Environment online monitoring method and system for VOC emission

Abstract

The invention discloses an environment online monitoring method and system for VOC emission, and relates to the technical field of environment monitoring. The method comprises the following steps: interactively arranging monitor arrays at M point locations of a target mine exhaust pipeline, and collecting historical monitoring logs to obtain a historical monitoring log set of the M point locations; traversing the historical monitoring log set of the M point locations to identify high and low monitoring frequencies, and determining a high and low monitoring frequency group set of the M point locations; continuously monitoring the M point locations to obtain a high-frequency monitoring data sequence set of the M point locations and a low-frequency monitoring data sequence set of the M point locations; and performing high and low frequency interactive fusion analysis on the M point location high frequency monitoring data sequence sets and the M point location low frequency monitoring data sequence sets, and determining M point location monitoring result sets. The technical problems of low mine VOC emission monitoring efficiency and insufficient monitoring data accuracy in the prior art are solved, and the technical effect of improving the monitoring efficiency and the monitoring data accuracy is achieved.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring technology, and specifically to an online environmental monitoring method and system for VOC emissions. Background Technology

[0002] In the industrial sector, particularly during mining operations, the emission of volatile organic compounds (VOCs) is receiving increasing attention. As a significant component of air pollutants, VOCs not only adversely affect the ecological environment but also pose a potential threat to the health of workers. Current online environmental monitoring methods for VOC emissions in mines mostly employ fixed-frequency, fixed-location monitoring, which suffers from technical shortcomings such as unreasonable monitoring frequency settings, low utilization of monitoring resources, insufficient response sensitivity, and limited data accuracy. These methods fail to meet the actual needs for efficient and accurate VOC emission monitoring under the complex operating conditions of mines. Summary of the Invention

[0003] This application provides an online environmental monitoring method and system for VOC emissions, which solves the technical problems of low efficiency and insufficient accuracy of monitoring data in existing mine VOC emission monitoring.

[0004] A first aspect of this application provides a method for online environmental monitoring of VOC emissions, the method comprising: An array of monitoring instruments deployed at M locations along the ventilation duct of a target mine collects historical monitoring logs, obtaining a set of historical monitoring logs for the M locations, where M is a positive integer. Using a preset set of environmental monitoring parameters as an index, the set of historical monitoring logs for the M locations is traversed to identify high and low monitoring frequencies, determining a set of high and low monitoring frequency groups for the M locations. The monitoring array is then invoked to continuously monitor the M locations based on the set of high and low monitoring frequency groups, obtaining a set of high-frequency monitoring data sequences and a set of low-frequency monitoring data sequences for the M locations. A high-low frequency interactive fusion analysis is performed on the set of high-frequency and low-frequency monitoring data sequences for the M locations to determine a set of monitoring results for the M locations.

[0005] A second aspect of this application provides an online environmental monitoring system for VOC emissions, the system comprising: The data acquisition module is used to interactively collect historical monitoring logs from an array of monitoring instruments deployed at M locations along the ventilation duct of the target mine, obtaining a set of historical monitoring logs for the M locations, where M is a positive integer. The frequency identification module is used to identify high and low monitoring frequencies by traversing the set of historical monitoring logs for the M locations using a preset set of environmental monitoring parameters as an index, thus determining a set of high and low monitoring frequency groups for the M locations. The monitoring module is used to invoke the monitoring array to continuously monitor the M locations based on the set of high and low monitoring frequency groups, obtaining a set of high-frequency monitoring data sequences and a set of low-frequency monitoring data sequences for the M locations. The fusion analysis module is used to perform high-low frequency interactive fusion analysis on the set of high-frequency monitoring data sequences and the set of low-frequency monitoring data sequences for the M locations, determining a set of monitoring results for the M locations.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, an array of monitoring instruments deployed at M locations along the exhaust duct of the target mine collects historical monitoring logs, obtaining a set of historical monitoring logs for M locations, where M is a positive integer. Next, using a preset set of environmental monitoring parameters as an index, the historical monitoring log sets for the M locations are traversed to identify high and low monitoring frequencies, determining a set of high and low monitoring frequency groups for the M locations. Then, the monitoring array is invoked to continuously monitor the M locations based on the set of high and low monitoring frequency groups, obtaining a set of high-frequency monitoring data sequences and a set of low-frequency monitoring data sequences for the M locations. Finally, a high-low frequency interactive fusion analysis is performed on the set of high-frequency and low-frequency monitoring data sequences for the M locations to determine the set of monitoring results for the M locations. This solves the technical problems of low efficiency and insufficient accuracy in mine VOC emission monitoring in existing technologies, achieving the technical effect of improving monitoring efficiency and data accuracy. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a schematic diagram of the process for online environmental monitoring of VOC emissions provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of an online environmental monitoring system for VOC emissions provided in an embodiment of this application.

