Substation hidden danger comprehensive monitoring method, system and equipment integrating multiple sensors and medium

Abstract

The invention discloses a transformer substation hidden danger comprehensive monitoring method, system and equipment integrating multiple sensors and a medium, and belongs to the technical field of transformer substation hidden danger monitoring. Cleaning and fusing the collected data, and constructing a unified data set; inputting the unified data set into an identification structure constructed by combining a convolutional neural network and a recurrent neural network, respectively processing image type and time sequence type data, and completing hidden danger type identification according to a preset training set proportion and training parameters; carrying out risk value calculation and grade judgment in combination with an identification result and hidden danger related characteristics; and early warning information is generated based on a judgment result, classified pushing and display are performed, early warning processing feedback information is received, and state updating and recording and archiving are completed. Fusion sensing and recognition of multi-source monitoring data are achieved, hidden danger early warning capacity and good environmental adaptability are achieved, and the operation safety and management efficiency of the transformer substation are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of substation hidden danger monitoring, and particularly relates to a substation hidden danger comprehensive monitoring method, system, device and medium integrated with multiple sensors. BACKGROUND

[0002] Traditional substation hidden danger monitoring mainly relies on manual patrol and simple equipment monitoring, which has the following problems:

[0003] The monitoring means is single and cannot comprehensively cover various hidden danger types of substations, such as water immersion, geological displacement, hanging objects, weather changes, etc.

[0004] The data collection and processing efficiency is low, real-time monitoring and rapid response cannot be achieved, and it is difficult to discover and handle hidden dangers in the early stage.

[0005] The adaptability to complex environments is poor, and the monitoring effect is not good in adverse weather or complex terrain conditions.

[0006] There is a lack of intelligent analysis and early warning functions, and it is difficult to provide effective decision support by deeply mining and risk assessment of monitoring data.

[0007] In view of the deficiencies of the prior art, the present application realizes all-around and real-time monitoring of substations and their surrounding environment, and timely discovers and handles hidden dangers through intelligent analysis and early warning functions, thereby improving the safety and reliability of substation operation. SUMMARY

[0008] In view of the above problems, the present application is proposed.

[0009] Therefore, the present application solves the technical problems of the prior art, i.e., how to solve the problems of single monitoring means of traditional methods, which cannot comprehensively cover various hidden danger types of substations, such as water immersion, geological displacement, hanging objects, weather changes, etc.

[0010] To solve the above technical problems, the application provides the following technical scheme: a substation hidden danger comprehensive monitoring method integrated with multiple sensors, which comprises the following steps: collecting multiple operation data of a substation and its surroundings; cleaning and fusing the collected data to construct a unified data set; inputting the unified data set into an identification structure constructed by combining a convolutional neural network and a recurrent neural network, which are respectively used for processing image type data and time series type data, and completing hidden danger type identification according to a pre-set training set proportion and training parameters; combining the identification result and hidden danger related features to perform risk value calculation and grade determination; generating early warning information based on the determination result, and performing classified pushing and display, receiving early warning processing feedback information, and completing state updating and record archiving.

[0011] As a preferred scheme of the substation hidden danger comprehensive monitoring method integrated with multiple sensors, the step of collecting multiple operation data of a substation and its surroundings comprises the following steps: obtaining the operation states of multiple monitoring areas, and transmitting the collected data to the same processing procedure for unified processing.

[0012] As a preferred scheme of the substation hidden danger comprehensive monitoring method integrated with multiple sensors, the step of inputting the unified data set into an identification structure constructed by combining a convolutional neural network and a recurrent neural network comprises the following steps: performing corresponding identification processing on different types of input data, and obtaining parameters of the identification structure through a training process.

[0013] As a preferred scheme of the substation hidden danger comprehensive monitoring method integrated with multiple sensors, the step of risk value calculation and grade determination comprises the following steps: performing comprehensive analysis based on multiple hidden danger related information, and generating a determination result according to corresponding evaluation logic.

[0014] As a preferred scheme of the substation hidden danger comprehensive monitoring method integrated with multiple sensors, the step of hidden danger type identification comprises the following steps: inputting image type data into a convolutional neural network model to extract spatial features including edges, textures and colors of images; inputting time series type data into a recurrent neural network model to extract change trends; comparing the identification result with a data label, and performing joint training based on a set training set, verification set and test set proportion until a pre-set accuracy threshold is met.

[0015] The preferred scheme can effectively support hidden danger automatic identification of different types of monitoring data by inputting image type data into a convolutional neural network to extract spatial features, inputting time series data into a recurrent neural network to extract change trends, and performing joint training by using a set proportion of training set, verification set and test set, and has a cross-modal input model training mechanism, which improves the identification ability and adaptability of the system to composite hidden dangers.

[0016] As a preferred embodiment of the integrated multi-sensor substation hazard monitoring method described in this invention, the step of performing risk value calculation and level determination includes: calling a risk assessment model to quantify multiple features of the identified hazard type, location, severity, and development trend, and assigning preset weights; using fuzzy comprehensive evaluation method for calculation, comparing the model output results with the level classification standard, and determining the level as high risk, medium risk, or low risk.

[0017] This preferred solution quantifies and weights the characteristics of identified hazards, such as type, location, severity, and development trend, and uses fuzzy comprehensive evaluation method for calculation. By comparing the results with the grading standards, it can achieve a unified quantitative grading judgment of hazard risk levels, ensuring that the risk assessment process is structured and logically consistent, and helping to support the controllable risk assessment process under multi-source heterogeneous data.

[0018] As a preferred embodiment of the integrated multi-sensor substation hazard monitoring method described in this invention, the step of generating early warning information based on the judgment results and classifying, pushing, and displaying it includes: generating a Level 1 early warning information when the monitoring data reaches the early warning threshold, or when the identified hazard is assessed as high-risk, and sending it simultaneously via SMS, email, and on-site audible and visual alarms; generating a Level 2 early warning information when the hazard is of medium risk, and sending it via SMS and email; generating a Level 3 early warning information when the hazard is of low risk, and sending it via email; receiving processing progress and result information from relevant personnel via mobile terminals and updating the corresponding early warning information status; and upgrading the early warning information level and expanding the notification scope when the hazard is not fully processed or when new anomalies occur during the processing.

[0019] This preferred solution generates warning information in a tiered manner based on the determined risk level and pushes it out in multiple ways, such as SMS, email, and audio-visual communication. At the same time, it dynamically updates the warning status based on the processing feedback information, which can realize real-time closed-loop management of the hidden danger alarm process, improve the efficiency of hidden danger response, and ensure the traceability and collaborative linkage of the entire processing process.

[0020] This invention provides a comprehensive monitoring system for potential hazards in substations that integrates multiple sensors.

[0021] To address the aforementioned technical problems, this invention provides the following technical solution: a multi-sensor integrated substation hazard monitoring system, comprising: a data acquisition unit, a data cleaning and fusion unit, a data processing module, a risk assessment model module, and a monitoring and early warning module; the data acquisition unit is used to collect various operational data from the substation and its surroundings; the data cleaning and fusion unit is used to clean and fuse the collected data to construct a unified dataset; the data processing module is used to input the unified dataset into a recognition structure constructed by combining convolutional neural networks and recurrent neural networks, respectively processing image-type and time-series data, and completing hazard type identification according to a pre-set training set ratio and training parameters; the risk assessment model module is used to perform risk value calculation and level determination based on the identification results and hazard-related features; the monitoring and early warning module is used to generate early warning information based on the determination results, and to classify, push, and display the information, receive early warning processing feedback information, and complete status updates and record archiving.

