Smart site monitoring visualization system and method based on multi-source heterogeneous data fusion

Abstract

The present invention relates to the field of station monitoring technology, specifically to a smart station monitoring visualization system and method based on multi-source heterogeneous data fusion, the system includes a data integration module, an anomaly detection module, a trend prediction module, a dynamic adjustment module, an intelligent alarm module and a learning optimization module. In the present invention, by using time series analysis and data cleaning, the accuracy and availability of data are improved, thereby ensuring the quality of monitoring data, applying long short-term memory networks to make anomaly detection more accurate, able to identify small data changes, thereby early warning of potential safety risks, through in-depth analysis of trends and probability calculations, can output more accurate future state prediction information, enhance the basis for decision support, dynamic adjustment strategies allow real-time adjustment of monitoring parameters, make resource allocation more flexible and efficient, and the response process adapts to different emergency situations through multi-level trigger conditions, thereby improving the efficiency of emergency management.

Description

Technical Field

[0001] The present invention relates to the field of site monitoring technology, and in particular to a smart site monitoring visualization system and method based on multi-source heterogeneous data fusion. Background Art

[0002] The field of site monitoring technology focuses on using various sensors and monitoring equipment to collect real-time data within the site to achieve real-time monitoring and management of its environment, safety and operating conditions. It covers a variety of means from video monitoring to environmental monitoring. Through real-time data analysis, site monitoring can warn of potential safety hazards, optimize the operating efficiency of the site, reduce maintenance costs, and improve the level of safety management.

[0003] Among them, the smart site monitoring visualization system based on multi-source heterogeneous data fusion involves using information from different data sources, integrating and analyzing data through data fusion technology, and building an intelligent site monitoring system. The system can not only monitor various physical and environmental parameters within the site, but also provide more accurate prediction and decision support through data fusion. The visualization part makes the monitoring information more intuitive and easy to understand, helping operation and maintenance personnel to quickly identify problems and respond.

[0004] Traditional systems lack advanced data fusion capabilities, which often leads to information dispersion and the inability to comprehensively evaluate the operating status of the station. They cannot effectively identify complex or subtle abnormal patterns when performing anomaly detection, resulting in untimely responses. The response strategies of existing systems are mostly static configurations, lack flexibility, and cannot be dynamically adjusted based on real-time data. They are insufficient to cope with various emergencies in a changing operating environment, which increases the complexity of operation and maintenance, increases manpower and costs, and reduces the safety management and operating efficiency of the station. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings existing in the prior art and to propose a smart site monitoring visualization system and method based on multi-source heterogeneous data fusion.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solution: A smart station monitoring visualization system based on multi-source heterogeneous data fusion includes:

[0007] The data integration module integrates the data streams for time series analysis based on the video streams and sensor data monitored by the station, and cleans and formats the data according to the operation data of the station to obtain a standard set of operation data;

[0008] The anomaly detection module, based on the standard set of operation data, determines the mutation points in the data through the long short-term memory network, analyzes the preliminary abnormal signals, identifies the cause and type of the abnormal state, and generates an abnormal state analysis log;

[0009] The trend prediction module performs abnormal trend analysis based on the abnormal state analysis log, calculates the probability distribution of the future state, verifies the predicted state, determines the reliability of the prediction, and generates a prediction accuracy verification result;

[0010] The dynamic adjustment module modifies the data sampling frequency and monitoring range based on the prediction accuracy verification result, refines the data collection point configuration, and reconfigures resources using the adjusted monitoring parameters to optimize the station operation efficiency and generate an adjustment efficiency optimization record;

[0011] The intelligent alarm module constructs multi-level alarm trigger conditions based on the regulation efficiency optimization record, implements an automated response process, adjusts the alarm level and response process in combination with the current monitoring data of the station, and generates an alarm response configuration;

[0012] The learning optimization module uses stochastic gradient descent based on the alarm response configuration to automatically adjust the long short-term memory network parameters, perform network retraining, and optimize the accuracy of anomaly detection and state prediction to generate continuous optimization feedback results.

[0013] As a further solution of the present invention, the abnormal status analysis log includes abnormal event identification, impact assessment results, and abnormal source analysis results; the prediction accuracy verification results include error rate calculation results, prediction confidence intervals, and calibration results; the adjustment efficiency optimization records include resource reconfiguration evaluation information, efficiency improvement index, and operation delay reduction information; the alarm response configuration includes trigger mechanism, response priority sorting, and alarm coverage.

[0014] As a further solution of the present invention, the data integration module includes a data acquisition submodule, a data cleaning submodule, and a data formatting submodule;

[0015] The data acquisition submodule performs real-time data acquisition based on the video stream and sensor data monitored by the site, collects information about the environment and equipment operation status, fuses the video and sensor data, performs preliminary time tagging, and generates an initial data fusion set;

[0016] The data cleaning submodule analyzes the data quality based on the initial data fusion set, identifies and removes noise and outliers in the data, and fills in missing values ​​of key data to obtain an optimized data set;

[0017] Based on the optimized data set, the data formatting submodule sorts and time-standardizes the data through time series analysis, adjusts the data timestamp, refines the data structure to match the analysis requirements, and establishes a standard set of operating data.

