Knowledge graph-based ozone precursor collaborative traceability method and system

Abstract

The invention relates to the technical field of ozone precursor traceability, in particular to an ozone precursor collaborative traceability method and system based on a knowledge graph, and the method comprises the following steps: obtaining precursor concentration change, recognizing an abnormal path, extracting path characteristics, carrying out standardized scoring, adjusting graph connection strength, and analyzing sequence offset to obtain collaborative nodes. And screening origin nodes in combination with time sequence meteorology to generate an origin node list. According to the method, pollution response channels are identified through precursor node concentration changes and path connection relations, focusing of key paths is enhanced, paths are scored based on multi-dimensional indexes, connection attributes are dynamically adjusted in combination with monitoring period ozone response intensity, map structure updating is achieved, and concentration response sequence offset is analyzed; according to the method, nodes with co-evolution characteristics are screened, the stable relation identification capability is improved, the path reasonability is evaluated by combining a release time sequence and meteorological conditions, the initial source positioning accuracy is improved, and the response speed and the identification precision of precursor tracing are integrally enhanced.

Description

Technical Field

[0001] This invention relates to the field of ozone precursor traceability technology, and in particular to a knowledge graph-based collaborative traceability method and system for ozone precursors. Background Technology

[0002] The field of ozone precursor source tracing technology belongs to the interdisciplinary field of environmental science and pollution control. It mainly studies the identification and tracking of precursor sources that cause ozone pollution at regional or urban scales. Its core issues include the emission characteristics analysis, source identification methods, spatiotemporal evolution laws, and coupling relationship analysis of key ozone-forming substances such as volatile organic compounds and nitrogen oxides. This technical field usually uses multi-source data fusion, inversion model construction, factor analysis, source apportionment, and statistical modeling to invert and trace the contribution of ozone and its precursors under different pollution scenarios. Among them, traditional ozone precursor source tracing methods refer to the analysis of pollution source contribution based on chemical transport models, positive definite matrix factor decomposition, or high temporal resolution monitoring data. Emission inventory data is usually used as input, combined with atmospheric observation data for numerical simulation or statistical modeling calculation. Common methods include using regional chemical transport models to conduct multi-scenario simulation analysis, using positive definite matrix factor decomposition for pollution factor identification, and using path inversion algorithms to trace pollution source trajectories. When dealing with the correlation between high-dimensional heterogeneous data, the above methods usually rely on manual rules to establish relationships or use a single model to determine causal relationships.

[0003] Existing technologies for tracing pollution sources rely on numerical simulations or statistical modeling using emission inventory and observational data. However, in practice, these methods suffer from significant drawbacks, including strong model input dependence and weak real-time update capabilities. Due to the long update cycle and limited data dimensionality of emission inventories, source tracing analysis often lags behind the pollution evolution process, failing to respond promptly to sudden pollution changes. Furthermore, when dealing with dynamic synergistic relationships between pollutants, they rely heavily on pre-defined rules or single models to statically describe the relationships between variables, lacking the ability to dynamically track pollution evolution paths. This leads to incomplete identification of synergistic pollutants and omissions or misjudgments in source tracing. Traditional methods lack flexibility in integrating multi-source, high-dimensional data, often requiring remodeling for different pollution scenarios, making it difficult to support frequently evolving pollution response processes. For example, simulations based on fixed emission inventories and path inversion methods suffer from amplified path tracking errors under frequent meteorological disturbances or rapid changes in pollutant composition, affecting the accuracy of pollution origin point location. Therefore, existing technologies struggle to meet the comprehensive requirements of response speed, path identification accuracy, and the rigor of origin source tracing logic in complex pollution evolution scenarios. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a knowledge graph-based collaborative tracing method and system for ozone precursors.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a knowledge graph-based collaborative tracing method for ozone precursors, comprising the following steps:

[0006] S1: Obtain the concentration change information of precursor nodes in the map under the pollution change monitoring scenario. Combine the path connection relationship between nodes to judge the trend of precursor response amplitude. Identify the path with the change amplitude exceeding the set proportion as the key channel of pollution response and obtain the pollution response abnormal path set.

[0007] S2: Based on the pollution response anomaly path set, collect the pollution factor change frequency, synergistic pollutant participation and concentration fluctuation range of the nodes connected by the path, extract the indicators and perform standardization processing to obtain the graph edge path scoring reference table;

[0008] S3: Based on the edge path scoring reference table of the graph, match the degree of ozone response change in the current monitoring period, adjust the connection strength attribute of the corresponding path in the graph, and write the updated value into the graph structure to obtain the real-time adjustment result of the graph structure;

[0009] S4: Based on the real-time adjustment results of the graph structure, extract the pollutant concentration change sequence and ozone concentration sequence of the upstream nodes, analyze the degree of offset between the pollutant concentration change sequence and the ozone concentration sequence in the time series, determine whether the response between nodes is continuously within the fluctuation range limit in the time period, and obtain the collaborative response offset trend node set.

[0010] As a further aspect of the present invention, the pollution response anomaly path set includes paths with significant changes in precursor concentration, inter-node connection trend paths, and key pollution response channels. The graph edge path scoring reference table includes the frequency of pollution factor changes, the participation of synergistic pollutants, and the concentration fluctuation range. The real-time adjustment result of the graph structure includes changes in path connection strength, adjusted graph edge attributes, and updated graph topology. The synergistic response offset trend node set includes nodes with stable concentration sequence offsets, nodes with continuous responses within a time period, and nodes with limited fluctuation amplitudes.

[0011] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0012] S101: Obtain the concentration change data of precursor nodes in the spectrum under the monitoring scenario, combine the path connection relationship between nodes, organize the concentration change information of adjacent nodes on each path, extract the response amplitude change of the corresponding nodes of the path, and generate the response amplitude ratio matrix.

[0013] S102: Based on the response amplitude changes of the paths in the response amplitude ratio matrix, and combined with the set response amplitude change ratio threshold, identify the paths whose change amplitude exceeds the threshold, and obtain a set of path sequences that exceed the threshold.

[0014] The threshold for the change in response amplitude is determined based on a statistical analysis method of historical monitoring data. Specifically, all response amplitude ratios recorded in the target area within a recent period (e.g., the last 3 months) are sampled, sorted in ascending order, and the value at the 90th percentile is selected as the default threshold reference (i.e., the paths with the highest change amplitudes are selected as abnormal paths). If sensitivity needs to be increased, it can be adjusted to the 85th percentile; if stability needs to be enhanced, it can be adjusted to the 95th percentile. The threshold can also be dynamically adjusted according to the pollution type, pollution scenario, and response granularity, either by preset or manual setting.

[0015] S103: Call the path sequence set that exceeds the threshold, identify adjacent connected path pairs in the graph topology, and determine whether their response amplitudes simultaneously exceed the set threshold. If the condition is met, the path pair is determined to be a path with an associated response. Extract the set of paths where the ratio of the response amplitudes of the path segments between two or more consecutive nodes all exceeds the threshold, as the pollution response abnormal path set.

[0016] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0017] S201: Based on the pollution response anomaly path set, collect the pollution factor change frequency, synergistic pollutant participation and concentration fluctuation range of the path connection nodes, extract the number of pollution factor changes, the number of synergistic pollutant types and the concentration fluctuation range of the nodes, and generate a node pollution characteristic parameter value set.