[0022] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the integrated multi-sensor substation hidden danger comprehensive monitoring method.

[0023] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the integrated multi-sensor substation hazard monitoring method.

[0024] The beneficial effects of this invention are as follows: By integrating multiple monitoring methods, this invention achieves multi-dimensional and comprehensive monitoring of substations and their surrounding environment, including water immersion, displacement, video, and meteorological aspects, ensuring comprehensive coverage of various types of potential hazards in substations. Simultaneously, by employing high-precision sensors and advanced deep learning algorithms, such as a combination of convolutional neural networks and recurrent neural networks, intelligent analysis and processing of monitoring data are performed, improving the accuracy and reliability of the monitoring data and enabling timely detection of subtle changes in potential hazards, thus achieving early warning.

[0025] Equipped with powerful intelligent early warning capabilities, the system utilizes deep learning algorithms to deeply mine and analyze monitoring data. Combined with a risk assessment model, it can quickly and accurately identify potential hazards and assess their risk levels. When monitoring data reaches or exceeds the early warning threshold, the monitoring and early warning module immediately notifies relevant personnel through multiple methods, ensuring that potential hazards are detected and addressed as soon as possible, effectively improving emergency response speed. In addition, the system is equipped with an intuitive visualization module, displaying complex monitoring data and early warning information in various intuitive forms such as charts, images, and maps on the monitoring screen or user terminal, facilitating users to quickly understand the situation and make decisions.

[0026] The system fully considers the complex and ever-changing operating environment of substations, possessing excellent environmental adaptability and the ability to operate stably under adverse weather conditions. It is also equipped with a backup power unit to ensure the continuity of monitoring operations. The system architecture has good scalability, allowing for easy integration with new monitoring equipment or other power system equipment management systems, thereby improving the overall safety and management efficiency of the power system. Simultaneously, through real-time monitoring and intelligent early warning, it reduces equipment maintenance costs and power outage losses, lowers labor costs, and demonstrates significant economic benefits and environmental friendliness. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a flowchart illustrating an integrated multi-sensor method for comprehensive monitoring of substation hazards, provided as an embodiment of the present invention.

[0029] Figure 2 This is a hazard category identification diagram for a substation hazard comprehensive monitoring method integrating multiple sensors, provided as an embodiment of the present invention.

[0030] Figure 3 This is a schematic diagram of a substation hazard monitoring system integrating multiple sensors, provided as an embodiment of the present invention.

[0031] Figure 4 This is a diagram of an integrated multi-sensor substation hazard monitoring method according to an embodiment of the present invention. Detailed Implementation

[0032] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0033] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a comprehensive monitoring method for potential hazards in substations integrating multiple sensors, including:

[0034] S1: Collect various operational data from the substation and its surrounding area.

[0035] S2. Clean and merge the collected data to construct a unified dataset.

[0036] S3. Input the unified dataset into the recognition structure constructed by combining convolutional neural networks and recurrent neural networks, which are used to process image data and time series data respectively, and complete the identification of hazard types according to the pre-set training set ratio and training parameters.

[0037] S4. Based on the identification results and the characteristics of potential hazards, calculate the risk value and determine the level.

[0038] S5. Generate early warning information based on the judgment results, and push and display it in categories. Receive early warning processing feedback information and complete status updates and record archiving.

[0039] It should be noted that substations may face multiple potential hazards simultaneously during operation, including water immersion in equipment areas, structural displacement, and external hanging objects. The data related to these hazards are characterized by uneven spatial distribution, rapid temporal changes, and heterogeneous types. Therefore, relying on traditional threshold-based methods for hazard identification often suffers from response delays and insufficient identification accuracy. Furthermore, the lack of effective grading and judgment standards in risk warning processes makes it difficult to guarantee subsequent classified responses and closed-loop management.

[0040] To address the aforementioned issues, this invention, through steps S1 to S5, unifies the modeling of multi-source data from different monitoring units, extracts spatial features and temporal evolution trends using a deep learning structure, and constructs a risk level judgment logic using a fuzzy comprehensive evaluation method. Ultimately, it achieves automatic identification of hidden dangers, risk assessment, and graded early warning, providing fundamental support for on-site safety assurance of substations.

[0041] Example 2, refer to Figure 2 As one embodiment of the present invention, based on the previous embodiment, a comprehensive monitoring method for substation hidden dangers integrating multiple sensors is provided, including:

[0042] In the embodiments of this application, step S1 involves collecting various operational data from the substation and its surrounding area.

[0043] The system acquires the operational status of multiple monitoring areas and transmits the collected data to the same processing flow for unified processing.

[0044] Specifically, each monitoring unit collects data in real time according to a set sampling frequency and transmits the collected data to the data acquisition unit via wired or wireless means. The data acquisition unit reads the data from each monitoring unit through the corresponding interface program and performs preliminary processing and storage. For example, the water immersion monitoring unit collects water level data once per minute and transmits it to the data acquisition unit via a 485 serial port; the displacement monitoring unit collects displacement data once per hour and transmits it to the data acquisition unit via a network port; the video monitoring unit collects image data in real time and transmits it to the data acquisition unit via a network port; and the meteorological monitoring unit collects meteorological data every 10 minutes and transmits it to the data acquisition unit via a 485 serial port.

[0045] In this embodiment of the application, step S2 involves cleaning and fusing the collected data to construct a unified dataset.

[0046] The received data undergoes preprocessing, including data cleaning and fusion. The data cleaning subunit performs noise reduction, deduplication, and data completion operations to remove outliers and erroneous data, improving data quality. For example, for water immersion monitoring data, if multiple consecutive data points exceed the sensor's measurement range, they are considered outliers and discarded; interpolation is then used to complete the missing data. For displacement monitoring data, if the rate of change exceeds physical limits, it is considered noisy and smoothed. The data fusion subunit merges data collected from different monitoring units, establishing correlations between the data to provide more comprehensive and accurate information for subsequent data analysis and hazard identification. For example, displacement monitoring data can be merged with meteorological monitoring data to analyze the impact of meteorological factors such as rainfall and wind speed on displacement changes.

[0047] In one optional implementation, the cleaning and fusion of the collected data may further include: after performing preliminary processing on image data and non-image sensor data respectively, using the target detection event extracted from the image frame as the key anchor point, selecting non-image data segments within the interval corresponding to the event occurrence time for matching, and completing the event-driven data fusion.

[0048] In another optional implementation, the cleaning and fusion of the collected data may also include: constructing a set of data quality assessment rules, scoring each type of monitoring data in real time, and removing low-confidence data records; at the same time, normalization processing is used to standardize the data of different monitoring units to a unified numerical range, and then constructing a unified data vector.

[0049] This invention enables the integrity restoration and synchronization alignment of time series data, facilitating subsequent unified modeling and processing.

[0050] In this embodiment of the application, in step S3, a unified dataset is input into the recognition structure constructed by combining a convolutional neural network and a recurrent neural network, which are used to process image-type and time series-type data respectively, and to complete the identification of hazard types according to the pre-set training set ratio and training parameters.

[0051] Different types of input data are processed accordingly, and the recognition structure obtains parameters through the training process.

[0052] Image-based data is input into a convolutional neural network model to extract spatial features including image edges, textures, and colors; time-series data is input into a recurrent neural network model to extract trends; the recognition results are compared with data labels, and joint training is performed based on a set ratio of training, validation, and test sets until a preset accuracy threshold is met. Figure 2 As shown.

