Intelligent monitoring method and system for air environment monitoring station

Abstract

The invention relates to an intelligent monitoring method and system for an air environment monitoring station, and belongs to the technical field of environment monitoring. The method comprises the following steps: acquiring multi-modal monitoring data, performing data association analysis on the multi-modal monitoring data by adopting a multi-source data association model, and establishing an association relationship among the multi-modal monitoring data; performing data exception monitoring on the multi-modal monitoring data, and triggering and starting an exception type diagnosis event when data exception is monitored; extracting a first data subset related to an exception type from the multi-modal monitoring data; extracting a second data subset associated with the first data subset from multi-modal monitoring data; and performing fault diagnosis based on the first data subset and the second data subset, and outputting an abnormality diagnosis result. According to the method, the multi-source data association model is utilized, the association between different types of data can be effectively analyzed, detailed fault analysis and diagnosis are carried out, the fault response time is shortened, and the operation and maintenance efficiency is improved.

Description

Technical Field

[0001] The present application belongs to the field of environmental monitoring technology, and specifically relates to an intelligent monitoring method and system for an air environment monitoring station. Background Art

[0002] Air pollution is a global public health issue. Pollutants such as particulate matter (PM2.5 and PM10), sulfur dioxide, and carbon monoxide have extremely serious impacts on human health. Air monitoring stations can provide key data, providing real-time data for urban planners and managers to help assess air quality and make scientific decisions.

[0003] At present, it is difficult to find problems in air environment monitoring stations at the first time. In terms of problem diagnosis, it usually relies on manual analysis of monitoring data, which is time-consuming and labor-intensive, and cannot detect and solve problems in a timely manner. The lack of intelligent diagnosis and analysis makes it impossible to accurately diagnose environmental problems, which is prone to misjudgment.

[0004] Therefore, how to overcome the defects of the existing air environment monitoring stations, such as difficult diagnosis, low efficiency and low accuracy, is a topic worthy of study for those skilled in the art. Summary of the invention

[0005] In view of the above analysis, the embodiments of the present invention aim to provide an intelligent monitoring method and system for an air environment monitoring station to solve the problems of difficult diagnosis, low efficiency and low accuracy of existing air environment monitoring stations.

[0006] In a first aspect of the present application, an intelligent monitoring method for an air environment monitoring station is provided, comprising: Acquire multimodal monitoring data, wherein the multimodal monitoring data includes real-time video monitoring data collected by video monitoring equipment, real-time gas concentration data collected by gas concentration monitoring equipment, real-time equipment operation status data collected by equipment operation status monitoring equipment, and real-time humidity data collected by humidity monitoring equipment; Using a multi-source data association model, performing data association analysis on the multi-modal monitoring data, and establishing association relationships between the multi-modal monitoring data; Performing data anomaly monitoring on the multimodal monitoring data, and triggering an abnormality type diagnostic event when data anomaly is detected; Extracting a first data subset related to the abnormal type from the multimodal monitoring data based on the data screening model; Based on the association relationship between the multimodal monitoring data, extracting a second data subset associated with the first data subset from the multimodal monitoring data, wherein the second data subset is data from a different source than the first data subset; Fault diagnosis is performed based on the first data subset and the second data subset, and an abnormality diagnosis result is output.

[0007] Optionally, it also includes: When data anomalies are detected, the first data subset and the second data subset are triggered to be associated and stored, and the associated data are stored in the evidence database.

[0008] Optionally, the evidence database is stored in a blockchain; After obtaining the first data subset and the second data subset, the data is logically verified, and the data that passes the verification is permanently recorded in the blockchain for storage, and each data point is added to the corresponding block of the blockchain as an immutable record.

[0009] Optionally, the adopting a multi-source data association model to perform data association analysis on the multi-modal monitoring data and establishing association relationships between the multi-modal monitoring data includes: An algorithm based on association rule mining is used to mine frequently occurring item sets and strong association rules from the multimodal monitoring data. Assume that the transaction database is , the item set is , the support threshold is , the confidence threshold is ; When the Apriori algorithm is used in association rule mining, all frequent 1-item sets are found. , generate candidate sets through connection and pruning operations , and calculate its support, filter out frequent - Itemset , repeat the process until no new frequent item sets can be generated; generate potential association rules from the frequent item sets, and determine the final strong association rules by calculating the confidence; When the FP-Growth algorithm is used for association rule mining, a frequent pattern tree is constructed and frequent patterns are mined from the frequent pattern tree; appropriate support and confidence thresholds are set to filter out strong association rules; Based on the mined strong association rules, the association relationship between multimodal monitoring data is established.

[0010] Optionally, before adopting the multi-source data association model to perform data association analysis on the multi-modal monitoring data and establishing the association relationship between the multi-modal monitoring data, the method further includes: Standardizing the data format of the multimodal monitoring data and integrating it into a multi-source database; Synchronizing the multimodal monitoring data using a timestamp to align the data in the same time point or time window; The multimodal monitoring data is spatially aligned according to physical location.

[0011] Optionally, extracting a first data subset related to the abnormal type from the multimodal monitoring data based on the data screening model includes: Pre-set different filtering conditions for different exception types: For real-time humidity data or real-time gas concentration data, filter the data according to whether it exceeds the set threshold and whether it meets the set change trend rules; when the data meets the filtering conditions, record the corresponding filtering data, which includes the timestamp and the location information of the monitoring device; For real-time video surveillance data, a deep learning target detection algorithm is used to extract feature vectors; the extracted feature vectors are compared with the preset interference behavior feature vectors, and the similarity between the two is calculated; if the similarity exceeds the set similarity threshold, the corresponding video frame is identified as having relevant interference behavior, and the relevant video clips are screened and extracted.

[0012] Optionally, the multimodal monitoring data further includes: equipment operation and maintenance data; Sorting the equipment operation and maintenance data in chronological order; Extract key information from the equipment operation and maintenance data, and associate and integrate the key information with real-time equipment operation status data and real-time video monitoring data; Use data clustering and classification algorithms to classify and archive the equipment operation and maintenance data, and group similar operation and maintenance events into one category; Establish data indexes and associated databases, store various types of sorted data in an associated manner according to multiple dimensions, and build a structured data system.