[0018] S202: Based on the set of node pollution characteristic parameters, integrate the difference in the frequency of pollution factor changes and the difference in the total amount and concentration fluctuation of synergistic pollutants between path connection nodes, normalize the three parameters respectively, and generate a path normalization characteristic coefficient table.

[0019] The difference in the frequency of pollution factor changes refers to the absolute value of the difference between the frequency of pollution factor changes per unit time between the starting node and the ending node of the path.

[0020] The total number of co-polluting pollutant types is the arithmetic sum of the number of pollutant types participating in co-polluting in the two nodes connected by the path;

[0021] The concentration fluctuation difference is the absolute value of the difference between the upper limits of the pollution factor concentration fluctuation ranges at the start and end nodes of the path within a given time monitoring period.

[0022] S203: Based on the path normalization characteristic coefficient table, the normalization coefficient groups of the frequency of change of pollutants, the types of synergistic pollutants and the concentration fluctuation values ​​are weighted and integrated to extract the score value of each path and generate a graph edge path score reference table.

[0023] The three indicators are weighted and summed using preset weighting coefficients, and the weight values ​​can be determined based on experience in pollution characteristic analysis.

[0024] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0025] S301: Based on the map edge path scoring reference table, call the ozone response change rate data in the current monitoring period, calculate the response change amplitude of each path, compare it with the set change amplitude benchmark value, identify paths with change amplitude exceeding the benchmark value as key paths, classify them according to the change amplitude level, and generate ozone change path intensity grading values.

[0026] S302: Based on the ozone change path intensity classification value, match the path connection strength parameters and adjustment coefficients in the map, update the connection strength parameters, and generate a path connection strength adjustment coefficient set;

[0027] S303: Call the path connection strength adjustment coefficient set, write the updated connection strength parameters into the graph structure path attribute, update the connection strength attribute of the graph path, and obtain the real-time connection strength value set of the graph structure.

[0028] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0029] S401: Based on the real-time connection intensity value set of the spectrum structure, obtain the pollutant concentration sequence of the upstream and midstream nodes and the ozone concentration sequence of the corresponding nodes, arrange the two types of sequences based on the time index and establish the correlation, identify the time difference characteristics of concentration changes between nodes, and generate pollutant-ozone time difference sequence values.

[0030] S402: Call the pollutant-ozone time difference sequence value, set the standard deviation threshold as the basis for judging the stability of the fluctuation, perform statistical analysis on the time difference sequence in a continuous time period, and if the standard deviation of the time difference value in the time period is less than the preset threshold, the time period is determined to be the stable interval of the response fluctuation, and the stable node sequence interval value of the response fluctuation is generated.

[0031] S403: Based on the interval values ​​of the stable node sequence of the response fluctuation, analyze the trend direction and consistency of the concentration time difference between nodes, screen the nodes with cooperative response offset relationship, and obtain the cooperative response offset trend node set.

[0032] As a further aspect of the present invention, the method further includes:

[0033] S5: Based on the collaborative response offset trend node set, read the pollutant release time and regional meteorological conditions of the nodes, combine the node spatial location information and wind field data, use the atmospheric diffusion model to estimate the shortest transmission path and time of pollutants from the release node to the monitoring point, compare the transmission time with the actual monitoring time, judge the temporal consistency and spatial rationality of the path, and screen the nodes that meet the following conditions as the starting source of the pollution event, and generate an origin node identification list in the map;

[0034] The origin node identification list within the map includes pollutant release time sequence nodes, nodes that meet meteorological conditions, and nodes with reasonable path sources.

[0035] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0036] S501: Obtain the pollutant release time and regional meteorological conditions in the cluster of nodes of the collaborative response offset trend. Combine the spatial location of the nodes with the characteristics of wind speed and direction to determine the dominant direction and path length of the pollutant diffusion and generate the diffusion path direction quantity and the initial path length value.

[0037] S502: Based on the diffusion path direction and initial path length, combined with the monitoring point location and pollutant detection time, analyze the changing trends of wind direction and wind speed in the release path, screen nodes that meet the wind field stability characteristics, and obtain a set of wind field consistency offset nodes.

[0038] S503: Based on the set of wind field consistency offset nodes, combined with the time interval and path length between the nodes and the monitoring points, determine the spatiotemporal coordination relationship, select nodes that meet the conditions as the starting source of the pollution event, and obtain the origin node identification list.

[0039] As a further aspect of the present invention, the set ratio refers to the threshold of the response amplitude change ratio obtained based on the statistical analysis of monitoring data. For example, by analyzing the distribution of the response amplitude ratio in the concentration change path, the 90th percentile ratio is determined as the threshold reference value, which is used to identify the path with the change amplitude exceeding the set ratio as the key channel for pollution response.

[0040] A knowledge graph-based collaborative tracing system for ozone precursors includes:

[0041] The precursor channel identification module acquires the concentration change information of precursor nodes in the pollution map under the pollution change monitoring scenario. Combined with the path connection relationship between nodes, it determines whether the response amplitude of precursors in the path exceeds the set change ratio. Paths with change amplitude exceeding the set change ratio are identified as key pollution response channels, and a set of abnormal pollution response paths is obtained.

[0042] The path scoring generation module collects the frequency of pollution factor changes, the participation of synergistic pollutants, and the concentration fluctuation range of path nodes based on the pollution response anomaly path set. After extracting the associated indicators, it performs standardization processing to obtain the graph edge path scoring reference table.

[0043] The intensity attribute adjustment module adjusts the connection intensity attribute of the spectrum path according to the spectrum edge path scoring reference table, matches the ozone response change degree in the current monitoring period, and writes it into the spectrum structure to obtain the real-time adjustment result of the spectrum structure.

[0044] The offset trend extraction module adjusts the results in real time according to the spectrum structure, extracts the pollutant concentration sequence and ozone concentration sequence of the upstream node, analyzes the degree of time offset between the two, and screens the nodes that meet the stability characteristics of the amplitude range to obtain the set of cooperative response offset trend nodes.

[0045] The origin node identification module reads the pollutant release time and regional meteorological conditions of the nodes based on the collaborative response offset trend node set, evaluates the temporal rationality of the path, and selects nodes that meet the conditions as the starting position of the pollution event to obtain the origin node identification list in the map.

[0046] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0047] In this invention, pollution response channels are identified by the relationship between precursor node concentration changes and path connectivity, enhancing the focus on key paths, eliminating interference from invalid information, scoring the degree of path influence based on multi-dimensional indicators, and dynamically adjusting path connectivity attributes in conjunction with ozone response intensity within the monitoring period to achieve real-time updates of the spectral structure. By analyzing concentration response sequence shifts, nodes with co-evolutionary characteristics are screened to improve the ability to identify stable response relationships. Combined with release timing and meteorological conditions, the rationality of the path is evaluated, improving the accuracy of locating the origin of pollution events. The above methods improve the response speed, path identification accuracy, and origin location reliability of precursor source tracing. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the steps of the present invention;

[0049] Figure 2 This is a flowchart of steps S1 of the present invention;

[0050] Figure 3 This is a flowchart of steps S2 of the present invention;

[0051] Figure 4 This is a flowchart of steps S3 of the present invention;

[0052] Figure 5 This is a flowchart of step S4 of the present invention;

[0053] Figure 6 This is a flowchart of steps S5 of the present invention;

[0054] Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0056] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0057] Please see Figure 1 A knowledge graph-based collaborative tracing method for ozone precursors includes the following steps:

[0058] S1: Obtain the concentration change information of precursor nodes in the map under the pollution change monitoring scenario. Combine the path connection relationship between nodes to judge the trend of precursor response amplitude. Identify the path with the change amplitude exceeding the set proportion as the key channel of pollution response and obtain the pollution response abnormal path set.