[0018] As a further solution of the present invention, the anomaly detection module includes an anomaly identification submodule, a signal analysis submodule, and a log generation submodule;

[0019] The abnormal identification submodule analyzes discontinuities and mutation points in the data stream through a long short-term memory network based on the standard set of operation data, determines abnormal signals in the data, classifies the abnormal signals, and generates a preliminary abnormal signal set;

[0020] The signal analysis submodule analyzes the characteristics and influencing factors of each signal based on the preliminary abnormal signal set, determines the abnormal category and potential cause in combination with changes in environmental and operating conditions, and obtains details of the abnormal factors;

[0021] The log generation submodule records the time, location, impact scope and potential risks of the abnormality based on the details of the abnormal factors, and organizes the recorded data to generate an abnormal status analysis log.

[0022] As a further solution of the present invention, the trend prediction module includes a trend analysis submodule, a probability calculation submodule, and a verification submodule;

[0023] The trend analysis submodule extracts the occurrence frequency and pattern of abnormal events based on the abnormal state analysis log, identifies abnormal development trends, predicts future potential abnormal patterns, and generates development trend prediction information;

[0024] The probability calculation submodule performs quantitative analysis on future abnormal events based on the development trend prediction information, calculates the potential for occurrence, and evaluates the risk level of different abnormal types to obtain a risk probability distribution map;

[0025] The verification submodule verifies the prediction results in real time based on the risk probability distribution diagram through historical comparative analysis, adjusts the prediction parameters and matches the new data, and generates a prediction accuracy verification record.

[0026] As a further solution of the present invention, the dynamic adjustment module includes a sampling adjustment submodule, a monitoring range adjustment submodule, and a resource configuration submodule;

[0027] The sampling adjustment submodule evaluates the current effect of the data sampling process based on the prediction accuracy verification record, adjusts the data sampling frequency according to the risk probability, optimizes the data collection time and quality, and obtains the sampling frequency optimization configuration;

[0028] The monitoring scope adjustment submodule is based on the sampling frequency optimization configuration, expands or reduces the monitoring scope, covers key risk points, optimizes the distribution and utilization of monitoring resources, and obtains a monitoring scope adjustment record;

[0029] The resource configuration submodule adjusts the monitoring scope record based on the monitoring scope, reconfigures the monitoring resources, including monitoring points and human resources, optimizes the monitoring response capability and efficiency, and generates an adjustment efficiency optimization record.

[0030] As a further solution of the present invention, based on the monitoring range adjustment record, the monitoring resources are reconfigured according to the formula: Calculate the regulation efficiency, where represents the efficiency of resource allocation, Represents the effect evaluation score of the monitoring point, represents the frequency weight of the monitoring point, is the total number of monitoring points. As a further solution of the present invention, the intelligent alarm module includes an alarm triggering condition submodule, a response process submodule, and an alarm level adjustment submodule;

[0031] The alarm trigger condition submodule analyzes the real-time data of the station based on the regulation efficiency optimization record, identifies key monitoring points, sets the trigger threshold of the multi-level alarm, and automatically triggers the logic according to the differential risk level to generate the multi-level alarm condition setting;

[0032] The response process submodule builds a matching automated response process based on the multi-level alarm condition settings, including emergency response and routine monitoring response, minimizes the impact of potential risks, and optimizes alarm response time and efficiency to obtain an optimized configuration of the response process;

[0033] The alarm level adjustment submodule optimizes the configuration according to the response process, dynamically adjusts the alarm level setting, updates the response priority, matches the monitoring environment and security requirements, and obtains the alarm response configuration.

[0034] As a further solution of the present invention, the learning optimization module includes a parameter adjustment submodule, a network retraining submodule, and a performance optimization submodule;

[0035] The parameter adjustment submodule uses stochastic gradient descent based on the alarm response configuration to adjust the learning rate and weight of the long short-term memory network, optimize the adaptability and predictive ability to new trends, and generate network parameter tuning information;

[0036] The network retraining submodule performs LSTM network retraining based on the network parameter tuning information, adjusts the network structure according to the collected data samples, and obtains the network structure optimization result;

[0037] The performance optimization submodule evaluates the performance in site monitoring based on the network structure optimization results, analyzes the efficiency of anomaly detection and state prediction, and generates continuous optimization feedback results.

[0038] The method for visualizing smart site monitoring based on multi-source heterogeneous data fusion is implemented based on the above-mentioned smart site monitoring visualization system based on multi-source heterogeneous data fusion, and includes the following steps:

[0039] S1: Based on the video stream and sensor data monitored by the station, integrate the data stream for time series analysis, extract the key trends and periodic factors of the time series, and perform data cleaning and formatting based on the station operation data to generate a standard set of operation data;

[0040] S2: Based on the standard set of operation data, use the long short-term memory network to identify abnormal mutation points in the data, determine the cause and type of the abnormal state by comparing and analyzing the data before and after the mutation point, and generate an abnormal state analysis log;

[0041] S3: Based on the abnormal state analysis log, analyze the abnormal event trend, calculate the probability distribution of future abnormal events, and verify the predicted state through historical data to generate a prediction accuracy verification result;

[0042] S4: Based on the prediction accuracy verification result, adjust the data sampling frequency and monitoring range, refine the data collection point configuration, and use the adjusted monitoring parameters to optimize the station operation efficiency and generate a regulation efficiency optimization record;

[0043] S5: Based on the regulation efficiency optimization record, the alarm level and response process are dynamically adjusted according to the real-time monitoring data, and the parameters of the long short-term memory network are adjusted by using stochastic gradient descent, the network is retrained, and the accuracy of anomaly detection and state prediction is optimized to generate continuous optimization feedback results.

