An industrial area atmospheric pollutant concentration detection method and system

Abstract

The application relates to the technical field of data processing, and discloses an industrial area atmospheric pollutant concentration detection method and system, which comprises the following steps: arranging detection nodes in an industrial area, collecting instantaneous wind direction, wind speed, vertical wind field at different height layers and pollutant concentration, and generating original wind field and concentration data; performing weighted calculation on a dominant wind direction vector, monitoring an included angle change and determining a rapid wind direction switching; extracting data in two time windows before and after switching, analyzing a concentration gradient, generating a plume shape vector, and superimposing plumes in space; based on a low-concentration area, a virtual background is constructed and a corrected concentration field is generated; when the overlapping area meets the conditions, the source is inverted based on connectivity, otherwise the trajectory is backtracked and inverted by reverse integration; and the corrected concentration field is fused to generate a continuous concentration distribution and output results. The application can accurately correct the background concentration and reliably invert the pollution source position under complex weather conditions such as rapid wind direction switching, so that the timeliness and accuracy of the industrial area atmospheric pollutant detection are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method and system for detecting the concentration of air pollutants in industrial areas. Background Technology

[0002] Online monitoring of air pollutants typically employs a fixed monitoring network. This involves collecting background concentration data at reference monitoring points deployed upwind, comparing this data with measured concentrations at downwind monitoring points, and then subtracting the background level to calculate pollutant increments and locate pollution sources. Under normal meteorological conditions, this method can accurately reflect the spatial distribution and propagation trends of pollutants.

[0003] However, in small areas with unique meteorological characteristics, such as coastal chemical industrial parks, river valley industrial zones, and high-density factory clusters, the prevailing wind direction is easily affected by the switching between sea and land breezes and the funneling effect, shifting rapidly within minutes to tens of minutes. Such sudden wind direction changes cause instantaneous shifts in the propagation direction and diffusion path of pollutant plumes, impacting the effectiveness of fixed upwind reference points. On one hand, when a change in wind direction causes a pollutant plume to cover the reference point, the concentration measured at that point no longer represents the true background value, resulting in distorted background subtraction results. On the other hand, the relative positional relationship between the reference point and downstream monitoring points after a sudden wind direction change loses its representativeness, making source inversion based on this relationship prone to significant bias.

[0004] Existing technologies lack dynamic reference point selection and real-time background correction mechanisms for situations involving rapid wind direction changes. This can lead to misjudgments or omissions in monitoring systems during critical periods of pollution events, thereby affecting the timely identification and accuracy of pollution events. Summary of the Invention

[0005] In view of this, the present invention proposes a method and system for detecting the concentration of air pollutants in industrial areas to solve the above problems.

[0006] On the one hand, the present invention proposes a method for detecting the concentration of air pollutants in industrial areas, comprising: Detection nodes are set up in the target industrial area. The detection nodes are used to collect the instantaneous wind direction, wind speed and vertical wind field information and pollutant concentration at different heights at the current location, and generate raw wind field data and raw concentration data. The dominant wind direction vector is obtained by weighted averaging the instantaneous wind direction of each detection node based on the original wind field data. The angle change between adjacent dominant wind direction vectors is calculated within a fixed time window. When the angle change is greater than or equal to the angle threshold set by the historical wind direction statistical analysis, a rapid wind direction switching event is determined to have occurred. When a fast switching event is triggered, data within the first and second time windows of the same duration before and after the fast switching event are extracted, and the original concentration data in the two windows are subjected to concentration gradient analysis to form plume morphology vector data. The plume morphology vector data of the two windows are spatially superimposed to obtain plume overlapping area data and plume non-overlapping area data. Based on the low concentration part of the non-overlapping area, virtual background concentration field data is constructed to generate corrected concentration field data. When the overlapping area of ​​the plumes meets the preset coverage conditions, source inversion is performed based on the spatial connectivity of the current area; when the overlapping area of ​​the plumes is insufficient, plume trajectory backtracking data is calculated by inverse integration of wind vectors based on the corrected concentration field data and the original wind field data, and source inversion is performed. The corrected concentration field data of the first time window and the second time window are weighted and fused in chronological order to obtain continuous pollutant concentration distribution data, and the pollutant concentration distribution data, source location and concentration spatiotemporal change map during the rapid switching event are output.

[0007] Furthermore, the process of obtaining the dominant wind direction vector by weighted averaging of the instantaneous wind direction at each detection node based on the original wind field data includes: Within the same sampling period, raw wind field data are statistically analyzed by ground detection nodes and detection nodes at different height levels, and data records marked as exceeding the range by the self-inspection of the detected nodes are removed. Based on the planar coordinates and height information of the detection nodes registered during deployment, the set of physically adjacent detection nodes for each detection node is determined, and wind direction jump records that only appear on a single detection node and whose set of physically adjacent detection nodes does not appear simultaneously are removed. Representative wind directions are determined in the ground detection node set and the data sets of each altitude layer to obtain the ground representative wind direction and the representative wind direction sequence of each altitude layer. The representative wind direction is the direction that minimizes the wind direction dispersion within the set. Compare the surface-represented wind direction with the wind direction sequence at each altitude level. If the two wind directions change in the same direction over time and have a continuous transition relationship at altitude, merge the surface-represented wind direction and the altitude-represented wind direction to obtain the dominant wind direction vector; otherwise, use the surface-represented wind direction as the dominant wind direction vector.

[0008] Furthermore, when the representative wind direction at the ground level and the representative wind direction at each altitude level change in the same direction over time and have a continuous transition relationship at altitude, merging the representative wind direction at the ground level and the representative wind direction at each altitude level yields the dominant wind direction vector, which includes: In multiple consecutive sampling periods, the direction of change of the ground-represented wind direction and the direction of change of the wind direction at each altitude layer are compared. When the direction of change of the two is consistent and the magnitude of change is within the same quadrant in the same sampling period, it is determined that they are changing in the same direction in time. Within the same sampling period, the ground-based representative wind direction and the representative wind direction of each height layer are arranged in order of height. If the wind direction difference between adjacent height layers is less than the predetermined continuity determination angle and there is no sudden change in the wind direction in the height direction, it is determined that there is a continuous transition relationship in height. When both the temporal direction of change and the height of continuous transition are satisfied, the representative wind direction at ground level and the representative wind direction at each height level are weighted and synthesized according to their corresponding wind speed ratios to obtain the dominant wind direction vector; if neither condition is satisfied, the representative wind direction at ground level is used as the dominant wind direction vector.

[0009] Furthermore, the step of calculating the angle change between the prevailing wind direction vectors at adjacent times within a fixed time window and determining the occurrence of a rapid wind direction switching event includes: Within the time window, the prevailing wind direction vectors at each moment are obtained in the order of the sampling period. The prevailing wind direction vectors of two adjacent moments are projected onto the same horizontal plane and the included angle is calculated. Positive and negative values ​​are calculated according to the clockwise or counterclockwise direction of the included angle to eliminate the zero-degree reversal error. Calculate the average wind speed for each sampling period within the time window, where the average wind speed is the arithmetic mean of the instantaneous wind speeds of all valid detection nodes within that sampling period; Calculate the vertical wind shear intensity for each sampling period within the time window. The vertical wind shear intensity is the absolute value of the wind direction difference between the wind direction represented by the highest altitude layer and the wind direction represented by the lowest altitude layer, and is obtained by weighted averaging the wind speed differences between each altitude layer. When the average wind speed of any sampling period is lower than the 20th percentile of the historical wind speed distribution of the target industrial area, and the vertical wind shear intensity is higher than the 80th percentile of the historical wind shear intensity distribution of the target industrial area, the absolute value of the angle between the sampling period and its adjacent sampling period is multiplied by the amplification factor obtained by back-calculation from the low wind speed ratio for correction. The corrected absolute value of the included angle is compared with the included angle threshold, which is the 95th percentile value of the set of included angle changes of the dominant wind direction vectors at adjacent moments in adjacent historical sampling periods of the historical representative wind direction sequence of the target industrial area. When the absolute value of the corrected angle at any adjacent moment is greater than or equal to the angle threshold, and the concentration change of at least one downwind detection node in the adjacent moment and subsequent sampling period exceeds the 90th percentile of the historical concentration change distribution of the detection node, a rapid wind direction switching event is determined to have occurred.