[0053] Specifically, the data processing module performs in-depth processing and analysis on the received data. First, the data is normalized to eliminate the influence of different dimensions and units. Then, signal processing methods such as wavelet transform and Fourier transform are used to extract features from the data, extracting time-domain, frequency-domain, and time-frequency-domain features to provide feature vectors for hazard identification. For example, for displacement monitoring data, features such as mean, variance, peak value, and frequency are extracted; for video monitoring data, features such as image edges, texture, and color are extracted.

[0054] The extracted feature vectors are input into a pre-trained deep learning model for hazard identification and classification. Based on the learned features and patterns, the model automatically identifies and classifies monitoring data, determining the presence and type of hazards. For example, for video monitoring data, the model can identify the shape, size, and color of hanging objects and classify them into different types such as plastic film, kites, and bird nests; for displacement monitoring data, the model can determine the trend and rate of displacement change and classify it into different states such as normal displacement, slow displacement, and rapid displacement. Simultaneously, the data processing module feeds back the identification results and related feature information to the visualization and monitoring early warning modules for subsequent display and early warning processing.

[0055] After receiving the data, the data center stores it in the corresponding database tables. The data processing module periodically reads newly collected data from the database for in-depth processing and analysis. For image data, a trained convolutional neural network model is used for target detection, identifying potential hazards such as hanging objects and abnormal human activity in the image, and determining their location and size. For displacement and meteorological data, a recurrent neural network model is used for time series analysis to predict data trends and determine whether there are any anomalies. For example, if displacement data shows a sudden change in a short period of time, exceeding a set threshold, it is determined to be abnormal displacement, which may indicate a potential geological hazard.

[0056] The data processing module feeds back the identified hazards to the visualization and monitoring / early warning modules. The visualization module displays the hazard identification results in real time on the corresponding pages, such as marking the location and type of hanging objects on the video surveillance screen and marking abnormal points on the displacement data curve; the monitoring / early warning module conducts risk assessments and issues early warnings based on the type and severity of the hazard.

[0057] In the embodiments of this application, step S4 combines the identification results and the characteristics related to the hidden dangers to perform risk value calculation and level determination.

[0058] Based on a comprehensive analysis of multiple potential hazard-related information, and in accordance with the corresponding assessment logic, a judgment result is generated.

[0059] The risk assessment model is invoked to quantify multiple characteristics of the identified hazards, including their type, location, severity, and development trend, and assign them preset weights. The fuzzy comprehensive evaluation method is used for calculation, and the model output is compared with the classification standard to determine whether it is a high-risk, medium-risk, or low-risk level.

[0060] Specifically, after the data processing module identifies a potential hazard, the monitoring and early warning module calls a pre-configured risk assessment model to assess the hazard's risk. Based on factors such as the hazard's type, location, size, and trend of change, as well as pre-set weights, the risk value of the hazard is calculated, and its risk level is determined. For example, for a landslide hazard located near a substation, based on factors such as its landslide area, landslide velocity, and distance from the substation equipment, the calculated risk value is 85, classifying it as a high-risk level.

[0061] The risk assessment model periodically reassesses and updates the risk level of potential hazards to reflect the latest changes. The update frequency is determined based on the type and urgency of the hazard: high-risk hazards are updated hourly; medium-risk hazards are updated every 6 hours; and low-risk hazards are updated daily. Simultaneously, changes in the risk assessment results are recorded in a database for historical data retrieval and analysis.

[0062] The visualization module on the homepage's risk assessment page displays the risk assessment results for each hazard in the form of a map and a list. The map uses different colored icons to represent the risk level of the hazard, such as red for high risk, yellow for medium risk, and blue for low risk. The list details the location, type, risk level, and assessment time of each hazard, and allows users to sort and filter by risk level, hazard type, and other criteria, enabling them to quickly understand the overall risk status of the substation.

[0063] In one optional implementation, the risk assessment method may further include: setting weight vectors for each risk factor based on the experience of on-site experts and the results of historical data analysis; using the analytic hierarchy process (AHP) to verify the consistency of expert scoring results and generate final weights; normalizing the hazard characteristic items and inputting them into the model for calculation; and matching the output values ​​with the grading thresholds.

[0064] In another optional implementation, the risk assessment method may further include: dynamically adjusting the weights of each factor in the assessment model according to the risk distribution characteristics of different types of hidden dangers, for example, increasing the location weight for structural hidden dangers and increasing the development trend weight for meteorological hidden dangers, and using a preset strategy to switch different weight combinations for scenario-based modeling and assessment.

[0065] This invention constructs a unified and reproducible risk assessment process, realizing structured hazard classification and judgment under multiple indicators.

[0066] In this embodiment of the application, step S5 generates early warning information based on the judgment result, and pushes and displays it in categories, receives early warning processing feedback information, and completes status updates and record archiving.

[0067] When monitoring data reaches the warning threshold, or when an identified hazard is assessed as high-risk, a Level 1 warning is generated and sent simultaneously via SMS, email, and on-site audible and visual alarms. When the hazard is at a medium-risk level, a Level 2 warning is generated and sent via SMS and email. When the hazard is at a low-risk level, a Level 3 warning is generated and sent via email. The system receives feedback from relevant personnel via mobile terminals regarding the processing progress and results, and updates the corresponding warning status accordingly. If a hazard is not fully resolved or a new anomaly occurs during the processing, the warning level is upgraded and the scope of notification is expanded.

[0068] When monitoring data reaches or exceeds the warning threshold, or when the risk level of a potential hazard is deemed high, the monitoring and warning module generates corresponding warning information based on preset warning rules. The warning information includes detailed information such as the location, type, severity, time of occurrence, and warning level of the hazard, along with corresponding handling suggestions and emergency plans. For example, when the water immersion monitoring unit detects a water level exceeding 10cm, it generates a water immersion warning with a level one warning. Handling suggestions include immediately dispatching personnel to the site, activating the drainage system, and implementing waterproofing measures for equipment.

[0069] The monitoring and early warning module selects the appropriate notification method based on the level and type of the early warning information and sends it to relevant personnel. For Level 1 early warning information, SMS, email, and audible / visual alarms are triggered simultaneously to ensure that relevant personnel receive the early warning information immediately. For Level 2 early warning information, SMS and email notifications are sent. For Level 3 early warning information, only email notifications are sent. The recipients of early warning information are configured according to user permissions and responsibilities, such as substation maintenance personnel, management personnel, and emergency response personnel, ensuring that early warning information is promptly delivered to those who need to take action.

[0070] Upon receiving the early warning information, relevant personnel will promptly take appropriate measures to address the potential hazard based on the suggested solutions. During the process, progress and results can be reported in real time via a mobile app or web page, allowing the monitoring and early warning module to update the status of the warning information promptly. If the hazard is not addressed within the specified time, or if new anomalies occur during the handling process, the monitoring and early warning module will escalate the warning information, expand the notification scope, and notify higher-level management personnel to intervene, ensuring that the hazard is effectively addressed and preventing accidents.

[0071] Example 3, referring to Figure 3 As one embodiment of the present invention, a substation hidden danger comprehensive monitoring system integrating multiple sensors is provided, including: a data acquisition unit, a data cleaning and fusion unit, a data processing module, a risk assessment model module, and a monitoring and early warning module.

[0072] The data acquisition unit is used to collect various operational data from the substation and its surrounding area.

[0073] The data cleaning and fusion unit is used to clean and fuse the collected data to build a unified dataset.

[0074] The data processing module is used to input a unified dataset into a recognition structure that combines convolutional neural networks and recurrent neural networks. It is used to process image-type and time-series data respectively, and to complete the identification of hazard types according to the pre-set training set ratio and training parameters.

[0075] The risk assessment model module is used to combine the identification results and hazard-related characteristics to perform risk value calculation and level determination.