[0044] Compared with the prior art, the advantages and positive effects of the present invention are:

[0045] In the present invention, by utilizing time series analysis and data cleaning, the accuracy and availability of data are improved, thereby ensuring the quality of monitoring data. The application of long short-term memory networks makes anomaly detection more accurate and can identify small data changes, thereby providing early warning of potential safety risks. Through in-depth analysis of trends and probability calculations, more accurate future state prediction information can be output to enhance the basis for decision support. Dynamic adjustment strategies allow real-time adjustment of monitoring parameters, making resource allocation more flexible and efficient. The response process adapts to different emergency situations through multi-level trigger conditions, thereby improving the efficiency of emergency management. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a system flow chart of the present invention;

[0047] Figure 2 It is a schematic diagram of the system framework of the present invention;

[0048] Figure 3 It is a flow chart of the data integration module of the present invention;

[0049] Figure 4 is a flow chart of the abnormality detection module of the present invention;

[0050] Figure 5 is a flow chart of the trend prediction module of the present invention;

[0051] Figure 6 It is a flow chart of the dynamic adjustment module of the present invention;

[0052] Figure 7 It is a flow chart of the intelligent alarm module of the present invention;

[0053] Figure 8 This is a flow chart of the learning optimization module of the present invention;

[0054] Fig. 9 It is a schematic diagram of the method steps of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0056] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0057] Example 1: Please refer to Figure 1 The present invention provides a technical solution: a smart station monitoring visualization system based on multi-source heterogeneous data fusion includes:

[0058] The data integration module integrates the data streams for time series analysis based on the video streams and sensor data monitored by the station, and cleans and formats the data according to the operation data of the station to obtain a standard set of operation data;

[0059] Based on 3DGIS maps, three-dimensional modeling, logical modeling, automatic model generation and digital twin modeling methods, the physical entities in the railway station are refinedly modeled and coupled simulated in terms of geometry, function and performance, so as to accurately simulate the shape, behavior and performance of physical entities such as tracks, parking areas, cargo areas, container areas, vehicles, loading and unloading tools, etc., and realize high-fidelity simulation, real-time dynamic control and risk warning of physical entities in the railway station;

[0060] Using the freight yard plan as a carrier, grid modeling is carried out to build infrastructure management such as container areas, container spaces, and tracks based on the Internet of Things, management of loading and unloading equipment such as gantry cranes, reach stackers, and container forklifts based on 5G transmission, and management of loading and unloading and transshipment operations plans based on the Internet of Things;

[0061] Establish an infrastructure collection network, based on the automatic collection of freight yard production resources and production links, and through unified system access and integrated calculations, build an efficient and accurate data perception and collection network to achieve real-time perception and seamless connection of people, vehicles, goods, and roads.

[0062] The anomaly detection module is based on the standard set of operation data, uses the long short-term memory network to determine the mutation points in the data, analyzes the preliminary abnormal signals, identifies the cause and type of the abnormal state, and generates an abnormal state analysis log;

[0063] The trend prediction module performs abnormal trend analysis based on abnormal state analysis logs, calculates the probability distribution of future states, verifies the predicted state, determines the reliability of the prediction, and generates prediction accuracy verification results;

[0064] Management of loading and unloading equipment such as gantry cranes, reach stackers, and container forklifts based on 5G; the cargo yard is equipped with corresponding gantry cranes, reach stackers, container forklifts, and other loading and unloading equipment. Sensors, cameras, and other information collection devices based on 5G, Beidou, and other technologies are installed on each piece of equipment according to its business characteristics to achieve real-time tracking and AI predictive analysis of the operating status of the loading and unloading equipment;

[0065] Loading, unloading and transshipment operation plan management based on the Internet of Things; according to the unified planning of container areas and container spaces, container transshipment needs, cargo yard loading and unloading operation progress and shunting plans, the Internet of Things, 5G, big data, artificial intelligence and other technologies are used to comprehensively consider the collaborative relationship between various operations inside and outside the cargo yard, realize the automatic preparation of production operation plans such as loading, on-site transshipment, unloading and stacking, intelligent scheduling and decision-making of the entire production process, and optimize the allocation of terminal resources.

