figure 1 This is a system block diagram of a substation auxiliary monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit this application.

[0020] like Figure 1 As shown, an embodiment of the present invention discloses a substation auxiliary monitoring system, comprising: The data acquisition module is used to synchronously collect equipment status information, real-time environmental parameters and personnel behavior data through a multimodal sensor network and dynamic tag management unit deployed in multiple areas of the substation, and to perform timestamp calibration and format unification. It is an adaptive processing module that outputs multimodal monitoring datasets and merges them to transmit them to edge computing nodes. The adaptive processing module is used to perform real-time cleaning and feature extraction on the multimodal monitoring data set through the built-in streaming data processing engine to obtain a multimodal monitoring feature set. It also integrates multi-sensor data fusion technology to perform multi-dimensional fusion on the multimodal monitoring feature set to obtain comprehensive monitoring features. Combined with a reinforcement learning model trained on a historical monitoring database, it dynamically optimizes the risk assessment model to obtain an improved risk assessment model. Based on the improved risk assessment model, it performs risk assessment on the comprehensive monitoring features and outputs a comprehensive risk assessment result including regional hazard level coefficient and personnel behavior risk index, which is then transmitted to the early warning and linkage module. The early warning and linkage module is used to classify the comprehensive risk assessment results according to the preset risk level classification rules and generate a graded alarm signal. At the same time, it links the ventilation equipment, access control system and visual interface in the substation to perform corresponding response operations and outputs linkage execution status feedback. The hardware platform is used to support the operation of the data acquisition module, the adaptive processing module, and the early warning and linkage module. It utilizes a cloud-edge collaborative architecture to realize data synchronization and task scheduling between the edge computing nodes and the cloud platform, performs real-time monitoring and fault tolerance processing of the system operation status, outputs system operation status reports, and updates the historical monitoring database.

[0021] In practical applications, the data acquisition module achieves full-dimensional data collection through a multi-modal sensor network deployed in different zones. Visual, infrared, radar, UWB positioning, and gas sensors are configured for key equipment areas, environmentally sensitive areas, and personnel access areas, respectively. A hardware interrupt synchronization mechanism controls all sensors to start collecting data simultaneously, ensuring that the timestamp error between equipment status, environmental parameters, and personnel behavior data is less than 1ms. A dynamic tag management unit is integrated synchronously, assigning unique encrypted RFID tags to all equipment and personnel to achieve real-time association between identity and location. After data collection, outlier removal, missing data imputation, and normalization are performed locally, outputting a standardized multi-modal monitoring dataset which is then merged and transmitted to the edge computing node, solving the problems of data asynchrony and single-dimensionality in existing technologies.

[0022] The adaptive processing module utilizes a built-in low-latency streaming data processing engine to perform real-time cleaning and feature extraction on the input multimodal data, generating a multimodal monitoring feature set that includes equipment status, environmental changes, and personnel behavior. Multi-sensor data fusion technology achieves spatiotemporal alignment and complementary fusion of data from different sensors, eliminating monitoring blind spots such as corners and low-light conditions. A reinforcement learning-optimized fuzzy logic risk assessment model is employed. The initial model is trained based on labeled data from a historical monitoring database. The membership function and weight coefficients are dynamically adjusted using reward signals generated from real-time running data, outputting a comprehensive risk assessment result that includes regional hazard coefficients and personnel behavior risk indices. This addresses the shortcomings of traditional fixed algorithms, such as high false alarm rates and poor adaptability.

[0023] The early warning and linkage module enables tiered response and coordinated handling of risks. Based on a pre-defined four-level risk classification rule, this module matches and determines the comprehensive risk assessment results, generating corresponding audible and visual alarm signals, SMS alerts, or platform alarm signals. Simultaneously, it sends linkage control commands to ventilation equipment, access control systems, and visual interfaces, enabling functions such as automatic ventilation of SF6 leaks, access control locking of hazardous areas, and visual display of equipment anomalies. It also provides real-time feedback on the linkage execution status, ensuring rapid response to emergencies.