[0059] S2: Based on the abnormal path set of pollution response, collect the frequency of pollution factor changes, the participation of synergistic pollutants and the concentration fluctuation range of the nodes connected by the path, extract the indicators and perform standardization processing to obtain the graph edge path scoring reference table.

[0060] S3: Based on the edge path scoring reference table of the graph, match the degree of ozone response change in the current monitoring period, adjust the connection strength attribute of the corresponding path in the graph, and write the updated value into the graph structure to obtain the real-time adjustment result of the graph structure;

[0061] S4: Based on the real-time adjustment results of the graph structure, extract the pollutant concentration change sequence and ozone concentration sequence of the upstream nodes, analyze the degree of offset of the pollutant concentration change sequence and ozone concentration sequence in the time series, determine whether the response between nodes is continuously within the fluctuation range in multiple time periods, screen nodes that meet the stable characteristics of the difference value interval, and obtain the set of nodes with collaborative response offset trend.

[0062] S5: Based on the collaborative response offset trend node set, read the pollutant release time and regional meteorological conditions of the nodes, combine the node spatial location information and wind field data, use the atmospheric diffusion model to estimate the shortest transmission path and time of pollutants from the release node to the monitoring point, compare the transmission time with the actual monitoring time, judge the temporal consistency and spatial rationality of the path, and select the nodes that meet the following conditions as the starting source of the pollution event, and generate an identification list of origin nodes in the map.

[0063] The pollution response anomaly path set includes paths with significant changes in precursor concentration, paths with inter-node connection trends, and key pollution response channels. The map edge path scoring reference table includes the frequency of pollutant changes, the participation of synergistic pollutants, and the range of concentration fluctuations. The real-time map structure adjustment results include changes in path connection strength, adjusted map edge attributes, and updated map topology. The synergistic response offset trend node set includes nodes with stable concentration sequence offsets, nodes with continuous responses within a time period, and nodes with limited fluctuation ranges. The map origin node identification list includes pollutant release time sequence nodes, nodes that meet meteorological conditions, and nodes with reasonable path sources.

[0064] Please see Figure 2 The specific steps of S1 are as follows:

[0065] S101: Obtain the concentration change data of precursor nodes in the spectrum under the monitoring scenario, combine the path connection relationship between nodes, organize the concentration change information of adjacent nodes on each path, extract the response amplitude change of the corresponding nodes of the path, and generate the response amplitude ratio matrix.

[0066] Concentration data for precursor nodes in the spectrum can be obtained through an environmental monitoring system. The concentration values ​​need to be arranged chronologically, and corresponding time series should be constructed for each node. Taking urban air monitoring as an example, if the detected substances are precursors such as NO, SO2, and VOCs, their concentrations at different time points, such as T0, T1, and T2, need to be recorded. For example, the NO concentration at T0 is 18 μg / m³. 3 At T1, it was 21 μg / m 3At T2, the concentration was 24 μg / m3; SO2 and VOCs were recorded in the same manner. Based on the connection paths of the nodes in the spectrum, such as NO→O3→HNO3, the concentration difference between each pair of adjacent nodes at consecutive time points needs to be organized according to the path order. The concentration difference can be obtained by comparing the current time point with the previous time point; for example, the difference between T1 and T0 is 3 μg / m3. 3 The difference between T2 and T1 is 3 μg / m 3 Then, using this difference, the ratio of response amplitude changes for each segment in the path is calculated. For example, if the NO concentration difference between T1 and T2 is 3, and the O3 concentration difference is 5, then the ratio for that segment is 5 divided by 3, which is 1.67. All paths are processed in this way, extracting the ratios between adjacent nodes segment by segment and summarizing them to form a matrix structure. Each row represents a path, and each column represents the response ratio of a certain node pair in the path. For example, the ratio for the NO→O3 segment is 1.67, and the ratio for the O3→HNO3 segment is 1.22. In the final generated matrix, each element is the actual ratio result, used for subsequent judgment of the path's change characteristics.

[0067] S102: Based on the changes in response amplitude of the paths in the response amplitude ratio matrix, and combined with the set response amplitude change ratio threshold, identify the paths whose change amplitude exceeds the threshold, and obtain a set of path sequences that exceed the threshold.

[0068] When processing the aforementioned ratio matrix, a threshold for response amplitude variation needs to be set. This threshold can be determined through statistical methods from historical monitoring data. For example, select the response ratios of the past 1000 paths, sort them in ascending order, and take the ratio corresponding to the 900th position as the 90th percentile reference value. If the corresponding ratio is 1.5, then use it as the current threshold. Subsequently, search the matrix for all segments with ratios greater than 1.5 and their corresponding paths. For example, if the ratio of the segment NO→O3 in a certain path is 1.67, which is greater than the threshold, then this path is marked as exceeding the threshold. As long as the ratio of any segment in the path exceeds the threshold, it can be included in the filter set. For example, in the path SO2→H2SO4→PM2.5, if the ratio of the first segment is 1.6 and the ratio of the second segment is 1.3, although the latter segment does not meet the threshold, the path is still included because the former segment has exceeded 1.5. Process the entire ratio matrix in this way to form a set of multiple paths exceeding the threshold, which serves as the input for the next step of analysis.

[0069] S103: Call the path sequence set that exceeds the threshold, identify the path pairs that are adjacent to each other in the graph topology, and determine whether their response amplitudes exceed the set threshold at the same time. If the condition is met, the path pair is determined to be a path with a correlated response. Extract the set of paths where the ratio of the response amplitude of the path segments between two or more consecutive nodes exceeds the threshold, as the pollution response abnormal path set.

[0070] Retrieving a path sequence set exceeding a threshold refers to extracting a set of paths with response amplitudes greater than a set threshold based on monitoring data. The threshold can be calculated and determined using the mean and standard deviation of historical response data. For example, if the response amplitudes of a path set are 12.4, 13.2, 12.9, 13.5, and 13.1 μV, the calculated mean is approximately 13.02 μV, and the standard deviation is approximately 0.38 μV. If the threshold is set to 13.78 μV, then all paths with response amplitudes greater than 13.78 μV are considered threshold-exceeding paths. Subsequently, the presence of adjacent connections between these paths is identified in the topological map. The starting and ending node numbers of the path are determined, and it is checked whether there are shared nodes. For example, if path P1 connects nodes N3-N4 and path P2 connects nodes N4-N5, then because the shared node N4 forms an adjacent path pair, if the response amplitudes of P1 and P2 are 14.2μV and 14.7μV respectively, both higher than the threshold, then they are considered a path pair whose response amplitudes synchronously exceed the threshold, satisfying the condition for an associative response path. Then, starting from any associative path in the graph structure, based on the node connection relationship, the path segments that form continuous connections are searched layer by layer along the path direction. At the same time, it is checked whether the ratio of the response amplitudes of each adjacent path within the path segment is... The path response ratio threshold can be set based on the maximum fluctuation amplitude of the background noise and the average response level. If the background fluctuation amplitude is 0.3μV and the average response is 13.0μV, the minimum effective response amplification ratio is set to 1.023. In practice, it can be set to 1.05 to improve screening accuracy, meaning the response amplitude of the next path needs to be at least 5% higher than the previous one. For example, if the path segment responses are 14.2, 15.0, and 15.9μV respectively, the adjacent ratios are 1.056 and 1.06, both higher than 1.05, indicating that the path segment meets the continuous enhancement characteristic and constitutes an abnormal pollution response path. If a certain ratio in a path segment is lower than the ratio threshold, for example, if the responses are 14.8, 15.3, and 15.7 μV, the ratios are 1.034 and 1.026 respectively, then the condition is not met and the path segment should be excluded. During the execution process, all candidate path segments are traversed and the response ratio is judged. Only path segments that meet the four requirements of continuous path structure, synchronous response exceeding the threshold, clear adjacent connection, and increasing response ratio are retained, and the final set of abnormal pollution response paths is formed. This set can be used in practical engineering applications such as pollution source propagation path identification, fault chain screening in sensor networks, and voltage anomaly detection.