[0066] The dynamic adjustment module modifies the data sampling frequency and monitoring range based on the prediction accuracy verification results, refines the data collection point configuration, and uses the adjusted monitoring parameters to reconfigure resources, optimize the station operation efficiency, and generate adjustment efficiency optimization records;

[0067] The intelligent alarm module optimizes records based on regulation efficiency, builds multi-level alarm trigger conditions, implements automated response processes, adjusts alarm levels and response processes based on current monitoring data from the station, and generates alarm response configurations;

[0068] Based on technologies such as IoT, 5G, AI, big data, and video recognition, we can realize the automated management of cargo loading and unloading operations based on IoT;

[0069] Gantry crane automation control. Using technologies such as the Internet of Things, LiDAR, Beidou / UWB positioning, 5G network, and image recognition, the gantry crane automatic control system can realize functions such as automatic identification and error correction and grabbing and releasing of containers, automatic positioning and deviation correction of gantry cranes, active anti-collision of large vehicles, anti-lifting of container trucks, intelligent alarm for container operations based on F-TR locks, anti-collision of containers, and safety protection of operating areas; reach stacker automation control. By installing 5G+AI cameras around the reach stacker, it is possible to identify the container number and monitor the operation process by back-end management personnel;

[0070] Comprehensively utilize multiple technologies such as the Internet of Things, 5G communication, Beidou positioning, digital twins, etc. to realize functions such as freight yard intrusion protection and vehicle and personnel positioning, further improve the safety management and control level of the Handan-Huanghuai Railway freight yard, and realize freight yard safety management and control. Based on BIM, GIS positioning, 5G, and AI intelligent identification, key areas are visually marked and displayed, and a holographic monitoring model is constructed. Through AI video recognition, 5G+ high-precision positioning work badges and other equipment, the system monitors the location information, movement trajectory and other information of operating personnel or vehicles in real time, and realizes security incident alarms such as illegal intrusion, illegal residence, and unauthorized entry. Through various sensors such as temperature and humidity, the environment and cargo information of key areas such as transshipment and warehousing in the freight yard are collected and monitored, and the on-duty personnel are notified in time through sound and light alarms. The safety of entering and leaving the freight yard and the intrusion monitoring of electronic fences are monitored. According to different security management requirements, different levels of electronic fences are set to realize functions such as intrusion protection against people and vehicles.

[0071] Based on the alarm response configuration, the learning optimization module uses stochastic gradient descent to automatically adjust the long short-term memory network parameters, perform network retraining, and optimize the accuracy of anomaly detection and state prediction, generating continuous optimization feedback results.

[0072] The abnormal status analysis log includes abnormal event identification, impact assessment results, and abnormal source analysis results. The prediction accuracy verification results include error rate calculation results, prediction confidence interval, and calibration results. The adjustment efficiency optimization record includes resource reconfiguration evaluation information, efficiency improvement index, and operation delay reduction information. The alarm response configuration includes trigger mechanism, response priority sorting, and alarm coverage.

[0073] See also Figure 2 and Figure 3 ,The data integration module includes a data acquisition submodule, a data cleaning submodule, and a data formatting submodule;

[0074] The data acquisition submodule performs real-time data acquisition based on the video stream and sensor data monitored by the site, collects information about the environment and equipment operation status, fuses the video and sensor data, performs preliminary time tagging, and generates an initial data fusion set;

[0075] Real-time data integration is performed using the video stream and sensor data monitored by the site. Information is collected using the status information of the environment and equipment operation. For this process, frame analysis of the video data is performed to identify key dynamic information, while the sensor data is processed through frequency filtering and other methods to remove environmental interference. The two parts of data are synchronously marked on the timeline to ensure time consistency. The marking process involves timestamp correction and time zone conversion to generate an initial data fusion set.

[0076] The main goal of multi-source data fusion is to integrate heterogeneous and dispersed data into a unified, coherent and accurate information source, which includes human, vehicle, equipment and environmental data in the real-time operation scenario of the yard;

[0077] 1) Data alignment: ensuring that data from different sources remain consistent in time and space;

[0078] 2) Feature extraction: Extract feature information from the data that helps describe and identify the target object, such as time information in the same space, spatial location information at the same time, and parameter information in both time and space;

[0079] 3) Data association: establish the corresponding relationship between objects in different data sources;

[0080] 4) Target recognition and classification: Identify and classify target objects based on the extracted features and associations;

[0081] 5) Decision fusion: Integrate information from various data sources to form a comprehensive and accurate description of the target object, and make fusion based on this.

[0082] The data cleaning submodule analyzes data quality based on the initial data fusion set, identifies and removes noise and outliers in the data, and fills in missing values ​​of key data to obtain an optimized data set;

[0083] Based on the initial data fusion set, fill in the missing values ​​of key data according to the formula: Get the filled data set, where represents the result of filling missing values ​​in the data. Represents the valid data values ​​in the data set, Represents the number of valid data.

[0084] In the data cleaning process, the missing values ​​of key data are filled by calculating the mean of valid data. Assuming that the valid data values ​​in the data set are 100, 200, and 300, the number of valid data is is 3, according to the formula:

[0085] It is calculated that the missing values ​​should be filled with 200, indicating that missing data can be effectively handled by filling in the mean value and ensuring data integrity.

[0086] The data formatting submodule sorts and standardizes the data based on the optimized data set through time series analysis, adjusts the data timestamp, refines the data structure to match the analysis requirements, and establishes a standard set of operational data;

[0087] The cleaned data set is sorted in chronological order, and then the timestamp is adjusted, including converting the time mark of each data point into a unified standard time format, such as UTC time, to ensure the consistency and accuracy of the data in the time series. Furthermore, the data structure is refined according to the analysis requirements to adapt to different analysis models, and a standard set of operational data is constructed to ensure the accurate application and efficient processing of the data. Finally, the timestamp and structure of the data are adjusted, providing a standardized and structured data set for subsequent data analysis and application.