[0013] Optionally, the performing data anomaly monitoring on the multimodal monitoring data includes: For real-time humidity data, calculate the rate of change of real-time humidity data r =( H t - H t-1 ) / H t-1 , where H t The current humidity value, H t-1 It is the humidity value at the last moment; Using standard deviation s To determine the real-time humidity data in a time window [tT , t]; Based on the average of historical humidity data for the same period m and standard deviation s h, if the rate of change r Exceeds the normal change rate range and the current humidity value H t Beyond m - ks h , m + ks h ]; k If it is a coefficient set according to the actual situation, the humidity data is judged to be abnormal; For real-time equipment operation status data and real-time gas concentration data, the real-time equipment operation status data and real-time gas concentration data are input into the fault diagnosis model, and the abnormality type is output; the fault diagnosis model is based on big data analysis and machine learning technology, and is obtained by training large-scale training data including normal and abnormal conditions.

[0014] Optionally, the performing fault diagnosis based on the first data subset and the second data subset and outputting an abnormal diagnosis result includes: A basic probability distribution function is defined for each data source, wherein the basic probability distribution function describes the degree of support for different hypotheses provided by the evidence provided by each data source; Assign corresponding weights to each data source based on its reliability and relevance; The evidence from multiple data sources is integrated through evidence synthesis rules to obtain the final abnormality diagnosis result.

[0015] In a second aspect of the present application, an intelligent monitoring system for an air environment monitoring station is provided, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, an intelligent monitoring method for an air environment monitoring station according to any one of the above-described methods is implemented.

[0016] The intelligent monitoring method for air environment monitoring stations provided in this application enhances the capabilities of the monitoring system by integrating multi-source data (such as video, gas concentration, equipment operating status, humidity, etc.) from different monitoring devices. Using the multi-source data association model, the correlation between different types of data can be effectively analyzed, so as to detect and identify potential anomalies and faults at an early stage. Timely fault detection helps to avoid serious damage to equipment and possible environmental risks, reduce maintenance costs and extend the service life of equipment. When data anomalies are detected, this application can automatically trigger abnormal type diagnostic events, quickly extract relevant data subsets from multimodal data, and perform detailed fault analysis and diagnosis. This rapid response mechanism greatly shortens the time from fault discovery to fault response and improves operation and maintenance efficiency. Through association analysis and abnormal monitoring, not only can the current abnormal situation be detected, but also future trends and potential problems can be predicted through historical data analysis, which enhances the adaptability and flexibility of air environment monitoring stations.

[0017] In addition, the present application also provides an intelligent monitoring device for an air environment monitoring station having the above-mentioned technical effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this specification. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0019] Figure 1 Schematic diagram of the intelligent monitoring method for air environment monitoring stations provided in this application; Figure 2 A flowchart of a specific implementation method of the air environment monitoring station intelligent monitoring method provided by this application; Figure 3 This is a structural block diagram of the intelligent monitoring system for the air environment monitoring station provided in this application. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. It should be noted that, in the absence of conflict, the embodiments in the present disclosure and the features in the embodiments can be combined, separated, interchanged and / or rearranged with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0021] The terms used here are for the purpose of describing specific embodiments, and are not intended to be restrictive. As used here, unless the context clearly indicates otherwise, the singular forms "one (kind, person)" and "said (the)" are also intended to include plural forms. In addition, when the terms "comprise" and / or "include" and their variations are used in this specification, it is explained that there are stated features, integral bodies, steps, operations, parts, assemblies and / or their groups, but it is not excluded that there are or add one or more other features, integral bodies, steps, operations, parts, assemblies and / or their groups. It should also be noted that, as used here, the terms "substantially", "approximately" and other similar terms are used as approximate terms and not as degree terms, so that they are used to explain the inherent deviations of the measured values, calculated values ​​and / or the values ​​provided that will be recognized by those of ordinary skill in the art.

[0022] It is difficult to detect problems in current air environment monitoring stations at the first time. In terms of problem diagnosis, it is usually dependent on manual analysis of monitoring data, which is inefficient and prone to misjudgment. Based on this, the present application provides an intelligent monitoring method for air environment monitoring stations, which has a highly integrated intelligent monitoring system. Using sensors and monitoring equipment, real-time monitoring of the environment inside and outside the station can be achieved.

[0023] like Figure 1 As shown in the schematic diagram of the intelligent monitoring method for air environment monitoring station provided in this application, the method uses sensors and monitoring equipment to realize real-time monitoring of the environment inside and outside the station house, and uses the diagnostic module to perform real-time intelligent diagnosis and analysis on the real-time monitoring data, including whether it exceeds the threshold, whether it is constant, whether it fluctuates suddenly, whether it is specially marked, whether it is in an abnormal state, etc. When a problem is found, evidence is collected by calling various relevant historical records, and the matching of indoor and outdoor video surveillance images with pre-established data models (such as environmental interference library models, equipment operation behavior library models, etc.), and finally the collected evidence and abnormal diagnosis results are automatically generated according to the template to generate a problem diagnosis report and make a summary report.

[0024] The sensing device may specifically include: Humidity sensing devices are distributed in multiple locations in the air environment monitoring station to monitor the ambient humidity of the air environment monitoring station. For example, multiple humidity sensing devices arranged outside the station building, together with humidity monitoring devices within 5-500 meters around the station building, form a humidity monitoring network, which can obtain real-time humidity changes in different areas around the station building 24 hours a day. These sensing devices are able to withstand severe weather conditions and provide accurate humidity readings.

[0025] Image acquisition devices are distributed at multiple locations outside the air environment monitoring station. The image acquisition devices can be high-resolution cameras and / or infrared cameras that operate around the clock to capture environmental changes or human activities that may be related to humidity anomalies.

[0026] Gas concentration monitoring equipment is distributed at multiple locations in the air environment monitoring station room to detect gas concentration data.

[0027] Equipment operation status monitoring equipment is distributed in multiple locations in the air environment monitoring station room to monitor the equipment in the monitoring station.

[0028] Through the Internet of Things technology, humidity monitoring, station building surrounding environment monitoring and various monitoring equipment inside the station building (such as gas concentration monitoring equipment, equipment operation status monitoring equipment, etc.) are organically connected.

[0029] In addition, servers and storage devices are also set up. The storage devices have sufficient data storage capacity to save the large amount of data collected from sensors and cameras. The servers have high performance to support real-time processing and analysis of data, supporting functions such as time series analysis, image recognition, and machine learning algorithms.