[0071] Please see Figure 3 The specific steps of S2 are as follows:

[0072] S201: Based on the abnormal path set of pollution response, collect the frequency of pollution factor changes, participation of synergistic pollutants and concentration fluctuation range of the path connection nodes, extract the number of pollution factor changes, number of synergistic pollutant types and concentration fluctuation range of the nodes, and generate a set of node pollution characteristic parameter values.

[0073] Identifying anomalous pollution response pathways requires first extracting concentration change pathway sequences between multiple nodes in the environmental monitoring network. By comparing historical concentration data with real-time sampling data, pathways exhibiting rapid concentration fluctuations within short periods can be identified and marked as anomalous pathways. Taking an anomalous NO2 response pathway as an example, it can be observed that a node experiences a concentration change from 35 μg / m³ within 12 hours. 3 Increased to 82 μg / m 3 And continue sampling for more than 3 sampling cycles to form a concentration mutation standard. For all connection nodes on this path, count the frequency of significant concentration changes of each pollutant per unit time, for example, setting a change threshold of ±10 μg / m³. 3 Within a 12-hour continuous sampling period, if five concentration fluctuations exceeding the threshold are detected, the frequency is 0.417 times / hour. Next, it is necessary to identify whether co-pollutants are involved, i.e., whether other pollutants (such as O3, CO, etc.) simultaneously fluctuate significantly when the target pollutant concentration changes. If the proportion of simultaneous changes exceeds 50% of the total fluctuations, and the trends of the two pollutants are strongly correlated (e.g., correlation coefficient greater than 0.6), then co-pollutants are recorded, and the number of types is the number of co-pollutants involved. Finally, the concentration fluctuation range of each pollutant within a given monitoring period is extracted, such as a maximum value of 90 μg / m³. 3 Minimum value 32 μg / m 3 The fluctuation range is 58 μg / m 3 The frequency of change, the number of co-pollutants, and the concentration fluctuation values ​​of each node are integrated to form a set of node pollution characteristic parameters, which are used for subsequent calculation of differences between paths.

[0074] S202: Based on the set of node pollution characteristic parameter values, integrate the difference in the frequency of pollution factor changes and the difference in the total amount and concentration fluctuation of synergistic pollutants between path connection nodes, normalize the three parameters respectively, and generate a path normalization characteristic coefficient table.

[0075] Based on the pollution characteristic parameter sets of each node in the path, it is necessary to integrate and calculate the parameter differences between the connecting nodes. First, the difference in the frequency of pollution factor changes between the start and end nodes is extracted. For example, if the starting frequency is 0.417 times / h and the ending frequency is 0.25 times / h, the difference is 0.167 times / h. The total number of synergistic pollutant types is the sum of the pollutant types involved by each node. If one end involves two types and the other end involves one type, the total is three types. The concentration fluctuation difference is the absolute difference between the maximum and minimum concentrations at both ends. For example, if the fluctuation value at one end is 58 μg / m³, the difference is calculated as follows: 3 The other end is 41 μg / m 3 The difference is 17 μg / m 3After obtaining the differences of all path parameters, the above three data points need to be normalized to fall within the standard range of 0 to 1. A maximum-minimum value range transformation method is used. Assuming the frequency difference range is 0.05 to 0.3, and the current difference is 0.167, the normalized value is approximately 0.47. If the concentration fluctuation difference is between 5 and 40 μg / m³, the normalized value will be approximately 0.47. 3 Between, the current value is 17 μg / m 3 The normalized value is approximately 0.34; the total amount of co-pollutants ranges from 1 to 5, currently there are 3, so the normalized value is 0.5. The normalized variation frequency, total amount of co-pollutants, and concentration fluctuations for all pathways are compiled into a unified table to record the normalization coefficients, providing a basis for subsequent scoring.

[0076] S203: Based on the path normalization characteristic coefficient table, the normalization coefficient groups of the frequency of change of pollutants, the types of synergistic pollutants and the concentration fluctuation values ​​are weighted and integrated to extract the score value of each path and generate a graph edge path score reference table.

[0077] For the normalized characteristic coefficient results of each path, the three parameter values ​​need to be further weighted and integrated to form a single path score. The weight ratios of the three parameters are set, for example, frequency of variation is weighted at 40%, number of co-pollutant species at 30%, and concentration fluctuation at 30%. Taking a path as an example, its normalized frequency of variation is 0.47, co-pollutant at 0.5, and concentration fluctuation at 0.34. The products of the three components can be calculated according to the above weights and summed to obtain a total score of approximately 0.44. This score represents the comprehensive pollution characteristic intensity of the path. By comparing and ranking the scores of all paths, a reference table for edge path scores in the graph is constructed. All scores are recorded in the scoring table to facilitate subsequent path selection and priority determination, and to build a basic dataset for pollution impact propagation models and distribution trend analysis.

[0078] Please see Figure 4 The specific steps of S3 are as follows:

[0079] S301: Based on the edge path scoring reference table, call the ozone response change rate data in the current monitoring period, calculate the response change amplitude of each path, compare it with the set change amplitude benchmark value, identify paths with change amplitude exceeding the benchmark value as key paths, classify them according to the change amplitude level, and generate ozone change path intensity grading values.

[0080] The specific calculation formula for identifying paths whose changes exceed the baseline value as priority paths is as follows:

[0081]

[0082] Calculate the characteristic value of path change intensity to generate ozone change path intensity classification value;

[0083] in, The standardized variation intensity characteristic value represents the i-th edge path. This represents the rate of change of ozone response for the i-th edge path at time t. represents the average ozone response change rate of the i-th edge path at all times t within the monitoring period, n represents the total number of time steps within the monitoring period, and t represents the monitoring time step number;

[0084] The ozone response change rate of path i at time t is expressed in ppb / h. It is calculated by dividing the difference in ozone concentration between adjacent time periods by the time interval after collecting ozone concentration data through a high-precision environmental monitoring station. The monitoring equipment used is a TEI49i ozone analyzer with a sampling frequency of once per hour and a concentration accuracy of ±1ppb.