[0009] Figure labeling: Data acquisition module 11, frequency identification module 12, monitoring module 13, fusion analysis module 14. Detailed Implementation

[0010] This application provides an online environmental monitoring method and system for VOC emissions, which solves the technical problems of low efficiency and insufficient accuracy of VOC emission monitoring data in existing technologies.

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0012] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0013] Example 1, as Figure 1 As shown, this application provides a method for online environmental monitoring of VOC emissions, wherein the method includes: An array of monitoring instruments is deployed at M locations along the ventilation duct of the target mine to collect historical monitoring logs, resulting in a set of historical monitoring logs for the M locations, where M is a positive integer.

[0014] In this embodiment of the invention, a structural and airflow path analysis is performed on the ventilation system of the target mine to determine the key nodes of the ventilation duct and the potential VOC concentration fluctuation areas. Based on the analysis results, M representative monitoring points are selected, where M is a positive integer set according to the mine size, ventilation duct length, and number of branches.

[0015] An environmental monitoring array is formed by deploying VOC detection devices at M locations. Each device includes a sensing and acquisition unit (such as a PID photoionization detector), a local data storage module, a communication module (supporting wired or wireless communication), and a power supply module. The monitoring devices maintain data synchronization with the central control system via the Internet of Things (IoT), enabling data exchange and remote control between the locations.

[0016] After deployment, the monitoring array is activated to continuously sample environmental parameters at each location. The collected data includes, but is not limited to, VOC concentration, temperature, pressure, gas flow rate, humidity, timestamps, and equipment operating status. After a period of continuous operation (e.g., 7 to 30 days), each monitoring unit generates a corresponding historical monitoring log. This log is stored in a structured data format, such as CSV or JSON, and includes complete parameter records for each sampling time point. Finally, the historical monitoring logs from M locations are collected to form a unified historical monitoring log set for M locations, serving as the foundational data source for subsequent high- and low-frequency identification and data fusion analysis.

[0017] Using a preset set of environmental monitoring parameters as an index, the high and low monitoring frequencies of the M points are identified by traversing the historical monitoring log sets of the M points, and the high and low monitoring frequency groups of the M points are determined.

[0018] Furthermore, the preset set of environmental monitoring parameters includes temperature, flue pressure, flow rate, and humidity.

[0019] In this embodiment of the invention, the system sets a set of preset environmental monitoring parameters as an important index for analyzing the fluctuation pattern of VOCs. The preset environmental monitoring parameter set includes temperature, flue pressure, flow rate and humidity, and each parameter is an important environmental factor affecting the change of VOC concentration.

[0020] Specifically, the system extracts historical data of the aforementioned parameters for each monitoring point from the historical monitoring log set of the M points. By traversing and analyzing the historical monitoring logs of each point, anomaly detection and duration assessment are performed on the time series data corresponding to each parameter. Preferably, a sliding window method, threshold detection method, or statistical distribution method can be used to identify abnormal fluctuations in the parameter data (such as a sharp increase in temperature, a sudden change in flow rate, etc.), and the interval duration of parameter anomalies is recorded. For each point, a parameter anomaly interval duration cluster is generated based on the anomaly interval duration of different parameters. Subsequently, the system performs normalization processing and cluster analysis on this cluster, selecting parameter groups with lower anomaly fluctuation frequencies as the basis for low monitoring frequencies, and simultaneously selecting parameter groups with higher fluctuation frequencies as the basis for high monitoring frequencies. Based on the analysis results, the system divides each monitoring point into high-monitoring frequency points and low-monitoring frequency points, generating corresponding high- and low-monitoring frequency groups. Finally, a set of high- and low-monitoring frequency groups for M points is formed, which is used to guide the subsequent monitoring instrument array to dynamically adjust the data acquisition frequency of each point, thereby achieving optimized resource allocation and accurate perception of key areas.