[0024] The hardware platform adopts a modular, plug-and-play design, supporting the operation of all the aforementioned modules. A cloud-edge collaborative architecture enables bidirectional data synchronization and task scheduling between edge computing nodes and the cloud platform, deploying multiple redundant edge nodes to achieve automatic fault migration and fault tolerance. It also monitors the system's operational status in real time, generates operational reports, and synchronously updates the historical monitoring database, improving the system's scalability and reliability.

[0025] In the specific implementation process, the adaptive processing module, which synchronously collects equipment status information, real-time environmental parameters, and personnel behavior data through a multimodal sensor network and dynamic tag management unit deployed in multiple areas of the substation, performs timestamp calibration and format unification, and outputs a multimodal monitoring dataset for merging and transmission to the edge computing node, includes: Based on the substation's topology and equipment distribution, a multimodal sensor network is deployed in key equipment areas, environmentally sensitive areas, and personnel access areas. The multimodal sensor network includes at least visual sensors, infrared sensors, radar, UWB positioning sensors, temperature sensors, humidity sensors, and SF6 gas concentration sensors. The hardware interrupt mechanism controls all sensors in the multimodal sensor network to start data acquisition at the same time. The system synchronously collects equipment status information including equipment aging degree, equipment density and equipment electrical load, real-time environmental parameters including temperature, humidity, SF6 gas concentration and external meteorological data, and personnel behavior data including personnel action images captured by the visual sensor and personnel location information obtained by radio frequency identification tags. The system outputs a time-synchronized set of raw monitoring data. The edge computing node preprocesses the original monitoring data set, including removing abnormal data, filling in missing data, and normalizing data, and outputs a multimodal monitoring data set. The dynamic tag management unit assigns unique encrypted RFID tags to all equipment or personnel in the substation, monitors the equipment or personnel information corresponding to the unique encrypted RFID tags in real time and dynamically updates the unique encrypted RFID tags, outputs tag dynamic management data and integrates it into the multimodal monitoring data set, and transmits the integrated multimodal monitoring data set to the adaptive processing module of the edge computing node.

[0026] The process involves assigning unique encrypted RFID tags to equipment or personnel within the substation through the dynamic tag management unit, monitoring the equipment or personnel information corresponding to the unique encrypted RFID tags in real time and dynamically updating the unique encrypted RFID tags, outputting tag dynamic management data and integrating it into the multimodal monitoring data set, including: The dynamic tag management unit assigns a unique encrypted RFID tag to the equipment or personnel in the substation, associates and binds the ID of the unique encrypted RFID tag with the equipment information or personnel information, and stores it in the RFID tag database of the cloud platform. The unique encrypted RFID tag is read in real time by RFID readers deployed at the entrance and exit of the substation and in key reading areas. When a change is detected in the equipment information or personnel information associated with the unique encrypted RFID tag, an RFID tag update request is automatically sent to the cloud platform. After receiving the RFID tag update request, the cloud platform performs identity verification and permission review on the RFID tag update request. After the identity verification and permission review are passed, the unique encrypted RFID tag is updated in real time, the tag dynamic management data is output and integrated into the multimodal monitoring data set; When the substation is expanded or upgraded, the unique encrypted RFID tag is automatically assigned to the newly added equipment or personnel, and the reading range of the unique encrypted RFID tag and the deployment location of the RFID reader are adjusted according to the topology of the expanded substation.

[0027] Understandably, differentiated sensor deployment is implemented based on the substation topology and equipment risk distribution. The substation is divided into critical equipment areas, environmentally sensitive areas, and personnel access areas, with targeted configuration of visual sensors, infrared sensors, radar, UWB positioning sensors, and environmental parameter sensors to construct a fully covered, blind-spot-free multimodal sensor network.

[0028] A hardware interrupt synchronization mechanism is employed to control all sensors to start acquiring data simultaneously, ensuring that the timestamp error of the three types of data is less than 1ms. This results in the output of a time-aligned set of raw monitoring data, resolving the correlation analysis error problem caused by asynchronous multi-source data in existing technologies. The acquired data specifically covers equipment status information such as equipment aging and electrical load, environmental parameters such as temperature and humidity and SF6 gas concentration, and personnel behavior data such as images of personnel movements and location information.