[0076] The monitoring and early warning module is used to generate early warning information based on the judgment results, and to push and display the information in a categorized manner, receive feedback information on early warning processing, and complete status updates and record archiving.

[0077] System architecture and deployment steps:

[0078] Step 1: Building the data center.

[0079] Hardware Configuration: Select servers with high-performance processors, large-capacity memory, and storage as the physical carriers of the data center to ensure they can process and store large-scale monitoring data. The server's processor frequency should be no less than 3.0GHz, memory capacity no less than 128GB, storage space no less than 10TB, and it should be equipped with redundant power supplies and cooling systems to ensure stable operation over long periods.

[0080] Software Environment: A Linux operating system will be installed due to its stability and security, making it suitable for a server environment. On top of this, a database management system, such as MySQL or Oracle, will be deployed to create and manage the substation hazard database. The database should be optimized, including setting appropriate character sets and encoding methods, and creating users and permissions to ensure secure data access.

[0081] Network settings: Connect the server to the substation's local area network, configure a static IP address, and set firewall rules to allow data communication between the monitoring device and other system modules' IP addresses, while blocking unauthorized access to ensure the network security of the data center.

[0082] Step 2: Development and integration of the data processing module.

[0083] Algorithm Selection: A combination of Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) from deep learning algorithms is used to identify and classify potential hazards. CNNs are primarily used to process image data, such as images acquired by video monitoring units, and can automatically extract features from images; RNNs are suitable for processing time-series data, such as data acquired by displacement monitoring units and meteorological monitoring units, and can capture the time dependencies in the data.

[0084] Model Training: Collect a large amount of sample data on substation hazards, including various normal and abnormal situations, to train the algorithm model. The training data should be representative, covering different hazard types, environmental conditions, and operating states. Divide the dataset into training, validation, and test sets in a 6:2:2 ratio to ensure the model's generalization ability and accuracy. Set appropriate training parameters, such as learning rate, number of iterations, and batch size. Through repeated training and adjustments, the model's recognition accuracy should reach over 90%.

[0085] Module Integration: The trained algorithm model is integrated into the data processing module, and corresponding interface programs are developed to enable the module to interact with the data center and other system modules. During the integration process, rigorous testing and debugging are conducted to ensure the stability and reliability of the data processing module.

[0086] Step 3: Configure the monitoring and early warning module.

[0087] Warning threshold setting: Based on the substation's safety standards and historical data, determine the warning thresholds for various potential hazards. For example, for water immersion monitoring, a water level exceeding 10cm is set as the warning threshold; for displacement monitoring, a horizontal displacement exceeding 5mm or a vertical settlement exceeding 3mm is set as the warning threshold; for meteorological monitoring, warnings are triggered by wind speeds exceeding 20m / s or rainfall exceeding 50mm within one hour. The warning thresholds should be regularly evaluated and adjusted according to actual conditions to adapt to changes in the substation environment.

[0088] Risk assessment model construction: A risk assessment model is established by comprehensively considering factors such as the type, severity, and probability of occurrence of potential hazards. Fuzzy comprehensive evaluation is used to quantify each factor and determine its corresponding weight. Through model calculation, the risk level of the hazard is determined, categorized into high, medium, and low levels, providing a basis for the sending and processing of early warning information.

[0089] Early warning configuration: Multiple early warning notification methods are integrated to ensure timely and accurate delivery of warning information to relevant personnel. An SMS modem and email server are configured to interface with mobile operators and email service providers, enabling automatic SMS and email sending. Audible and visual alarms are installed in the substation control room and relevant office areas; when a high-risk warning occurs, the alarms are triggered to alert on-site personnel to take timely measures. Simultaneously, a mobile app and web page are developed, allowing users to receive warning information in real time and view detailed hazard information and handling suggestions.

[0090] Step 4: Design and implementation of the visualization module.

[0091] Display Interface Layout Planning: Based on user operating habits and needs, the display interface layout of the visualization modules is designed. A multi-page structure is adopted, including a homepage, monitoring data page, early warning information page, and history page. The homepage displays the overall hidden danger situation of the substation, showing the location of the substation and its surrounding environment in map form, and intuitively representing the existence status and risk level of various hidden dangers through icons and colors; the monitoring data page displays detailed real-time data collected by each monitoring unit, presented in the form of charts, tables, and curves, making it easy for users to view data change trends; the early warning information page lists existing early warning information, sorted by time order and risk level, and provides a query function for early warning details; the history page saves past monitoring data and early warning information, supporting users to query and statistically analyze data by time range, hidden danger type, and other conditions.

[0092] Data visualization technologies selected: Mainstream visualization libraries such as ECharts and D3.js are used for data visualization development. These libraries provide rich chart types and interactive functions, meeting the display needs of different data types. For example, line charts are used to display the changing trends of displacement monitoring data, bar charts to display the statistics of meteorological data, and heat maps to display the distribution density of potential hazards around substations. Simultaneously, HTML5, CSS3, and JavaScript technologies are combined to implement dynamic page effects and interactive operations, improving the user experience.

[0093] User interaction functionality development: This includes developing user login, registration, and access control features to ensure only authorized users can access the system, and displaying appropriate functions and data based on user permission levels. Data filtering, querying, and export functions are also included to allow users to easily obtain the information they need. For example, users can filter data by date range, monitoring unit type, and hazard level, and export the results to Excel or PDF format for further analysis and reporting. Simultaneously, the system supports user confirmation, processing, and feedback of early warning information, achieving closed-loop management of early warning processing.

[0094] Step 5: Development of the space-ground collaborative monitoring algorithm.

[0095] Satellite Data Acquisition and Preprocessing: Collaborate with satellite data providers to acquire high-resolution imagery data from visible light satellites and radar satellites, ensuring the spatiotemporal resolution of the data meets monitoring requirements. Visible light satellite resolution should be no less than 1 meter, and radar satellite resolution no less than 3 meters, with data acquisition intervals not exceeding one week. Preprocess the received satellite data, including radiometric calibration, geometric correction, and noise reduction, to improve data quality and usability. Utilize professional remote sensing image processing software, such as ENVI or ERDAS, for preprocessing and save the processed data for subsequent analysis.

[0096] Satellite remote sensing monitoring model establishment: Based on deep learning-based target detection algorithms, such as Faster R-CNN or YOLO, the model identifies and classifies potential hazards around substations in satellite imagery, including geological disasters and hanging objects. A large amount of satellite imagery sample data is collected, and the location and category of potential hazards are labeled to train the target detection model. Simultaneously, time-series analysis techniques are used to compare satellite imagery from different times to establish a model of hazard changes around substations, enabling the detection and early warning of hazard changes. By analyzing historical and real-time data, changes in hazards can be identified in a timely manner, providing a scientific basis for hazard management at substations.

[0097] Implementation of Space-Ground Collaborative Monitoring Algorithm: A space-ground collaborative monitoring algorithm was developed to fuse satellite monitoring data with data from ground monitoring devices. Data assimilation technology and a weighted average algorithm were employed, comprehensively considering factors such as the accuracy and timeliness of both satellite and ground data to determine the weights and methods for data fusion. Through the fusion analysis of multi-source data, the accuracy and comprehensiveness of hazard monitoring are improved, achieving space-ground collaborative monitoring and early warning. For example, when satellite imagery detects a landslide hazard near a substation, data from ground displacement monitoring units is combined to further determine the landslide's extent, speed, and trend, enabling timely issuance of early warning information and providing more accurate decision support for the substation's emergency response.

