Pollution diffusion path modeling method and system based on multiple sensors

Abstract

The invention relates to the technical field of computer modeling, and provides a pollution diffusion path modeling method and system based on multiple sensors, which are used for improving the accuracy and adaptability of pollution diffusion path modeling. The method comprises the following steps: synchronously acquiring original observation data through a multi-source sensor deployed in a pollution monitoring area, and establishing a data association index based on a spatial arrangement position and a data acquisition timestamp of the multi-source sensor; extracting spatio-temporal characteristics representing gas migration in combination with the data association index and the physical diffusion characteristics of the pollution gas; establishing a basic model structure of a gas migration model based on spatial-temporal characteristics, analyzing and determining a correlation weight in combination with characteristic responsivity, and generating a parameter self-learning strategy; starting a parameter self-learning strategy, and iteratively updating the parameters by using a particle swarm search space to obtain optimized parameters; and importing the optimized parameters into a gas migration model for inversion calculation, determining initial distribution of path nodes in combination with topographic features and meteorological conditions, performing iterative updating, and outputting a pollution diffusion inversion path.

Description

Technical Field

[0001] This application belongs to the field of computer modeling technology, specifically relating to a pollution diffusion path modeling method and system based on multiple sensors. Background Technology

[0002] With the acceleration of industrial production and urbanization, frequent incidents of pollutant gas leaks pose a serious threat to the ecological environment and human health. Therefore, accurately constructing pollution diffusion pathways has become a key research focus in the field of pollution prevention and control.

[0003] Currently, pollution diffusion path modeling mainly relies on multi-sensor data acquisition and model simulation technologies. For multi-sensor data acquisition, existing technologies typically deploy multiple sensors in the pollution monitoring area, such as gas concentration sensors, temperature and humidity sensors, and wind speed and direction sensors, to simultaneously collect data on pollutant gas concentration, ambient temperature and humidity, wind speed and direction. The collected data is transmitted to a data processing center for storage and analysis via wireless communication technology. Regarding model simulation technologies, commonly used gas migration models include Gaussian diffusion models and Eulerian-Lagrange models. These models are based on fluid mechanics and atmospheric diffusion theory, combined with the physical diffusion characteristics of pollutants. By inputting environmental parameters collected by sensors and initial gas concentrations, they simulate the diffusion process of pollutants and predict pollution diffusion paths.

[0004] However, in practical applications, existing technologies suffer from weak spatiotemporal correlation of data collected by multi-source sensors, making it difficult to effectively support accurate model simulation. Furthermore, the fixed parameters of gas migration models result in poor adaptability to complex pollution diffusion scenarios, leading to low accuracy in pollution diffusion path inversion. Summary of the Invention

[0005] This application provides a multi-sensor-based method and system for modeling pollution diffusion paths, which improves the accuracy and adaptability of pollution diffusion path modeling.

[0006] This application provides a multi-sensor-based pollution diffusion path modeling method, applied to a pollution diffusion path modeling system. The method includes: Raw observation data is collected synchronously by multi-source sensors deployed in the pollution monitoring area. A data association index is established based on the spatial deployment location of the multi-source sensors and the data acquisition timestamp. The data association index is used to establish the spatiotemporal correspondence between different raw observation data. By combining the data association index and the physical diffusion characteristics of polluting gases, spatiotemporal features characterizing gas migration are extracted from the original observation data. These spatiotemporal features include the spatial distribution gradient and temporal variation law of gas concentration. Based on the aforementioned spatiotemporal features, a basic model structure for the gas migration model is established. The mapping relationship between the original observation data and the spatiotemporal features is associated through the basic model structure. The correlation weight between the input and output of the gas migration model is determined by combining feature responsivity analysis, and a parameter self-learning strategy for the gas migration model is generated. The parameter self-learning strategy is initiated and the particle swarm search space is initialized. The spatiotemporal features are input into the gas migration model to obtain the initial parameter combination. The parameter fitness value is calculated by iteratively updating the position of the particle swarm and the target parameters of the gas migration model are updated based on the parameter fitness value. After multiple iterations, the optimal parameter combination is obtained as the optimization parameters. The optimized parameters are imported into the gas migration model to drive the inversion calculation of the current pollution diffusion path. The initial distribution of path nodes is determined by combining the topographic features and meteorological conditions of the pollution monitoring area. The node positions and correlations are adjusted by node concentration error feedback to iteratively update the path nodes in the current pollution diffusion path. The pollution diffusion inversion path is output based on the updated path node information. The current pollution diffusion path and the pollution diffusion inversion path both correspond to the target monitoring period.

[0007] This application provides a pollution diffusion path modeling system, which includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the above-described method.

[0008] This application provides a computer-readable storage medium including a computer program. When the computer program is run on a pollution diffusion path modeling system, the computer program is used to cause the pollution diffusion path modeling system to perform the steps of the above-described method.