[0088] See also Figure 2 and Figure 4 ,The anomaly detection module includes an anomaly identification submodule, a signal analysis submodule, and a log generation submodule;

[0089] The anomaly identification submodule is based on the standard set of operation data, analyzes the discontinuities and mutation points in the data stream through the long short-term memory network, determines the abnormal signals in the data, classifies the abnormal signals, and generates a preliminary abnormal signal set;

[0090] Based on the standard set of operational data, determine the abnormal signals in the data according to the formula: Get the initial abnormal signal set, where Represents abnormal signal detection information, represents a data point, represents the average of the data points, represents the weight factor, represents the threshold value, Represents the number of data points.

[0091] There are three data points ,average value , the weight factor is , threshold , the calculation process is: because Less than threshold ,This data group is not considered abnormal, in order to adapt to different situations, the detection efficiency is optimized by adjusting the weights and thresholds.

[0092] The signal analysis submodule analyzes the characteristics and influencing factors of each signal based on the preliminary abnormal signal set, determines the abnormal category and potential cause in combination with the changes in environmental and operating conditions, and obtains the details of the abnormal factors;

[0093] Based on the preliminary abnormal signal set, analyze the characteristics and influencing factors of each signal, including the measurement of the amplitude, frequency and duration of the abnormal signal. The data feature extraction of each signal is carried out according to its performance in the time series. For each type of abnormal signal, it is compared with the changes in environmental and operating conditions. The external variables associated with the abnormal signal are identified through statistical analysis methods, including temperature fluctuations, operating errors or equipment failures. According to the data, the abnormal category and potential causes are determined, and a detailed report on the abnormal factors is generated.

[0094] The log generation submodule records the time, location, impact scope and potential risks of the abnormality based on the details of the abnormal factors, and organizes the recorded data to generate an abnormal status analysis log;

[0095] According to the details of the abnormal factors, the time, location, scope of impact and potential risks of the abnormality are recorded. It is required to be able to access the real-time data stream and record the data synchronously. Subsequently, the recorded data is sorted and formatted to generate an abnormal status analysis log that is easy to understand and analyze. It not only involves the data entry process, but also includes data archiving and query optimization to ensure that the details of the abnormal event can be quickly extracted and reviewed when needed. The whole process emphasizes the safe storage and efficient retrieval of data. The generated log records the comprehensive information of each abnormal event in detail.

[0096] See also Figure 2 and Figure 5 ,The trend prediction module includes a trend analysis submodule, a probability calculation submodule, and a verification submodule;

[0097] The trend analysis submodule extracts the frequency and pattern of abnormal events based on abnormal status analysis logs, identifies abnormal development trends, and predicts potential abnormal patterns in the future to generate development trend forecast information;

[0098] Extract key information from abnormal status analysis logs, including the frequency and specific patterns of abnormal events. Through statistical analysis methods, determine the distribution and occurrence patterns of various types of abnormal events to identify potential abnormal development trends. In this way, it is possible to predict abnormal patterns that may occur in the future, so that preparations can be made in advance and risk management measures can be optimized and adjusted. The generated development trend forecast information provides data support for subsequent decision-making.

[0099] The probability calculation submodule conducts quantitative analysis on future abnormal events based on the development trend forecast information, calculates the potential for occurrence, and evaluates the risk level of different abnormal types to obtain a risk probability distribution map;

[0100] Based on the development trend forecast information, quantitative analysis of future abnormal events is carried out according to the formula: Calculate the risk level, where represents the probability of abnormal events. represents the occurrence rate of events within a time interval, Represents a time interval.

[0101] Considering the randomness and rarity of abnormal events, the exponential distribution is used to model the probability of abnormal occurrence. There is a system that records an average of 2 abnormalities per month, that is, , we want to calculate the probability of at least one anomaly occurring in the next month, according to the formula: The results show that the probability of at least one anomaly occurring within the next month is 86.5%. Based on this analysis, the risk level of different types of anomalies can be assessed separately and preventive measures can be adjusted accordingly.

[0102] The verification submodule verifies the prediction results in real time based on the risk probability distribution diagram, through historical comparative analysis, adjusts the prediction parameters and matches the new data, and generates prediction accuracy verification records;

[0103] According to the risk probability distribution diagram, the prediction results are verified by comparing historical data with real-time data, including extracting records of similar events from historical data, comparing them with current prediction results, and evaluating the prediction accuracy. If it is found that the prediction parameters deviate significantly from the actual data, necessary adjustments will be made, including retraining or parameter optimization, to improve the accuracy of the prediction. At the same time, it is also necessary to ensure that the new data can match the existing predictions to generate accuracy verification records.

[0104] See also Figure 2 and Figure 6 ,The dynamic adjustment module includes a sampling adjustment submodule, a monitoring range adjustment submodule, and a resource configuration submodule;

[0105] The sampling adjustment submodule verifies the records based on the prediction accuracy, evaluates the current effect of the data sampling process, adjusts the data sampling frequency according to the risk probability, optimizes the data collection time and quality, and obtains the optimal configuration of the sampling frequency;

[0106] The effectiveness of the data sampling process is evaluated by analyzing the prediction accuracy verification records, which involves complex data analysis and verification. The purpose is to determine the effectiveness and efficiency of the current sampling strategy. Based on the risk probability distribution, the data sampling frequency is adjusted, which usually includes increasing the sampling frequency in high-risk areas and reducing the frequency in low-risk areas, so as to optimize the time and quality of data collection and generate a sampling frequency optimization configuration. The configuration will directly affect the global strategy of data collection.