[0010] Furthermore, the step of performing concentration gradient analysis on the raw concentration data within the first and second time windows and forming plume morphology vector data includes: In the first and second time windows respectively, the raw concentration data of all detection nodes are acquired according to the sampling period, and the raw concentration data is paired with the prevailing wind direction vector and wind speed of the corresponding sampling period. In each time window, with the prevailing wind direction vector as the reference direction, the concentration gradient components of each detection node in the prevailing wind direction and the vertical direction are calculated. The concentration gradient components are obtained by the ratio of the concentration difference between adjacent detection nodes to the distance between nodes. The concentration gradient components in the same time window are spatially interpolated to obtain the two-dimensional concentration gradient field of that time window. In the two-dimensional concentration gradient field, a continuous connected region that is consistent with the prevailing wind direction vector and whose concentration increases in the prevailing wind direction vector direction is extracted as the main body of the plume. The geometric center, length, width and direction angle from the centroid to each boundary of the current region are recorded to form the corresponding plume morphology vector data.

[0011] Furthermore, spatially overlaying the plume morphology vector data from the first and second time windows to generate corrected concentration field data includes: The main plume regions of the first and second time windows are geometrically superimposed in a unified planar coordinate system to obtain the intersection of the two regions as the plume overlapping region and the non-intersection as the plume non-overlapping region. In the plume non-overlapping region, continuous low-concentration regions along the opposite direction of the prevailing wind are extracted. The low-concentration region is a set of detection nodes whose concentration values ​​are lower than the 30th percentile of the historical concentration distribution of the plume non-overlapping region. A virtual background concentration field of the plume non-overlapping region is generated by spatial interpolation. The virtual background concentration field is compared point by point in space with the original concentration field of the corresponding time window, and the part of the original concentration field that is lower than the background value is replaced by the corresponding position value of the virtual background concentration field to obtain the corrected concentration field data. The corrected concentration field data of the first and second time windows are stored and output respectively.

[0012] Furthermore, the source inversion based on spatial connectivity when the overlapping region of plumes meets the coverage condition includes: Under a unified coordinate system, the ratio of the overlapping area of ​​the plume to the average area of ​​the main plume region in the first time window and the main plume region in the second time window is calculated, and this ratio is taken as the coverage. The coverage is compared with the 85th percentile value based on the historical plume overlap ratio distribution of the target industrial area. When the coverage is greater than or equal to the 85th percentile value, it is determined that the coverage condition is met. Under the condition that the coverage requirement is met, the corrected concentration values ​​of each detection node in the overlapping area of ​​the plume are used as node weights, and a weighted adjacency matrix is ​​constructed based on the Euclidean distance between nodes: The weighted adjacency matrix is ​​subjected to maximum connected subgraph extraction to determine the spatial connectivity components of the overlapping plume region. Among the connected components, the node with the highest concentration weight that is adjacent to the upstream boundary of the wind direction is selected as the initial inversion point. From the initial inversion point, the process is gradually traced back to the adjacent node along the prevailing wind direction until the point of change of concentration gradient sign or the starting point of the backtracking path at the boundary of the detection node is determined as the location of the pollution source.

[0013] Furthermore, the step of weightedly fusing and outputting the corrected concentration field data of the first time window and the second time window in chronological order includes: corresponding the corrected concentration field data of the first time window and the second time window point by point in a unified coordinate system; assigning time weights to the two windows in chronological order, wherein the time weights are calculated according to the time difference between the window center time and the time of the rapid switching event using an exponential decay function; The corrected concentration values ​​at the corresponding locations are weighted and summed according to time weights to generate continuous time-series pollutant concentration distribution data; the continuous time-series concentration distribution data is matched with the pollution source location, and a concentration spatiotemporal variation map including pollution source location markings, pollutant concentration contour lines, fast switching event markers, and plume trajectories is generated; the concentration distribution data, pollution source location, and concentration spatiotemporal variation map are output as detection results.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: By extracting concentration data from the same time window before and after the switch, and constructing a virtual background concentration field by combining overlapping and non-overlapping areas of the plume morphology, background subtraction distortion caused by pollution plume covering the reference point can be avoided, thus improving the accuracy of background correction. The source tracing mode based on spatial connectivity or wind vector inverse integration is automatically switched according to the degree of plume overlap, maintaining high source inversion accuracy even if sudden wind changes cause the relative positional relationship between the original reference point and downstream monitoring points to become invalid. By fusing surface and multi-level wind field information and introducing wind shear and low-wind-speed correction mechanisms, the prevailing wind direction calculation is made more consistent with actual diffusion conditions, thus accurately identifying wind direction changes and pollutant propagation paths under conditions such as sea-land breeze switching, valley wind channels, or valley winds in factory areas. By performing time-weighted fusion of the corrected concentration fields before and after the switch, continuous and dynamic concentration distribution data and visualization maps are generated, providing a reliable basis for rapid assessment and emergency decision-making in pollution events.

[0015] On the other hand, the present invention proposes an industrial zone air pollutant concentration detection system, comprising: The data acquisition module is configured to set up multiple detection nodes in the target industrial area to collect the instantaneous wind direction, wind speed and vertical wind field information at different heights and pollutant concentrations at the current location, and generate raw wind field data and raw concentration data. The dominant wind direction calculation module is configured to perform a weighted average of the instantaneous wind direction of each detection node based on the original wind field data to obtain the dominant wind direction vector, and calculate the angle change of the dominant wind direction vector between adjacent times within a fixed time window. When the angle change is greater than or equal to the angle threshold set by the historical wind direction statistical analysis, a rapid wind direction switching event is determined to have occurred. The plume morphology analysis module is configured to extract data from the first and second time windows of the same duration before and after the rapid switching event when a rapid switching event is triggered, perform concentration gradient analysis on the raw concentration data of the two windows, and generate corresponding plume morphology vector data. The background correction module is configured to spatially superimpose the plume morphology vector data of the two windows to obtain plume overlapping area data and plume non-overlapping area data, and construct virtual background concentration field data based on the low concentration part of the non-overlapping area to generate corrected concentration field data. The source inversion module is configured to perform source inversion based on the spatial connectivity of the current region when the coverage condition is met in the overlapping area of ​​the plumes; when the overlapping area of ​​the plumes is insufficient, it calculates the plume trajectory backtracking data by inverse integration of wind vectors based on the corrected concentration field data and the original wind field data, and performs source inversion. The result fusion and output module is configured to perform weighted fusion of the corrected concentration field data of the first time window and the second time window in chronological order to obtain continuous pollutant concentration distribution data, and output the pollutant concentration distribution data, source location, and concentration spatiotemporal change map during the rapid switching event.

[0016] It should be noted that the method for detecting atmospheric pollutant concentration in industrial areas according to the present invention has the same beneficial effects as its system, and will not be described in detail here. Attached Figure Description

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of a method for detecting atmospheric pollutant concentration in an industrial area, provided as an embodiment of the present invention.

[0018] Figure 2 This is a functional block diagram of an industrial area air pollutant concentration detection system provided in an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] See Figure 1 As shown, this embodiment of the invention provides a method for detecting the concentration of air pollutants in an industrial area, including: S1: Set up detection nodes in the target industrial area. The detection nodes are used to collect the instantaneous wind direction, wind speed and vertical wind field information and pollutant concentration at different heights at the current location, and generate raw wind field data and raw concentration data. S2: Based on the original wind field data, the instantaneous wind direction of each detection node is weighted and averaged to obtain the dominant wind direction vector. Within a fixed time window, the angle change between adjacent dominant wind direction vectors is calculated. When the angle change is greater than or equal to the angle threshold set by the historical wind direction statistical analysis, a rapid wind direction switching event is determined to have occurred. S3: When a fast switching event is triggered, extract the data within the first time window and the second time window of the same duration before and after the fast switching event, and perform concentration gradient analysis on the original concentration data of the two windows to form plume morphology vector data; S4: Spatially superimpose the plume morphology vector data from the two windows to obtain plume overlapping region data and plume non-overlapping region data, and construct virtual background concentration field data based on the low-concentration portion of the non-overlapping region to generate corrected concentration field data. S5: When the overlapping area of ​​the plumes meets the preset coverage conditions, source inversion is performed based on the spatial connectivity of the current area; when the overlapping area of ​​the plumes is insufficient, plume trajectory backtracking data is calculated by inverse integration of wind vectors based on the corrected concentration field data and the original wind field data, and source inversion is performed. S6: The corrected concentration field data of the first time window and the second time window are weighted and fused in chronological order to obtain continuous pollutant concentration distribution data, and output pollutant concentration distribution data, source location and concentration spatiotemporal change map during the rapid switching event.