[0030] You can also set up communication equipment and establish a stable wireless network system to ensure seamless data transmission from sensors and cameras to the server. You can also set up network security equipment: deploy necessary network security measures to protect information security and privacy during data transmission.

[0031] Through the Internet of Things monitoring network, high-precision sensors are used to collect the external humidity of the station, the operating status of the monitoring equipment inside the station, and the concentration of various monitoring data. The data is transmitted to the core monitoring center in real time using advanced transmission technology to ensure the timeliness and accuracy of the data. For example, the data collected by the sensor equipment is transmitted using Modbus RTU to ensure the integrity and real-time nature of the data. The video data collected by the image acquisition device is obtained through high-definition network cameras and Ethernet transmission to ensure image quality and transmission stability.

[0032] Intelligent diagnostic algorithms are used to track and analyze real-time data and equipment status in real time. When there are data abnormalities such as humidity outside the station exceeding the threshold, abnormal equipment operation status, or abnormal monitoring data concentration (such as exceeding the threshold, constant value, sudden fluctuation, special mark, etc.), abnormal diagnosis events and evidence collection events are automatically triggered. Among them, the diagnostic algorithm is based on big data analysis and machine learning technology. By learning a large amount of historical data, it has established a variety of abnormal situation recognition models, which can accurately determine the occurrence of problems.

[0033] A flowchart of a specific implementation of the intelligent monitoring method for an air environment monitoring station provided in this application is as follows Figure 2 As shown, the method specifically includes: S201: Acquire multimodal monitoring data, wherein the multimodal monitoring data includes real-time video monitoring data collected by video monitoring equipment, real-time gas concentration data collected by gas concentration monitoring equipment, real-time equipment operation status data collected by equipment operation status monitoring equipment, and real-time humidity data collected by humidity monitoring equipment.

[0034] Install and configure various types of monitoring equipment at the monitoring site, such as video monitoring equipment, gas concentration monitoring equipment, equipment operation status monitoring equipment, and humidity monitoring equipment. All equipment is configured to transmit data to the intelligent monitoring system in real time. This includes video streams, gas concentration readings, equipment operation parameters, and humidity data.

[0035] S202: using a multi-source data association model to perform data association analysis on the multi-modal monitoring data, and establish association relationships between the multi-modal monitoring data.

[0036] This step builds a comprehensive and accurate multi-source data association model based on big data analysis technology, and deeply explores the intrinsic connection between different data sources based on spatiotemporal characteristics. In terms of data sources, it covers various types of sensor equipment (multimodal monitoring data provided by high-precision humidity sensors, gas concentration sensors, equipment operation status sensors, etc.) and real-time video monitoring data. In addition, it can further include equipment operation logs (recording equipment operation and operation status history), operation and maintenance records (reflecting equipment maintenance activity details) and operation records (recording personnel's operation behavior on equipment) and other data, which are not limited here. Real-time data collection and transmission are realized through Internet of Things technology, and standardized data interfaces are used to ensure data integrity and consistency.

[0037] For monitoring data from different sources, before performing data association analysis on the multimodal monitoring data and establishing the association relationship between the multimodal monitoring data, it also includes: using key attributes of the data such as timestamp, spatial location and data type to align the data, which is important for subsequent data association analysis.

[0038] The multimodal monitoring data is standardized in data format and integrated into a multi-source database.

[0039] Use timestamps to synchronize the multimodal monitoring data so that the data at the same time point or time window are aligned. Based on timestamps, a strict time synchronization mechanism is established to associate data from different data sources at the same time point or in a specific time window, and to align the gas concentration monitoring data at a certain moment. The equipment operation status data at the same time And monitoring video frames Matching is performed to reveal the synchronization and causal relationships of events in the time dimension.

[0040] The multimodal monitoring data is spatially aligned according to the physical location. The spatial location information is used to associate the data of the surrounding related sensor devices. Through the geographic information system (GIS) technology or the pre-set spatial topological relationship, the relative position relationship between different sensor devices and monitoring devices and video surveillance devices is determined, so as to integrate and analyze the spatially related data. For example, analyzing the humidity sensor data near a specific monitoring device (in is the spatial coordinate) and the monitoring data of the device The relationship between the two, as well as the image information captured by the video surveillance equipment in the area The connection between.

[0041] An algorithm based on association rule mining is used to mine frequently occurring item sets and strong association rules from the multimodal monitoring data. Assume that the transaction database is , the item set is , the support threshold is , the confidence threshold is Specifically, algorithms based on association rule mining, such as the Apriori algorithm or the FP-Growth algorithm, can be used to mine frequently occurring data patterns and association relationships from a large amount of multi-source data. Appropriate support and confidence thresholds can be set to screen out meaningful association rules for data association analysis and the construction of evidence chains.

[0042] Specifically, when the association rule mining algorithm uses the Apriori algorithm, all frequent 1-item sets are found. , generate candidate sets through connection and pruning operations , and calculate its support, filter out frequent - Itemset , repeat the process until no new frequent item sets can be generated; generate potential association rules from the frequent item sets, and determine the final strong association rules by calculating the confidence.

[0043] Specifically, when the association rule mining algorithm adopts the FP-Growth algorithm, a frequent pattern tree is constructed, and frequent patterns are mined from the frequent pattern tree; appropriate support and confidence thresholds are set to screen out strong association rules.

[0044] Based on the mined strong association rules, the association relationship between multimodal monitoring data is established.

[0045] For example, when it is found that the frequency of abnormal changes in real-time humidity data collected by humidity sensing equipment and equipment operation and maintenance operation records within a specific time period exceeds the set support threshold, and there is a clear causal relationship between the two (confidence is higher than the threshold), it can be determined as a valuable association rule for constructing an evidence chain for data association analysis. This evidence chain can serve as an evidence clue to provide accurate and comprehensive evidence for problem diagnosis, thereby achieving accurate monitoring and analysis of the station environment and equipment operating conditions.

[0046] S203: Perform data anomaly monitoring on the multimodal monitoring data, and when data anomaly is detected, trigger and start an abnormal type diagnosis event.

[0047] The data anomaly detection model is built to monitor and identify abnormal patterns in data in real time. Once an anomaly is detected, an abnormal type diagnosis event is automatically triggered.