[0009] This application embodiment synchronously collects raw observation data using multi-source sensors deployed in the pollution monitoring area. A data association index is established by combining the spatial deployment location of the multi-source sensors and the data acquisition timestamps, integrating multi-dimensional and multi-temporal observation data. This solves the problems of scattered data and weak spatiotemporal correlation in traditional pollution monitoring. Furthermore, by combining the data association index with the physical diffusion characteristics of pollutant gases, spatiotemporal features characterizing gas migration are extracted, ensuring the relevance and accuracy of feature extraction. Based on these spatiotemporal features, a basic model structure for the gas migration model is established. The correlation weights between input and output are determined through feature response analysis, and a parameter self-learning strategy is generated. The design achieves dynamic optimization of model parameters, improving the model's adaptability to complex pollution diffusion scenarios. A parameter self-learning strategy is initiated, and parameters are iteratively updated using a particle swarm optimization space. Fitness values ​​drive target parameter optimization, ensuring optimal model parameters and enhancing prediction accuracy. Finally, the optimized parameters are imported into a gas migration model for inversion calculations. The initial distribution of path nodes is determined by combining terrain features and meteorological conditions, and nodes are iteratively updated through error feedback, outputting accurate pollution diffusion inversion paths. This achieves efficient and accurate inversion of pollution diffusion paths during the target monitoring period, providing a scientific basis for pollution source tracing and risk prevention. This design improves the accuracy, efficiency, and adaptability of pollution diffusion path modeling. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating a multi-sensor-based pollution diffusion path modeling method provided in an embodiment of this application.

[0011] Figure 2 This is a schematic diagram of a pollution diffusion path modeling system provided in an embodiment of this application.

[0012] Figure 3 This is a functional block diagram of a pollution diffusion path modeling system provided in an embodiment of this application. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this application. Obviously, the described embodiments are only some embodiments of the technical solutions of this application, and not all embodiments. Based on the embodiments recorded in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the technical solutions of this application.

[0014] See Figure 1 This is a pollution diffusion path modeling method based on multiple sensors provided in the embodiments of this application. This method can be applied to a pollution diffusion path modeling system. The specific process is as follows: steps 110-150.

[0015] Step 110: The pollution diffusion path modeling system synchronously collects raw observation data through multi-source sensors deployed in the pollution monitoring area. Based on the spatial deployment location of the multi-source sensors and the data acquisition timestamp, a data association index is established. The data association index is used to establish the spatiotemporal correspondence between different raw observation data.

[0016] The pollution monitoring area is a closed monitoring area within an urban industrial park. Multi-source sensors, including gas concentration sensors, temperature and humidity sensors, and wind speed and direction sensors, are deployed in a grid pattern within the area. The spatial layout of each type of sensor is based on a preset coordinate origin within the area and is identified by coordinate codes. For example, the coordinate codes for gas concentration sensors are S1 to S100, and each sensor corresponds to a unique spatial coordinate. The data acquisition timestamp records the time when the sensor collects data in milliseconds. For example, a gas concentration sensor collects concentration data at time T1, and another wind speed and direction sensor collects wind speed and direction data at time T1.

[0017] After collecting raw observation data, the pollution diffusion path modeling system binds the spatial deployment coordinates of each sensor with the data acquisition timestamp to establish a data association index. Each record in the index structure contains the sensor type, spatial coordinates, timestamp, and corresponding raw observation data. Through this index, sensor data at different spatial locations at the same timestamp can be quickly queried, or observation data of the same sensor at different timestamps can be obtained, thereby clarifying the spatiotemporal correspondence between different raw observation data. For example, the gas concentration data, wind speed data, and temperature and humidity data of sensors S1 to S100 at time T1 can be obtained through the index, and the spatial distribution association of the above data at the same time point can be determined.

[0018] Step 120: The pollution diffusion path modeling system combines data association indexes and the physical diffusion characteristics of pollutant gases to extract spatiotemporal features characterizing gas migration from the original observation data. The spatiotemporal features include the spatial distribution gradient and temporal variation law of gas concentration.

[0019] The physical diffusion characteristics of pollutants include their diffusion coefficient, density, and mixing properties with air. The pollution diffusion path modeling system, based on a data association index, first extracts gas concentration data from different spatial locations at the same time stamp. By calculating the difference in gas concentration between adjacent spatial locations, the spatial distribution gradient of gas concentration is obtained. For example, calculating the gas concentration difference between sensors S1 and S2 at time T1 reflects the concentration gradient in that direction. Next, gas concentration data from sensors at the same spatial location at consecutive time stamps is extracted. By analyzing the trend of concentration data over time, the temporal variation pattern is obtained. For example, analyzing the concentration data of sensor S1 at consecutive time stamps from T1 to T10 determines whether the concentration is increasing, decreasing, or remaining stable. The extracted spatiotemporal features are stored in the form of a multidimensional array. Each array element contains spatial coordinates, a time stamp, a spatial distribution gradient value, and a temporal variation feature value, forming a set of spatiotemporal features characterizing gas migration.

[0020] Step 130: The pollution diffusion path modeling system establishes the basic model structure of the gas migration model based on spatiotemporal characteristics. The mapping relationship between the original observation data and spatiotemporal characteristics is associated through the basic model structure. The correlation weight between the input and output of the gas migration model is determined by combining feature response analysis, and the parameter self-learning strategy of the gas migration model is generated.