[0107] The monitoring scope adjustment submodule optimizes the configuration based on the sampling frequency, expands or reduces the monitoring scope, covers key risk points, optimizes the distribution and utilization of monitoring resources, and obtains the monitoring scope adjustment record;

[0108] Operations are performed based on the optimized configuration of the sampling frequency, and the deployment of monitoring resources is adjusted according to the risk assessment results, such as expanding the monitoring coverage of high-risk areas or reducing the coverage of low-risk areas. Through this adjustment, the distribution and utilization of monitoring resources are optimized, and the monitoring efficiency and responsiveness are further improved. The generated monitoring scope adjustment records provide resource allocation data, which provides a basis for future adjustments and ensures that monitoring activities can efficiently cover all key risk points.

[0109] The resource configuration submodule adjusts the records based on the monitoring scope, reconfigures the monitoring resources, including monitoring points and human resources, optimizes the monitoring response capability and efficiency, and generates the adjustment efficiency optimization records;

[0110] Based on the monitoring scope adjustment records, reconfigure the monitoring resources according to the formula: Calculate the regulation efficiency, where represents the efficiency of resource allocation, Represents the effect evaluation score of each monitoring point, represents the frequency weight of the monitoring point, is the total number of monitoring points.

[0111] If there are three monitoring points, the effect evaluation score is , and the corresponding frequency weight is , then the resource allocation efficiency is calculated as: It shows that the average efficiency of the adjusted resource allocation is 83.75, ensuring that resource allocation maximizes its effect and efficiency and optimizes monitoring response capabilities.

[0112] See also Figure 2 and Figure 7 ,The intelligent alarm module includes an alarm trigger condition submodule, a response process submodule, and an alarm level adjustment submodule;

[0113] The alarm trigger condition submodule optimizes records based on regulation efficiency, analyzes real-time data of the station, identifies key monitoring points, sets trigger thresholds for multi-level alarms, and automatically triggers logic based on differential risk levels to generate multi-level alarm condition settings;

[0114] Based on the regulation efficiency optimization record, set the trigger threshold of the multi-level alarm according to the formula: Calculate the alarm trigger condition, where represents the trigger threshold, Represents the average value of key monitoring point data, represents the standard deviation, Represents the coefficient corresponding to the risk level.

[0115] When setting alarm trigger thresholds, consider the variability and risk level of the data. For example, if the average data value of a monitoring point is , the standard deviation is , and considering the higher risk level, choose , the trigger threshold is calculated according to the formula: This means that when the monitoring data reaches or exceeds 130, an alarm will be automatically triggered to ensure that the alarm sensitivity matches the risk level of the site.

[0116] The response process submodule builds a matching automated response process based on multi-level alarm condition settings, including emergency response and routine monitoring response, minimizing the impact of potential risks, optimizing alarm response time and efficiency, and obtaining an optimized configuration of the response process;

[0117] Matching automated response processes, covering the full range of responses from emergency situations to routine monitoring. Implementation details include configuring corresponding response measures for different levels of alarms. For example, high-level alarms directly trigger on-site intervention or emergency shutdown, while low-level alarms only need to notify maintenance personnel for inspection, ensuring that response measures are both fast and appropriate, reducing unnecessary operational delays. The generated response process optimizes the configuration and records the preset responses for the alarm levels.

[0118] The alarm level adjustment submodule optimizes the configuration according to the response process, dynamically adjusts the alarm level settings, updates the response priority, matches the monitoring environment and security requirements, and obtains the alarm response configuration;

[0119] Update response priorities based on the actual monitoring environment and security requirements. Adjustments can more flexibly match the current monitoring status and potential risks. The optimized alarm response configuration includes setting the alarm level for each monitoring point to ensure that the alarm is neither over-responded nor under-reported. The resulting alarm response configuration provides a more accurate prevention and response strategy for risk events, greatly improving overall security management efficiency.

[0120] See also Figure 2 and Figure 8 ,The learning optimization module includes a parameter adjustment submodule, a network retraining submodule, and a performance optimization submodule;

[0121] The parameter adjustment submodule uses stochastic gradient descent based on the alarm response configuration to adjust the learning rate and weights of the long short-term memory network, optimize the adaptability and predictive ability to new trends, and generate network parameter tuning information;

[0122] Based on the alert response configuration, stochastic gradient descent is used to adjust the learning rate and weights according to the formula: Calculate the adjusted weights , where represents the weight parameter, represents the learning rate, Represents the loss function J on weight The gradient of is the current weight. The adjustment of learning rate and weight is achieved through gradient descent. The learning rate is 0.5. is 0.01, the gradient of the loss function is 0.1, according to the formula, the new weight is calculated as: This indicates that the weights have dropped slightly, reflecting the process of fine-tuning the model through the gradient descent algorithm, which aims to find the minimum value of the loss function and optimize the performance of the network.