[0048] For real-time humidity data, a multi-factor comprehensive diagnosis algorithm based on time series analysis is used. First, the real-time humidity data is preprocessed to remove noise and outliers (using methods such as median filtering). Then, the rate of change of the real-time humidity data is calculated. r =( H t - H t-1 ) / H t-1 , where H t The current humidity value, H t-1 It is the humidity value at the previous moment.

[0049] At the same time, the standard deviation s To determine the real-time humidity data in a time window [tT , t] within the period. Based on the average of historical humidity data for the same period m and standard deviation s h (Historical humidity data for the same period is also calculated using the same time window). If the rate of change r The current humidity value exceeds the normal change rate range (determined based on historical data statistics) and H t Beyond m - k s h , m + ks h ]; k If it is a coefficient set according to the actual situation, it is judged that the humidity data is abnormal.

[0050] For real-time equipment operation status data and real-time gas concentration data, a classification algorithm based on deep learning is used. A fault diagnosis model consisting of a multi-layer neural network is constructed, and the input includes real-time equipment operation status data (such as equipment power consumption, sensor response time, etc.), real-time gas concentration data and its time series characteristics, and the output is the abnormal type. The fault diagnosis model is based on big data analysis and machine learning technology. By training large-scale training data including normal and abnormal conditions, the model learns the characteristic patterns of different abnormal conditions. During real-time diagnosis, real-time data is input into the model, and the model outputs the corresponding abnormal type (such as equipment fault type, concentration data abnormal type, etc.).

[0051] S204: Extracting a first data subset related to the abnormal type from the multimodal monitoring data based on the data screening model.

[0052] The data screening model is used to extract the first data subset directly related to the detected abnormal type from the total data set. In this step, the data screening model sets a series of rule-based data screening algorithms to screen out the first data subset related to the abnormal type from the original data according to the set rules and conditions.

[0053] For humidity anomalies, consider the humidity threshold range (set the upper limit and lower threshold ) and humidity trend rules, at specific time intervals Humidity variation range Exceeding the preset value (by calculating the humidity change rate to quantify), in order to filter out abnormal humidity data and its corresponding timestamp and location information of humidity monitoring equipment.

[0054] For video surveillance data, we screen them based on the output of the image recognition model and the preset interference behavior feature rules. These feature rules cover multiple dimensions such as object shape, color, and motion mode. Through image processing algorithms and target detection algorithms, we extract and compare the features of objects in the video frames, thereby screening out video clips containing specific interference behaviors (artificial spray water flow shape, fog cannon vehicle outline and spray form, etc.).

[0055] The process specifically includes: Set different screening conditions for different exception types in advance.

[0056] For real-time humidity data or real-time gas concentration data, filtering is performed based on whether the data exceeds a set threshold and whether it complies with a set change trend rule; when the data meets the filtering conditions, the corresponding filtering data is recorded, and the filtering data includes a timestamp and monitoring device location information.

[0057] For real-time video surveillance data, a deep learning target detection algorithm is used to extract feature vectors; the extracted feature vectors are compared with the preset interference behavior feature vectors, and the similarity between the two is calculated; if the similarity exceeds the set similarity threshold, the corresponding video frame is identified as having relevant interference behavior, and the relevant video clips are screened and extracted.

[0058] This process can use a screening algorithm based on conditional judgment and filters to define a series of screening conditions and regular expressions to judge and screen the original data line by line or frame by frame. For example, for humidity data, use an if-then statement to determine whether the humidity value exceeds the set range ( ) and whether the changing trend conforms to the rules ( ), extract the qualified data. For video surveillance data, use image processing algorithms and target detection algorithms to filter video frames according to preset interference behavior features (such as object shape, color, motion mode, etc.) and extract video clips with relevant interference behaviors. Let the object feature vector be , combined with the preset interference behavior feature vector , calculate the similarity between the two ( ), determine whether the video frame contains relevant interference behavior (similarity is greater than the set threshold ), and screen out the video frames that meet the conditions to form video clips containing specific interference behaviors, providing a focused and effective data basis for subsequent analysis.

[0059] S205: extracting a second data subset associated with the first data subset from the multimodal monitoring data based on the association relationship between the multimodal monitoring data, wherein the second data subset is data from a different source than the first data subset.

[0060] Use the previously established data association model to analyze what relevant data exists in the context of the first data subset. For example, if the first data subset includes a high temperature alarm for a device, the second data subset may include video surveillance data and environmental humidity data that match that time point. Based on the association rules and time synchronization data, extract data that is temporally or logically related to the first data subset from different data sources as the second data subset.

[0061] S206: Perform fault diagnosis based on the first data subset and the second data subset, and output abnormality diagnosis results.

[0062] Specifically, a deep learning model can be used to extract key features from the first data subset and the second data subset, and the data is annotated as training data. The training data is input into the multimodal fusion large model for model training to obtain an abnormal diagnosis model. The first data subset and the second data subset obtained in real time are input into the abnormal diagnosis model, and the first data subset and the second data subset are combined for comprehensive analysis to obtain and output the abnormal diagnosis result.

[0063] In the embodiment of the present application, the diagnostic rules for performing data anomaly monitoring on multimodal monitoring data may specifically include but are not limited to the following: Exceeding the threshold: If the threshold exceeds the fixed range, it is diagnosed as abnormal. The threshold includes the upper threshold and the lower threshold. For example, for a certain type of gas concentration monitoring, the upper threshold is set to X ppm, the lower threshold is Y ppm, when the real-time monitoring data is greater than X ppm or less Y ppm, it is judged as exceeding the threshold value abnormality.

[0064] Constant value: When the monitoring value remains constant for more than a preset number of sampling cycles (for example, 2 to 3), it is diagnosed as abnormal and no specific parameters need to be configured. T The values ​​monitored at time are Z , and the fluctuation is within a very small range (such as ± ∊ , which is a very small allowable error value), it is judged as a constant value abnormality.

[0065] Sudden fluctuation: when the difference of the monitored value changes suddenly and exceeds a certain range, set a specific reference line, such as the hourly average, daily average, previous real-time value, and the change range (percentage value). For example, if the reference line is the previous real-time value and the change range is 5%, then when the new real-time value is greater than or less than ±5% of the previous real-time value, it is diagnosed as abnormal. The specific calculation method can be: Set the previous real-time value to V prev , the new real-time value is V new , if | V new - V prev | / V prev >0.05, it is judged as a sudden fluctuation abnormality.