[0021] The pollution diffusion path modeling system first constructs a neural network model, including an input layer, hidden layers, and an output layer, based on the spatial distribution gradient and temporal variation patterns contained in the spatiotemporal features. The input layer receives spatiotemporal feature data, the hidden layer uses several neurons to perform feature transformation, and the output layer outputs the predicted results related to gas migration. Then, using the basic model structure, the gas concentration data from the original observation data is used as the output label, and the spatiotemporal features are used as the input features. The model is trained to establish a mapping relationship between the two; for example, inputting the spatiotemporal features at time T1 into the model will output the corresponding predicted gas concentration value. The model parameters are adjusted by comparing them with the actual concentration values ​​in the original observation data. Then, feature response analysis is performed by changing the value of each feature in the spatiotemporal features one by one and observing the degree of change in the model output. For example, by changing the spatial distribution gradient value, the change in the concentration prediction value of the model output is recorded. The correlation weight between the feature and the output is determined according to the degree of change, and the weight value is assigned to the corresponding input feature. Finally, based on the correlation weight, a parameter self-learning strategy is generated. The strategy specifies the update method, learning rate, and number of iterations of the model parameters. For example, when the correlation weight of a certain feature is high, the adjustment range of the model parameters corresponding to that feature is increased during the parameter update process.

[0022] Step 131: The pollution diffusion path modeling system decomposes the spatiotemporal characteristics into spatial distribution characteristics and temporal variation characteristics. The spatial distribution characteristics reflect the spatial gradient and distribution pattern of the concentration of pollutants in the pollution monitoring area, while the temporal variation characteristics reflect the fluctuation pattern and trend of gas concentration at the calibration location over time.

[0023] The pollution diffusion path modeling system decomposes the spatiotemporal features extracted in step 120. Spatial distribution features include gas concentration gradient values ​​at different spatial locations at the same time stamp, concentration distribution uniformity indicators, and the distribution range of high-concentration areas. For example, the concentration gradient values ​​of sensors S1 to S100 at time T1 constitute part of the spatial distribution features. Temporal variation features include the concentration change rate of sensors at the same spatial location at consecutive time stamps, the time of concentration peak occurrence, and the duration of concentration stability. For example, the concentration change rate of sensor S1 from time T1 to T10 constitutes part of the temporal variation features. The decomposed spatial distribution features and temporal variation features are stored separately.

[0024] Step 132: The pollution diffusion path modeling system classifies the raw observation data according to the collection time period and sensor type, and establishes a two-way mapping relationship between the classified raw observation data and spatial distribution characteristics and temporal change characteristics.

[0025] The pollution diffusion path modeling system categorizes raw observation data into time periods such as T1-T5 and T6-T10, and into sensor types such as gas concentration sensor data and wind speed and direction sensor data. Then, for each type of data, a bidirectional mapping is established between spatial distribution characteristics and temporal variation characteristics. For example, gas concentration sensor data for time period T1-T5 is mapped to the spatial distribution characteristics and temporal variation characteristics of that time period. This allows for quick querying of corresponding spatiotemporal characteristics using raw observation data, and conversely, tracing back to the corresponding raw observation data using spatiotemporal characteristics. For instance, querying gas concentration data for time period T1-T5 simultaneously yields the spatial distribution gradient and temporal variation patterns of that time period.

[0026] Step 133: The pollution diffusion path modeling system constructs the basic model structure of the gas migration model based on the bidirectional mapping relationship. The basic model structure includes a feature input layer, a parameter calculation layer, and a result output layer. The feature input layer is used to receive spatiotemporal features, the parameter calculation layer is used to perform model calculations, and the result output layer is used to output pollution diffusion-related parameters.

[0027] The model comprises several layers. The feature input layer has multiple input nodes, each corresponding to different dimensions of spatial distribution and temporal variation features. For example, nodes receive spatial distribution gradient values ​​and concentration uniformity indices. The parameter computation layer contains several neurons connected by weights, performing linear transformations and nonlinear activation on the input spatiotemporal features. For instance, matrix multiplication transforms the input features, followed by activation using the sigmoid function. The output layer has output nodes that output pollution diffusion-related parameters, such as the velocity, direction, and predicted concentration distribution of gas diffusion. The basic model structure ensures accurate correlation between the input spatiotemporal features and the original observation data through a bidirectional mapping relationship, and the model calculations are based on real observation data.

[0028] Step 134: The pollution diffusion path modeling system adjusts the spatiotemporal features as model input values ​​one by one using the control variable method, monitors the changes in the predicted values ​​of pollutant gas concentration distribution output by the gas migration model, normalizes the degree of change in the predicted values ​​caused by adjusting each feature individually, and obtains the normalized influence coefficients corresponding to each spatiotemporal feature. Based on the normalized influence coefficients, the correlation weights between each spatiotemporal feature and the model output are determined.

[0029] The pollution diffusion path modeling system selects a specific spatiotemporal feature, such as the spatial distribution gradient value. While keeping other features constant, it adjusts the value of this feature, for example, from the initial value to 1.2 times the initial value. It monitors the change in the predicted value of pollutant gas concentration distribution output by the model and calculates the ratio of the change to the initial predicted value to obtain the influence degree of the feature. The above operation is repeated for each spatiotemporal feature. After obtaining the influence degree of all features, normalization is performed to convert the influence degree into a normalized influence coefficient between 0 and 1. For example, if the influence degree of a feature is 0.5, the normalized coefficient is 0.5. Finally, the normalized influence coefficient is used as the correlation weight between each spatiotemporal feature and the model output. The higher the weight value, the greater the influence of the feature on the model output. For example, the normalized influence coefficient of the spatial distribution gradient is 0.6, indicating that it has a large influence on the concentration distribution prediction value.