[0021] Specifically, step S1: Detection node deployment and data collection: Several detection nodes are set up in the target industrial area according to the predetermined spatial layout. The detection nodes can be deployed on the ground, building roofs, factory fences, chimney platforms, communication towers, etc., to form a monitoring network covering different functional areas of the industrial area, different downwind channels and surrounding areas.

[0022] Each detection node includes: a meteorological measurement unit for acquiring instantaneous wind direction and speed at the current location; a vertical wind profile monitoring unit for collecting instantaneous wind direction and speed data at multiple altitude levels (such as the ground level, lower atmosphere, middle atmosphere, and upper atmosphere); a pollutant concentration acquisition unit for monitoring the concentration of target pollutants such as SO2, NOx, VOCs, and PM2.5; and a data identification and transmission unit for attaching timestamps, node identifiers, and altitude level identifiers to the collected data and transmitting it to the central processing unit.

[0023] The detection nodes collect data at a set sampling period T (e.g., 10 seconds, 30 seconds, or 1 minute) to obtain raw wind field data (including instantaneous wind direction and speed at the ground and at each height level) and raw concentration data (including pollutant concentration values ​​at the corresponding location and height). In some implementations, the node location can be optimized through computational fluid dynamics (CFD) simulation to reduce the impact of building obstruction or local airflow anomalies on the data.

[0024] Step S2: Dominant Wind Direction Vector Calculation and Rapid Switching Event Determination: Based on the received raw wind field data, the central processing unit performs a weighted average of the instantaneous wind direction at each detection node to obtain the dominant wind direction vector for the current sampling period. During the weighting process, factors such as wind speed and data integrity are considered, and data from measurement points with abnormal jumps are removed. Within a fixed time window, the angle change between the dominant wind direction vectors of adjacent sampling periods is calculated and compared with an angle threshold set through historical wind direction statistical analysis. When any angle change value is greater than or equal to this threshold, a rapid wind direction switching event is determined to have occurred at the current moment.

[0025] Step S3: Plume Morphology Vector Data Extraction: After detecting a rapid switching event, the raw concentration data within the first and second time windows before and after the event are extracted. For the concentration data within the two time windows, combined with the prevailing wind direction vector and wind speed information of the corresponding sampling period, the concentration gradient components in the prevailing wind direction and the vertical direction are calculated to form a two-dimensional concentration gradient field. A continuous region consistent with the prevailing wind direction and whose concentration increases along that direction is extracted from the gradient field as the plume body, and its geometric center, length, width, and boundary direction angle are recorded to obtain the plume morphology vector data.

[0026] Step S4: Background Correction and Generation of Corrected Concentration Field: The plume morphology vector data from the first and second time windows are spatially superimposed in a unified coordinate system to obtain overlapping and non-overlapping plume regions. Within the non-overlapping plume regions, low-concentration areas (below the 30th percentile of the historical concentration distribution in that region) are extracted in the opposite direction of the prevailing wind, and a virtual background concentration field is constructed based on these areas. The virtual background concentration field is compared point-by-point with the original concentration field of the corresponding window, and the portion of the original field below the background value is replaced with the background field value, thereby generating the corrected concentration field data.

[0027] Step S5: Pollution Source Inversion: When the coverage of the overlapping region of the plumes (the ratio of the intersection area to the average area of ​​the two main plumes) is greater than or equal to the 85th percentile value based on historical data, the coverage condition is met, and pollution source inversion is performed based on the spatial connectivity of the overlapping region. If the coverage is insufficient, the plume trajectory is traced back using the corrected concentration field data and the original wind field data by the method of inverse integration of wind vectors, and pollution source inversion is performed accordingly. During the inversion process, the source location can be determined by combining the concentration gradient sign change points or the monitoring network boundary.

[0028] Step S6: Result Fusion and Output: The corrected concentration field data of the first time window and the second time window are matched point by point in a unified coordinate system, and time weights are assigned according to time order (calculated according to the time difference between the window center time and the event time according to exponential decay). The concentration values ​​are weighted and summed to generate continuous pollutant concentration distribution data.

[0029] The final output includes: continuous concentration distribution data, estimated pollution source locations, and a spatiotemporal variation map of concentration including pollution source labels, pollutant contour lines, fast-switching event markers, and plume trajectories.

[0030] This embodiment achieves high-precision and rapid location of pollutant diffusion paths and pollution source locations within industrial areas by synchronously collecting wind field information and pollutant concentrations at multiple altitudes, combined with steps such as wind direction rapid switching event identification, plume morphology analysis, background correction, and source inversion. This significantly improves adaptability and source tracing accuracy under complex meteorological conditions.

[0031] In some embodiments of this application, the dominant wind direction vector is obtained by weighted averaging of the instantaneous wind direction of each detection node based on the original wind field data, including: Within the same sampling period, raw wind field data are statistically analyzed by ground detection nodes and detection nodes at different height levels, and data records marked as exceeding the range by the self-inspection of the detected nodes are removed. Based on the planar coordinates and height information of the detection nodes registered during deployment, the set of physically adjacent detection nodes for each detection node is determined, and wind direction jump records that only appear on a single detection node and whose set of physically adjacent detection nodes does not appear simultaneously are removed. Representative wind directions are determined in the ground detection node set and the data sets of each height layer, respectively, to obtain the ground representative wind direction and the representative wind direction sequence of each height layer. The representative wind direction is the direction that minimizes the wind direction dispersion within the set. Compare the surface-represented wind direction with the wind direction sequences at each altitude level. If the two wind directions change in the same direction over time and have a continuous transition relationship at altitude, merge the surface-represented wind direction with the altitude-represented wind direction to obtain the dominant wind direction vector; otherwise, use the surface-represented wind direction as the dominant wind direction vector.

[0032] It should be noted that wind direction dispersion refers to the degree of dispersion of wind direction values ​​within a set. The smaller the dispersion, the more concentrated the wind direction is within the set.

[0033] In this embodiment, the process of obtaining the dominant wind direction vector by weighted averaging of the instantaneous wind direction of each detection node based on the original wind field data is as follows: First, within the same sampling period, the central processing unit classifies and statistically analyzes the raw wind field data collected by ground-based detection nodes and detection nodes at different altitudes. Each detection node carries a self-checking flag when collecting data to indicate the sensor's operating status and data validity. When the wind direction or wind speed measurement value of a detection node exceeds the sensor's range or is determined to be abnormal by an internal diagnostic algorithm, its output record will be marked as "out of range." During data processing, all wind field data records marked as "out of range" are discarded to avoid outliers causing deviations in subsequent calculations.

[0034] Next, based on the planar coordinates and height information registered during the deployment of the detection nodes, the central processing unit establishes a set of physically adjacent detection nodes for each node. This set typically includes other nodes whose two-dimensional planar distance and vertical height are both within a preset proximity threshold range. For example, nodes with a planar distance of less than 50 meters and a height difference of less than 5 meters can be considered physically adjacent nodes. This set of adjacent nodes is used to identify local anomalies in subsequent calculations: when, within a certain sampling period, only one detection node experiences a wind direction change that is significantly inconsistent with its surrounding nodes, and no identical change record appears in its set of physically adjacent detection nodes, this change is determined to be an isolated anomaly and is removed. This effectively eliminates interference from occasional factors such as local eddies and transient sensor distortion.