[0021] Furthermore, using a preset set of environmental monitoring parameters as an index, the historical monitoring log sets of the M locations are traversed to identify high and low monitoring frequencies, thus determining a set of high and low monitoring frequency groups for the M locations, including: Using the preset set of environmental monitoring parameters as an index, the abnormal interval duration of parameters is extracted from the historical monitoring log sets of the M points to obtain clusters of abnormal interval durations of parameters at the M points; the clusters of abnormal interval durations of parameters at the M points are filtered to obtain sets of low monitoring frequencies at the M points and sets of high monitoring frequencies at the M points; the sets of low monitoring frequencies at the M points and sets of high monitoring frequencies at the M points are mapped and combined to obtain sets of high and low monitoring frequency groups at the M points.

[0022] For each monitoring point, time-series data of temperature, flue gas pressure, flow rate, and humidity are extracted from its corresponding historical monitoring logs. Outliers are marked using predefined parameter anomaly identification rules (e.g., setting fluctuation thresholds, mutation rates, or deviation rates based on historical averages). For each parameter, the time points of occurrence of abnormal events are statistically analyzed, and the time interval between two adjacent abnormal events is calculated to obtain the abnormal interval duration sequence for that parameter. The abnormal interval duration sets of all parameters at the same monitoring point are merged to form the parameter abnormal interval duration cluster for that point.

[0023] The abnormal interval duration clusters of parameters at each monitoring point are analyzed. Preferably, extreme value extraction, statistical quantile method, or K-means clustering algorithm are used to extract the maximum abnormal interval (representing sparse anomalies, suitable for low-frequency monitoring) and the minimum abnormal interval (representing dense anomalies, suitable for high-frequency monitoring) of each parameter. This allows the parameters at each point to be divided into a low-monitoring-frequency set (containing parameters with longer abnormal interval durations) and a high-monitoring-frequency set (containing parameters with shorter abnormal interval durations). The low-monitoring-frequency set and the high-monitoring-frequency set are mapped to form the high-low monitoring frequency group for that monitoring point. This process is repeated for M points, ultimately forming a set of M high-low monitoring frequency groups, providing a strategic basis for subsequent differentiated monitoring configuration and fusion analysis.

[0024] Furthermore, the abnormal interval duration clusters of the M location parameters are filtered to obtain a set of M locations with low monitoring frequencies and a set of M locations with high monitoring frequencies, including: From the cluster of abnormal interval durations of the parameters at the M locations, the maximum value of the abnormal interval duration of different environmental monitoring parameters is extracted as a set of low monitoring frequencies for the M locations; from the cluster of abnormal interval durations of the parameters at the M locations, the minimum value of the abnormal interval duration of different environmental monitoring parameters is extracted as a set of high monitoring frequencies for the M locations.

[0025] For each monitoring point, the system iterates through all environmental monitoring parameters (including temperature, flue pressure, flow rate, and humidity) within its parameter anomaly interval duration cluster. It extracts the maximum anomaly interval duration for each parameter throughout the entire monitoring period. This maximum value represents the sparsity of anomalies for that parameter, indicating that the parameter fluctuates little and changes slowly at that point. The system assigns the monitoring dimension identifiers corresponding to these extracted maximum value parameters to the low-monitoring-frequency set for that point. Simultaneously, for each parameter item within the parameter anomaly interval duration cluster for that point, the system further extracts its corresponding minimum anomaly interval duration. This minimum value represents the frequency of anomalies for that parameter, indicating that the parameter fluctuates significantly and changes drastically at that point. The system assigns the monitoring dimension identifiers corresponding to these extracted minimum value parameters to the high-monitoring-frequency set for that point.

[0026] The monitoring array is invoked to continuously monitor the M points according to the high and low monitoring frequency group set of the M points, thereby obtaining a high frequency monitoring data sequence set of the M points and a low frequency monitoring data sequence set of the M points.