[0085] The arithmetic mean of the ozone response change rate at n time points represents path i, which is directly calculated from the n sampled data.

[0086] Setting n to 6 indicates that a total of 6 hours of ozone response data will be collected within the current monitoring period, with the sampling time being from 07:00 to 12:00, for a continuous 6 hours;

[0087] The following data were obtained from the measured data of urban stations in July 2024 published by the Shanghai Municipal Bureau of Ecology and Environment. The sampling path number is i=12, and the response change rate data for times t=1 to 6 are as follows (unit: ppb / h):

[0088]

[0089] Calculate the mean:

[0090]

[0091] Calculate the first part, namely the average relative rate of change:

[0092] List them one by one (t=2 to 6, since there is no preceding term for t=1):

[0093]

[0094] Take the average of the above 5 items:

[0095]

[0096] Calculate the second part, namely the square root term of the normalized variance:

[0097]

[0098]

[0099] Calculate the average and take the square root:

[0100]

[0101] Substitute into the complete formula expression:

[0102]

[0103] The results show that the standardized change intensity characteristic value of path No. 12 in the current monitoring period is 0.0992. This value indicates that the relative response change fluctuation of the path is relatively low but there is a certain trend stability. According to the intensity classification standard, this value can be used as an input parameter for path classification and subsequent classification processes such as ozone path level division and evolution trend ranking.

[0104] This formula uses the ozone response change rate of a path as the core variable. It extracts the average response change trend of the path within the monitoring period by calculating the relative difference between the change rates at adjacent times and taking the mean. Then, it constructs a standardized variance by analyzing the deviation of the current change rate from the overall mean, and takes its square root to reflect the overall volatility. These two terms form the basis of the formula's two operations. Structurally, the first term is a mean operation, reflecting the direction and rate of change of the path response over time. The second term is the square root of the mean of the squared deviation, reflecting the amplitude of the path's fluctuations within the period. Subtracting the two results and taking the absolute value measures the difference between the path's trend and instability, revealing its response intensity characteristics, and thus serving as the basis for calculating the path intensity classification value. Addition and subtraction are used for difference comparison; multiplication, division, and standardization are used to eliminate the influence of dimensions; squaring is used to amplify fluctuation differences; and square rooting is used to restore scale consistency. Finally, the absolute value is used to unify the directional output.

[0105] The path change intensity characteristic value is used to characterize the difference between the changing trend and fluctuation amplitude of a certain ozone response path within the monitoring period. This value comprehensively reflects the dynamic behavior of the path response over time. Specifically, the larger the value, the more obvious the changing trend of the path is in the overall response process, accompanied by strong response fluctuations, indicating higher uncertainty and the possibility of sudden changes. If the value is close to zero, it indicates that the path change is relatively stable, the response rate and fluctuation amplitude are basically consistent, and it shows the characteristics of stable trend and small disturbance. This characteristic value can be used as a quantitative indicator in path screening, classification and early warning models to identify key changing paths and their evolution intensity in the pollution control process.

[0106] S302: Based on the ozone change path intensity classification value, match the path connection strength parameters and adjustment coefficients in the map, update the connection strength parameters, and generate a set of path connection strength adjustment coefficients;

[0107] Based on the path change levels obtained in the previous step, the original connectivity strength values ​​of each path need to be extracted from the graph structure. These values ​​are the influence weights given during the modeling phase and are generally between 0 and 1; for example, the original connectivity strength of path P003 is 0.5. The system sets adjustment coefficients corresponding to different levels of change, such as ±0.05 for mild, ±0.15 for moderate, and ±0.25 for severe. The adjustment direction is determined by the direction of the response change; if the response change is positive, the connectivity strength increases, and vice versa. For example, if path P004 is determined to have a moderate positive change, its connectivity strength increases by 0.15, resulting in an adjusted value of 0.65; if path P005 has a severe negative change, its original value of 0.7 is adjusted to 0.45. This process is executed on all paths requiring updates and generates a structured parameter set containing the path number, original connectivity strength, adjustment magnitude, and updated connectivity strength value.

[0108] S303: Call the path connection strength adjustment coefficient set, write the updated connection strength parameters into the graph structure path attribute, update the connection strength attribute of the graph path, and obtain the real-time connection strength value set of the graph structure.

[0109] Based on the updated connectivity strength dataset, the attribute fields of all paths in the graph structure are traversed, matching the corresponding path numbers, reading the original connectivity strength value field, and writing the new updated value. If the graph structure is organized as a graph database, the update can be executed through path attribute modification commands. The update operation requires precise location of path relationships and replacement of the original connectivity strength values. After the update is completed, the graph path structure needs to be regenerated, and the edge path relationship calculation and update are completed using a graph traversal algorithm. The final output structure is a dataset containing all path numbers and their latest connectivity strength values, such as P001 corresponding to a value of 0.75 and P002 corresponding to 0.45, representing the connectivity strength status of each path in the current period, for subsequent analysis or system calls.

[0110] Please see Figure 5 The specific steps of S4 are as follows:

[0111] S401: Based on the real-time connection of intensity value set of the graph structure, obtain the pollutant concentration sequence of the middle and upstream nodes and the ozone concentration sequence of the corresponding nodes, arrange the two types of sequences based on the time index and establish the correlation, identify the time difference characteristics of concentration changes between nodes, and generate pollutant-ozone time difference sequence values.

[0112] The specific calculation formula for arranging two types of sequences based on time index and establishing association is as follows:

[0113]

[0114] Calculate the pollutant-ozone time difference characteristic value to generate a pollutant-ozone time difference sequence value;

[0115] Where, Δt ij C represents the characteristic value of the pollutant-ozone time difference between upstream node i and downstream node j. p,i,k w represents the pollutant concentration value of node i at time k. i,k C represents the graph connection strength value between node i and node j at time k (based on the node graph structure, the connection strength value ranges from [0,1]). o,j,k+τ This represents the ozone concentration at node j at time k+τ. The value represents the average pollutant concentration of node i over all time points, n represents the length of the time series of pollutant concentration versus ozone concentration, and ∈′ represents a minimal constant to prevent the denominator from being zero (recommended to be set to 10). -6 ), where τ is the time lag step of the pollutant's impact on ozone;

[0116] C p,i,k The NOx concentration at node i at time k is 52.3, as collected by air quality monitoring equipment.

[0117] w i,k The graph connection strength between node i and node j at time k is 0.78, generated by normalizing the meteorological wind field model and traffic flow network. This value changes proportionally with the increase of wind speed.

[0118] C o,j,k+τ The ozone concentration at node j at time k+τ was 64.1, as collected by an ultraviolet spectroscopy ozone detection device.

[0119] The average NOx concentration at node i was 52.04 after 20 minutes of continuous sampling.

[0120] The minimum coefficient is 1×10 -6 The result after introducing a smoothing term is 0.00005204.

[0121] n: The sampling length of the continuous time series is 10;

[0122] τ: Determined by the maximum mutual information lag window, with a lag step size of 3;

[0123] Substituting into the formula, the calculation is as follows:

[0124] Molecular part:

[0125] 52.3 * 0.78 = 40.794;

[0126] 40.794 - 64.1 = -23.306;

[0127] Denominator part:

[0128] |52.3-52.04|=0.26;

[0129] 0.26 + 0.00005204 = 0.26005204;

[0130]

[0131] Single value:

[0132]

[0133] Summary average:

[0134] If the remaining 9 results have the same structure, then:

[0135]

[0136] The results show that there is a significant difference between the pollutant concentration change at node i and the ozone concentration change at node j within the time delay window. The value 45.1 represents the characteristic time difference between the two types of pollution time series responses, which is used to generate pollutant-ozone time difference sequence values, and is used as the step result data to enter the subsequent spatial analysis stage.