[0098] System functions and operating procedures:

[0099] Hazard identification function:

[0100] Data Acquisition and Transmission: The monitoring device collects data from each monitoring unit according to a set sampling frequency and transmits the data to the data center in real time via a wireless communication module. For example, the water immersion monitoring unit collects water level data every minute, the displacement monitoring unit collects displacement data every hour, the video monitoring unit collects image data in real time, and the meteorological monitoring unit collects meteorological data every 10 minutes. The wireless communication module uses 4G / 5G networks for data transmission and employs SSL / TLS encryption protocols to ensure data security.

[0101] Data Processing and Analysis: After receiving data, the data center stores it in the corresponding database tables. The data processing module periodically reads newly collected data from the database for in-depth processing and analysis. For image data, a trained CNN model is used for target detection, identifying potential hazards such as hanging objects and abnormal human activity, and determining their location and size. For displacement and meteorological data, an RNN model is used for time series analysis to predict data trends and determine if any anomalies exist. For example, if displacement data shows a sudden change exceeding a set threshold within a short period, it is considered abnormal displacement, potentially indicating a geological hazard.

[0102] Identification Result Feedback: The data processing module feeds back the identified hazards to the visualization and monitoring / early warning modules. The visualization module displays the hazard identification results in real time on the corresponding pages, such as marking the location and type of hanging objects on the video surveillance screen and marking abnormal points on the displacement data curve; the monitoring / early warning module conducts risk assessment and early warning processing based on the type and severity of the hazard.

[0103] Risk assessment function:

[0104] Risk assessment model invocation: After the data processing module identifies a potential hazard, the monitoring and early warning module invokes a pre-configured risk assessment model to assess the hazard's risk. Based on factors such as the hazard's type, location, size, and trend of change, as well as pre-set weights, the risk value of the hazard is calculated, and its risk level is determined. For example, for a landslide hazard located near a substation, based on factors such as its landslide area, landslide velocity, and distance from the substation equipment, the calculated risk value is 85, classifying it as a high-risk level.

[0105] Risk assessment results updates: The risk assessment model periodically reassesses and updates the risk level of hazards to reflect the latest changes in hazard status. The update frequency is determined based on the type and urgency of the hazard: high-risk hazards are updated hourly; medium-risk hazards are updated every 6 hours; and low-risk hazards are updated daily. Simultaneously, changes in risk assessment results are recorded in a database for historical data retrieval and analysis.

[0106] Risk Assessment Results Display: The visualization module on the homepage displays the risk assessment results for each hazard in the form of a map and a list. The map uses different colored icons to represent the risk level of the hazard, such as red for high risk, yellow for medium risk, and blue for low risk. The list details the location, type, risk level, and assessment time of each hazard, and allows users to sort and filter by risk level, hazard type, and other criteria, facilitating a quick understanding of the substation's overall risk status.

[0107] Early warning function:

[0108] Early Warning Information Generation: When monitoring data reaches or exceeds the early warning threshold, or when the risk level of a potential hazard is high, the monitoring and early warning module generates corresponding early warning information according to preset early warning rules. The early warning information includes detailed information such as the location, type, severity, time of occurrence, and warning level of the hazard, along with corresponding handling suggestions and emergency plans. For example, when the water immersion monitoring unit detects a water level exceeding 10cm, it generates a water immersion early warning information at level one, with handling suggestions including immediately dispatching personnel to the site, activating the drainage system, and implementing waterproofing measures for equipment.

[0109] Warning Information Sending: The monitoring and warning module selects the appropriate notification method to send the warning to relevant personnel based on the warning level and type. For Level 1 warnings, SMS, email, and audible / visual alarms are triggered simultaneously to ensure relevant personnel receive the warning information immediately. For Level 2 warnings, SMS and email notifications are sent. For Level 3 warnings, only email notifications are sent. The recipients of warning information are configured according to user permissions and responsibilities, such as substation maintenance personnel, management personnel, and emergency response personnel, ensuring that warning information is promptly delivered to those requiring action.

[0110] Early Warning Handling and Feedback: Upon receiving an early warning, relevant personnel will promptly take appropriate measures to address the potential hazard based on the suggested actions. During the handling process, progress and results can be reported in real-time via a mobile app or web page, enabling the monitoring and early warning module to update the status of the warning information promptly. If the hazard is not addressed within the specified time, or if new anomalies occur during the handling process, the monitoring and early warning module will escalate the warning information and expand the notification scope, notifying higher-level management personnel to intervene and ensure that the hazard is effectively addressed, preventing accidents from occurring.

[0111] Integrating satellite monitoring, ground monitoring devices, and data processing and analysis technologies, this system achieves comprehensive and multi-dimensional monitoring of the substation and its surrounding environment. It can promptly detect various potential hazards, improving the comprehensiveness and accuracy of monitoring and providing strong support for the safe operation of the substation. Employing advanced deep learning algorithms and risk assessment models, it intelligently analyzes and evaluates monitoring data, accurately identifying potential hazards and issuing timely warnings. This provides ample time for hazard mitigation and effectively reduces the probability of substation accidents. Through a visualization module, complex monitoring data and warning information are presented to users in an intuitive and visual way, allowing them to quickly understand the potential hazards at the substation and improving decision-making efficiency and emergency response capabilities.

[0112] With a wide range of applications, this system is suitable for substations of various sizes and types, from large hub substations to small distribution substations. It can be configured and deployed according to actual needs to monitor and warn of potential substation hazards. Furthermore, it can be integrated with other power system equipment management systems and dispatch automation systems to support the overall safe operation of the power system.

[0113] Example 4, refer to Figure 4 As one embodiment of the present invention, a substation hidden danger comprehensive monitoring device integrating multiple sensors is provided, comprising:

[0114] Device composition and installation steps:

[0115] Step 1: Installation of the water immersion monitoring unit.

[0116] Site selection principle: Based on the substation's terrain and drainage system, select areas prone to water accumulation for the installation of water immersion monitoring units, such as the lowest point of cable trenches, corners of basements, and the bottom of outdoor terminal boxes. These locations are typically high-risk areas for water immersion and can promptly reflect the substation's water immersion status.

[0117] Installation method: Use a fixed bracket to install the water immersion sensor at the selected location, ensuring the sensor maintains an appropriate distance from the ground or water surface, generally 1-2 cm, for accurate measurement of water level changes. The water immersion sensor is connected to the data acquisition unit via a waterproof cable. During connection, pay attention to the cable routing and securing it to avoid contact between the cable and live equipment to prevent danger. Simultaneously, ensure waterproofing at the cable connection points to prevent moisture infiltration and short circuits.

[0118] Parameter Settings: Based on the actual conditions of the substation and the water resistance of the equipment, set the alarm thresholds for the water immersion monitoring unit. For example, for ordinary equipment areas, set a water level exceeding 5cm as the warning threshold and exceeding 10cm as the alarm threshold; for important equipment areas, such as the main transformer room and control room, set a water level exceeding 2cm as the warning threshold and exceeding 5cm as the alarm threshold. The sampling frequency of the water immersion sensor is generally set to once per minute to ensure timely detection of rapid changes in water level.

[0119] Step 2: Installation of the displacement monitoring unit.

[0120] Site selection principle: Displacement monitoring units are mainly used to monitor the displacement of substation foundations and surrounding geology. Therefore, they should be installed in areas with significant changes in geological conditions, such as the foundations of the main substation buildings, support structures, perimeter walls, and surrounding areas. These locations can reflect the stability of the substation and its surrounding geology, and promptly detect abnormal displacement.

[0121] Installation method: Chemical anchors are used to fix the displacement sensor to the selected foundation or geological point, ensuring that the sensor is firmly and reliably installed and unaffected by external factors. The displacement sensor is connected to the data acquisition unit via a shielded cable. The cable routing should avoid strong electromagnetic interference sources, such as high-voltage lines and transformers, to reduce interference with signal transmission. Simultaneously, moisture-proof and corrosion-proof treatments should be applied to the cable connections to extend the cable's service life.