[0030] Step 135: The pollution diffusion path modeling system generates parameter adjustment rules by combining correlation weights and bidirectional mapping relationships, and generates an initial parameter self-learning strategy for the gas migration model by combining the parameter adjustment rules.

[0031] The parameter adjustment rules specify how model parameters are adjusted based on correlation weights. For example, model parameters corresponding to features with high correlation weights will be adjusted more significantly during parameter updates. A bidirectional mapping relationship ensures that parameter adjustments are based on the accurate correlation between the original observation data and spatiotemporal features. For instance, when adjusting model parameters corresponding to a certain feature, the original observation data can be traced back through bidirectional mapping to verify the rationality of the parameter adjustment. The pollution diffusion path modeling system generates an initial parameter self-learning strategy based on these rules. The strategy includes the frequency of parameter adjustments, the maximum magnitude of each adjustment, and the parameter verification method. For example, the strategy stipulates that model parameters are adjusted once every 10 iterations, with each adjustment not exceeding 0.1 times the current parameter value. After adjustment, the effectiveness of the parameters is verified using the original observation data.

[0032] Step 136: The pollution diffusion path modeling system uses some original observation data and corresponding spatiotemporal characteristics as validation data, inputs them into the gas migration model, and starts the initial parameter self-learning strategy. Based on the deviation between the model output after the strategy is executed and the actual monitoring results, the amplitude coefficient in the parameter adjustment rule is corrected to generate the parameter self-learning strategy of the gas migration model.

[0033] The pollution diffusion path modeling system selects 20% of the total original observation data as validation data, such as the original observation data from time T11 to T20 and the corresponding spatiotemporal characteristics. The validation data is input into the gas migration model, and the initial parameter self-learning strategy is initiated to adjust the parameters. After the strategy is executed, the deviation between the predicted value of the pollutant gas concentration distribution output by the model and the actual monitoring results in the validation data is calculated, such as the mean square error between the predicted value and the actual value. The magnitude coefficient in the parameter adjustment rule is corrected according to the magnitude of the deviation. If the deviation is large, the magnitude coefficient of the parameter adjustment is increased; if the deviation is small, the magnitude coefficient is decreased. The corrected parameter adjustment rule is combined with other contents in the initial strategy to generate the final parameter self-learning strategy, ensuring that the self-learning process of the model parameters can effectively reduce the prediction deviation.

[0034] Step 140: The pollution diffusion path modeling system starts the parameter self-learning strategy and initializes the particle swarm search space. It inputs the spatiotemporal features into the gas migration model to obtain the initial parameter combination. It calculates the parameter fitness value by iteratively updating the position of the particle swarm and drives the update of the target parameters of the gas migration model based on the parameter fitness value. After multiple rounds of iteration, the optimal parameter combination is obtained as the optimization parameters.

[0035] The pollution diffusion path modeling system first initiates a parameter self-learning strategy. Based on the parameter range specified in the strategy, it initializes the particle swarm search space. Each particle in the swarm represents a parameter combination of a gas migration model, and the particle's position coordinates correspond to the parameter values. For example, the position coordinates of particle P1 correspond to the weight values ​​of neurons in the hidden layer of the model. Next, the spatiotemporal features are input into the gas migration model, and the model calculates based on the initial parameter combination to obtain the output results corresponding to the initial parameter combination. Then, through iterative updates of the particle swarm's position, the fitness value of each particle is calculated. The fitness value is determined based on the deviation between the model output and the original observation data; the smaller the deviation, the higher the fitness value. The fitness value drives the update of the target parameters. For example, the parameter combination corresponding to the particle with the high fitness value is selected to update the model's target parameters. After multiple iterations, when the fitness value no longer significantly improves, the parameter combination corresponding to the particle with the highest fitness value is selected as the optimization parameter.

[0036] Step 141: The pollution diffusion path modeling system starts the parameter self-learning strategy, reads the parameter adjustment rules and feature association weights contained in the parameter self-learning strategy, initializes the particle swarm search space based on the parameter value range of the gas migration model, each particle corresponds to a set of model parameter combinations, and the position coordinates of the particles are mapped to parameter values.

[0037] The parameter adjustment rules in the parameter self-learning strategy include the adjustment method and magnitude of the parameters. The feature association weights are the association weights between each spatiotemporal feature and the model output determined in step 134. The parameter value range of the gas migration model is set according to the model type and actual application requirements. For example, the weight value range of the hidden layer neurons is -1 to 1. Based on this value range, the pollution diffusion path modeling system initializes the particle swarm search space. The particle swarm size is set to N, and the position coordinates of each particle are randomly generated within the value range. For example, the position coordinates of particle P1 are [w1, w2, ..., wn], where wi is the value of the i-th parameter in the model, corresponding to the weight value of the hidden layer neurons, thereby realizing the mapping between particle position coordinates and parameter values.

[0038] Step 142: The pollution diffusion path modeling system inputs the spatiotemporal characteristics into the gas migration model, calls the initial parameter combination to perform calculations, obtains the initial pollution diffusion simulation results, and extracts the concentration distribution data and time series change data from the initial pollution diffusion simulation results as the initial output.