[0066] Special marking: when the monitoring device is in an abnormal state, the monitoring data is marked as abnormal data, diagnosed as abnormal, and the specific judgment is set as abnormal state. For example, when the device has a fault code or sensor calibration is abnormal, the monitoring data is specially marked as abnormal.

[0067] It can be seen that the present application adopts a multi-factor comprehensive diagnosis method, which comprehensively considers multiple factors such as the numerical changes, change trends, change frequencies, and historical data patterns of the monitoring data. For example, when judging that the humidity outside the station building is abnormal, in addition to comparing the real-time humidity data with the set threshold, the rate of change and fluctuation of humidity over a period of time are also analyzed. If the rate of change of humidity exceeds the normal range in a short period of time, it will be judged as abnormal even if the current humidity value does not exceed the threshold. At the same time, combined with the pattern analysis of the historical humidity data for the same period, the interference of normal factors such as seasonality is eliminated, which greatly improves the accuracy of diagnosis.

[0068] The embodiment of the present application can also automatically adjust the threshold of the monitoring indicator based on the statistical analysis of historical data and real-time environmental factors (such as season, weather, surrounding environment changes, etc.).

[0069] The specific determination method is: Collect historical environmental data .

[0070] The normal distribution model is used to analyze the distribution law of data and determine the threshold values ​​of monitoring indicators under different seasons, different time periods and different environmental conditions. Suppose the normal distribution probability density function is ,in, is the mean, is the standard deviation, and the value of the monitoring indicator threshold is determined by the mean and standard deviation.

[0071] Obtain the current environmental multivariate data, and determine the change value of the monitoring indicator threshold based on the current environmental multivariate data and historical environmental data .

[0072] use Adjust the value of the monitoring indicator threshold. is the adjusted monitoring indicator threshold.

[0073] Taking the hot and humid summer season as an example, by analyzing the distribution of historical summer humidity data, the upper threshold limit of humidity anomaly is appropriately increased. ( Based on historical data distribution and real-time environmental factors), it can avoid false alarms caused by normal seasonal humidity increases, effectively improving the accuracy and effectiveness of diagnosis. For stations close to pollution sources, the threshold of gas concentration monitoring can be dynamically adjusted according to the type and emission pattern of the pollution source to achieve more accurate diagnosis.

[0074] This application can perform real-time correlation analysis on monitoring data. The real-time correlation analysis between image recognition results and other monitoring data (such as real-time humidity data, real-time gas concentration data, and real-time equipment operation status data) is further elaborated below. As a specific implementation, when real-time video monitoring data identifies the presence of interference behavior, the information is immediately correlated and compared with other monitoring data in the corresponding time period. For example, while identifying the operation of a fog cannon vehicle, combined with the sharp rise in humidity monitoring data and the short-term fluctuation in the concentration of a certain gas, the degree and scope of the impact of the fog cannon vehicle operation on the environmental monitoring data can be accurately assessed, providing more comprehensive evidence support for problem diagnosis. A timestamp-based correlation algorithm is used to match the timestamp of the real-time video monitoring data with the timestamp of the monitoring data, and the data of the corresponding time period is found for correlation analysis. At the same time, the spatial location information (such as the relative position of the video monitoring equipment) is used to further calibrate and fuse the data to improve the accuracy of the correlation analysis.

[0075] In addition, the spatiotemporal information (timestamp and geographic location information) of the image can be used for comprehensive analysis with other monitoring data. The spatial relationship between the location where the interference behavior occurs and the video monitoring device can be accurately determined by combining the location and shooting angle information of the video monitoring device with the layout of the sensors inside and outside the station. At the same time, the temporal influence of the interference behavior on the monitoring data can be analyzed based on the timestamp of the image and the time series of the monitoring data. For example, by analyzing the changes in the images of multiple video monitoring devices and the data of the corresponding position sensors within a certain period of time, the movement path of the interference source and the scope of its impact on the surrounding environment can be accurately tracked. Using geographic information system (GIS) technology and spatiotemporal data analysis algorithms, the geographic location information of the video monitoring device and the sensor layout information are integrated into the GIS map to realize the visualization and analysis of spatial data. Through the modeling and analysis of time series data, the autoregressive moving average model (ARMA) is used to predict the future trend of interference behavior on monitoring data, providing a basis for taking countermeasures in advance.

[0076] In addition, the prior art also lacks effective evidence collection methods, making it difficult to determine the real cause of the problem, and unable to automatically collect evidence of the cause of the problem. The evidence collection method is cumbersome and easy to miss key information. In response to this problem, the present application also includes: when data anomalies are detected, triggering the associated storage of the first data subset and the second data subset, and storing the associated data in the evidence database.

[0077] In this way, once data anomalies are found, this application can automatically retrieve the data associated with the data anomaly problem and store it as evidence for subsequent analysis. For example, when judging whether the monitored concentration data is abnormal, the application can collect evidence comprehensively by combining the video image to see whether there are people operating the equipment or whether there are interference behaviors such as manual spraying and fog cannon operations in the surrounding area, as well as querying the equipment's operation history, operation and maintenance history, and operation history.

[0078] Using the association analysis algorithm based on big data analysis, the association relationship between multi-source data is automatically established according to the data's timestamp, spatial location, data type and other attributes. When abnormal concentration data is monitored, the equipment operation status data and real-time video monitoring data at the same time point are associated through the timestamp, and the data of surrounding related sensor equipment are associated through spatial location information. Through data type analysis and matching with historical operation and maintenance records and operation records, the key factors that may cause abnormal concentration are quickly screened out, providing accurate evidence clues for problem diagnosis.

[0079] The evidence database can be stored in the blockchain. After obtaining the first data subset and the second data subset, the data is logically verified, and the verified data is permanently recorded in the blockchain for storage, and each data point is added to the corresponding block of the blockchain as an immutable record.

[0080] The core feature of blockchain is that data cannot be changed once it is written. Once each block is added to the blockchain, the information in it cannot be modified or deleted. This immutability ensures the integrity of the evidence stored on the blockchain and greatly reduces the risk of data tampering. Blockchain also provides a clear tracking path for auditing and compliance. Any relevant party can review historical transactions and data changes, enhancing the transparency and trust of the system.