[0035] Then, after removing outlier records, statistical methods are used to determine the representative wind direction for each set of ground-based and multi-level detection nodes. The representative wind direction is defined as the direction that minimizes the wind direction dispersion (e.g., azimuth standard deviation) among all detection nodes within the set. In practice, a vector averaging method can be used to convert the wind direction of each node into a unit vector form, sum them by wind speed weights, and then calculate the azimuth to obtain the optimal representative wind direction. This determined representative wind direction reflects the dominant flow characteristics of the overall wind field at that level or at the ground.

[0036] Subsequently, the central processing unit compares the representative wind direction at ground level with the representative wind direction sequences at each altitude level, analyzing their consistency in both time and altitude dimensions. In the time dimension, if the representative wind direction at ground level and the representative wind direction at each altitude level change in the same direction within a continuous sampling period, i.e., their trends of increase and decrease over time are consistent, it is determined that they "change in the same direction over time." In the altitude dimension, if the difference between the representative wind directions arranged in altitude order shows a gradual change without abrupt changes at each adjacent altitude level, it is determined that they "have a continuous transitional relationship at altitude."

[0037] When both conditions are met—that the wind direction changes in the same direction over time and that there is a continuous transition relationship in altitude—the representative wind direction at ground level and the representative wind direction at each altitude level are weighted and synthesized according to the wind speed ratio at the corresponding altitude level to obtain the dominant wind direction vector for the current sampling period. If neither condition is met, the representative wind direction at ground level is directly used as the dominant wind direction vector for that sampling period.

[0038] By using the above method, while eliminating local outliers, we can integrate ground and multi-level wind field information to obtain a more stable dominant wind direction vector that reflects the overall trend of the vertical wind field, providing a highly reliable input for subsequent wind direction change analysis and pollutant plume tracking.

[0039] In some embodiments of this application, when the ground-represented wind direction and the wind directions at each altitude level change in the same direction over time and have a continuous transition relationship in altitude, the ground-represented wind direction and the wind directions at each altitude level are merged to obtain the dominant wind direction vector, including: In multiple consecutive sampling periods, the direction of change of the ground-represented wind direction and the direction of change of the wind direction at each altitude layer are compared. When the direction of change of the two is consistent and the magnitude of change is within the same quadrant in the same sampling period, it is determined that they are changing in the same direction in time. Within the same sampling period, the ground-based representative wind direction and the representative wind direction of each height layer are arranged in order of height. If the wind direction difference between adjacent height layers is less than the predetermined continuity determination angle and there is no abrupt change in the height direction, it is determined that there is a continuous transition relationship in height. When both temporal and altitude-based unidirectional changes and continuous transitions are satisfied, the representative wind direction at ground level and the representative wind direction at each altitude level are weighted and synthesized according to their corresponding wind speed ratios to obtain the dominant wind direction vector; if neither condition is satisfied, the representative wind direction at ground level is used as the dominant wind direction vector.

[0040] When the surface wind direction and the wind direction at each altitude level change in the same direction over time and have a continuous transition relationship in the altitude dimension, the two are fused to obtain a more accurate dominant wind direction vector. The processing procedure is as follows: First, within multiple consecutive sampling periods, the changing directions of the representative wind direction at the ground level and at each altitude level are compared one by one. The direction of change is determined by calculating the difference between the representative wind directions of two adjacent sampling periods: when the signs of the difference are the same, it indicates that the changing trends are the same, that is, both are changing in a clockwise or counterclockwise direction. To avoid misjudgments caused by minor fluctuations, the magnitude of the change is also mapped to the four quadrants of the azimuth angle, and it is determined whether the two fall within the same quadrant. For example, when the representative wind direction at the ground level changes from 80° to 100°, while the representative wind direction at a certain altitude level changes from 95° to 115°, the changing directions of the two are the same and the magnitude of the change is within the 90° to 180° quadrant, then they can be considered to be changing in the same direction over time.

[0041] Next, within the same sampling period, the representative wind direction at ground level and the representative wind direction at each altitude level are arranged in ascending order of measurement altitude to form an altitude-wind direction profile. The difference in representative wind direction between adjacent altitude levels is calculated one by one and compared with a pre-set continuity judgment angle. If the wind direction difference between all adjacent altitude levels is less than the judgment angle, and there are no abrupt changes in the altitude direction (i.e., there is no significant deviation between the difference between a single altitude level and the levels above and below), then the profile is determined to have a continuous transition relationship in the altitude dimension. The abrupt change judgment here can be achieved by comparing with the average difference of adjacent layers; a change is considered to occur when the difference exceeds a certain multiple.

[0042] When both time and altitude conditions are met, the representative wind direction at ground level and the representative wind direction at each altitude level are weighted and synthesized according to the wind speed ratio at the corresponding altitude level to obtain the dominant wind direction vector for that sampling period. Specifically, the representative wind direction at each altitude level is converted into a unit vector form, and the vector is weighted using the ratio of the wind speed at that altitude level to the sum of the wind speeds across the entire profile. Finally, the azimuth angle of the synthesized vector is calculated, which is the dominant wind direction vector.

[0043] If any condition is not met, such as inconsistent time change direction or abrupt changes in the height profile, the ground representative wind direction is directly used as the dominant wind direction vector to ensure that the dominant wind direction calculation results are not skewed by outliers under complex or unstable wind field conditions.

[0044] By employing the above methods, while ensuring data stability, we can fully utilize the overall structural information of the vertical wind field, improve the representativeness and accuracy of the dominant wind direction vector, and provide reliable input for subsequent monitoring of wind direction changes and prediction of pollutant diffusion paths.

[0045] In some embodiments of this application, calculating the angle change between adjacent dominant wind direction vectors within a fixed time window and determining the occurrence of a rapid wind direction switching event includes: Within the time window, the prevailing wind direction vectors at each moment are obtained in the order of the sampling period. The prevailing wind direction vectors of two adjacent moments are projected onto the same horizontal plane and the included angle is calculated. Positive and negative values ​​are calculated according to the clockwise or counterclockwise direction of the included angle to eliminate the zero-degree reversal error. The average wind speed for each sampling period within the calculation time window is the arithmetic mean of the instantaneous wind speeds of all valid detection nodes within that sampling period. The vertical wind shear intensity is calculated for each sampling period within the calculation time window. The vertical wind shear intensity is the absolute value of the wind direction difference between the wind direction represented by the highest altitude layer and the wind direction represented by the lowest altitude layer, and is obtained by weighted averaging the wind speed differences between each altitude layer. When the average wind speed of any sampling period is lower than the 20th percentile of the historical wind speed distribution of the target industrial area, and the vertical wind shear intensity is higher than the 80th percentile of the historical wind shear intensity distribution of the industrial area, the absolute value of the angle between the sampling period and its adjacent sampling period is multiplied by the amplification factor obtained by back-calculation of the low wind speed ratio for correction. The corrected absolute value of the included angle is compared with the included angle threshold, which is the 95th percentile value of the set of included angle changes of the dominant wind direction vectors at adjacent times in adjacent historical sampling periods of the historical representative wind direction sequence of the target industrial area. A rapid wind direction switching event is determined to have occurred when the absolute value of the corrected angle between any two adjacent moments is greater than or equal to the angle threshold, and the concentration change of at least one downwind detection node in the adjacent moment and subsequent sampling period exceeds the 90th percentile of the historical concentration change distribution of that detection node.

[0046] Specifically, the process of calculating the angle change between the prevailing wind direction vectors at adjacent times within a fixed time window, and determining whether a rapid wind direction switching event has occurred based on this, is as follows: Following a preset sampling period, the prevailing wind direction vector is sequentially acquired from the monitoring data within the time window. To avoid the influence of spatial tilt caused by the three-dimensional wind field data, the prevailing wind direction vectors of two adjacent sampling times are projected onto the same horizontal plane before angle calculation. The angle is calculated using the vector angle formula, and a positive or negative sign is assigned based on the rotation direction (clockwise or counterclockwise) of the angle change, thereby eliminating possible reversal errors when crossing 0°. For example, if the prevailing wind direction is 350° at one moment and 10° at the next moment, the positive or negative sign indicates that the actual change is 20° rather than 340°, ensuring a true reflection of wind direction changes.