[0123] The network retraining submodule performs LSTM network retraining based on the network parameter tuning information, adjusts the network structure according to the collected data samples, and obtains the network structure optimization result;

[0124] Using the network parameter tuning information, the LSTM network structure is retrained. By analyzing the collected data samples, the network's various layers and connections are adjusted to ensure that the network can learn patterns and trends in the data more effectively. In this process, attention is paid to the optimization of memory units and connection weights in the network to improve its adaptability to new trends and prediction accuracy. The network structure optimization results obtained will directly affect the performance of the LSTM network in practical applications, so as to more accurately predict future states or identify potential anomalies.

[0125] The performance optimization submodule evaluates the performance in station monitoring based on the network structure optimization results, analyzes the efficiency of anomaly detection and status prediction, and generates continuous optimization feedback results;

[0126] Based on the results of network structure optimization, the performance in site monitoring is evaluated, focusing on the efficiency of the LSTM network in anomaly detection and status prediction. Through evaluation, necessary adjustments are identified and implemented to further improve response speed and accuracy. The generated continuous optimization feedback results include analysis reports and improvement suggestions on the performance of the LSTM network. Feedback is a core part of the optimization process, ensuring that the LSTM network adapts to new data environments and monitoring needs over time.

[0127] See also Fig. 9 ,The smart station monitoring visualization method based on multi-source heterogeneous data fusion includes the following steps:

[0128] S1: Based on the video stream and sensor data monitored by the station, integrate the data stream for time series analysis, extract the key trends and periodic factors of the time series, and perform data cleaning and formatting based on the station operation data to generate a standard set of operation data;

[0129] S2: Based on the standard set of operation data, use the long short-term memory network to identify abnormal mutation points in the data, determine the cause and type of the abnormal state by comparing and analyzing the data before and after the mutation point, and generate an abnormal state analysis log;

[0130] S3: Based on the abnormal status analysis log, analyze the abnormal event trend, calculate the probability distribution of future abnormal events, verify the predicted status through historical data, and generate the prediction accuracy verification result;

[0131] S4: Based on the prediction accuracy verification results, adjust the data sampling frequency and monitoring range, refine the data collection point configuration, and use the adjusted monitoring parameters to optimize the station operation efficiency and generate regulation efficiency optimization records;

[0132] S5: Based on the regulation efficiency optimization record, the alarm level and response process are dynamically adjusted according to the real-time monitoring data, and the parameters of the long short-term memory network are adjusted by stochastic gradient descent, the network is retrained, and the accuracy of anomaly detection and state prediction is optimized to generate continuous optimization feedback results.