[0029] The raw data is preprocessed locally at the edge computing nodes. Outliers are removed using the 3σ criterion, and missing data is filled using a combination of linear interpolation and historical data prediction. Data of different dimensions are normalized in the [0,1] interval to output a standardized multimodal monitoring dataset, thereby improving the accuracy and efficiency of subsequent analysis.

[0030] The dynamic tag management unit enables full lifecycle tag management for equipment and personnel. It assigns unique encrypted RFID tags to all objects and completes information association and binding, monitors tag status in real time, and dynamically updates tag IDs and permissions. It also supports automatic tag system adaptation during substation expansion. The dynamic tag management data is integrated into a standardized monitoring data set and transmitted to the adaptive processing module.

[0031] The improved risk assessment model is obtained by dynamically optimizing the risk assessment model using a reinforcement learning model trained on a historical monitoring database. Based on this improved model, a risk assessment is performed on the comprehensive monitoring features, outputting a comprehensive risk assessment result including a regional hazard coefficient and a personnel behavior risk index, and transmitting it to the early warning and linkage module. This includes: Historical operation data, historical fault records, and historical safety event data of the substation within a historical time period are extracted from the historical monitoring database of the cloud platform to construct a risk assessment training dataset. The risk assessment training dataset is then labeled and classified, and the labeled risk assessment training dataset is output. The risk assessment model based on fuzzy logic rules is trained based on the labeled risk assessment training dataset. The membership function and weight coefficient of the fuzzy logic rules are determined. At the same time, the reinforcement learning model is trained as a reinforcement learning optimizer, and the initial risk assessment model and the reinforcement learning optimizer are output. The comprehensive monitoring features are input into the initial risk assessment model to obtain a preliminary risk assessment result. The preliminary risk assessment result is compared with the current security event data to generate a reward signal, which is then input into the reinforcement learning optimizer. The membership function and weight coefficient of the fuzzy logic rule are dynamically adjusted through the reinforcement learning optimizer to output the improved risk assessment model. The comprehensive monitoring features are input into the improved risk assessment model to generate a comprehensive risk assessment result that includes the regional risk level coefficient and the personnel behavior risk index, and then transmitted to the early warning and linkage module.

[0032] Historical operation data, historical fault records, and historical safety event data of the substation within a preset historical time period are extracted from the historical monitoring database of the cloud platform. After data cleaning and deduplication, the data is labeled and classified according to event types such as equipment failure, environmental anomaly, and personnel violation, and a set of labeled risk assessment training data is output.

[0033] An initial risk assessment model based on fuzzy logic rules is trained using a labeled risk assessment training dataset. The shape of the membership function and the initial weight coefficients corresponding to the fuzzy logic rules are determined by gradient descent. At the same time, a reinforcement learning model is trained as a reinforcement learning optimizer. The matching degree between the risk assessment results and actual events is used as the optimization objective. The initial risk assessment model and the reinforcement learning optimizer are output, which solves the defect of traditional fixed algorithms that cannot adapt to the dynamic operating environment of substations.

[0034] The real-time generated comprehensive monitoring features are input into the initial risk assessment model to obtain preliminary risk assessment results. These preliminary results are compared with current actual safety event data, and a corresponding reward signal is generated based on the matching degree and input into a reinforcement learning optimizer. Through iterative updates of the reinforcement learning optimizer, the membership function parameters and weight coefficients of the fuzzy logic rules are dynamically adjusted, outputting an improved risk assessment model that continuously adapts to the substation's operating environment.

[0035] The comprehensive monitoring features are input into the improved risk assessment model to generate the regional risk level coefficient and personnel behavior risk index in the [0,1] interval, which together form the comprehensive risk assessment results and are transmitted to the early warning and linkage module.