[0039] The spatiotemporal features are input into the gas migration model in the form of a multidimensional array. The model performs calculations on the input features based on the weights and biases in the initial parameter combination, such as through matrix multiplication and activation functions, to obtain the initial pollution diffusion simulation results. The simulation results include gas concentration distribution data at different spatial locations and timestamps, as well as the temporal variation data of concentration over time. The pollution diffusion path modeling system extracts the above data as the initial output, for example, extracting the concentration distribution data of each spatial location from time T1 to T10 to form the initial output dataset.

[0040] Step 143: The pollution diffusion path modeling system uses the original observation data as a benchmark to calculate the errors in spatial concentration distribution and temporal variation patterns between the initial output and the original observation data. The spatial concentration distribution error and the temporal variation pattern error are normalized to obtain spatial error index and temporal error index. Based on the spatial error index, the temporal error index and their respective weights determined by the characteristic response degree analysis, a comprehensive fitness function is constructed and the initial fitness value is calculated.

[0041] The spatial concentration distribution error is obtained by calculating the mean square error between the concentration data at each spatial location in the initial output and the corresponding concentration data at the original observation data. The temporal variation error is obtained by calculating the deviation between the temporal variation trend of the concentration data at each spatial location in the initial output and the corresponding trend in the original observation data. These two errors are normalized to convert the error values ​​into indices between 0 and 1, resulting in spatial error indices and temporal error indices. The weights determined by the feature response analysis are the spatial error index weight W1 and the temporal error index weight W2. The comprehensive fitness function is F = 1 - (W1 × spatial error index + W2 × temporal error index). The initial fitness value is calculated using this function. The closer the fitness value is to 1, the smaller the deviation between the model output and the original observation data.

[0042] Step 144: The pollution diffusion path modeling system determines the iteration direction and step size of the particle swarm based on the adjustment rules in the parameter self-learning strategy and the difference between the initial fitness value and the preset standard, and updates the position coordinates of each particle to obtain a new parameter combination.

[0043] The adjustment rules in the parameter self-learning strategy include the correspondence between fitness values ​​and iteration directions. For example, when the initial fitness value is lower than the preset standard, the iteration direction is to increase the parameter value to improve the fitness value. The difference between the initial fitness value and the preset standard is obtained by calculating the difference between the two. For example, if the preset standard is 0.8 and the initial fitness value is 0.6, the difference is 0.2. The step size is determined according to the difference, and the larger the difference, the larger the step size. The pollution diffusion path modeling system updates the position coordinates of each particle based on the iteration direction and step size. For example, the position coordinates of particle P1 move by the step size distance in the iteration direction to obtain new position coordinates, which correspond to a new parameter combination.

[0044] Step 1441: The pollution diffusion path modeling system analyzes the adjustment rules in the parameter self-learning strategy, extracts the parameter adjustment direction, adjustment ratio and constraint conditions corresponding to different fitness value ranges, and establishes a matching table between fitness values ​​and parameter adjustment rules.

[0045] The parameter self-learning strategy divides fitness values ​​into multiple intervals, such as 0 to 0.5, 0.5 to 0.8, and 0.8 to 1. For each interval, the adjustment direction is extracted; for example, the adjustment direction for the 0 to 0.5 interval is to increase all parameter values, while the adjustment direction for the 0.5 to 0.8 interval is to fine-tune some parameter values. The adjustment ratio is the magnitude of parameter adjustment; for example, the adjustment ratio for the 0 to 0.5 interval is 0.2, meaning the parameter value increases by 20%. The constraint condition is the range of values ​​after parameter adjustment; for example, the parameter value must not exceed the preset maximum value or fall below the preset minimum value. The pollution diffusion path modeling system establishes a matching table based on the above information. Each record in the table contains the fitness value interval, adjustment direction, adjustment ratio, and constraint condition. The matching table allows for quick lookup of the adjustment rules corresponding to different fitness values.

[0046] Step 1442: The pollution diffusion path modeling system calculates the difference between the initial fitness value and the preset standard. Based on the interval where the difference is located, it matches the corresponding parameter adjustment direction and adjustment ratio from the matching table. Combined with the parameter constraints of the gas migration model, the adjustment ratio is corrected to generate the target adjustment amount for each parameter.

[0047] The preset standard is 0.8, the initial fitness value is 0.6, the difference is 0.2, and the difference range is 0.5 to 0.8. The adjustment direction corresponding to this range matched from the matching table is to fine-tune the values ​​of some parameters, with an adjustment ratio of 0.1. The parameter constraint condition of the gas migration model is that the parameter value range is -1 to 1. For example, if a parameter is currently 0.9, and is adjusted by an adjustment ratio of 0.1, the value will become 0.99, which does not exceed the constraint range, and the adjustment ratio does not need to be corrected. If a parameter is currently 0.95, and is adjusted by an adjustment ratio of 0.1, it will exceed 1. At this time, the adjustment ratio is corrected, for example, the adjustment ratio is changed to 0.05, so that the adjusted parameter value is 0.995, which meets the constraint condition. Based on the corrected adjustment ratio, the target adjustment amount for each parameter is generated. For example, the target adjustment amount for parameter w1 is the current value multiplied by the adjustment ratio.

[0048] Step 1443: The pollution diffusion path modeling system calculates the inertia weight of the current iteration based on the inertia weight formula of the particle swarm optimization algorithm and the initial fitness value. The inertia weight reflects the degree to which the particles retain their historical position information.