[0137] The formula's operational logic is based on the extraction process of the time misalignment characteristics between pollutant concentration changes and ozone response. First, the pollutant concentration value C... p,i,k Connection strength w with the spectrum i,k Multiplying these values ​​allows for a weighted calculation of the contribution of node i to downstream node j at time k, reflecting the modulation effect of the spectral structure on the pollutant transport path. Subsequently, this weighted concentration is multiplied by the ozone concentration C at node j at a delay time point k+τ. o,j,k+τ Subtraction is performed to capture the difference between the input of precursor pollutants and the subsequent ozone formation; in the denominator, the current pollutant concentration and its time average are calculated. The absolute value of the difference between them characterizes the dynamic intensity of the change in pollutant concentration, and a minimum coefficient term is superimposed. To control the stability of the denominator, a global square root operation is performed to ensure that the influence of the dynamic disturbance term on the result exhibits a non-linear convergence trend. Finally, the standardized concentration difference term at all time points is summed and normalized to the average value. An absolute value operation is then performed to ensure that this feature is expressed as an intensity-based indicator, free from directional negative interference. The output Δt... ij It can be used for subsequent time zone pattern recognition and spatial node response analysis tasks;

[0138] The pollutant-ozone time difference characteristic value represents the average time delay in the impact of pollutant concentration changes at upstream nodes on ozone concentration at downstream nodes after being transmitted through the spectral path. This value comprehensively considers the weighted connection strength of pollutants during spatial propagation, the degree of concentration difference between pollutants and ozone response, and the volatility of pollutant concentration changes. Through standardization and normalization, it quantifies the dynamic trend of different nodes' effects on ozone formation. The larger the value, the stronger and more delayed the effect of upstream pollution sources on downstream ozone formation. It can be used to characterize the spatiotemporal coupling characteristics of pollution transmission between regions and identify key influence channels.

[0139] S402: Call the pollutant-ozone time difference sequence value, set the standard deviation threshold as the basis for judging the stability of the fluctuation, perform statistical analysis on the time difference sequence within a continuous time period, and if the standard deviation of the time difference value within the time period is less than the preset threshold, the time period is determined to be the stable interval of the response fluctuation, and the stable node sequence interval value of the response fluctuation is generated.

[0140] When calling the time difference sequence values ​​generated in the previous process, a judgment criterion needs to be set, namely the fluctuation amplitude threshold, to assess whether the changes in time differences are stable over a continuous period of time. This threshold can be set based on the historical range or statistical distribution of the time difference sequence. For example, twice the standard deviation can be used as the judgment boundary. If the standard deviation of a certain set of historical time difference sequences is 5 minutes, then the fluctuation amplitude threshold can be set to 10 minutes. Then, slide the time window one by one to check whether the difference between the maximum and minimum time differences in each window is within this range. For example, if the time differences in a certain period of time are 20, 22, 19, 21, and 20 minutes respectively... If the time difference is 3 minutes, which is less than the set threshold of 10 minutes, it indicates that the time difference change is stable within that time period. This judgment is performed for each monitoring node. When a node meets the stable fluctuation condition in multiple time periods, its stable intervals are summarized and marked. For example, if node X meets the condition between 9:00 and 10:30 and between 11:00 and 12:00, then these two time periods are recorded respectively, and a mapping record of node-stable time interval is constructed. After traversing all nodes, a set containing multiple nodes and their respective stable response time periods is obtained, providing interval basis for trend consistency analysis.

[0141] S403: Based on the interval values ​​of the stable node sequence of response fluctuations, analyze the trend direction and consistency of the concentration time difference between nodes, screen the nodes with cooperative response offset relationship, and obtain the set of cooperative response offset trend nodes.

[0142] Based on the set of nodes within stable intervals, further analysis is needed to determine the trend of time differences between nodes within their respective stable intervals. This requires using methods such as data fitting or moving averages to extract the trend direction. If the time difference continues to increase over a continuous period, the trend is upward; if it decreases, it is downward; if the change is insignificant, it can be determined that there is no obvious trend. For example, if a node's time differences are 15, 17, 19, 21, and 22 minutes between 10:00 and 11:00, this indicates a continuous upward trend. Similarly, if another node's time differences are 16, 18, 20, 22, and 23 minutes within the same interval, the trend is also upward. The two nodes are identified as having the same trend direction. Then, the similarity index of the degree of trend change between the two nodes is calculated, such as by comparing the slope difference. If the difference in the trend slope of the two nodes is less than a set limit, such as 1 unit / hour, they can be considered consistent. All node combinations with consistent trends are selected and a set of collaborative response offset nodes is constructed. For example, if the trends of nodes M and N are both upward and the difference in the rate of change is less than the limit, they can be identified as collaborative offset nodes. Finally, multiple node pairs with collaborative trend responses are obtained and recorded as offset trend node combinations for subsequent response relationship modeling and spatial expansion research.

[0143] Please see Figure 6 The specific steps of S5 are as follows:

[0144] S501: Obtain the pollutant release time and regional meteorological conditions in the cluster of nodes of the collaborative response offset trend. Combine the spatial location of the nodes with the characteristics of wind speed and direction to determine the dominant direction and path length of pollutant diffusion, and generate the diffusion path direction quantity and initial path length value.

[0145] When obtaining pollutant release times and regional meteorological conditions involved in the coordinated response offset trend node set, it is necessary to first collect pollutant concentration time series and spatial coordinate information from multiple monitoring points, and simultaneously collect meteorological parameters within the region during the corresponding time period, including wind speed, wind direction, temperature, and air pressure. Meteorological data can be obtained from ground automatic weather stations or high-resolution remote sensing data. The time axis of the monitoring points is synchronized with the meteorological records to construct a spatiotemporal data structure. The time points of sudden concentration increases in each monitoring point are identified. For example, at a certain monitoring point, if the PM2.5 average value in the previous hour was 55 and the current value is 85, the increase is 30, exceeding the set sudden change threshold of 20, and this time can be marked as a possible release node. The system summarizes and statistically analyzes the abrupt changes at multiple nodes, taking the time with the highest frequency as the starting point of pollution release. Further, it extracts the average wind speed and direction information for approximately one hour before and after this time, and combines this with the node coordinates to calculate the projected path length of each node along the wind direction. The longest projected direction is selected as the dominant diffusion direction, while the minimum projected distance in that direction is recorded as the initial diffusion path length value. For example, if a node has coordinates of 100 in the X direction and 50 in the Y direction, assuming the dominant wind direction is 45 degrees, and the source point coordinates are 90 in the X direction and 40 in the Y direction, the path length is estimated to be approximately 14 meters based on spatial projection relationships. The final output is the wind direction of 45 degrees and the initial path length of 14 meters.

[0146] S502: Based on the diffusion path direction and initial path length, combined with the monitoring point location and pollutant detection time, analyze the changing trends of wind direction and wind speed in the release path, screen nodes that meet the wind field stability characteristics, and obtain a set of wind field consistency offset nodes.