[0122] Parameter Settings: Based on the substation's design requirements and geological conditions, set the alarm thresholds for the displacement monitoring unit. For horizontal displacement of the substation foundation, a warning threshold is generally set for a displacement velocity exceeding 1 mm / month or a cumulative displacement exceeding 5 mm, and an alarm threshold is set for exceeding 10 mm. For vertical settlement, a warning threshold is set for a settlement velocity exceeding 2 mm / month or a cumulative settlement exceeding 10 mm, and an alarm threshold is set for exceeding 20 mm. The sampling frequency of the displacement sensor can be adjusted according to the rate of change of geological conditions. It is generally set to once per hour. Under relatively stable geological conditions, the sampling frequency can be appropriately reduced, such as once every 6 hours; during periods of rapid geological change or the rainy season, the sampling frequency can be increased, such as once every 30 minutes.

[0123] Step 3: Installation of the video monitoring unit.

[0124] Site Selection Principles: Video monitoring units are used to monitor the surrounding environment of substations in real time and identify potential hazards such as hanging objects and abnormal personnel activities. Therefore, they should be installed at various entrances and exits of the substation, around the perimeter walls, in equipment areas, areas prone to hanging objects (such as near construction sites, farmland, billboards, etc.), and at high points to ensure comprehensive coverage of the substation and its surrounding key areas.

[0125] Installation method: High-resolution cameras should be installed at the selected location using pole or wall mounting. The installation height should generally be no less than 3 meters to avoid obstruction of the view and to ensure the monitoring range meets requirements. The camera is connected to the data acquisition unit via a network cable. The network cable should be waterproof and rodent-proof armored, and properly routed and secured to prevent damage from wind, rain, or other factors. Simultaneously, the camera lens should be cleaned and calibrated regularly to ensure image clarity and monitoring effectiveness.

[0126] Parameter Settings: Configure the video monitoring unit parameters according to the substation's safety requirements and the size of the monitored area. The camera resolution should be 1080P or higher, and the frame rate 25fps or higher to ensure video image clarity and smoothness. Image recognition algorithm parameters should be optimized based on actual conditions. For example, for recognizing hanging objects, set appropriate size, shape, color, and other feature parameters to improve recognition accuracy; for recognizing abnormal personnel activity, set parameters such as activity range, speed, and behavior patterns to achieve automatic detection and alarm for suspicious individuals. The image transmission frame rate of the video monitoring unit can be adjusted according to network bandwidth and storage capacity, generally set to 5-10fps to reduce data traffic and storage space usage while ensuring monitoring effectiveness.

[0127] Step 4: Installation of the meteorological monitoring unit.

[0128] Site Selection Principles: Since meteorological monitoring units are used to monitor meteorological parameters around substations, they should be installed in open, well-ventilated areas to avoid obstruction and interference from buildings, trees, or other obstacles. Typically, meteorological sensors are installed on independent meteorological monitoring poles within the substation or at the edge of building rooftops.

[0129] Installation Method: Mount the weather sensors (temperature, humidity, wind speed, rainfall, etc.) on dedicated brackets. The brackets should be securely fixed to ensure stable operation of the sensors under various weather conditions. The weather sensors are connected to the data acquisition unit via signal cables. Shielded cables should be used, and proper wiring and securing should be implemented to prevent signal transmission interruption or interference due to wind, rain, or other factors. Regular calibration and maintenance of the weather sensors are essential to ensure data accuracy and reliability.

[0130] Parameter Settings: Based on local meteorological conditions and substation operational requirements, set alarm thresholds for the meteorological monitoring unit. For example, set a wind speed exceeding 15 m / s as a warning threshold and exceeding 20 m / s as an alarm threshold; set rainfall exceeding 30 mm within one hour as a warning threshold and exceeding 50 mm as an alarm threshold; set temperatures below -10℃ or above 40℃ as warning thresholds and below -15℃ or above 45℃ as alarm thresholds. The sampling frequency of the meteorological sensor is generally set to once every 10 minutes. Under special meteorological conditions, such as heavy rain or strong winds, the sampling frequency can be appropriately increased, such as once every 5 minutes, to reflect changes in meteorological parameters in a timely manner.

[0131] Step 5: Installation and configuration of the data acquisition unit.

[0132] Installation Location Selection: As the core component of the entire monitoring device, the data acquisition unit should be installed in a safe and reliable location within the substation, such as a control room or equipment room. This location should have good ventilation, heat dissipation, and dust protection, while also facilitating equipment maintenance and management.

[0133] Hardware Configuration and Connection: The data acquisition unit utilizes an industrial-grade computer with a high-performance processor, large-capacity memory, and storage to meet the needs of multi-channel data acquisition and processing. The computer should be equipped with multiple serial and network ports for connecting to each monitoring unit. Connect the signal cables from each monitoring unit to the corresponding ports on the data acquisition unit and ensure a secure connection to guarantee stable data transmission. Simultaneously, the data acquisition unit should be equipped with an uninterruptible power supply (UPS) to provide at least two hours of power support in the event of an external power failure, ensuring continuous acquisition and transmission of monitoring data.

[0134] Software Configuration and Debugging: Install data acquisition and processing software on the data acquisition unit. This software should have functions such as multi-channel data acquisition, data preprocessing, data storage, and data transmission. Configure the software's acquisition parameters, such as serial port baud rate, data bits, and stop bits, according to the communication protocol and data format of each monitoring unit to ensure correct reading and parsing of data from each monitoring unit. Debug and optimize the software to ensure stable operation and achieve real-time data acquisition and transmission. Simultaneously, develop a data cleaning and fusion subunit to clean the acquired data, removing outliers, duplicates, and other invalid data, and to fuse different types of data to improve data quality and usability.

[0135] Step 6: Installation and configuration of the wireless communication module.

[0136] Communication Method Selection: Based on the substation's network coverage and data transmission requirements, select an appropriate wireless communication method. Common wireless communication methods include 4G / 5G networks, NB-IoT, and LoRa. For monitoring devices with large data volumes and high real-time requirements, 4G / 5G networks can be selected for data transmission; for monitoring devices with smaller data volumes and lower power consumption requirements, NB-IoT or LoRa networks can be selected.

[0137] Equipment Installation and Connection: Install the wireless communication module inside or outside the data acquisition unit, and connect and configure it according to the equipment's installation manual. The wireless communication module should have good signal reception and transmission capabilities to ensure stable data transmission in the complex electromagnetic environment of the substation. During installation, pay attention to the antenna's installation position and direction to avoid signal obstruction and interference as much as possible.

[0138] Network Configuration and Security Settings: Configure the wireless communication module to set parameters such as APN, username, and password to enable it to access the corresponding wireless network. Simultaneously, employ encryption technologies such as SSL / TLS and VPN to encrypt data transmission, ensuring data security and confidentiality. During data transmission, regularly check network connectivity and data transmission quality, promptly addressing network failures and data loss issues to ensure timely and accurate transmission of monitoring data to the data center.

[0139] Installation and configuration of backup power unit:

[0140] Power supply equipment selection: Based on the power consumption and backup time requirements of the monitoring device, select a DC power supply system (DCS) or uninterruptible power supply (UPS) of appropriate capacity as the backup power unit. The backup power supply should have characteristics such as high stability, high reliability, and long lifespan, and be able to meet the power needs of the monitoring device in the event of an external power failure.