[0049] The inertia weight formula for the particle swarm optimization algorithm is W = Wmax - (Wmax - Wmin) × (Fcurrent / Fmax), where Wmax is the maximum inertia weight, Wmin is the minimum inertia weight, Fcurrent is the initial fitness value, and Fmax is the preset maximum fitness value. For example, if Wmax is set to 0.9, Wmin to 0.4, Fmax to 1, and the initial fitness value is 0.6, the inertia weight for the current iteration is calculated as W = 0.9 - (0.9 - 0.4) × (0.6 / 1) = 0.6. The larger the inertia weight, the higher the degree to which the particle retains its historical position information, and the more likely it is to maintain its original direction of motion during iteration; conversely, it is more susceptible to the influence of the group's optimal position.

[0050] Step 1444: The pollution diffusion path modeling system updates the position coordinates of each particle based on the inertia weight, iteration direction, and target adjustment amount, combined with the particle's own historical best position and the global best position of the particle swarm. The updated particle position coordinates are then converted into corresponding parameter values ​​to obtain a new parameter combination.

[0051] In this system, the historical best position of a particle is the coordinate of the particle when its fitness value was highest in a previous iteration, and the global best position of the particle swarm is the coordinate of the position of all particles when their fitness value was highest in a previous iteration. The pollution diffusion path modeling system first calculates the velocity update of the particles, which is determined based on the inertia weight, the particle's current velocity, the difference between the historical best position and the current position, and the difference between the global best position and the current position. Then, the position coordinates of the particles are updated according to the velocity update and the iteration direction. Finally, the updated position coordinates are converted into corresponding parameter values, such as each element in the position coordinates corresponding to a parameter in the model, thus obtaining a new parameter combination.

[0052] Step 145: The pollution diffusion path modeling system substitutes the new parameter combination into the gas migration model for recalculation, obtains new pollution diffusion simulation results, calculates the corresponding fitness value, and records the optimal fitness value and corresponding parameter combination in each iteration.

[0053] The pollution diffusion path modeling system inputs the new parameter combination obtained in step 144 into the gas migration model. The model calculates based on the new parameters to obtain new pollution diffusion simulation results. The simulation results include new concentration distribution data and time-series change data. Then, the fitness value corresponding to the simulation result is calculated according to the method in step 143. The fitness value in each iteration is recorded, and the highest fitness value and the corresponding parameter combination are selected and saved. For example, the fitness value F1 is obtained in the nth iteration, and the fitness value F2 is obtained in the (n+1)th iteration. If F2>F1, the optimal fitness value is updated to F2, and the corresponding parameter combination is saved.

[0054] Step 146: When the variation range of the optimal fitness value in multiple consecutive iterations is less than the preset variation range, or when the number of iterations reaches the set upper limit, the pollution diffusion path modeling system stops iterating and determines the final recorded optimal parameter combination as the optimization parameter.

[0055] The preset variation range is 0.01, and the upper limit is set to 100 iterations. After each iteration, the pollution diffusion path modeling system calculates the difference between the current optimal fitness value and the optimal fitness value of the previous iteration. If the difference is less than 0.01 for 5 consecutive iterations, or the number of iterations reaches 100, the iteration stops. After stopping the iteration, the parameter combination corresponding to the final recorded optimal fitness value is retrieved and determined as the optimization parameters. For example, if the final optimal fitness value is 0.95, the corresponding parameter combination is the optimal parameter of the model.

[0056] Step 1461: The pollution diffusion path modeling system sets dual judgment conditions for iteration termination. The first condition is that the change range of the optimal fitness value for a consecutive preset number of iterations is less than the preset change range. The second condition is that the number of iterations reaches the preset maximum number of iterations.

[0057] The preset number of iterations is set to 5, the preset change range is set to 0.01, and the maximum number of iterations is set to 100. The first condition requires that the difference between the optimal fitness value of each iteration and the optimal fitness value of the previous iteration is less than 0.01 in 5 consecutive iterations. The second condition requires that the number of iterations reaches 100. When either of these conditions is met, the iteration is terminated.

[0058] Step 1462: After each iteration, the pollution diffusion path modeling system records the optimal fitness value of the current iteration and the corresponding iteration number, and calculates the absolute difference between the current optimal fitness value and the optimal fitness value of the previous iteration. The absolute difference is the magnitude of the change in fitness value.

[0059] In this process, after the first iteration, the optimal fitness value F1 and the iteration number 1 are recorded; after the second iteration, the optimal fitness value F2 and the iteration number 2 are recorded, and |F2-F1| is calculated as the change in fitness value; this operation is repeated for each subsequent iteration, recording the optimal fitness value, the iteration number and calculating the change.

[0060] Step 1463: The pollution diffusion path modeling system counts the number of times the fitness value changes by a factor less than the preset value. If this number reaches the preset number, the first termination condition is triggered, and the iteration is marked as stopped. Simultaneously, it checks whether the current iteration count has reached the maximum iteration count. If it has, the second termination condition is triggered, and the iteration is marked as stopped.

[0061] The preset change range is 0.01, and the preset number of iterations is 5. After each calculation of the fitness value change range, the pollution diffusion path modeling system increments the count by 1 if the change range is less than 0.01, otherwise the count is reset to zero. When the count reaches 5, the first termination condition is triggered. At the same time, it checks whether the current iteration count has reached 100. If it has, the second termination condition is triggered. When any condition is triggered, the iteration can be stopped.