[0147] Based on the dominant diffusion direction and initial path length, several sampling profile segments with 10-meter intervals are sequentially divided along the diffusion direction from the pollution source location. For each intersecting monitoring point within a profile segment, pollutant detection time points and corresponding wind speed and direction data are extracted. The wind direction and speed information are arranged chronologically, and the wind direction and speed changes between adjacent monitoring points are compared pairwise. If the wind direction change is less than 10 degrees and the wind speed change is less than 0.5 m / s, the wind field in that segment is considered stable and can be identified as a wind field consistency node. This process is performed on all profile segments. Finally, all nodes meeting the above conditions are selected as the set of wind field consistency offset nodes. Taking two nodes as an example, one has a wind direction of 50 degrees and a wind speed of 2.0 m / s, and the other has a wind direction of 55 degrees and a wind speed of 2.4 m / s. The wind direction difference is 5 degrees and the wind speed difference is 0.4 m / s, both less than the set standard. Therefore, the second node can be identified as a consistency node. The entire diffusion path is screened segment by segment, forming a continuous and wind field consistent offset path point set.

[0148] S503: Based on the set of wind field consistency offset nodes, combined with the time interval and path length between nodes and monitoring points, determine the spatiotemporal coordination relationship, select nodes that meet the conditions as the starting source of pollution events, and obtain the origin node identification list;

[0149] For each node in the wind field consistency offset node set, the spatial distance between it and the downstream monitoring point is calculated sequentially, and the time difference between the node and the monitoring point is extracted. Based on the average wind speed between the two points, the time required for pollutants to spread in the wind is estimated. This propagation time estimate is compared with the actual detection time difference. If the error between the two is within the allowable range, such as within 2 minutes, then the node can be considered to match the target monitoring point in time and space and has the potential to be the starting point of the pollution source. It is recorded as a candidate node for the origin of the pollution event. For example, if a node is about 32 meters away from the downstream point and the average wind speed between the two points is about 2.1 m / s, the estimated time required for diffusion to the point is about 15 seconds. However, the monitoring data shows that the detection time of the node is about 14 seconds earlier. The difference between the two is 1 second, which is less than the allowable range. Therefore, it is confirmed that the node can be added to the list of pollution origin node. Finally, all nodes that meet the dual matching conditions of distance and time are selected to form the identification result set.

[0150] The dominant direction of pollutant diffusion refers to the main direction of pollutant propagation during its diffusion in the air, which is influenced by factors such as wind speed, wind direction, and geographical location.

[0151] The diffusion path length refers to the distance from the pollutant release node to the monitored node along the prevailing wind direction.

[0152] Please see Figure 7 A knowledge graph-based collaborative tracing system for ozone precursors includes:

[0153] The precursor channel identification module acquires the concentration change information of precursor nodes in the pollution map under the pollution change monitoring scenario. Combined with the path connection relationship between nodes, it determines whether the response amplitude of precursors in the path exceeds the set change ratio. Paths with change amplitude exceeding the set change ratio are identified as key pollution response channels, and a set of abnormal pollution response paths is obtained.

[0154] The path scoring generation module collects the frequency of pollution factor changes, the participation of synergistic pollutants, and the concentration fluctuation range of path nodes based on the abnormal path set of pollution response. After extracting the associated indicators, it performs standardization processing to obtain the graph edge path scoring reference table.

[0155] The intensity attribute adjustment module adjusts the connection intensity attribute of the spectrum path according to the spectrum edge path scoring reference table, matches the ozone response change degree in the current monitoring period, and writes it into the spectrum structure to obtain the real-time adjustment result of the spectrum structure.

[0156] The offset trend extraction module adjusts the results in real time according to the spectral structure, extracts the pollutant concentration sequence and ozone concentration sequence of the upstream node, analyzes the degree of time offset between the two, and screens the nodes that meet the stability characteristics of the amplitude range to obtain the set of cooperative response offset trend nodes.

[0157] The origin node identification module reads the pollutant release time and regional meteorological conditions of the nodes based on the collaborative response offset trend node set, evaluates the temporal rationality of the path, and selects nodes that meet the conditions as the starting position of the pollution event to obtain the origin node identification list in the map.

[0158] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A knowledge graph-based collaborative source tracing method for ozone precursors, characterized in that, Includes the following steps: S1: Obtain the concentration change information of precursor nodes in the map under the pollution change monitoring scenario. Combine the path connection relationship between nodes to judge the trend of precursor response amplitude. Identify the path with the change amplitude exceeding the set proportion as the key channel of pollution response and obtain the pollution response abnormal path set. S2: Based on the pollution response anomaly path set, collect the pollution factor change frequency, synergistic pollutant participation and concentration fluctuation range of the nodes connected by the path, extract the indicators and perform standardization processing to obtain the graph edge path scoring reference table; S3: Based on the edge path scoring reference table of the graph, match the degree of ozone response change in the current monitoring period, adjust the connection strength attribute of the corresponding path in the graph, and write the updated value into the graph structure to obtain the real-time adjustment result of the graph structure; S4: Based on the real-time adjustment results of the graph structure, extract the pollutant concentration change sequence and ozone concentration sequence of the upstream nodes, analyze the degree of offset between the pollutant concentration change sequence and the ozone concentration sequence in the time series, determine whether the response between nodes is continuously within the fluctuation range limit in the time period, and obtain the collaborative response offset trend node set.

2. The knowledge graph-based collaborative tracing method for ozone precursors according to claim 1, characterized in that, The pollution response anomaly path set includes paths with significant changes in precursor concentration, inter-node connection trend paths, and key pollution response channels. The graph edge path scoring reference table includes the frequency of pollutant changes, the participation of synergistic pollutants, and the concentration fluctuation range. The real-time adjustment results of the graph structure include changes in path connection strength, adjusted graph edge attributes, and updated graph topology. The synergistic response offset trend node set includes nodes with stable concentration sequence offsets, nodes with continuous responses within a time period, and nodes with limited fluctuation ranges.

3. The knowledge graph-based collaborative tracing method for ozone precursors according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Obtain the concentration change data of precursor nodes in the spectrum under the monitoring scenario, combine the path connection relationship between nodes, organize the concentration change information of adjacent nodes on each path, extract the response amplitude change of the corresponding nodes of the path, and generate the response amplitude ratio matrix. S102: Based on the response amplitude changes of the paths in the response amplitude ratio matrix, and combined with the set response amplitude change ratio threshold, identify the paths whose change amplitude exceeds the threshold, and obtain a set of path sequences that exceed the threshold. The threshold for the change in response amplitude is determined based on a statistical analysis method of historical monitoring data. Specifically, all response amplitude ratios recorded in the target area within a recent period (e.g., the last 3 months) are sampled, sorted in ascending order, and the value at the 90th percentile is selected as the default threshold reference (i.e., the paths with the highest change amplitudes are selected as abnormal paths). If sensitivity needs to be increased, it can be adjusted to the 85th percentile; if stability needs to be enhanced, it can be adjusted to the 95th percentile. The threshold can also be dynamically adjusted according to the pollution type, pollution scenario, and response granularity, either by preset or manual setting. S103: Call the path sequence set that exceeds the threshold, identify adjacent connected path pairs in the graph topology, and determine whether their response amplitudes simultaneously exceed the set threshold. If the condition is met, the path pair is determined to be a path with an associated response. Extract the set of paths where the ratio of the response amplitudes of the path segments between two or more consecutive nodes all exceeds the threshold, as the pollution response abnormal path set.