[0133] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. The intelligent station monitoring visualization system based on multi-source heterogeneous data fusion is characterized by: The system comprises: The data integration module integrates the data streams for time series analysis based on the video streams and sensor data monitored by the station, and cleans and formats the data according to the operation data of the station to obtain a standard set of operation data; The anomaly detection module, based on the standard set of operation data, determines the mutation points in the data through the long short-term memory network, analyzes the preliminary abnormal signals, identifies the cause and type of the abnormal state, and generates an abnormal state analysis log; The trend prediction module performs abnormal trend analysis based on the abnormal state analysis log, calculates the probability distribution of the future state, verifies the predicted state, determines the reliability of the prediction, and generates a prediction accuracy verification result; The dynamic adjustment module modifies the data sampling frequency and monitoring range based on the prediction accuracy verification result, refines the data collection point configuration, and reconfigures resources using the adjusted monitoring parameters to optimize the station operation efficiency and generate an adjustment efficiency optimization record; The intelligent alarm module constructs multi-level alarm trigger conditions based on the regulation efficiency optimization record, implements an automated response process, adjusts the alarm level and response process in combination with the current monitoring data of the station, and generates an alarm response configuration; The learning optimization module automatically adjusts the long short-term memory network parameters based on the alarm response configuration using stochastic gradient descent, performs network retraining, and optimizes the accuracy of anomaly detection and state prediction, generating continuous optimization feedback results; The data integration module includes a data acquisition submodule, a data cleaning submodule, and a data formatting submodule; The data acquisition submodule performs real-time data acquisition based on the video stream and sensor data monitored by the site, collects information about the environment and equipment operation status, fuses the video and sensor data, performs preliminary time tagging, and generates an initial data fusion set; The data cleaning submodule analyzes the data quality based on the initial data fusion set, identifies and removes noise and outliers in the data, and fills in missing values ​​of key data to obtain an optimized data set; The data formatting submodule sorts and time-standardizes the data based on the optimized data set through time series analysis, adjusts the data timestamp, refines the data structure to match the analysis requirements, and establishes a standard set of operation data; The anomaly detection module includes an anomaly identification submodule, a signal analysis submodule, and a log generation submodule; The abnormal identification submodule analyzes discontinuities and mutation points in the data stream through a long short-term memory network based on the standard set of operation data, determines abnormal signals in the data, classifies the abnormal signals, and generates a preliminary abnormal signal set; The signal analysis submodule analyzes the characteristics and influencing factors of each signal based on the preliminary abnormal signal set, determines the abnormal category and potential cause in combination with changes in environmental and operating conditions, and obtains details of the abnormal factors; The log generation submodule records the time, location, impact scope and potential risks of the abnormality based on the details of the abnormal factors, and organizes the recorded data to generate an abnormal status analysis log; The trend prediction module includes a trend analysis submodule, a probability calculation submodule, and a verification submodule; The trend analysis submodule extracts the occurrence frequency and pattern of abnormal events based on the abnormal state analysis log, identifies abnormal development trends, predicts future potential abnormal patterns, and generates development trend prediction information; The probability calculation submodule performs quantitative analysis on future abnormal events based on the development trend prediction information, calculates the potential for occurrence, and evaluates the risk level of different abnormal types to obtain a risk probability distribution map; The verification submodule verifies the prediction results in real time based on the risk probability distribution diagram, through historical comparative analysis, adjusts the prediction parameters and matches the new data, and generates a prediction accuracy verification record; The dynamic adjustment module includes a sampling adjustment submodule, a monitoring range adjustment submodule, and a resource configuration submodule; The sampling adjustment submodule evaluates the current effect of the data sampling process based on the prediction accuracy verification record, adjusts the data sampling frequency according to the risk probability, optimizes the data collection time and quality, and obtains the sampling frequency optimization configuration; The monitoring scope adjustment submodule is based on the sampling frequency optimization configuration, expands or reduces the monitoring scope, covers key risk points, optimizes the distribution and utilization of monitoring resources, and obtains a monitoring scope adjustment record; The resource configuration submodule reconfigures monitoring resources, including monitoring points and human resources, based on the monitoring scope adjustment record, and optimizes monitoring response capability and efficiency, and generates an adjustment efficiency optimization record; Based on the monitoring range adjustment record, the monitoring resources are reconfigured according to the formula: Calculate the regulation efficiency, where represents the efficiency of resource allocation, Represents the effect evaluation score of the monitoring point, represents the frequency weight of the monitoring point, is the total number of monitoring points; The intelligent alarm module includes an alarm trigger condition submodule, a response process submodule, and an alarm level adjustment submodule; The alarm trigger condition submodule analyzes the real-time data of the station based on the regulation efficiency optimization record, identifies key monitoring points, sets the trigger threshold of the multi-level alarm, and automatically triggers the logic according to the differential risk level to generate the multi-level alarm condition setting; The response process submodule builds a matching automated response process based on the multi-level alarm condition settings, including emergency response and routine monitoring response, minimizes the impact of potential risks, and optimizes alarm response time and efficiency to obtain an optimized configuration of the response process; The alarm level adjustment submodule optimizes the configuration according to the response process, dynamically adjusts the alarm level setting, updates the response priority, matches the monitoring environment and security requirements, and obtains the alarm response configuration; The learning optimization module includes a parameter adjustment submodule, a network retraining submodule, and a performance optimization submodule; The parameter adjustment submodule uses stochastic gradient descent based on the alarm response configuration to adjust the learning rate and weight of the long short-term memory network, optimize the adaptability and predictive ability to new trends, and generate network parameter tuning information; The network retraining submodule performs LSTM network retraining based on the network parameter tuning information, adjusts the network structure according to the collected data samples, and obtains the network structure optimization result; The performance optimization submodule evaluates the performance in site monitoring based on the network structure optimization results, analyzes the efficiency of anomaly detection and state prediction, and generates continuous optimization feedback results.

2. According to claim 1, the intelligent station monitoring visualization system based on multi-source heterogeneous data fusion is characterized in that: The abnormal status analysis log includes abnormal event identification, impact assessment results, and abnormal source analysis results; the prediction accuracy verification results include error rate calculation results, prediction confidence intervals, and calibration results; the adjustment efficiency optimization record includes resource reconfiguration evaluation information, efficiency improvement index, and operation delay reduction information; the alarm response configuration includes trigger mechanism, response priority sorting, and alarm coverage.

3. A smart station monitoring visualization method based on multi-source heterogeneous data fusion, characterized in that: The smart site monitoring visualization system based on multi-source heterogeneous data fusion according to any one of claims 1 to 2 comprises the following steps: Based on the video stream and sensor data monitored by the station, the data stream is integrated for time series analysis, key trends and periodic factors of the time series are extracted, and data is cleaned and formatted according to the station operation data to generate a standard set of operation data; Based on the standard set of operation data, using a long short-term memory network to identify abnormal mutation points in the data, determining the cause and type of the abnormal state by comparing and analyzing the data before and after the mutation point, and generating an abnormal state analysis log; Based on the abnormal state analysis log, analyze the abnormal event trend, calculate the probability distribution of future abnormal events, and verify the predicted state through historical data to generate a prediction accuracy verification result; Based on the prediction accuracy verification results, the data sampling frequency and monitoring range are adjusted, the data collection point configuration is refined, and the adjusted monitoring parameters are used to optimize the station operation efficiency and generate a regulation efficiency optimization record; Based on the regulation efficiency optimization record, the alarm level and response process are dynamically adjusted according to the real-time monitoring data, and the parameters of the long short-term memory network are adjusted by using stochastic gradient descent, the network is retrained, and the accuracy of anomaly detection and state prediction is optimized to generate continuous optimization feedback results.

Images