[0062] Step 1464: When any termination condition is triggered, the pollution diffusion path modeling system stops the particle swarm iteration process, retrieves all the optimal fitness values ​​and corresponding parameter combinations recorded during the iteration process, and selects the parameter combination with the highest fitness value.

[0063] When the first or second termination condition is triggered, the pollution diffusion path modeling system stops updating the particle swarm position and calculating the fitness value; then it retrieves the optimal fitness value and corresponding parameter combination recorded during the iteration process, for example, it retrieves the optimal fitness value and parameter combination from the 1st to the 100th iteration; by comparing the above fitness values, it selects the parameter combination corresponding to the fitness value with the largest value, for example, it selects the parameter combination corresponding to the fitness value of 0.95.

[0064] Step 1465: The pollution diffusion path modeling system substitutes the selected parameter combinations into the gas migration model for verification calculation, inputs spatiotemporal characteristics to obtain verification simulation results, and calculates the final error between the verification simulation results and the original observation data.

[0065] The pollution diffusion path modeling system inputs the selected parameter combinations into the gas migration model. The model calculates the spatiotemporal characteristics based on the parameter combinations to obtain the verification simulation results. The verification simulation results include the predicted gas concentration values ​​at different spatial locations and timestamps. Then, the mean square error between the predicted values ​​and the concentration values ​​at the corresponding locations and timestamps in the original observation data is calculated to obtain the final error, for example, the final error is 0.05.

[0066] Step 1466: If the final error of the pollution diffusion path modeling system is within the preset tolerance range, then the parameter combination is determined as the optimization parameter; if the final error exceeds the preset tolerance range, then the termination condition is adjusted, the number of iterations is increased, and the iteration process is re-executed until a parameter combination that meets the preset tolerance range is obtained as the optimization parameter.

[0067] The preset tolerance range is 0 to 0.06. If the final error is 0.05, which is within the tolerance range, the parameter combination is determined as the optimized parameter. If the final error is 0.07, which exceeds the tolerance range, the termination condition is adjusted, the maximum number of iterations is increased from 100 to 150, and the particle swarm iteration process is restarted until a parameter combination with a final error in the range of 0 to 0.06 is obtained, which is then determined as the optimized parameter.

[0068] Step 150: The pollution diffusion path modeling system imports the optimized parameters into the gas migration model to drive the inversion calculation of the current pollution diffusion path. It determines the initial distribution of path nodes by combining the topographic features and meteorological conditions of the pollution monitoring area. It adjusts the node positions and correlations through node concentration error feedback to iteratively update the path nodes in the current pollution diffusion path. Based on the updated path node information, it outputs the pollution diffusion inversion path. Both the current pollution diffusion path and the pollution diffusion inversion path correspond to the target monitoring period.

[0069] The target monitoring period is from T1 to T100. The pollution diffusion path modeling system first imports the optimized parameters into the gas migration model, which drives the inversion calculation of the current pollution diffusion path based on these parameters. Then, combined with the topographic features of the pollution monitoring area, such as the distribution of buildings and topographic relief, as well as meteorological conditions, such as wind speed and direction, the initial distribution of path nodes is determined. The initial path nodes are set on possible paths of pollution diffusion, such as setting nodes along the prevailing wind direction. Next, the concentration prediction value of each initial path node is calculated and compared with the measured concentration value in the original observation data to obtain the concentration error. The node positions and correlations are adjusted according to the error feedback, for example, moving nodes with larger errors to the direction of higher measured concentration values ​​and correcting the correlation weights between nodes. After multiple iterations, the update stops when the concentration error of all nodes is less than a preset threshold. Finally, based on the updated path node information, the nodes are connected in chronological order to output the pollution diffusion inversion path, which reflects the diffusion trajectory of pollutant gas during the target monitoring period.

[0070] Step 151: The pollution diffusion path modeling system imports the optimized parameters into the gas migration model and completes the parameter configuration. It reads the terrain feature data and real-time meteorological condition data of the pollution monitoring area. The terrain feature data includes the elevation distribution, obstacle distribution and surface roughness of the pollution monitoring area, and the meteorological condition data includes meteorological conditions and temperature and humidity information.

[0071] The pollution diffusion path modeling system imports the weight values ​​and bias values ​​from the optimization parameters into the corresponding parameter positions of the gas migration model to complete the parameter configuration. Then, it reads the terrain feature data. The elevation distribution represents the changes in altitude in the region in the form of contour lines. The obstacle distribution includes the location and size of objects such as buildings and trees that hinder gas diffusion. The surface roughness reflects the resistance of the surface to gas diffusion. The meteorological condition data includes the weather status, such as sunny, cloudy, or rainy. The temperature and humidity information is the temperature and humidity values ​​of the current region. All of the above data is stored in digital form. For example, the elevation distribution data is an array containing the altitude of different coordinate points, and the obstacle distribution data is a list containing the location coordinates and size of obstacles.

[0072] Step 152: The pollution diffusion path modeling system combines terrain feature data and meteorological condition data to analyze the potential diffusion direction and obstructed areas of pollutants, and sets initial path nodes in the dominant diffusion direction to form the initial distribution of path nodes; wherein, the spacing of the initial path nodes is adjusted according to the terrain complexity.