[0036] The comprehensive risk assessment results are classified and determined according to a preset risk level classification rule to generate a graded alarm signal. Simultaneously, the ventilation equipment, access control system, and visual interface within the substation are linked to execute corresponding response operations, and the linked execution status feedback is output, including: Based on the range of values ​​for the regional risk level coefficient and the personnel behavior risk index in the comprehensive risk assessment results, the comprehensive risk is divided into four levels: Level 1, Level 2, Level 3, and Level 4. Alarm methods, response procedures, and a list of linked devices are formulated for each risk level, and a hierarchical early warning rule library is output. The comprehensive risk assessment results are matched with the hierarchical early warning rule base to determine the current risk level and generate corresponding alarm signals, including at least audible and visual alarms, SMS alarms and platform pop-up alarms. At the same time, the time and location of the alarm are recorded and the alarm event record is output. Based on the alarm event records and the hierarchical early warning rule base, control commands are sent to the corresponding linked devices. When an SF6 gas leak is detected, the ventilation equipment is activated to start air exchange. When personnel are detected entering a dangerous area, the access control system is activated to close the passage. When an equipment malfunction is detected, the visualization interface is activated to display equipment information, and the linkage execution status feedback is output.

[0037] Based on the range of values ​​for the regional hazard coefficient and the personnel behavior risk index in the comprehensive risk assessment results, the substation operation risk is divided into four levels, from Level 1 to Level 4. The alarm trigger threshold, alarm method, emergency response process and linkage equipment list corresponding to each level are clearly defined, forming a configurable hierarchical early warning rule base.

[0038] The real-time comprehensive risk assessment results output by the adaptive processing module are matched with thresholds in the hierarchical early warning rule base to automatically determine the current risk level. Based on the matching results, multi-dimensional alarm signals corresponding to the level are generated, including on-site audible and visual alarms, SMS alarms for maintenance personnel, and pop-up alarms on the monitoring center platform. At the same time, the precise time, geographical coordinates, and risk type of the alarm are recorded, and a structured alarm event record is output.

[0039] Based on alarm event records and a tiered early warning rule base, standardized control commands are issued to designated linked devices. In the case of SF6 gas leaks, the linked ventilation equipment initiates tiered air exchange; in the case of unauthorized personnel entering hazardous areas, the linked access control system locks the passage; in the case of abnormal equipment operation, the linked visual interface automatically retrieves real-time equipment parameters, historical operating curves, and maintenance manuals, achieving automated and precise risk handling.

[0040] Real-time collection of operational status data from various linked devices verifies the execution effect of control commands, generates linkage execution status feedback including command issuance time, device response time, and execution result, and synchronously uploads it to the historical monitoring database of the cloud platform.

[0041] The visualization interface in the early warning and linkage module provides contextual analysis functionality for the personnel behavior data, specifically including: Collect contextual information of the personnel behavior data, including the time and location of the personnel behavior, personnel identity, current task type and surrounding equipment operating status, extract personnel behavior features corresponding to the personnel behavior data and associate them with the contextual information, and output a personnel behavior dataset with context labels; The context-labeled personnel behavior dataset is input into a pre-trained AI behavior classification model. The AI ​​behavior classification model identifies the differences in contextual features between normal behavior data and abnormal behavior data, thus distinguishing between the normal behavior data and the abnormal inspection data. The abnormal behavior data is verified, and the duration, movement trajectory and impact range of the abnormal behavior data are analyzed in combination with historical personnel behavior records to eliminate misjudgments caused by environmental interference or normal operation, and output true abnormal inspection data. An abnormal behavior analysis report is generated based on the actual abnormal behavior data and pushed to the visualization interface in real time.

[0042] The system synchronously collects contextual data such as timestamps of personnel behavior, spatial coordinates, personnel identification, current task work order information, and real-time operating status of surrounding related equipment. It also extracts behavioral features such as action sequences, movement speed, and dwell time corresponding to personnel behavior. Through a timestamp alignment mechanism, the behavioral features are bound one by one with the context information, and a personnel behavior dataset with structured context labels is output, which solves the problem of misjudgment that is easy to occur in traditional isolated behavior recognition.

[0043] The dataset of personnel behavior with context labels is input into a pre-trained AI behavior classification model. By learning from a large number of labeled samples, the model has mastered the differences in contextual features between normal inspection behaviors (such as equipment inspection along a predetermined route and standardized data recording) and abnormal behaviors (such as illegal entry into restricted areas and unauthorized operation of equipment), and automatically distinguishes between normal behavior data and the initially determined abnormal inspection data.