[0027] Specifically, for each monitoring point, the system configures different sampling time intervals based on its corresponding high-frequency monitoring parameter set and low-frequency monitoring parameter set. For example, for parameters marked as high-frequency monitoring (such as frequent fluctuations in flow rate), the monitor is set to a shorter sampling interval (such as once every 1 minute); for parameters marked as low-frequency monitoring (such as slow changes in humidity), a longer sampling interval is set (such as once every 10 minutes or 30 minutes).

[0028] Data is collected by calling the monitoring array. The data collected by the high-frequency monitoring parameters are collected into M sets of high-frequency monitoring data sequences based on the location; the data collected by the low-frequency monitoring parameters are collected into M sets of low-frequency monitoring data sequences based on the location.

[0029] A high-low frequency interactive fusion analysis is performed on the set of high-frequency monitoring data sequences of the M locations and the set of low-frequency monitoring data sequences of the M locations to determine the set of monitoring results for the M locations.

[0030] After collecting high-frequency and low-frequency environmental monitoring data from M locations, a high-low frequency interactive fusion analysis is performed on the high-frequency monitoring data sequence set and the low-frequency monitoring data sequence set from the M locations, and the monitoring results set from the M locations is output.

[0031] Furthermore, a high-low frequency interactive fusion analysis is performed on the set of high-frequency monitoring data sequences from the M locations and the set of low-frequency monitoring data sequences from the M locations to determine the set of monitoring results for the M locations, including: A first feature extraction branch and a second feature extraction branch are pre-constructed. The first feature extraction branch is used to perform feature convolution analysis on the set of high-frequency monitoring data sequences at M locations to determine a set of high-frequency time-series vectors at M locations. The second feature extraction branch is used to perform feature convolution analysis on the set of low-frequency monitoring data sequences at M locations to determine a set of low-frequency spatial vectors at M locations. The set of high-frequency time-series vectors and the set of low-frequency spatial vectors at M locations are mapped and fused, and the attention weights of the second feature extraction branches are updated to obtain M updated second feature extraction branches. The M updated second feature extraction branches are used to perform feature convolution analysis on the set of low-frequency monitoring data sequences at M locations to determine a set of updated low-frequency spatial vectors at M locations. A fully connected network layer is used to analyze the set of updated low-frequency spatial vectors and the set of high-frequency time-series vectors at M locations to obtain a set of monitoring results at M locations.

[0032] Specifically, two functionally independent neural network branches are pre-constructed. The first feature extraction branch is suitable for processing high-frequency time-series data, employing a one-dimensional convolutional neural network (1D-CNN) or a temporal convolutional network (TCN) with multi-scale convolutional kernels to extract local temporal dynamic features from the data. The second feature extraction branch is suitable for modeling low-frequency spatial data, using a shallow convolutional network with a smaller receptive field to focus on extracting the overall trend, steady-state characteristics, and long-term behavioral features of the data. The set of high-frequency monitoring data sequences from M locations is input into the first feature extraction branch, and feature convolution operations are performed on each location sequentially, outputting a corresponding set of M high-frequency time-series vectors to describe the dynamic changes in VOC emissions at each location over a short timescale. The set of low-frequency monitoring data sequences from M locations is input into the second feature extraction branch, which extracts their global steady-state information and outputs a corresponding set of M low-frequency spatial vectors to describe the stable emission characteristics of each location over a longer period. A point-by-point mapping and interactive fusion process is performed on the high-frequency time-series vector set and the low-frequency spatial vector set of M locations. Specifically, the similarity of each pair of high-frequency and low-frequency vectors is calculated by embedding a mapping function. The similarity results are then input into a Softmax function for normalization to form attention distribution weights. These weights are used to redistribute the network channels in the second feature extraction branch, thus dynamically adjusting the second feature extraction branch. Finally, M updated second feature extraction branches are obtained. The low-frequency monitoring data sequence set of the M locations is then input back into the updated second feature extraction branch for re-feature extraction, outputting a fused and enhanced set of M updated low-frequency spatial vectors. This set more comprehensively reflects the fusion results of the macroscopic steady-state characteristics and microscopic dynamic anomalies of the locations. The M updated low-frequency spatial vector sets are concatenated or input in parallel with the corresponding high-frequency time-series vector sets into a fully connected neural network layer (Dense Layer) to perform nonlinear transformation, feature aggregation, and result regression or classification operations. Finally, a set of monitoring results for M locations is output, which may include VOC concentration predictions, environmental risk levels, and parameter anomaly identification labels.