[0081] Based on any of the foregoing embodiments, performing fault diagnosis based on the first data subset and the second data subset and outputting an abnormality diagnosis result includes: A basic probability assignment function is defined for each data source, which describes the degree to which the evidence provided by each data source supports different hypotheses.

[0082] Assign weights to each data source based on its reliability and relevance. The reliability and relevance of different data sources may vary, and weights are assigned to reflect this difference. Different weights are assigned to data sources based on their historical performance, accuracy, or sensitivity to a particular situation.

[0083] The evidence from multiple data sources is integrated through evidence synthesis rules to obtain the final abnormality diagnosis result.

[0084] Use combination rules to fuse the weighted basic probability allocation functions to obtain a comprehensive basic probability allocation function. Based on the synthesized basic probability allocation function, output the final abnormal diagnosis result. Determine the fault hypothesis with the highest synthetic probability value as the diagnosis result, and provide a probability distribution to show the possibility of all considered fault types. By integrating information from multiple data sources, the possible misjudgment caused by relying on a single data source can be reduced. This method not only provides decision results, but also gives a quantitative evaluation of various possibilities, increasing the transparency and trust of decision-making.

[0085] In addition, in order to make the filtered data easier to analyze and understand, the present application also includes: a process of collating and formatting the filtered data using an intelligent data sorting algorithm.

[0086] Equipment operation logs and operation and maintenance records are sorted in chronological order to ensure the time sequence of data. Then, information extraction technology is used to extract key information (such as operation time, operator, maintenance content, etc.) from the original data and associate it with the corresponding monitoring data and video surveillance data. Fault alarm events during equipment operation are associated with equipment operating parameters (such as temperature) at the same time point. ,pressure , Current The causal relationship between the data and the development process of events are clearly visible.

[0087] For surveillance video data, video summarization and key frame extraction are performed to generate video thumbnails and key event markers to facilitate quick browsing and finding important information. The shot switching points in the video are identified through the shot boundary detection algorithm. Suppose the video frame sequence is , calculate the difference between adjacent frames (Methods such as structural similarity measurement can be used). When the difference exceeds the set threshold, it is determined as a shot switching point. Based on the content similarity calculation method, the most representative key frame in each shot is selected, and the key frame extraction function is set to , calculate content similarity based on features such as pixel information and color histogram within the frame. Add accurate time tags to key frames and a detailed description of the event , generate video thumbnails to facilitate quick browsing and locating important events.

[0088] For example, the key frames of the video during the time period when the equipment failure occurred are associated with the data on the changes in the equipment operating parameters to make the evidence more intuitive and clear. Use data clustering and classification algorithms to organize equipment operation logs and operation and maintenance records, and classify and archive similar operation and maintenance events. For video data, use key frame extraction algorithms, such as methods based on shot boundary detection and content similarity, to extract representative key frames and add time tags and event descriptions to them. At the same time, establish data indexes and associated databases, associate and store the sorted data with relevant monitoring data and other evidence, and facilitate subsequent queries and analysis.

[0089] Multimodal monitoring data can also include: equipment operation and maintenance data. For equipment operation and maintenance data, data clustering and classification algorithms can be used to classify and archive them, and similar operation and maintenance events can be grouped together to facilitate statistical analysis and pattern mining of data.

[0090] The specific implementation process can be: sorting the equipment operation and maintenance data in chronological order; extracting key information from the equipment operation and maintenance data, and associating and integrating the key information with real-time equipment operation status data and real-time video monitoring data; using data clustering and classification algorithms to classify and archive the equipment operation and maintenance data, and classify similar operation and maintenance events into one category; establishing data indexes and associated databases, and associating and storing the sorted data in multiple dimensions to build a structured data system. The structured data system is easy to query and analyze, providing an efficient data support environment for subsequent in-depth research and problem diagnosis.

[0091] The existing technology still has the problem of not being able to automatically generate problem diagnosis reports, the report generation method is not unified, and it is difficult to conduct effective data analysis and comparison. However, this application will summarize and display the real-time monitoring data, diagnostic analysis and evidence collection and solidification, and generate a detailed report. The report content includes: (1) Description of the problem phenomenon (such as the time, location, and indicators involved when the abnormal data occurred): The report shall provide a detailed and accurate description of the problem phenomenon, including the precise time (accurate to the second) and location (specific to the station location and monitoring equipment number) of the abnormal data, the monitoring indicators involved (such as the specific value of the real-time humidity data exceeding the threshold, the type and value change of the abnormal gas concentration, etc.), and the trend chart of the abnormal data (such as the curve of humidity change over time, the fluctuation of gas concentration in different time periods, etc.), so that regulators can intuitively and clearly understand the severity and development process of the problem.

[0092] (2) Diagnostic analysis process (algorithms used, sources of data for analysis, etc.): The system's diagnostic analysis process for the problem should be described in detail, including the specific diagnostic algorithm used (such as the weight allocation and analysis method of each factor in the multi-factor comprehensive diagnostic algorithm, the calculation process of the adaptive threshold adjustment, etc.), the source of the data for analysis (clearly listing the specific sensor device data retrieved, the monitoring video time period, the equipment operation log records, etc.) and the matching status with the data model library (explaining which models in the environmental interference library model and the equipment operation behavior library model match, and the degree of matching, etc.), so that regulators can understand how the system reaches a diagnostic conclusion.

[0093] (3) Historical data curves (humidity, gas concentration, etc.): The report contains a wealth of historical data curves. In addition to the change curves of conventional monitoring indicators such as humidity and gas concentration, it also includes historical curves of equipment operating status parameters (such as equipment power consumption, sensor response time, etc.). These curves mark abnormal points and key time nodes, and provide statistical analysis information of historical data (such as average value, standard deviation, maximum value, minimum value, etc.), which facilitates regulators to conduct in-depth analysis and research and understand the historical background and potential laws of the problem.

[0094] (4) Key video attachments and image recognition analysis results: Attach key surveillance video attachments and annotate and explain important events in the video. At the same time, provide detailed image recognition analysis results, including the identified interference behavior type, occurrence time, location, and impact analysis on monitoring data. For example, for the identified fog cannon operation, explain its operation time, the relative position relationship between the operation area and the station building, and the specific impact value and impact trend on the humidity and gas concentration monitoring data to enhance the credibility and persuasiveness of the report.