[0044] By combining the historical behavior records of personnel in the historical monitoring database of the cloud platform, the preliminary abnormal data is verified in multiple dimensions. The duration of abnormal behavior, the rationality of movement trajectory and the scope of impact are analyzed. Misidentification caused by interference such as changes in ambient light and object obstruction is eliminated, as well as normal business behaviors such as authorized temporary maintenance and emergency operations. The verified real abnormal inspection data is output.

[0045] Based on real-world anomaly inspection data, an anomaly behavior analysis report is automatically generated, which includes the time and location of the anomaly, the identity of the personnel, the description of the behavior, the risk level, and the handling suggestions. The report is then synchronized to the visualization interface of the monitoring center in real time via a message push mechanism.

[0046] The hardware platform is used to support the operation of the data acquisition module, the adaptive processing module, and the early warning and linkage module. It utilizes a cloud-edge collaborative architecture to achieve data synchronization and task scheduling between the edge computing nodes and the cloud platform, performs real-time monitoring and fault tolerance handling of the system's operating status, outputs a system operating status report, and updates the historical monitoring database, including: The hardware platform is built using a plug-and-play modular design, with the data acquisition module, the adaptive processing module, and the early warning and linkage module designed as independent functional modules. The number and type of the functional modules are flexibly configured according to the scale and requirements of the substation. The rotatable monitor in the hardware platform is equipped with a non-contact ultrasonic self-cleaning device. The non-contact ultrasonic self-cleaning device generates high-frequency vibrations through an ultrasonic generator, causing dust, moisture and dirt to fall off the surface of the monitor lens. It automatically adjusts the frequency and working duration of the ultrasonic waves according to the ambient dust concentration and provides real-time feedback of cleaning control signals and lens cleaning status data. The system adopts a hybrid power supply method that combines solar power and grid power. The hardware platform is equipped with solar panels and energy storage batteries. The system energy consumption is optimized through low-power chips and dynamic power management technology. The power supply mode is automatically switched according to the light intensity and system load, and energy consumption management data and power status information are fed back in real time. Multiple edge computing nodes are deployed in the hardware platform. Through a cloud-edge collaborative architecture, the edge computing nodes and the cloud platform achieve bidirectional data synchronization. When an edge computing node fails, its tasks are automatically migrated to other normally operating edge computing nodes, and the fault tolerance processing results and system reliability data are fed back in real time.

[0047] It also includes a privacy protection unit, which processes the personnel behavior data, specifically including: The edge computing node performs differential privacy desensitization processing on the collected personnel behavior data, which masks sensitive behavioral details by adding noisy data and outputs desensitized personnel behavior data. Based on the desensitized personnel behavior data, the personnel identity information, personnel facial features and personnel location information are anonymized in the visualization interface, and anonymized display data is output. Establish a role-based data access permission hierarchical mechanism, classify personnel into operation and maintenance level, management level and security level, assign different data access permissions to personnel of different levels, record all personnel's data access operations and generate data access logs, and output permission management results and access audit reports; The personnel behavior data is encrypted using national cryptographic algorithms during processing, and is stored on the cloud platform using block encryption and distributed storage methods.

[0048] The collected raw personnel behavior data is processed using a Laplace algorithm to add noise, masking sensitive behavioral details and precise trajectory information without altering the risk identification characteristics of the personnel behavior, thus outputting anonymized personnel behavior data. This processing is completed locally immediately after data collection, and the original sensitive data is not uploaded to the cloud platform, preventing data leakage risks at the source.

[0049] Based on the desensitized personnel behavior data, irreversible anonymization processing of personnel identity information, facial features, and precise location information is performed in the visualization interface of the monitoring center. A unique number randomly generated by the system is used to replace personnel names and employee numbers, and regional grid markers are used to replace centimeter-level precise coordinates. Sensitive information that can identify individuals is hidden, and anonymized display data is output, taking into account both monitoring needs and personnel privacy rights.