[0033] Furthermore, the receptive field of the first feature extraction branch is larger than that of the second feature extraction branch.

[0034] The first feature extraction branch is used to process high-frequency monitoring data sequences. It employs a multi-layer convolutional or dilated convolutional structure with a large kernel size and a deep network layer to expand the receptive field, thereby capturing temporal variation features over a larger time span. For example, with a kernel size of 5, 4 layers, and a stride of 1, its effective receptive field can be extended to multiple time steps, suitable for modeling dynamic features such as rapid fluctuations and short-period anomalies. In contrast, the second feature extraction branch is primarily designed for processing low-frequency monitoring data sequences. It uses shallower convolutional layers and a smaller kernel size (e.g., 3), resulting in a narrower receptive field. Its aim is to extract steady-state features, trend information, or periodic gradual changes at points over a longer sampling period.

[0035] By setting the receptive field of the first feature extraction branch to be larger than that of the second feature extraction branch, it is helpful to model the differences between high-frequency and low-frequency data at the information expression level, ensuring that the two types of features have complementarity and synergy in the subsequent interactive fusion analysis, thereby improving the accuracy and discriminative power of the monitoring results.

[0036] Furthermore, the set of high-frequency temporal vectors of the M points and the set of low-frequency spatial vectors of the M points are mapped and fused, and the attention weights of the second feature extraction branches are updated respectively to obtain M updated second feature extraction branches, including: Extract a first high-frequency time-series vector set and a first low-frequency spatial vector set from the M high-frequency time-series vector sets and the M low-frequency spatial vector sets; obtain the attention weights of the second feature extraction branch; perform interactive fusion analysis on the first high-frequency time-series vector set and the first low-frequency spatial vector set using a mapping interaction fusion function to obtain a first updated attention weight; update the second feature extraction branch using the first updated attention weight to obtain a first updated second feature extraction branch; perform mapping interaction analysis on the M high-frequency time-series vector sets and the M low-frequency spatial vector sets to update the attention weights of the second feature extraction branch to obtain M updated second feature extraction branches.

[0037] From the set of M high-frequency temporal vectors and the set of M low-frequency spatial vectors, for each specific point, extract its corresponding first high-frequency temporal vector set and first low-frequency spatial vector set; obtain the attention weight of the second feature extraction branch in the current state, and call the preset mapping interaction fusion function to perform interactive fusion analysis on the first high-frequency temporal vector set and the first low-frequency spatial vector set to obtain the first updated attention weight; use the first updated attention weight to perform weighted update of the second feature extraction branch, that is, dynamically modify the channel selection, feature fusion method or convolution kernel activation intensity in the network structure to construct the updated first updated second feature extraction branch; repeat the above operation steps to perform mapping interaction analysis on each high-frequency temporal vector set and the low-frequency spatial vector set in turn, and finally obtain M updated second feature extraction branches.

[0038] Furthermore, the mapping interaction fusion function is: ;in, To update the attention weights first, For attention weights, Let i be the i-th low-frequency spatial vector of the first point in the set of low-frequency spatial vectors of the first point. This is the normalized value of the similarity between the low-frequency spatial vector of the i-th first point and the high-frequency temporal vector of the i-th first point. This is the transpose of the high-frequency timing vector at the first point. Let be the dimension of the i-th high-frequency time-series vector of the first point, and m be the number of low-frequency spatial vectors of the first point in the set of low-frequency spatial vectors of the first point.

[0039] The mapping interaction fusion function is implemented based on an attention mechanism. It calculates the similarity between low-frequency spatial vectors and high-frequency temporal vectors and combines this with the current attention weights. By dynamically adjusting the attention distribution, effective modeling of the interaction relationships between multi-frequency features can be achieved.