[0141] Installation and Connection: Install the backup power unit near the data acquisition unit and make electrical connections to the data acquisition unit and other monitoring units. The output voltage of the backup power supply should match the input voltage of the monitoring devices to ensure normal operation. Simultaneously, equip the backup power unit with a battery management system (BMS) to monitor and control battery charging and discharging in real time, extending battery life and ensuring safe battery operation.

[0142] Parameter Setting and Testing: Based on actual needs, set parameters such as charging current, discharging current, undervoltage protection, and overvoltage protection for the backup power unit to ensure that the backup power supply can be put into operation promptly in the event of an external power failure and automatically switch back to normal power supply mode after power is restored. Perform charge / discharge tests and switching tests on the backup power unit to check its performance and reliability under different operating conditions, ensuring that the backup power supply can provide stable power support for the monitoring device at critical moments.

[0143] Device workflow and operating procedures:

[0144] Step 1: Monitoring data collection and preprocessing.

[0145] Data Acquisition: Each monitoring unit collects data in real time according to a set sampling frequency and transmits the collected data to the data acquisition unit via wired or wireless means. The data acquisition unit reads the data from each monitoring unit through the corresponding interface program and performs preliminary processing and storage. For example, the water immersion monitoring unit collects water level data once per minute and transmits it to the data acquisition unit via a 485 serial port; the displacement monitoring unit collects displacement data once per hour and transmits it to the data acquisition unit via a network port; the video monitoring unit collects image data in real time and transmits it to the data acquisition unit via a network port; the meteorological monitoring unit collects meteorological data every 10 minutes and transmits it to the data acquisition unit via a 485 serial port.

[0146] Data Preprocessing: The data acquisition unit preprocesses the received data, including data cleaning and fusion. The data cleaning subunit performs operations such as noise reduction, deduplication, and data completion to remove outliers and erroneous data, improving data quality. For example, for water immersion monitoring data, if multiple consecutive data points exceed the sensor's measurement range, they are considered abnormal data and discarded; interpolation is then used to complete the missing data. For displacement monitoring data, if the rate of change exceeds physical limits, it is considered noisy data and smoothed. The data fusion subunit merges data collected from different monitoring units, establishing correlations between the data to provide more comprehensive and accurate information for subsequent data analysis and hazard identification. For example, displacement monitoring data can be merged with meteorological monitoring data to analyze the impact of meteorological factors such as rainfall and wind speed on displacement changes.

[0147] Step 2: Data transmission and storage.

[0148] Data transmission: The data acquisition unit transmits the pre-processed data to the data center via the wireless communication module. During transmission, the data acquisition unit encapsulates the data according to a preset data transmission protocol and format to ensure data integrity and accuracy. The wireless communication module selects an appropriate time and method to send data based on network conditions and data priority. For example, when network bandwidth is sufficient, video monitoring data and displacement monitoring data with high real-time requirements are sent first; when network bandwidth is limited, the data is compressed to prioritize the transmission of critical data. Simultaneously, the data acquisition unit records the status and time of data transmission and retransmits failed data to ensure timely and accurate transmission to the data center.

[0149] Data Storage: After receiving data from the monitoring devices, the data center stores the data according to a pre-defined database table structure. Data storage should follow certain rules and strategies, such as storing data in chronological order and creating indexes for easy querying and analysis; classifying and storing data from different monitoring units to facilitate data management and access. Simultaneously, the data center should regularly back up and archive the stored data, employing redundant storage technologies such as disk arrays and optical disc storage to ensure data security and reliability. The data backup cycle can be set according to the data volume and importance, generally performing a full backup weekly and an incremental backup daily.

[0150] Step 3: Hazard identification and analysis.

[0151] Data Processing and Feature Extraction: The data processing module in the data center performs in-depth processing and analysis on the received data. First, the data is normalized to eliminate the influence of different units and dimensions. Then, signal processing methods such as wavelet transform and Fourier transform are used to extract features from the data, including time-domain, frequency-domain, and time-frequency-domain features, providing feature vectors for hazard identification. For example, for displacement monitoring data, features such as mean, variance, peak value, and frequency are extracted; for video monitoring data, features such as edge, texture, and color are extracted.

[0152] Hazard Identification and Classification: The extracted feature vectors are input into a pre-trained deep learning model for hazard identification and classification. Based on the learned features and patterns, the model automatically identifies and classifies monitoring data to determine the presence and type of hazards. For example, for video monitoring data, the model can identify the shape, size, and color of hanging objects and classify them into different types such as plastic film, kites, and bird nests; for displacement monitoring data, the model can determine the trend and rate of displacement change and classify it into different states such as normal displacement, slow displacement, and rapid displacement. Simultaneously, the data processing module feeds back the identification results and related feature information to the visualization and monitoring early warning modules for subsequent display and early warning processing.

[0153] Step 4: Generating and sending early warning information.

[0154] Risk Assessment and Early Warning Level Determination: After receiving the hazard identification results, the monitoring and early warning module invokes the risk assessment model. Combining factors such as the hazard's type, location, and severity, it assesses the hazard's risk and determines its risk level. The risk assessment model calculates the hazard's risk value based on preset weights and assessment rules, classifying it into high, medium, and low levels. For example, a rapidly displacing hazard located in the core area of ​​a substation might have a high risk value and be assessed as high-risk; a minor hanging object hazard located at the substation's edge might have a low risk value and be assessed as low-risk. Based on the risk level and preset early warning rules, the monitoring and early warning module determines whether to generate an early warning message and the urgency level of the warning.

[0155] Early Warning Information Generation and Dissemination: For high-risk hazards, the monitoring and early warning module immediately generates a Level 1 early warning, which includes the hazard's location, type, risk level, occurrence time, and recommended handling. This information is simultaneously communicated to relevant personnel via SMS, email, and audible / visual alarms. SMS messages are concise and clear, containing key hazard information and emergency contact details; emails are detailed and comprehensive, including relevant data and image evidence; and audible / visual alarms emit strong signals on-site to alert personnel to take immediate action. For medium-risk hazards, the monitoring and early warning module generates a Level 2 early warning, disseminating information via SMS and email. While the SMS and email messages are relatively concise, they contain sufficient information to inform relevant personnel about the hazard and enable them to take appropriate measures. For low-risk hazards, the monitoring and early warning module generates a Level 3 early warning, primarily disseminating information via email. Emails are primarily informational, reminding relevant personnel to monitor changes in the hazard. The dissemination scope of the early warning information is configured based on the hazard type and location to ensure that relevant maintenance personnel, management personnel, and emergency response personnel are notified.

[0156] Step 5: Feedback and evaluation of processing results.

[0157] Feedback on handling results: After receiving the early warning information, relevant personnel will promptly take corresponding measures to address the potential hazard based on the suggested handling methods. During the handling process, the progress and results will be reported to the monitoring and early warning module in real time via a mobile app or web page. Feedback information includes the personnel involved, the time of handling, the measures taken, and the results, enabling the monitoring and early warning module to understand the status of the hazard handling and update the status of the early warning information. For example, regarding a flooding hazard, the personnel will report that they have arrived on site, activated the drainage system, and the water level has begun to drop. After the handling is completed, they will report that the water level has returned to normal and the hazard has been eliminated.

[0158] Processing Result Evaluation and Recording: The monitoring and early warning module evaluates and records the received processing result feedback information. Based on the feedback information, it assesses whether the hazard has been effectively addressed. If the hazard has been eliminated, the corresponding early warning information is deactivated, and the processing result is recorded in the database for historical data storage. If the hazard has not been completely eliminated or new anomalies occur, the early warning information is escalated, the notification scope is expanded, and higher-level management personnel are notified to intervene. Simultaneously, detailed records and statistical analyses are performed on each early warning information, processing procedure, and result, providing data support for subsequent hazard management, equipment maintenance, and system optimization.