[0050] Establish a role-based, refined access control system, dividing system users into three categories: operation and maintenance level, management level, and security level. Assign differentiated data access scope and operation permissions to different roles, strictly restrict the subjects who can access sensitive data, and record all data access operations of all users throughout the process, automatically generate structured data access logs, and regularly output access control results and access audit reports to achieve traceability and auditability of data access.

[0051] Personnel behavior data is encrypted using national cryptographic algorithms throughout the entire transmission and processing process to prevent data theft or tampering during transmission. In the cloud platform storage phase, a combination of block encryption and distributed storage is employed, distributing encrypted data blocks across different nodes to further enhance data storage security and resistance to attacks, ensuring the secure and controllable flow of personnel behavior data throughout the entire process.

[0052] This application also discloses an electronic device.

[0053] Specifically, the electronic device includes a memory and a processor. The memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor performs the functions of a substation auxiliary monitoring system.

[0054] This invention also discloses a readable storage medium.

[0055] Specifically, the readable storage medium stores a computer program that, when executed by a processor, is used to implement the functions of a substation auxiliary monitoring system.

[0056] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A substation auxiliary monitoring system, characterized in that, include: The data acquisition module is used to synchronously collect equipment status information, real-time environmental parameters and personnel behavior data through a multimodal sensor network and dynamic tag management unit deployed in multiple areas of the substation, and to perform timestamp calibration and format unification. It is an adaptive processing module that outputs multimodal monitoring datasets and merges them to transmit them to edge computing nodes. The adaptive processing module is used to perform real-time cleaning and feature extraction on the multimodal monitoring data set through the built-in streaming data processing engine to obtain a multimodal monitoring feature set. It also integrates multi-sensor data fusion technology to perform multi-dimensional fusion on the multimodal monitoring feature set to obtain comprehensive monitoring features. Combined with a reinforcement learning model trained on a historical monitoring database, it dynamically optimizes the risk assessment model to obtain an improved risk assessment model. Based on the improved risk assessment model, it performs risk assessment on the comprehensive monitoring features and outputs a comprehensive risk assessment result including regional hazard level coefficient and personnel behavior risk index, which is then transmitted to the early warning and linkage module. The early warning and linkage module is used to classify the comprehensive risk assessment results according to the preset risk level classification rules and generate a graded alarm signal. At the same time, it links the ventilation equipment, access control system and visual interface in the substation to perform corresponding response operations and outputs linkage execution status feedback. The hardware platform is used to support the operation of the data acquisition module, the adaptive processing module, and the early warning and linkage module. It utilizes a cloud-edge collaborative architecture to realize data synchronization and task scheduling between the edge computing nodes and the cloud platform, performs real-time monitoring and fault tolerance processing of the system operation status, outputs system operation status reports, and updates the historical monitoring database.

2. The substation auxiliary monitoring system according to claim 1, characterized in that, The adaptive processing module, which synchronously collects equipment status information, real-time environmental parameters, and personnel behavior data through a multimodal sensor network and dynamic tag management unit deployed in multiple areas of the substation, performs timestamp calibration and format unification, and outputs a multimodal monitoring dataset for merging and transmission to the edge computing node, includes: Based on the substation's topology and equipment distribution, a multimodal sensor network is deployed in key equipment areas, environmentally sensitive areas, and personnel access areas. The multimodal sensor network includes at least visual sensors, infrared sensors, radar, UWB positioning sensors, temperature sensors, humidity sensors, and SF6 gas concentration sensors. The hardware interrupt mechanism controls all sensors in the multimodal sensor network to start data acquisition at the same time. The system synchronously collects equipment status information including equipment aging degree, equipment density and equipment electrical load, real-time environmental parameters including temperature, humidity, SF6 gas concentration and external meteorological data, and personnel behavior data including personnel action images captured by the visual sensor and personnel location information obtained by radio frequency identification tags. The system outputs a time-synchronized set of raw monitoring data. The edge computing node preprocesses the original monitoring data set, including removing abnormal data, filling in missing data, and normalizing data, and outputs a multimodal monitoring data set. The dynamic tag management unit assigns unique encrypted RFID tags to all equipment or personnel in the substation, monitors the equipment or personnel information corresponding to the unique encrypted RFID tags in real time and dynamically updates the unique encrypted RFID tags, outputs tag dynamic management data and integrates it into the multimodal monitoring data set, and transmits the integrated multimodal monitoring data set to the adaptive processing module of the edge computing node.