[0040] Based on the mapping interaction fusion function, the semantic correlation between observation data of different frequencies can be fully explored, making the updated attention weights more in line with the actual feature expression needs of the target monitoring points, and providing a more discriminative guidance mechanism for the subsequent feature extraction module.

[0041] In summary, the embodiments of this application have at least the following technical effects: First, an array of monitoring instruments deployed at M locations along the exhaust duct of the target mine collects historical monitoring logs, obtaining a set of historical monitoring logs for M locations, where M is a positive integer. Next, using a preset set of environmental monitoring parameters as an index, the historical monitoring log sets for the M locations are traversed to identify high and low monitoring frequencies, determining a set of high and low monitoring frequency groups for the M locations. Then, the monitoring array is invoked to continuously monitor the M locations based on the set of high and low monitoring frequency groups, obtaining a set of high-frequency monitoring data sequences and a set of low-frequency monitoring data sequences for the M locations. Finally, a high-low frequency interactive fusion analysis is performed on the set of high-frequency and low-frequency monitoring data sequences for the M locations to determine the set of monitoring results for the M locations. This solves the technical problems of low efficiency and insufficient accuracy in mine VOC emission monitoring in existing technologies, achieving the technical effect of improving monitoring efficiency and data accuracy.

[0042] Example 2, based on the same inventive concept as the online environmental monitoring method for VOC emissions in the foregoing examples, such as... Figure 2 As shown, this application provides an online environmental monitoring system for VOC emissions, wherein the system includes: The data acquisition module 11 is used to interactively collect historical monitoring logs from the monitoring array at M locations on the target mine's ventilation duct, obtaining a set of historical monitoring logs for the M locations, where M is a positive integer. The frequency identification module 12 is used to identify high and low monitoring frequencies by traversing the set of historical monitoring logs for the M locations using a preset set of environmental monitoring parameters as an index, and determining a set of high and low monitoring frequency groups for the M locations. The monitoring module 13 is used to call the monitoring array to continuously monitor the M locations based on the set of high and low monitoring frequency groups for the M locations, obtaining a set of high-frequency monitoring data sequences and a set of low-frequency monitoring data sequences for the M locations. The fusion analysis module 14 is used to perform high-low frequency interactive fusion analysis on the set of high-frequency monitoring data sequences and the set of low-frequency monitoring data sequences for the M locations, and determine a set of monitoring results for the M locations.

[0043] Furthermore, the frequency identification module 12 is used to perform the following method: The preset set of environmental monitoring parameters includes temperature, flue pressure, flow rate, and humidity.

[0044] Furthermore, the frequency identification module 12 is used to perform the following method: Using the preset set of environmental monitoring parameters as an index, the abnormal interval duration of parameters is extracted from the historical monitoring log sets of the M points to obtain clusters of abnormal interval durations of parameters at the M points; the clusters of abnormal interval durations of parameters at the M points are filtered to obtain sets of low monitoring frequencies at the M points and sets of high monitoring frequencies at the M points; the sets of low monitoring frequencies at the M points and sets of high monitoring frequencies at the M points are mapped and combined to obtain sets of high and low monitoring frequency groups at the M points.

[0045] Furthermore, the frequency identification module 12 is used to perform the following method: From the cluster of abnormal interval durations of the parameters at the M locations, the maximum value of the abnormal interval duration of different environmental monitoring parameters is extracted as a set of low monitoring frequencies for the M locations; from the cluster of abnormal interval durations of the parameters at the M locations, the minimum value of the abnormal interval duration of different environmental monitoring parameters is extracted as a set of high monitoring frequencies for the M locations.

[0046] Furthermore, the fusion analysis module 14 is used to perform the following methods: A first feature extraction branch and a second feature extraction branch are pre-constructed. The first feature extraction branch is used to perform feature convolution analysis on the set of high-frequency monitoring data sequences at M locations to determine a set of high-frequency time-series vectors at M locations. The second feature extraction branch is used to perform feature convolution analysis on the set of low-frequency monitoring data sequences at M locations to determine a set of low-frequency spatial vectors at M locations. The set of high-frequency time-series vectors and the set of low-frequency spatial vectors at M locations are mapped and fused, and the attention weights of the second feature extraction branches are updated to obtain M updated second feature extraction branches. The M updated second feature extraction branches are used to perform feature convolution analysis on the set of low-frequency monitoring data sequences at M locations to determine a set of updated low-frequency spatial vectors at M locations. A fully connected network layer is used to analyze the set of updated low-frequency spatial vectors and the set of high-frequency time-series vectors at M locations to obtain a set of monitoring results at M locations.