4. The knowledge graph-based collaborative tracing method for ozone precursors according to claim 3, characterized in that, The specific steps of S2 are as follows: S201: Based on the pollution response anomaly path set, collect the pollution factor change frequency, synergistic pollutant participation and concentration fluctuation range of the path connection nodes, extract the number of pollution factor changes, the number of synergistic pollutant types and the concentration fluctuation range of the nodes, and generate a node pollution characteristic parameter value set. S202: Based on the set of node pollution characteristic parameters, integrate the difference in the frequency of pollution factor changes and the difference in the total amount and concentration fluctuation of synergistic pollutants between path connection nodes, normalize the three parameters respectively, and generate a path normalization characteristic coefficient table. The difference in the frequency of pollution factor changes refers to the absolute value of the difference between the frequency of pollution factor changes per unit time between the starting node and the ending node of the path. The total number of co-polluting pollutant types is the arithmetic sum of the number of pollutant types participating in co-polluting in the two nodes connected by the path; The concentration fluctuation difference is the absolute value of the difference between the upper limits of the pollution factor concentration fluctuation ranges at the start and end nodes of the path within a given time monitoring period. S203: Based on the path normalization characteristic coefficient table, the normalization coefficient groups of the frequency of change of pollutants, the types of synergistic pollutants and the concentration fluctuation values ​​are weighted and integrated to extract the score value of each path and generate a graph edge path score reference table. The three indicators are weighted and summed using preset weighting coefficients, and the weight values ​​can be determined based on experience in pollution characteristic analysis.

5. The knowledge graph-based collaborative tracing method for ozone precursors according to claim 4, characterized in that, The specific steps for S3 are as follows: S301: Based on the map edge path scoring reference table, call the ozone response change rate data in the current monitoring period, calculate the response change amplitude of each path, compare it with the set change amplitude benchmark value, identify paths with change amplitude exceeding the benchmark value as key paths, classify them according to the change amplitude level, and generate ozone change path intensity grading values. S302: Based on the ozone change path intensity classification value, match the path connection strength parameters and adjustment coefficients in the map, update the connection strength parameters, and generate a path connection strength adjustment coefficient set; S303: Call the path connection strength adjustment coefficient set, write the updated connection strength parameters into the graph structure path attribute, update the connection strength attribute of the graph path, and obtain the real-time connection strength value set of the graph structure.

6. The knowledge graph-based collaborative tracing method for ozone precursors according to claim 5, characterized in that, The specific steps of S4 are as follows: S401: Based on the real-time connection intensity value set of the spectrum structure, obtain the pollutant concentration sequence of the upstream and midstream nodes and the ozone concentration sequence of the corresponding nodes, arrange the two types of sequences based on the time index and establish the correlation, identify the time difference characteristics of concentration changes between nodes, and generate pollutant-ozone time difference sequence values. S402: Call the pollutant-ozone time difference sequence value, set the standard deviation threshold as the basis for judging the stability of the fluctuation, perform statistical analysis on the time difference sequence in a continuous time period, and if the standard deviation of the time difference value in the time period is less than the preset threshold, the time period is determined to be the stable interval of the response fluctuation, and the stable node sequence interval value of the response fluctuation is generated. S403: Based on the interval values ​​of the stable node sequence of the response fluctuation, analyze the trend direction and consistency of the concentration time difference between nodes, screen the nodes with cooperative response offset relationship, and obtain the cooperative response offset trend node set.

7. The knowledge graph-based collaborative tracing method for ozone precursors according to claim 6, characterized in that, The method further includes: S5: Based on the collaborative response offset trend node set, read the pollutant release time and regional meteorological conditions of the nodes, combine the node spatial location information and wind field data, use the atmospheric diffusion model to estimate the shortest transmission path and time of pollutants from the release node to the monitoring point, compare the transmission time with the actual monitoring time, judge the temporal consistency and spatial rationality of the path, and screen the nodes that meet the following conditions as the starting source of the pollution event, and generate an origin node identification list in the map; The origin node identification list within the map includes pollutant release time sequence nodes, nodes that meet meteorological conditions, and nodes with reasonable source paths.

8. The knowledge graph-based collaborative tracing method for ozone precursors according to claim 7, characterized in that, The specific steps of S5 are as follows: S501: Obtain the pollutant release time and regional meteorological conditions in the cluster of nodes of the collaborative response offset trend. Combine the spatial location of the nodes with the characteristics of wind speed and direction to determine the dominant direction and path length of the pollutant diffusion and generate the diffusion path direction quantity and the initial path length value. S502: Based on the diffusion path direction and initial path length, combined with the monitoring point location and pollutant detection time, analyze the changing trends of wind direction and wind speed in the release path, screen nodes that meet the wind field stability characteristics, and obtain a set of wind field consistency offset nodes. S503: Based on the set of wind field consistency offset nodes, combined with the time interval and path length between the nodes and the monitoring points, determine the spatiotemporal coordination relationship, select nodes that meet the conditions as the starting source of the pollution event, and obtain the origin node identification list.

9. The knowledge graph-based collaborative tracing method for ozone precursors according to claim 1, characterized in that, The set ratio refers to the threshold of the response amplitude change ratio obtained based on the statistical analysis of monitoring data. For example, by analyzing the distribution of the response amplitude ratio in the concentration change path, the 90th percentile ratio is determined as the threshold reference value, which is used to identify the path with the change amplitude exceeding the set ratio as the key channel for pollution response.

10. A knowledge graph-based collaborative source tracing system for ozone precursors, characterized in that, The knowledge graph-based collaborative tracing method for ozone precursors according to any one of claims 1-9, wherein the system comprises: The precursor channel identification module acquires the concentration change information of precursor nodes in the pollution map under the pollution change monitoring scenario. Combined with the path connection relationship between nodes, it determines whether the response amplitude of precursors in the path exceeds the set change ratio. Paths with change amplitude exceeding the set change ratio are identified as key pollution response channels, and a set of abnormal pollution response paths is obtained. The path scoring generation module collects the frequency of pollution factor changes, the participation of synergistic pollutants, and the concentration fluctuation range of path nodes based on the pollution response anomaly path set. After extracting the associated indicators, it performs standardization processing to obtain the graph edge path scoring reference table. The intensity attribute adjustment module adjusts the connection intensity attribute of the spectrum path according to the spectrum edge path scoring reference table, matches the ozone response change degree in the current monitoring period, and writes it into the spectrum structure to obtain the real-time adjustment result of the spectrum structure. The offset trend extraction module adjusts the results in real time according to the spectrum structure, extracts the pollutant concentration sequence and ozone concentration sequence of the upstream node, analyzes the degree of time offset between the two, and screens the nodes that meet the stability characteristics of the amplitude range to obtain the set of cooperative response offset trend nodes. The origin node identification module reads the pollutant release time and regional meteorological conditions of the nodes based on the collaborative response offset trend node set, evaluates the temporal rationality of the path, and selects nodes that meet the conditions as the starting position of the pollution event to obtain the origin node identification list in the map.

Images