[0047] The average wind speed is calculated for each sampling period within the time window. The average wind speed is defined as the arithmetic mean of the instantaneous wind speeds at all valid detection nodes within that period, used to characterize the overall wind field intensity during that time period. Simultaneously, the vertical wind shear intensity is also calculated for each sampling period. The calculation method for vertical wind shear intensity is as follows: first, the representative wind directions at the highest and lowest altitude levels are taken, and the absolute value of their azimuth angle difference is calculated. Then, a weighted average is performed, combining the wind speed differences between different altitude levels, to reflect the comprehensive degree of wind direction change between different altitude levels.

[0048] When the average wind speed of a sampling period is lower than the 20th percentile of the historical wind speed distribution of the target industrial area (indicating low wind speed conditions and a high likelihood of unstable wind direction changes), and simultaneously its vertical wind shear intensity is higher than the 80th percentile of the historical shear intensity distribution of the industrial area (indicating significant inter-layer inconsistencies in the height wind field), the absolute value of the angle between this sampling period and its adjacent periods is corrected. The correction method involves multiplying this angle value by an amplification factor, which is calculated based on the proportion of low wind speeds. This appropriately amplifies the influence weight of wind direction changes under conditions of low wind speed and severe wind shear.

[0049] The corrected absolute value of the included angle is compared with the included angle threshold. The included angle threshold is set based on historical statistical results. Specifically, the set of included angle changes of the dominant wind direction vectors in adjacent sampling periods is extracted from the historical representative wind direction sequence of the target industrial area, and its 95th percentile value is taken as the threshold, so as to ensure that the threshold is sufficient to filter out the few drastic wind direction change events in history.

[0050] During the comparison process, if the absolute value of the corrected angle at any adjacent time point is greater than or equal to the angle threshold, and within that adjacent time point and its subsequent sampling period, the pollutant concentration change at at least one downwind detection node exceeds the 90th percentile of the historical concentration change distribution for that node, then a rapid wind direction switching event is determined to have occurred. By simultaneously introducing the concentration change condition, false signals caused solely by wind field fluctuations but without significantly affecting pollutant dispersion can be effectively eliminated, thereby improving the accuracy of rapid switching event identification.

[0051] The above method can accurately capture rapid wind direction switching under complex and multi-level wind field conditions, and perform dual verification by combining it with significant changes in pollutant concentration, thereby providing a reliable triggering basis for subsequent plume tracking and source inversion.

[0052] In some embodiments of this application, the concentration gradient analysis of the raw concentration data within the first and second time windows to form plume morphology vector data includes: In the first and second time windows respectively, the raw concentration data of all detection nodes are acquired according to the sampling period, and the concentration data is paired with the prevailing wind direction vector and wind speed of the corresponding sampling period. In each time window, with the prevailing wind direction vector as the reference direction, the concentration gradient components of each detection node in the prevailing wind direction and the vertical direction are calculated. The concentration gradient components are obtained by the ratio of the concentration difference between adjacent detection nodes to the distance between nodes. The concentration gradient components in the same time window are spatially interpolated to obtain the two-dimensional concentration gradient field of that time window. In the two-dimensional concentration gradient field, a continuous connected region that is consistent with the prevailing wind direction vector and whose concentration increases in this direction is extracted as the main body of the plume. The geometric center, length, width and direction angle from the centroid to each boundary of the region are recorded to form the corresponding plume morphology vector data.

[0053] In this embodiment, the process of performing concentration gradient analysis on the raw concentration data within the first and second time windows and generating plume morphology vector data includes: Within the first and second time windows, raw concentration data for all detection nodes are acquired according to the sampling period, and mapped one-to-one with the prevailing wind direction vector and wind speed data for the same sampling period. To ensure the reliability of the calculation results, all data undergo time synchronization and outlier removal before processing, eliminating invalid records caused by sensor over-range, self-test anomalies, or communication errors.

[0054] Within each sampling period, a local coordinate system is established with the prevailing wind direction vector as the reference direction, mapping the spatial positions of the detection nodes to this reference system. The prevailing wind direction corresponds to the direction of the prevailing wind direction vector, while the crosswind direction is perpendicular to the prevailing wind direction. In this coordinate system, the concentration gradient components of each node in the prevailing wind direction and crosswind direction are calculated based on the concentration difference between adjacent detection nodes and the projected distance between the two nodes in the prevailing wind direction or crosswind direction. Adjacent nodes are determined by constructing adjacency relationships based on the monitoring point locations, thus ensuring that the gradient calculation remains spatially representative even when the node distribution is irregular.

[0055] The concentration gradient components of each detection node are transformed into a two-dimensional concentration gradient field covering the entire monitoring area using spatial interpolation methods. The interpolation method can be selected according to the deployment environment, such as inverse distance weighting, Kriging interpolation, or other algorithms suitable for irregularly distributed points, so that the gradient field has spatial continuity.

[0056] In the generated two-dimensional concentration gradient field, continuous regions with increasing concentration trends are extracted along the prevailing wind direction. This trend must be higher than a set threshold for the overall field gradient statistics, while the gradient change in the crosswind direction must be small to ensure that the extracted regions have smooth upper lateral boundaries without significant abrupt changes. Spatial connectivity analysis is used to aggregate eligible grid cells or interpolation points into continuous plume main regions, and isolated regions with excessively small areas are removed to avoid interference from local noise.

[0057] For the obtained plume body region, its geometric features are further extracted, including the region's center location, length along the prevailing wind direction, width perpendicular to the prevailing wind direction, the bounding rectangle of the region's shape, the boundary distribution direction, and the directional relationship from the centroid to each boundary. These feature parameters constitute the plume morphology vector data for this time window, which is used in subsequent processing steps such as spatial overlay, background correction, and pollution source inversion.

[0058] In some embodiments of this application, the spatial overlay of plume morphology vector data from the first time window and the second time window to generate corrected concentration field data includes: The main plume regions of the first and second time windows are geometrically superimposed in a unified planar coordinate system. The intersection of the two regions is taken as the plume overlapping region, and the non-intersecting region is taken as the plume non-overlapping region. In the plume non-overlapping region, continuous low-concentration regions along the opposite direction of the prevailing wind are extracted. The low-concentration region is the set of detection nodes whose concentration value is lower than the 30th percentile of the historical concentration distribution of the region. A virtual background concentration field of the non-overlapping region is generated by spatial interpolation. The virtual background concentration field is compared with the original concentration field of the corresponding time window point by point in space. The part of the original concentration field that is lower than the background value is replaced by the corresponding position value of the virtual background concentration field to obtain the corrected concentration field data. The corrected concentration field data of the first and second time windows are stored and output respectively.

[0059] Specifically, the process of spatially superimposing the plume morphology vector data from the first and second time windows to generate corrected concentration field data includes: First, the plume main body regions extracted within the first and second time windows are projected onto the same planar coordinate system, which is aligned with the geographic reference coordinate system of the monitoring area to ensure the comparability of spatial data between the two time windows. During this projection process, coordinate transformation is performed on the detection node locations, and interpolation or re-gridization is used as necessary to map the plume regions of different time windows to the same spatial resolution.

[0060] In a unified coordinate system, the main plume regions of the first and second time windows are geometrically superimposed to obtain their intersection, which is designated as the plume overlap region. This region represents an anomaly in pollutant concentration that exists both before and after the rapid switching event, with consistent plume flow direction, and serves as a crucial basis for subsequent pollution source inversion. Simultaneously, the non-intersecting portions are marked as the plume non-overlapping region. This region reflects anomalies in concentration that exist in one time window but disappear in the other, potentially including background disturbances or contributions from short-term pollution sources.