[0047] Furthermore, the fusion analysis module 14 is used to perform the following methods: The receptive field of the first feature extraction branch is larger than that of the second feature extraction branch.

[0048] Furthermore, the fusion analysis module 14 is used to perform the following methods: Extract a first high-frequency time-series vector set and a first low-frequency spatial vector set from the M high-frequency time-series vector sets and the M low-frequency spatial vector sets; obtain the attention weights of the second feature extraction branch; perform interactive fusion analysis on the first high-frequency time-series vector set and the first low-frequency spatial vector set using a mapping interaction fusion function to obtain a first updated attention weight; update the second feature extraction branch using the first updated attention weight to obtain a first updated second feature extraction branch; perform mapping interaction analysis on the M high-frequency time-series vector sets and the M low-frequency spatial vector sets to update the attention weights of the second feature extraction branch to obtain M updated second feature extraction branches.

[0049] Furthermore, the fusion analysis module 14 is used to perform the following methods: The mapping interaction fusion function is: ;in, To update the attention weights first, For attention weights, Let i be the i-th low-frequency spatial vector of the first point in the set of low-frequency spatial vectors of the first point. This is the normalized value of the similarity between the low-frequency spatial vector of the i-th first point and the high-frequency temporal vector of the i-th first point. This is the transpose of the high-frequency timing vector at the first point. Let be the dimension of the i-th high-frequency time-series vector of the first point, and m be the number of low-frequency spatial vectors of the first point in the set of low-frequency spatial vectors of the first point.

[0050] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0051] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0052] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for online environmental monitoring of VOC emissions, characterized in that, The method includes: An array of monitoring instruments is deployed at M locations along the ventilation duct of the target mine to collect historical monitoring logs, resulting in a set of historical monitoring logs for the M locations, where M is a positive integer. Using a preset set of environmental monitoring parameters as an index, the historical monitoring log sets of the M points are traversed to identify high and low monitoring frequencies, and the set of high and low monitoring frequency groups of the M points is determined. The monitoring array is invoked to continuously monitor the M points according to the high and low monitoring frequency group set of the M points, thereby obtaining a high frequency monitoring data sequence set of the M points and a low frequency monitoring data sequence set of the M points; A high-low frequency interactive fusion analysis is performed on the set of high-frequency monitoring data sequences of the M locations and the set of low-frequency monitoring data sequences of the M locations to determine the set of monitoring results for the M locations.

2. The method for online environmental monitoring of VOC emissions as described in claim 1, characterized in that, The preset set of environmental monitoring parameters includes temperature, flue pressure, flow rate, and humidity.

3. The method for online environmental monitoring of VOC emissions as described in claim 1, characterized in that, Using a preset set of environmental monitoring parameters as an index, the historical monitoring log sets of the M locations are traversed to identify high and low monitoring frequencies, thus determining a set of high and low monitoring frequency groups for the M locations, including: Using the preset environmental monitoring parameter set as an index, extract the parameter abnormality interval duration from the historical monitoring log set of the M points to obtain the parameter abnormality interval duration cluster of the M points; The abnormal interval duration clusters of the M point parameters are filtered to obtain a set of M points with low monitoring frequency and a set of M points with high monitoring frequency. The sets of M low monitoring frequencies and M high monitoring frequencies are mapped and combined to obtain a set of M high and low monitoring frequency groups.

4. The method for online environmental monitoring of VOC emissions as described in claim 3, characterized in that, The abnormal interval duration clusters of the M location parameters are filtered to obtain a set of M locations with low monitoring frequencies and a set of M locations with high monitoring frequencies, including: From the cluster of abnormal interval durations of parameters at the M locations, the maximum values ​​of the abnormal interval durations of different environmental monitoring parameters are extracted as the set of low monitoring frequencies at the M locations. From the cluster of abnormal interval durations of parameters at the M locations, the minimum value of the abnormal interval duration of different environmental monitoring parameters is extracted as the set of high monitoring frequencies at the M locations.