3. The substation auxiliary monitoring system according to claim 2, characterized in that, The process involves assigning unique encrypted RFID tags to equipment or personnel within the substation through the dynamic tag management unit, monitoring the equipment or personnel information corresponding to the unique encrypted RFID tags in real time and dynamically updating the unique encrypted RFID tags, outputting tag dynamic management data and integrating it into the multimodal monitoring data set, including: The dynamic tag management unit assigns a unique encrypted RFID tag to the equipment or personnel in the substation, associates and binds the ID of the unique encrypted RFID tag with the equipment information or personnel information, and stores it in the RFID tag database of the cloud platform. The unique encrypted RFID tag is read in real time by RFID readers deployed at the entrance and exit of the substation and in key reading areas. When a change is detected in the equipment information or personnel information associated with the unique encrypted RFID tag, an RFID tag update request is automatically sent to the cloud platform. After receiving the RFID tag update request, the cloud platform performs identity verification and permission review on the RFID tag update request. After the identity verification and permission review are passed, the unique encrypted RFID tag is updated in real time, the tag dynamic management data is output and integrated into the multimodal monitoring data set; When the substation is expanded or upgraded, the unique encrypted RFID tag is automatically assigned to the newly added equipment or personnel, and the reading range of the unique encrypted RFID tag and the deployment location of the RFID reader are adjusted according to the topology of the expanded substation.

4. The substation auxiliary monitoring system according to claim 1, characterized in that, The improved risk assessment model is obtained by dynamically optimizing the risk assessment model using a reinforcement learning model trained on a historical monitoring database. Based on this improved model, a risk assessment is performed on the comprehensive monitoring features, outputting a comprehensive risk assessment result including a regional hazard coefficient and a personnel behavior risk index, and transmitting it to the early warning and linkage module. This includes: Historical operation data, historical fault records, and historical safety event data of the substation within a historical time period are extracted from the historical monitoring database of the cloud platform to construct a risk assessment training dataset. The risk assessment training dataset is then labeled and classified, and the labeled risk assessment training dataset is output. The risk assessment model based on fuzzy logic rules is trained based on the labeled risk assessment training dataset. The membership function and weight coefficient of the fuzzy logic rules are determined. At the same time, the reinforcement learning model is trained as a reinforcement learning optimizer, and the initial risk assessment model and the reinforcement learning optimizer are output. The comprehensive monitoring features are input into the initial risk assessment model to obtain a preliminary risk assessment result. The preliminary risk assessment result is compared with the current security event data to generate a reward signal, which is then input into the reinforcement learning optimizer. The membership function and weight coefficient of the fuzzy logic rule are dynamically adjusted through the reinforcement learning optimizer to output the improved risk assessment model. The comprehensive monitoring features are input into the improved risk assessment model to generate a comprehensive risk assessment result that includes the regional risk level coefficient and the personnel behavior risk index, and then transmitted to the early warning and linkage module.

5. A substation auxiliary monitoring system according to claim 1, characterized in that, The comprehensive risk assessment results are classified and determined according to a preset risk level classification rule to generate a graded alarm signal. Simultaneously, the ventilation equipment, access control system, and visual interface within the substation are linked to execute corresponding response operations, and the linked execution status feedback is output, including: Based on the range of values ​​for the regional risk level coefficient and the personnel behavior risk index in the comprehensive risk assessment results, the comprehensive risk is divided into four levels: Level 1, Level 2, Level 3, and Level 4. Alarm methods, response procedures, and a list of linked devices are formulated for each risk level, and a hierarchical early warning rule library is output. The comprehensive risk assessment results are matched with the hierarchical early warning rule base to determine the current risk level and generate corresponding alarm signals, including at least audible and visual alarms, SMS alarms and platform pop-up alarms. At the same time, the time and location of the alarm are recorded and the alarm event record is output. Based on the alarm event records and the hierarchical early warning rule base, control commands are sent to the corresponding linked devices. When an SF6 gas leak is detected, the ventilation equipment is activated to start air exchange. When personnel are detected entering a dangerous area, the access control system is activated to close the passage. When an equipment malfunction is detected, the visualization interface is activated to display equipment information, and the linkage execution status feedback is output.