[0095] The report provides preliminary diagnostic analysis results and response recommendations (such as remediation suggestions).

[0096] According to the needs and usage scenarios of different users, this application pre-sets a variety of personalized report templates. For regulatory authorities, the report template focuses on the diagnostic results, scope of impact, risk assessment and response recommendations of the problem, and uses concise and clear charts and text descriptions to facilitate regulators to make decisions quickly. For example, the surrounding environment of the station building and the location of the interference source are displayed in the form of a map, and the changes in monitoring data in different time periods are compared with bar charts, and red warning signs are used to highlight serious abnormalities. For operation and maintenance personnel, the report template pays more attention to the detailed operating status, fault analysis, maintenance suggestions and operating guides of the equipment, providing the technical parameters of the equipment, interpretation of fault codes and detailed descriptions of maintenance steps, which is convenient for operation and maintenance personnel to carry out equipment maintenance and management.

[0097] At the same time, users can customize reports according to their own special needs through the system's template customization interface, flexibly selecting report content display order, chart style, font size and other parameters. Use template engine-based technologies, such as Velocity or Freemarker, to define different report templates. The template contains placeholders and variables for dynamically inserting specific data and analysis results. Users can select different template elements and parameter settings through a graphical interface, and the system generates corresponding customized report templates based on the user's selection.

[0098] The report supports multiple output formats, including PDF (suitable for printing and archiving), Excel (convenient for further data analysis and processing), Word (convenient for text editing and reading), and other common formats. Users can view and download online through the system's web interface, or set up automatic sending of reports to a specified mailbox or integrate with other management systems. For example, after integration with the company's asset management system, when a station equipment fails, the relevant diagnostic report is automatically sent to the asset management system, triggering the equipment maintenance process, and displaying relevant prompt information on the interface of the asset management system to achieve data sharing and interaction and improve work efficiency. Use the document generation library and email sending library to achieve multiple output methods for reports. For integration with other systems, a Web Service or message queue-based method is used for data interaction and report delivery.

[0099] The report is presented in an intuitive and clear manner, allowing regulators to quickly understand the full picture of the problem and make decisions.

[0100] This application will automatically generate a problem diagnosis report based on a template using the collected evidence and analysis results, which can provide decision support for regulators.

[0101] In addition, the present application also provides an intelligent monitoring system for air environment monitoring stations, such as Figure 3 As shown in the structural block diagram of the intelligent monitoring system for an air environment monitoring station, the system includes a memory 31 and a processor 32. The memory 31 stores a computer program. When the computer program is executed by the processor 32, the intelligent monitoring method for an air environment monitoring station described above is implemented.

[0102] As a specific implementation, the system may adopt a layered architecture of a data collection layer, a data processing layer, an intelligent diagnosis layer, and an application display layer. Data collection layer: It is configured to collect raw data and adopts a distributed collection method to ensure the comprehensiveness and accuracy of the data. For example, multiple sensor devices are arranged at different locations in the station building, and the data is transmitted to the aggregation node through the wireless sensor network Zigbee technology, and then uniformly transmitted to the data processing layer. Data processing layer: It is configured to clean, organize and pre-process the collected data, use data cleaning algorithms to remove outliers and noise, and use data normalization algorithms to unify data of different dimensions to the same scale to ensure data quality. Intelligent diagnosis layer: It is configured to use advanced algorithms and models for real-time analysis and diagnosis, such as fault diagnosis models based on deep learning. This layer is deployed on high-performance servers and uses parallel computing technology to improve diagnostic efficiency. Application display layer: It is configured to display system operation status and diagnostic results through a visual interface and report form. It uses HTML5, CSS3 and JavaScript to achieve data visualization, including charts, maps, etc., to provide a friendly user interaction experience.

[0103] The system can also be divided into humidity monitoring module, gas concentration monitoring module, and equipment operation status monitoring module according to function. By adopting modular design, each module can be operated and upgraded independently. Each module has its own interface and specification, which is convenient for system maintenance and expansion. For example, the humidity monitoring module can be flexibly configured according to the needs of different station rooms, increase or decrease the number of sensors, adjust the monitoring accuracy, etc. Modules communicate through message queues or middleware to ensure reliable data transmission and system stability.

[0104] The system also establishes a unified data interface and standard to achieve seamless integration and interaction of different types of data. Through data fusion algorithms, the relevance and complementarity of multi-source data are fully utilized for comprehensive analysis. At the same time, the system supports data interaction with other related systems, and can transmit report data to the superior supervision platform or enterprise management system, using a standardized data interface RESTful API to achieve data sharing and interaction.

[0105] Compared with the prior art, this application has the following technical advantages: (1) High degree of automation: This application realizes a fully automated process from data collection, diagnostic analysis, evidence collection to report generation, without the need for human intervention, greatly improving work efficiency. Compared with the traditional method that relies on manual analysis and processing, it saves a lot of manpower and time costs. For example, in previous station environment monitoring, manual analysis of monitoring data may take hours or even days to discover problems and make preliminary diagnoses, while this application can monitor in real time and complete diagnostic analysis and report generation within minutes.

[0106] (2) High diagnostic accuracy: Intelligent diagnostic technology, through multi-factor comprehensive analysis and adaptive threshold adjustment algorithms, can accurately determine the cause of the problem and avoid misjudgment. Compared with traditional single threshold judgment or empirical diagnosis methods, it greatly improves the accuracy of diagnosis and effectively reduces unnecessary resource waste and losses caused by misjudgment.

[0107] (3) Comprehensive evidence collection: It can automatically collect various types of relevant evidence, including data records, change curves, surveillance video clips, image recognition analysis results, etc., providing strong support for problem diagnosis. Compared with traditional evidence collection methods, it is more comprehensive, accurate and less likely to miss key information. Through the integration and correlation analysis of multi-source data, the system can provide a richer chain of evidence, which helps to gain a deeper understanding of the nature and process of the problem.

[0108] (4) Convenient report generation: It can automatically generate problem diagnosis reports with rich content and standardized format, providing regulators with a clear basis for decision-making. Compared with the traditional manual report writing method, it not only saves time but also improves the quality and consistency of reports. The visual charts and detailed analysis results in the report enable regulators to understand the problem more intuitively, so as to make accurate decisions more quickly.