[0061] In non-overlapping plume regions, low-concentration features are extracted in the opposite direction to the prevailing wind. Low-concentration regions are identified based on their historical concentration distribution; nodes with concentrations below the 30th percentile of the historical statistical distribution are marked as low-concentration node sets. To avoid spatial breaks caused by uneven node distribution, isolated and excessively small low-concentration sub-regions are removed through spatial connectivity analysis, preserving spatially continuous low-concentration bands.

[0062] After obtaining the set of low-concentration nodes, virtual background concentration field data for this low-concentration area is generated using spatial interpolation methods (such as inverse distance weighting, Kriging interpolation, or other interpolation algorithms suitable for irregular monitoring point distributions). This virtual background concentration field spatially covers the entire non-overlapping area of ​​the plume and can reflect the concentration baseline level when there is no short-term pollution source influence.

[0063] The virtual background concentration field is compared point-by-point in space with the original concentration field of the corresponding time window. When the concentration value at a certain location in the original concentration field is lower than the value of the virtual background concentration field at that location, the value of the virtual background concentration field is used to replace the concentration value of the original concentration field at that location. This corrects the local abnormally low concentration caused by instantaneous meteorological fluctuations or monitoring errors, resulting in more stable and continuous corrected concentration field data.

[0064] After the above processing, the corrected concentration field data corresponding to the first time window and the second time window are obtained respectively, and they are stored in a structured manner, retaining the corrected concentration value, time label and spatial coordinate information corresponding to each monitoring point, for subsequent use in pollution source inversion, concentration distribution fusion and visualization output.

[0065] In some embodiments of this application, source inversion based on spatial connectivity when the overlapping region of plumes meets the coverage condition includes: Under a unified coordinate system, the ratio of the overlapping area of ​​the plume to the average area of ​​the main plume region in the first time window and the main plume region in the second time window is calculated, and this ratio is taken as the coverage. The coverage is compared with the 85th percentile value based on the historical plume overlap ratio distribution of the target industrial area. When the coverage is greater than or equal to this percentile value, it is determined that the coverage condition is met. Under the condition that the coverage requirement is met, the corrected concentration values ​​of each detection node in the overlapping area of ​​the plume are used as node weights, and a weighted adjacency matrix is ​​constructed based on the Euclidean distance between nodes: The weighted adjacency matrix is ​​subjected to maximum connected subgraph extraction to determine the spatial connectivity components of the overlapping plume region. Among the connected components, the node with the highest concentration weight that is adjacent to the upstream boundary of the wind direction is selected as the initial inversion point. From the initial inversion point, the process is gradually traced back to the adjacent node along the prevailing wind direction until the point of change of concentration gradient sign or the starting point of the backtracking path of the detection node is reached, which is then determined as the location of the pollution source.

[0066] In this embodiment, when the overlapping area of ​​the plumes meets the coverage condition, the process of pollution source inversion based on spatial connectivity includes: First, the plume main regions extracted from the corrected concentration field data of the first and second time windows are uniformly projected onto the same planar coordinate system, which is aligned with the geographic reference coordinates of the target industrial area to ensure the accuracy of spatial overlay and area calculation. Under this unified coordinate system, the geometric area of ​​the plume overlap region is calculated, and the areas of the plume main regions in the first and second time windows are calculated separately. The average of the two is taken as the reference area. Subsequently, the ratio of the plume overlap area to the reference area is defined as the coverage, which is used to measure the degree of spatial overlap of the plume morphology between the two time windows.

[0067] Next, the coverage is compared with the 85th percentile of the historical plume overlap distribution of the target industrial area. When the coverage is greater than or equal to this percentile, it is determined that the spatial change of the plume under the rapid switching event has sufficient continuity and stability, and source inversion analysis based on spatial connectivity can be carried out; if this condition is not met, it is considered that the spatial overlap is insufficient, and other inversion strategies such as trajectory backtracking need to be adopted.

[0068] Under the condition that the coverage requirement is met, each detection node within the overlapping area of ​​the plume is used as a graph structure node for inversion analysis, and the corrected concentration value at the corresponding location is used as the node weight to reflect the importance of the node in tracing the pollution source. Based on the spatial distance relationship between nodes, a weighted adjacency matrix is ​​constructed using Euclidean distance, where the matrix element value is a function of the distance between the corresponding nodes, and the node concentration value is introduced as a multiplication factor in the weight calculation to increase the influence weight of high-concentration areas in connectivity analysis.

[0069] Next, a maximum connected subgraph is extracted from the weighted adjacency matrix. This step aims to identify spatially connected concentration anomaly regions within the overlapping plume region, filtering out isolated high-value points or noise points that are not connected to the main region. The resulting maximum connected subgraph represents the spatially continuous core portion of the plume that persists in both time windows.

[0070] Within the largest connected subgraph, the set of nodes adjacent to the upstream boundary of the wind direction is first identified, and the node with the highest corrected concentration value is found in this set and determined as the initial inversion point. This point is usually located in the downstream near-field of the pollution source, with a significant concentration peak, and has strong traceability indication significance.

[0071] Starting from the initial inversion point, the path is traced back step by step in the opposite direction of the prevailing wind. During the path tracing process, each step selects the node adjacent to the current node and the closest node upstream in the wind direction, while prioritizing candidate nodes with higher concentration weights to ensure the physical rationality and data stability of the inversion path. During the path extension, the change in the sign of the concentration gradient between adjacent nodes is continuously monitored. When a change from a positive to a negative concentration gradient is detected, or when the tracing reaches the boundary of the detection node deployment, the current node is determined as the termination point of the tracing path, and this point is identified as the pollution source location.

[0072] The above method can utilize the spatial connectivity of plume morphology in two time windows to achieve stable and accurate pollution source localization under high coverage conditions, avoiding the error accumulation and instability problems commonly encountered in single-time or single-point inversion.

[0073] In some embodiments of this application, the weighted fusion and output of the corrected concentration field data of the first time window and the second time window in chronological order includes: corresponding the corrected concentration field data of the first time window and the second time window point by point in a unified coordinate system; assigning time weights to the two windows in chronological order, wherein the time weights are calculated according to the time difference between the window center time and the time when the fast switching event occurs using an exponential decay function; The corrected concentration values ​​at the corresponding locations are weighted and summed according to time weights to generate continuous time-series pollutant concentration distribution data; the continuous time-series concentration distribution data is matched with the pollution source locations, and a spatiotemporal variation map of concentration including pollution source location markings, pollutant concentration contour lines, fast switching event markers, and plume trajectories is generated; the concentration distribution data, pollution source locations, and spatiotemporal variation map of concentration are output as detection results.

[0074] A concentration spatiotemporal variation map is a chart that visually displays the changes in pollutant concentrations at different times and locations. It typically includes information such as contour lines, time axes, event markers, and source locations.

[0075] In this embodiment, when the overlapping area of ​​the plumes meets the coverage condition, pollution source inversion analysis can be performed based on spatial connectivity.

[0076] The "plume overlap region" refers to the spatial intersection formed by geometrically superimposing the main plume regions of two time windows (the first and second time windows) before and after a rapid switching event, within the same geographic coordinate system. The main plume region can be understood as a continuous high-concentration area in a two-dimensional concentration gradient field that is consistent with the prevailing wind direction and shows an increasing trend. Its spatial morphology can reflect the transport path and diffusion range of pollutants within a certain time scale.

[0077] "Coverage" is an indicator used to quantify the degree of spatial overlap of plume morphology between two time windows. The calculation method is as follows: Under a unified coordinate reference, calculate the actual area of ​​the overlapping area of ​​the plumes, and obtain the area values ​​of the main plume area of ​​the first time window and the main plume area of ​​the second time window respectively. Take the average of these two areas as the reference area, and then divide the area of ​​the overlapping area by the reference area to obtain the ratio value. This ratio is the coverage.

[0078] In this embodiment, the coverage is compared with the 85th percentile of the historical plume overlap ratio distribution of the target industrial area. Here, the "historical plume overlap ratio distribution" refers to the statistical distribution formed by the plume overlap ratio data corresponding to multiple rapid switching events in a long-term observation record. If the coverage of the current event is greater than or equal to the 85th percentile of this distribution, it is determined that the coverage condition is met, that is, the spatial overlap of the plumes in the current two time windows is high, the plume morphology changes are continuous and stable, and the premise for inversion based on spatial connectivity is met.