6. The substation auxiliary monitoring system according to claim 1, characterized in that, The visualization interface in the early warning and linkage module provides contextual analysis functionality for the personnel behavior data, specifically including: Collect contextual information of the personnel behavior data, including the time and location of the personnel behavior, personnel identity, current task type and surrounding equipment operating status, extract personnel behavior features corresponding to the personnel behavior data and associate them with the contextual information, and output a personnel behavior dataset with context labels; The context-labeled personnel behavior dataset is input into a pre-trained AI behavior classification model. The AI ​​behavior classification model identifies the differences in contextual features between normal behavior data and abnormal behavior data, thus distinguishing between the normal behavior data and the abnormal inspection data. The abnormal behavior data is verified, and the duration, movement trajectory and impact range of the abnormal behavior data are analyzed in combination with historical personnel behavior records to eliminate misjudgments caused by environmental interference or normal operation, and output true abnormal inspection data. An abnormal behavior analysis report is generated based on the actual abnormal behavior data and pushed to the visualization interface in real time.

7. A substation auxiliary monitoring system according to claim 1, characterized in that, The hardware platform is used to support the operation of the data acquisition module, the adaptive processing module, and the early warning and linkage module. It utilizes a cloud-edge collaborative architecture to achieve data synchronization and task scheduling between the edge computing nodes and the cloud platform, performs real-time monitoring and fault tolerance handling of the system's operating status, outputs a system operating status report, and updates the historical monitoring database, including: The hardware platform is built using a plug-and-play modular design, with the data acquisition module, the adaptive processing module, and the early warning and linkage module designed as independent functional modules. The number and type of the functional modules are flexibly configured according to the scale and requirements of the substation. The rotatable monitor in the hardware platform is equipped with a non-contact ultrasonic self-cleaning device. The non-contact ultrasonic self-cleaning device generates high-frequency vibrations through an ultrasonic generator, causing dust, moisture and dirt to fall off the surface of the monitor lens. It automatically adjusts the frequency and working duration of the ultrasonic waves according to the ambient dust concentration and provides real-time feedback of cleaning control signals and lens cleaning status data. The system adopts a hybrid power supply method that combines solar power and grid power. The hardware platform is equipped with solar panels and energy storage batteries. The system energy consumption is optimized through low-power chips and dynamic power management technology. The power supply mode is automatically switched according to the light intensity and system load, and energy consumption management data and power status information are fed back in real time. Multiple edge computing nodes are deployed in the hardware platform. Through a cloud-edge collaborative architecture, the edge computing nodes and the cloud platform achieve bidirectional data synchronization. When an edge computing node fails, its tasks are automatically migrated to other normally operating edge computing nodes, and the fault tolerance processing results and system reliability data are fed back in real time.

8. A substation auxiliary monitoring system according to claim 1, characterized in that, It also includes a privacy protection unit, which processes the personnel behavior data, specifically including: The edge computing node performs differential privacy desensitization processing on the collected personnel behavior data, and adds noise data to mask sensitive behavioral details of personnel, and outputs desensitized personnel behavior data; Based on the desensitized personnel behavior data, the personnel identity information, personnel facial features and personnel location information are anonymized in the visualization interface, and anonymized display data is output. Establish a role-based data access permission hierarchical mechanism, classify personnel into operation and maintenance level, management level and security level, assign different data access permissions to personnel of different levels, record all personnel's data access operations and generate data access logs, and output permission management results and access audit reports; The personnel behavior data is encrypted using national cryptographic algorithms during processing, and is stored on the cloud platform using block encryption and distributed storage methods.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor runs the computer program stored in the memory, the processor performs the functions of a substation auxiliary monitoring system as described in any one of claims 1-8.

10. A readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it is used to implement the functions of a substation auxiliary monitoring system as described in any one of claims 1-8.

Images