[0159] The device boasts advantages, integrating multiple monitoring methods such as water immersion monitoring, displacement monitoring, video monitoring, and meteorological monitoring. This enables comprehensive and multi-dimensional monitoring of the substation and its surrounding environment, allowing for timely detection of various potential hazards and improving the comprehensiveness and accuracy of monitoring. For example, during a heavy rainfall event, the monitoring device not only detected the rising water level in the substation's cable trench but also monitored changes in the surrounding soil displacement and potential landslides outside the perimeter wall, providing strong support for the substation's flood control and emergency response efforts. The displacement monitoring unit achieves millimeter-level accuracy, video monitoring recognition accuracy is ≥91%, and water level monitoring error is ≤1cm, enabling timely detection of even minor changes in potential hazards and providing strong support for early warning and response. For instance, long-term displacement monitoring of the substation foundation revealed a gradually accelerating settlement rate. Analysis indicated changes in the geological conditions beneath the foundation, allowing for timely reinforcement measures to prevent potential equipment tilting and damage. Advanced image recognition algorithms and data processing technologies are employed to achieve automatic identification and analysis of potential hazards, reducing manual intervention and improving monitoring efficiency. For example, the video monitoring unit can automatically identify plastic film hanging around the substation and issue timely warnings. Staff can then quickly arrive at the scene to clear the film, preventing accidents such as short circuits caused by the hanging debris to substation equipment. The monitoring device has excellent environmental adaptability and can operate stably in the complex operating environment of a substation. For example, the device can operate normally in harsh environments such as high temperature, high humidity, and high altitude, providing reliable safety monitoring for substations in different regions.

[0160] At a substation in a coastal area, the substation equipment is constantly exposed to sea breezes and typhoons, making it susceptible to threats such as salt spray corrosion and impacts from falling debris. After installing this monitoring device, analysis of video monitoring data promptly identified multiple potential debris hazards and notified maintenance personnel via an early warning system to remove them, effectively preventing short circuits caused by debris. Simultaneously, the meteorological monitoring unit monitored wind speed changes in real time before typhoons arrived, issuing early warnings so the substation could take timely windproofing measures, such as reinforcing outdoor equipment and closing doors and windows, reducing typhoon damage to the substation.

[0161] At a substation in a mountainous area, the surrounding geological conditions are complex, posing risks of landslides and mudslides. The displacement monitoring unit and water immersion monitoring unit of this monitoring device played a crucial role. After a heavy rain, the displacement monitoring unit detected a significant increase in the displacement velocity of the mountain behind the substation, and the water immersion monitoring unit detected a sharp rise in the water level in the cable trench. The monitoring and early warning system immediately issued a high-risk warning, notifying substation personnel to evacuate immediately and activating the emergency response plan. On-site investigation revealed that a landslide had damaged part of the cable trench; if not detected and addressed promptly, it could have caused a cable short circuit and a substation power outage. The application of this monitoring device successfully prevented a major safety accident.

[0162] Example 5 is an embodiment of the present invention. This embodiment also provides an electronic device applicable to a comprehensive monitoring method for substation hazards integrating multiple sensors, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the comprehensive monitoring method for substation hazards integrating multiple sensors as proposed in the above embodiments.

[0163] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements a comprehensive monitoring method for substation hazards integrating multiple sensors as proposed in the above embodiments.

[0164] The storage medium proposed in this embodiment and the method for comprehensive monitoring of substation hidden dangers integrating multiple sensors proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0165] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0166] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A comprehensive monitoring method for potential hazards in substations integrating multiple sensors, characterized in that: include, Collect various operational data from the substation and its surrounding area; The collected data is cleaned and merged to construct a unified dataset; A unified dataset is input into a recognition structure that combines convolutional neural networks and recurrent neural networks to process image-type and time-series data respectively, and to complete the identification of hazard types according to the pre-set training set ratio and training parameters. Based on the identification results and the characteristics of potential hazards, risk values ​​are calculated and risk levels are determined. Based on the judgment results, early warning information is generated, categorized, pushed and displayed, and feedback information on early warning processing is received. Status updates and record archiving are also completed.

2. The integrated multi-sensor substation hazard monitoring method as described in claim 1, characterized in that: The collection of various operational data from the substation and its surroundings includes acquiring the operational status of multiple monitoring areas and transmitting the collected data to the same processing flow for unified processing.

3. The integrated multi-sensor substation hazard monitoring method as described in claim 2, characterized in that: The unified dataset is input into the recognition structure constructed by combining convolutional neural networks and recurrent neural networks. include, Different types of input data are processed accordingly, and the recognition structure obtains parameters through the training process.

4. The integrated multi-sensor substation hazard monitoring method as described in claim 3, characterized in that: The risk value calculation and level determination include a comprehensive analysis based on multiple hazard-related information, and the generation of determination results according to the corresponding assessment logic.

5. The integrated multi-sensor substation hazard monitoring method as described in claim 4, characterized in that: The identification of the types of hazards includes, Inputting image data into a convolutional neural network model extracts spatial features including image edges, textures, and colors. Input time series data into a recurrent neural network model to extract trends; The identification results are compared with the data labels, and joint training is performed based on the set ratio of training set, validation set and test set until the preset accuracy threshold is met.

6. The integrated multi-sensor substation hazard monitoring method as described in claim 5, characterized in that: The execution risk value calculation and level determination include, The risk assessment model is invoked to quantify multiple characteristics of the identified hazards, including their type, location, severity, and development trend, and to assign preset weights. The fuzzy comprehensive evaluation method is used for calculation. The model output results are compared with the classification standard to determine whether the risk level is high, medium or low.

7. The integrated multi-sensor substation hazard monitoring method as described in claim 6, characterized in that: The system generates early warning information based on the judgment results, and then categorizes, pushes, and displays it. include, When the monitoring data reaches the warning threshold, or when the identified hidden danger is assessed as high-risk, a level one warning message is generated and sent simultaneously via SMS, email and on-site audible and visual alarms. When the potential hazard is classified as medium risk, a Level II early warning message is generated and sent via SMS and email. When the potential hazard is classified as low-risk, a Level 3 early warning message is generated and sent via email. Receive information on processing progress and results from relevant personnel via mobile terminals, and update the status of corresponding warning information; If a potential hazard is not fully addressed or a new anomaly occurs during the handling process, the warning information level will be upgraded and the scope of notification recipients will be expanded.

8. A substation hazard comprehensive monitoring system integrating multiple sensors, employing the substation hazard comprehensive monitoring method integrating multiple sensors as described in any one of claims 1 to 7, characterized in that, include: Data acquisition unit, data cleaning and fusion unit, data processing module, risk assessment model module, and monitoring and early warning module; The data acquisition unit is used to collect various operational data of the substation and its surrounding area; The data cleaning and fusion unit is used to clean and fuse the collected data to construct a unified dataset; The data processing module is used to input a unified dataset into a recognition structure that combines convolutional neural networks and recurrent neural networks, respectively for processing image-type and time-series data, and to complete the identification of hazard types according to the pre-set training set ratio and training parameters. The risk assessment model module is used to combine the identification results and hazard-related characteristics to perform risk value calculation and level determination. The monitoring and early warning module is used to generate early warning information based on the judgment results, and to classify, push and display the information, receive early warning processing feedback information, and complete status updates and record archiving.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the integrated multi-sensor substation hazard monitoring method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the integrated multi-sensor substation hazard monitoring method according to any one of claims 1 to 7.

Images