[0079] When the coverage condition is met, the corrected concentration values ​​of all detection nodes (i.e., environmental monitoring points deployed in this area) in the plume overlap area are used as the node weights. The so-called "corrected concentration value" refers to the pollutant concentration value obtained after data correction processing such as replacing low-value areas with virtual background concentration fields on the original concentration data. It can more realistically reflect the actual pollution level at the node location.

[0080] Based on the positional relationships of the detected nodes, a weighted adjacency matrix is ​​constructed using Euclidean distance. Euclidean distance refers to the straight-line distance between two nodes in a two-dimensional plane coordinate system, and this distance value reflects the spatial proximity between nodes. When constructing the weighted adjacency matrix, not only the spatial distance between nodes is considered, but also the adjusted concentration value is introduced into the weight calculation, so that nodes with higher concentrations and closer distances have a greater correlation strength in connectivity analysis.

[0081] Subsequently, the "maximum connected subgraph extraction" operation is performed on the constructed weighted adjacency matrix. A connected subgraph is a subset of nodes in a set where every two nodes are connected by a path; the maximum connected subgraph is the one with the most nodes and the strongest spatial continuity among all connected subgraphs. In this technical solution, the maximum connected subgraph reflects the core pollution zone with spatially close connections and abnormal concentrations within the plume overlap area, thereby removing isolated high-value points or interfering data that are not connected to the main area.

[0082] After obtaining the most connected subgraph, the first step is to identify the set of nodes on its upstream boundary along the prevailing wind direction. These nodes are closer to the pollution source and are more likely to become the starting point for inversion. Within this set of nodes, the node with the highest corrected concentration value is selected as the "initial inversion point." This node is generally located downstream near the source on the pollutant transport path, with a significant concentration peak.

[0083] The inversion process starts from the initial inversion point and gradually backtracks in the opposite direction of the prevailing wind. Each time, the node that is adjacent to the current node and is closest to it in the upstream direction is selected as the next backtracking node. Candidate nodes with higher concentration weights are given priority to ensure that the inversion path is consistent with the actual transport characteristics of pollutants.

[0084] During the path backtracking process, the changes in the "concentration gradient sign" between adjacent nodes are monitored in real time. When the concentration gradient changes from positive (concentration increases with the upstream direction) to negative (concentration decreases with the upstream direction), or when the path extends to the outermost boundary of the detection nodes, the current node position is determined as the end point of the inversion path, and the coordinates of this point are taken as the location of the pollution source.

[0085] Through the above process, the technical solution of this application can achieve high-precision pollution source location by relying on data from only two time windows under the condition of high spatial continuity of plume, utilizing spatial connectivity and concentration distribution characteristics, thus avoiding the path drift and positioning error problems that are prone to occur in traditional single-moment estimation.

[0086] See Figure 2 As shown, this embodiment of the invention provides an industrial zone air pollutant concentration detection system, comprising: The data acquisition module is configured to set up multiple detection nodes in the target industrial area to collect the instantaneous wind direction, wind speed and vertical wind field information at different heights and pollutant concentrations at the current location, and generate raw wind field data and raw concentration data. The dominant wind direction calculation module is configured to perform a weighted average of the instantaneous wind direction of each detection node based on the original wind field data to obtain the dominant wind direction vector, and calculate the angle change between adjacent dominant wind direction vectors within a fixed time window. When the angle change is greater than or equal to the angle threshold set by the historical wind direction statistical analysis, a rapid wind direction switching event is determined to have occurred. The plume morphology analysis module is configured to extract data from the first and second time windows of the same duration before and after the rapid switching event when a rapid switching event is triggered, perform concentration gradient analysis on the raw concentration data of the two windows, and generate corresponding plume morphology vector data. The background correction module is configured to spatially superimpose the plume morphology vector data of the two windows to obtain plume overlapping area data and plume non-overlapping area data, and construct virtual background concentration field data based on the low concentration part of the non-overlapping area, thereby generating corrected concentration field data. The pollution source inversion module is configured to perform source inversion based on the spatial connectivity of the current area when the coverage condition is met in the overlapping area of ​​the plumes; when the overlapping area of ​​the plumes is insufficient, it calculates the plume trajectory backtracking data by inverse integration of wind vectors based on the corrected concentration field data and the original wind field data, and performs source inversion. The results fusion and output module is configured to perform weighted fusion of the corrected concentration field data of the first time window and the second time window in chronological order to obtain continuous pollutant concentration distribution data, and output pollutant concentration distribution data, source location, and concentration spatiotemporal change map during the rapid switching event.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting the concentration of air pollutants in an industrial area, characterized in that, include: Detection nodes are set up in the target industrial area. The detection nodes are used to collect the instantaneous wind direction, wind speed and vertical wind field information and pollutant concentration at different heights at the current location, and generate raw wind field data and raw concentration data. The dominant wind direction vector is obtained by weighted averaging the instantaneous wind direction of each detection node based on the original wind field data. The angle change between adjacent dominant wind direction vectors is calculated within a fixed time window. When the angle change is greater than or equal to the angle threshold set by the historical wind direction statistical analysis, a rapid wind direction switching event is determined to have occurred. When a fast switching event is triggered, data within the first and second time windows of the same duration before and after the fast switching event are extracted, and the original concentration data in the two windows are subjected to concentration gradient analysis to form plume morphology vector data. The plume morphology vector data of the two windows are spatially superimposed to obtain plume overlapping area data and plume non-overlapping area data. Based on the low concentration part of the non-overlapping area, virtual background concentration field data is constructed to generate corrected concentration field data. When the overlapping area of ​​the plumes meets the preset coverage conditions, source inversion is performed based on the spatial connectivity of the current area; when the overlapping area of ​​the plumes is insufficient, plume trajectory backtracking data is calculated by inverse integration of wind vectors based on the corrected concentration field data and the original wind field data, and source inversion is performed to obtain the source location. The corrected concentration field data of the first time window and the second time window are weighted and fused in chronological order to obtain continuous pollutant concentration distribution data, and the pollutant concentration distribution data, source location and concentration spatiotemporal change map during the rapid switching event are output. The concentration spatiotemporal variation map is generated by matching the continuous pollutant concentration distribution data with the location of the pollution source; the concentration spatiotemporal variation map includes pollution source location markings, pollutant concentration contour lines, fast switching event markers, and plume trajectories.

2. The method for detecting air pollutant concentrations in industrial areas according to claim 1, characterized in that, The dominant wind direction vector obtained by weighting the instantaneous wind direction of each detection node based on the original wind field data includes: Within the same sampling period, raw wind field data are statistically analyzed by ground detection nodes and detection nodes at different height levels, and data records marked as exceeding the range by the self-inspection of the detected nodes are removed. Based on the planar coordinates and height information of the detection nodes registered during deployment, the set of physically adjacent detection nodes for each detection node is determined, and wind direction jump records that only appear on a single detection node and whose set of physically adjacent detection nodes does not appear simultaneously are removed. Representative wind directions are determined in the ground detection node set and the data sets of each altitude layer to obtain the ground representative wind direction and the representative wind direction sequence of each altitude layer. The representative wind direction is the direction that minimizes the wind direction dispersion within the set. When the surface wind direction and the wind direction at each altitude level change in the same direction over time and have a continuous transition relationship in altitude, merging the surface wind direction and the wind direction at each altitude level yields the dominant wind direction vector, which includes: In multiple consecutive sampling periods, the direction of change of the ground-represented wind direction and the direction of change of the wind direction at each altitude layer are compared. When the direction of change of the two is consistent and the magnitude of change is within the same quadrant in the same sampling period, it is determined that they are changing in the same direction in time. Within the same sampling period, the ground-based representative wind direction and the representative wind direction of each height layer are arranged in order of height. If the wind direction difference between adjacent height layers is less than the predetermined continuity determination angle and there is no sudden change in the wind direction in the height direction, it is determined that there is a continuous transition relationship in height. When both the temporal direction of change and the height of continuous transition are satisfied, the representative wind direction at ground level and the representative wind direction at each height level are weighted and synthesized according to their corresponding wind speed ratios to obtain the dominant wind direction vector; if neither condition is satisfied, the representative wind direction at ground level is used as the dominant wind direction vector.