5. The method for online environmental monitoring of VOC emissions as described in claim 1, characterized in that, A high-low frequency interactive fusion analysis is performed on the set of high-frequency monitoring data sequences from the M locations and the set of low-frequency monitoring data sequences from the M locations to determine the set of monitoring results for the M locations, including: Preconstruct the first feature extraction branch and the second feature extraction branch; The first feature extraction branch is used to perform feature convolution analysis on the set of high-frequency monitoring data sequences of the M points respectively to determine the set of high-frequency time-series vectors of the M points. The second feature extraction branch is used to perform feature convolution analysis on the set of low-frequency monitoring data sequences at M points to determine the set of low-frequency spatial vectors at M points. The set of high-frequency time-series vectors of the M points and the set of low-frequency spatial vectors of the M points are mapped and fused, and the attention weights of the second feature extraction branches are updated respectively to obtain the updated M second feature extraction branches. The M updated second feature extraction branches are used to perform feature convolution analysis on the M sets of low-frequency monitoring data sequences at the M locations to determine the set of low-frequency spatial vectors at the M updated locations. The low-frequency spatial vector set and the high-frequency temporal vector set of the M update points are analyzed using a fully connected network layer to obtain a set of monitoring results for the M points.

6. The method for online environmental monitoring of VOC emissions as described in claim 5, characterized in that, The receptive field of the first feature extraction branch is larger than that of the second feature extraction branch.

7. The method for online environmental monitoring of VOC emissions as described in claim 5, characterized in that, The set of high-frequency time-series vectors and the set of low-frequency spatial vectors of the M points are mapped and fused, and the attention weights of the second feature extraction branches are updated respectively to obtain M updated second feature extraction branches, including: Extract the first high-frequency time-series vector set and the first low-frequency spatial vector set from the M high-frequency time-series vector set and the M low-frequency spatial vector set; Obtain the attention weights of the second feature extraction branch, and perform interactive fusion analysis on the high-frequency temporal vector set and the low-frequency spatial vector set of the first point in combination with the mapping interaction fusion function to obtain the first updated attention weights; The second feature extraction branch is updated using the first updated attention weight to obtain the first updated second feature extraction branch; A mapping and interaction analysis is performed on the set of high-frequency time-series vectors of the M points and the set of low-frequency spatial vectors of the M points to update the attention weights of the second feature extraction branch and obtain M updated second feature extraction branches.

8. The method for online environmental monitoring of VOC emissions as described in claim 7, characterized in that, The mapping interaction fusion function is: ； in, To update the attention weights first, For attention weights, This refers to the i-th low-frequency spatial vector at the first point in the set of low-frequency spatial vectors at the first point. This is the normalized value of the similarity between the low-frequency spatial vector of the i-th first point and the high-frequency temporal vector of the i-th first point. This is the transpose of the high-frequency timing vector at the first point. Let be the dimension of the i-th high-frequency time-series vector of the first point, and m be the number of low-frequency spatial vectors of the first point in the set of low-frequency spatial vectors of the first point.

9. An online environmental monitoring system for VOC emissions, characterized in that, The system for implementing the online environmental monitoring method for VOC emissions according to any one of claims 1-8, the system comprising: The data acquisition module is used to interact with the monitoring array at M points deployed in the ventilation duct of the target mine, collect historical monitoring logs, and obtain a set of historical monitoring logs for M points, where M is a positive integer; The frequency identification module is used to identify high and low monitoring frequencies by traversing the historical monitoring log sets of the M points using a preset set of environmental monitoring parameters as an index, and to determine the set of high and low monitoring frequency groups for the M points. The monitoring module is used to call the monitoring array to continuously monitor the M points according to the set of high and low monitoring frequencies of the M points, and obtain a set of high frequency monitoring data sequences and a set of low frequency monitoring data sequences of the M points. The fusion analysis module is used to perform high-low frequency interactive fusion analysis on the set of high-frequency monitoring data sequences of the M points and the set of low-frequency monitoring data sequences of the M points to determine the set of monitoring results for the M points.

Images