[0109] The professionals should also be further aware that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0110] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0111] The specific implementation methods described above further illustrate the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only the specific implementation method of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. An intelligent monitoring method for an air environment monitoring station, characterized in that: include: Acquire multimodal monitoring data, wherein the multimodal monitoring data includes real-time video monitoring data collected by video monitoring equipment, real-time gas concentration data collected by gas concentration monitoring equipment, real-time equipment operation status data collected by equipment operation status monitoring equipment, and real-time humidity data collected by humidity monitoring equipment; Using a multi-source data association model, performing data association analysis on the multi-modal monitoring data, and establishing association relationships between the multi-modal monitoring data; Performing data anomaly monitoring on the multimodal monitoring data, and triggering an abnormality type diagnostic event when data anomaly is detected; Extracting a first data subset related to the abnormal type from the multimodal monitoring data based on the data screening model; Based on the association relationship between the multimodal monitoring data, extracting a second data subset associated with the first data subset from the multimodal monitoring data, wherein the second data subset is data from a different source than the first data subset; Fault diagnosis is performed based on the first data subset and the second data subset, and an abnormality diagnosis result is output.

2. The intelligent monitoring method of the air environment monitoring station according to claim 1, characterized in that: Also includes: When data anomalies are detected, the first data subset and the second data subset are triggered to be associated and stored, and the associated data are stored in the evidence database.

3. The intelligent monitoring method of the air environment monitoring station according to claim 2, characterized in that: The evidence database is stored in the blockchain; After obtaining the first data subset and the second data subset, the data is logically verified, and the data that passes the verification is permanently recorded in the blockchain for storage, and each data point is added to the corresponding block of the blockchain as an immutable record.

4. The intelligent monitoring method for an air environment monitoring station according to any one of claims 1 to 3, characterized in that: The adopting of a multi-source data association model to perform data association analysis on the multi-modal monitoring data and establishing association relationships between the multi-modal monitoring data includes: An algorithm based on association rule mining is used to mine frequently occurring item sets and strong association rules from the multimodal monitoring data. Assume that the transaction database is , the item set is , the support threshold is , the confidence threshold is ; When the Apriori algorithm is used in association rule mining, all frequent 1-item sets are found. , generate candidate sets through connection and pruning operations , and calculate its support, filter out frequent - Itemset , repeat the process until no new frequent item sets can be generated; generate potential association rules from the frequent item sets, and determine the final strong association rules by calculating the confidence; When the FP-Growth algorithm is used for association rule mining, a frequent pattern tree is constructed and frequent patterns are mined from the frequent pattern tree; appropriate support and confidence thresholds are set to filter out strong association rules; Based on the mined strong association rules, the association relationship between multimodal monitoring data is established.

5. The intelligent monitoring method of the air environment monitoring station according to claim 4, characterized in that: Before adopting the multi-source data association model to perform data association analysis on the multi-modal monitoring data and establishing the association relationship between the multi-modal monitoring data, the method further includes: Standardizing the data format of the multimodal monitoring data and integrating it into a multi-source database; Synchronizing the multimodal monitoring data using a timestamp to align the data in the same time point or time window; The multimodal monitoring data is spatially aligned according to physical location.

6. The intelligent monitoring method for an air environment monitoring station according to any one of claims 1 to 3, characterized in that: The step of extracting a first data subset related to the abnormal type from the multimodal monitoring data based on the data screening model includes: Pre-set different filtering conditions for different exception types: For real-time humidity data or real-time gas concentration data, filter the data according to whether it exceeds the set threshold and whether it meets the set change trend rules; when the data meets the filtering conditions, record the corresponding filtering data, which includes the timestamp and the location information of the monitoring device; For real-time video surveillance data, a deep learning target detection algorithm is used to extract feature vectors; the extracted feature vectors are compared with the preset interference behavior feature vectors, and the similarity between the two is calculated; if the similarity exceeds the set similarity threshold, the corresponding video frame is identified as having relevant interference behavior, and the relevant video clips are screened and extracted.

7. The intelligent monitoring method for an air environment monitoring station according to any one of claims 1 to 3, characterized in that: The multimodal monitoring data also includes: equipment operation and maintenance data; Sorting the equipment operation and maintenance data in chronological order; Extract key information from the equipment operation and maintenance data, and associate and integrate the key information with real-time equipment operation status data and real-time video monitoring data; Use data clustering and classification algorithms to classify and archive the equipment operation and maintenance data, and group similar operation and maintenance events into one category; Establish data indexes and associated databases, store various types of sorted data in an associated manner according to multiple dimensions, and build a structured data system.

8. The intelligent monitoring method of an air environment monitoring station according to claim 1, characterized in that: The performing data anomaly monitoring on the multimodal monitoring data includes: For real-time humidity data, calculate the rate of change of real-time humidity data r =( H t - H t-1 ) / H t-1 , where H t The current humidity value, H t-1 It is the humidity value at the last moment; Using standard deviation σ To determine the real-time humidity data in a time window [tT , t]; Based on the average of historical humidity data for the same period μ and standard deviation σ h , if the rate of change r Exceeds the normal change rate range and the current humidity value H t Beyond μ - kσ h , μ + kσ h ]; k If it is a coefficient set according to the actual situation, the humidity data is judged to be abnormal; For real-time equipment operation status data and real-time gas concentration data, the real-time equipment operation status data and real-time gas concentration data are input into the fault diagnosis model, and the abnormality type is output; the fault diagnosis model is based on big data analysis and machine learning technology, and is obtained by training large-scale training data including normal and abnormal conditions.

9. The intelligent monitoring method of air environment monitoring station according to claim 1, characterized in that: The performing fault diagnosis based on the first data subset and the second data subset and outputting an abnormal diagnosis result comprises: A basic probability distribution function is defined for each data source, wherein the basic probability distribution function describes the degree of support for different hypotheses provided by the evidence provided by each data source; Assign corresponding weights to each data source based on its reliability and relevance; The evidence from multiple data sources is integrated through evidence synthesis rules to obtain the final abnormality diagnosis result.

10. An intelligent monitoring system for an air environment monitoring station, characterized in that: It comprises a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the intelligent monitoring method for an air environment monitoring station according to any one of claims 1 to 9 is implemented.

Images