3. The method for detecting air pollutant concentration in industrial areas according to claim 2, characterized in that, The step of calculating the angle change between adjacent prevailing wind direction vectors within a fixed time window and determining the occurrence of a rapid wind direction switching event includes: Within the time window, the prevailing wind direction vectors at each moment are obtained in the order of the sampling period. The prevailing wind direction vectors of two adjacent moments are projected onto the same horizontal plane and the included angle is calculated. Positive and negative values ​​are calculated according to the clockwise or counterclockwise direction of the included angle to eliminate the zero-degree reversal error. Calculate the average wind speed for each sampling period within the time window, where the average wind speed is the arithmetic mean of the instantaneous wind speeds of all valid detection nodes within that sampling period; Calculate the vertical wind shear intensity for each sampling period within the time window. The vertical wind shear intensity is the absolute value of the wind direction difference between the wind direction represented by the highest altitude layer and the wind direction represented by the lowest altitude layer, and is obtained by weighted averaging the wind speed differences between each altitude layer. When the average wind speed of any sampling period is lower than the 20th percentile of the historical wind speed distribution of the target industrial area, and the vertical wind shear intensity is higher than the 80th percentile of the historical wind shear intensity distribution of the target industrial area, the absolute value of the angle between the sampling period and its adjacent sampling period is multiplied by the amplification factor obtained by back-calculation from the low wind speed ratio for correction. The corrected absolute value of the included angle is compared with the included angle threshold, which is the 95th percentile value of the set of included angle changes of the dominant wind direction vectors at adjacent moments in adjacent historical sampling periods of the historical representative wind direction sequence of the target industrial area. When the absolute value of the corrected angle at any adjacent moment is greater than or equal to the angle threshold, and the concentration change of at least one downwind detection node in the adjacent moment and subsequent sampling period exceeds the 90th percentile of the historical concentration change distribution of the detection node, a rapid wind direction switching event is determined to have occurred.

4. The method for detecting air pollutant concentration in industrial areas according to claim 3, characterized in that, The step of performing concentration gradient analysis on the raw concentration data within the first and second time windows and generating plume morphology vector data includes: In the first and second time windows respectively, the raw concentration data of all detection nodes are acquired according to the sampling period, and the raw concentration data is paired with the prevailing wind direction vector and wind speed of the corresponding sampling period. In each time window, with the prevailing wind direction vector as the reference direction, the concentration gradient components of each detection node in the prevailing wind direction and the vertical direction are calculated. The concentration gradient components are obtained by the ratio of the concentration difference between adjacent detection nodes to the distance between nodes. The concentration gradient components in the same time window are spatially interpolated to obtain the two-dimensional concentration gradient field of that time window. In the two-dimensional concentration gradient field, a continuous connected region that is consistent with the prevailing wind direction vector and whose concentration increases in the prevailing wind direction vector direction is extracted as the main body of the plume. The geometric center, length, width and direction angle from the centroid to each boundary of the current region are recorded to form the corresponding plume morphology vector data.

5. The method for detecting air pollutant concentration in industrial areas according to claim 4, characterized in that, Spatially overlaying the plume morphology vector data from the first and second time windows to generate corrected concentration field data includes: The main plume regions of the first and second time windows are geometrically superimposed in a unified planar coordinate system to obtain the intersection of the two regions as the plume overlapping region and the non-intersection as the plume non-overlapping region. In the plume non-overlapping region, continuous low-concentration regions along the opposite direction of the prevailing wind are extracted. The low-concentration region is a set of detection nodes whose concentration values ​​are lower than the 30th percentile of the historical concentration distribution of the plume non-overlapping region. A virtual background concentration field of the plume non-overlapping region is generated by spatial interpolation. The virtual background concentration field is compared point by point in space with the original concentration field of the corresponding time window, and the part of the original concentration field that is lower than the background value is replaced by the corresponding position value of the virtual background concentration field to obtain the corrected concentration field data. The corrected concentration field data of the first and second time windows are stored and output respectively.

6. The method for detecting air pollutant concentration in industrial areas according to claim 5, characterized in that, The process of source inversion based on spatial connectivity when the overlapping area of ​​plumes meets the coverage condition includes: Under a unified coordinate system, the ratio of the overlapping area of ​​the plume to the average area of ​​the main plume region in the first time window and the main plume region in the second time window is calculated, and this ratio is taken as the coverage. The coverage is compared with the 85th percentile value based on the historical plume overlap ratio distribution of the target industrial area. When the coverage is greater than or equal to the 85th percentile value, it is determined that the coverage condition is met. Under the condition that the coverage requirement is met, the corrected concentration values ​​of each detection node in the overlapping area of ​​the plume are used as node weights, and a weighted adjacency matrix is ​​constructed based on the Euclidean distance between nodes: The weighted adjacency matrix is ​​subjected to maximum connected subgraph extraction to determine the spatial connectivity components of the overlapping plume region. Among the connected components, the node with the highest concentration weight that is adjacent to the upstream boundary of the wind direction is selected as the initial inversion point. From the initial inversion point, the process is gradually traced back to the adjacent node along the prevailing wind direction until the point of change of concentration gradient sign or the boundary of the detection node is reached. The starting point of the backtracking path is determined to be the location of the pollution source.

7. The method for detecting air pollutant concentration in industrial areas according to claim 6, characterized in that, The step of weightedly fusing and outputting the corrected concentration field data of the first time window and the second time window in chronological order includes: mapping the corrected concentration field data of the first time window and the second time window point by point in a unified coordinate system; assigning time weights to the two windows in chronological order, wherein the time weights are calculated according to the time difference between the window center time and the time of the rapid switching event using an exponential decay function; The concentration distribution data, pollution source location, and concentration spatiotemporal variation map are output as the detection results.

8. An industrial zone air pollutant concentration detection system, used to implement the method according to any one of claims 1-7, characterized in that, The system includes: The data acquisition module is configured to set up multiple detection nodes in the target industrial area to collect the instantaneous wind direction, wind speed and vertical wind field information at different heights and pollutant concentrations at the current location, and generate raw wind field data and raw concentration data. The dominant wind direction calculation module is configured to perform a weighted average of the instantaneous wind direction of each detection node based on the original wind field data to obtain the dominant wind direction vector, and calculate the angle change of the dominant wind direction vector between adjacent times within a fixed time window. When the angle change is greater than or equal to the angle threshold set by the historical wind direction statistical analysis, a rapid wind direction switching event is determined to have occurred. The plume morphology analysis module is configured to extract data from the first and second time windows of the same duration before and after the rapid switching event when a rapid switching event is triggered, perform concentration gradient analysis on the raw concentration data of the two windows, and generate corresponding plume morphology vector data. The background correction module is configured to spatially superimpose the plume morphology vector data of the two windows to obtain plume overlapping area data and plume non-overlapping area data, and construct virtual background concentration field data based on the low concentration part of the non-overlapping area to generate corrected concentration field data. The pollution source inversion module is configured to perform source inversion based on the spatial connectivity of the current area when the coverage condition of the plume overlap area is met; when the plume overlap area is insufficient, it calculates the plume trajectory backtracking data by inverse integration of wind vectors based on the corrected concentration field data and the original wind field data, and performs source inversion to obtain the source location. The result fusion and output module is configured to perform weighted fusion of the corrected concentration field data from the first and second time windows in chronological order to obtain continuous pollutant concentration distribution data, and output the pollutant concentration distribution data, source location, and spatiotemporal variation map of concentration during the rapid switching event. The concentration spatiotemporal variation map is generated by matching the continuous pollutant concentration distribution data with the location of the pollution source; the concentration spatiotemporal variation map includes pollution source location markings, pollutant concentration contour lines, fast switching event markers, and plume trajectories.

Images