[0073] The pollution diffusion path modeling system first determines the dominant diffusion direction of pollutants based on wind speed and direction data from meteorological conditions. For example, if the dominant wind direction is east, the dominant diffusion direction is from west to east. Then, it analyzes the obstructed areas in the potential diffusion direction by combining topographic feature data. For example, if there are buildings in the east wind direction, the area where the buildings are located is an obstructed area. Next, initial path nodes are set in the dominant diffusion direction. The spacing between the initial nodes is adjusted according to the topographic complexity, which is measured by indicators such as obstacle density and elevation change rate. For example, the node spacing is set to 5 meters in areas with complex topography and 10 meters in areas with simple topography. The initial distribution of the initial path nodes is represented in the form of coordinate points. For example, in the dominant diffusion direction from west to east, a node is set every 5 or 10 meters starting from the initial pollution location.

[0074] Step 1521: The pollution diffusion path modeling system digitizes the terrain feature data, extracts the elevation change curve, obstacle outline coordinates and surface roughness coefficient, constructs a digital terrain model, and identifies the terrain elements that affect gas diffusion in the digital terrain model.

[0075] The pollution diffusion path modeling system first processes the elevation distribution data in the terrain feature data, and calculates the elevation change curve through interpolation. The curve reflects the altitude change of different coordinate points. Then, it extracts the outline coordinates of obstacles in the obstacle distribution data, such as the coordinates of the four corner points of a building. The surface roughness coefficient is determined according to the surface type. For example, the roughness coefficient of cement ground is 0.1, and that of grassland is 0.2. Next, it constructs a digital terrain model. The model represents the terrain in the region in three-dimensional coordinate form. The model identifies terrain elements that affect gas diffusion, such as places with significant elevation changes and the location of obstacles. For example, obstacle areas are marked with different colors in the model.

[0076] Step 1522: The pollution diffusion path modeling system performs time-series stability analysis on meteorological condition data, determines the prevailing wind direction and average wind speed, determines the dominant diffusion direction of pollutants based on the prevailing wind direction, and draws the initial diffusion axis along the prevailing wind direction, starting from the initial location of pollution occurrence. The initial diffusion axis is the core path direction of pollution diffusion.

[0077] The pollution diffusion path modeling system analyzes wind speed and direction data at different timestamps in meteorological condition data to determine the prevailing wind direction. For example, if the east wind occurs most frequently during the target monitoring period, then the prevailing wind direction is east. The system calculates the average wind speed during this period. Then, starting from the initial location of pollution occurrence, such as coordinate point (X0, Y0), the system draws the initial diffusion axis along the prevailing wind direction. The axis is a straight line that represents the core path direction of pollution diffusion. For example, the axis direction extends eastward from (X0, Y0).

[0078] Step 1523: The pollution diffusion path modeling system overlays the digital terrain model with the initial diffusion axis to identify obstructed areas on and around the initial diffusion axis. Obstructed areas are areas that hinder gas diffusion, including areas with abrupt changes in elevation or where obstacles are located.

[0079] The pollution diffusion path modeling system spatially overlays the digital terrain model with the initial diffusion axis and determines the position of the axis in the model through coordinate matching. Then, it identifies elevation abrupt changes on and around the axis. For example, if the elevation suddenly increases by 10 meters when the axis passes through a certain area, that area is an elevation abrupt change. At the same time, it identifies areas where obstacles are located. For example, if there are buildings near the axis, the area where the buildings are located is an obstructed area. These areas are marked as obstructed areas and are avoided or adjusted in subsequent node settings.

[0080] Step 1524: The pollution diffusion path modeling system performs terrain complexity scoring on the area along the initial diffusion axis based on terrain complexity evaluation indicators including elevation change rate, obstacle density, and surface roughness coefficient. The system determines the spacing rules for the initial path nodes based on the terrain complexity score. For complex terrain areas where the terrain complexity score reaches the first preset value, the node spacing is set as the first spacing; for areas where the terrain complexity score is between the first and second preset values, the node spacing is set as the second spacing; for flat areas where the terrain complexity score is lower than the second preset value, the node spacing is set as the third spacing, with the first spacing being smaller than the second spacing, and the second spacing being smaller than the third spacing.

[0081] The terrain complexity evaluation indicators include elevation change rate, obstacle density, and surface roughness coefficient. Each indicator is scored from 0 to 10 points, for a total of 30 points. The first preset score is 20 points, and the second preset score is 10 points. The first spacing is 5 meters, the second spacing is 8 meters, and the third spacing is 10 meters. The pollution diffusion path modeling system calculates the elevation change rate, obstacle density, and surface roughness coefficient of each region along the initial diffusion axis, scores them, and sums them to obtain the terrain complexity score. For example, if a region has an elevation change rate score of 8 points, an obstacle density score of 7 points, and a surface roughness coefficient score of 6 points, for a total score of 21 points, reaching the first preset score, the node spacing is set to 5 meters. If a region has a total score of 15 points, which is between 10 and 20 points, the node spacing is set to 8 meters. If a region has a total score of 5 points, which is below 10 points, the node spacing is set to 10 meters.

[0082] Step 1525: The pollution diffusion path modeling system sequentially deploys initial path nodes along the initial diffusion axis according to the set spacing rules, and adds additional nodes before and after the obstructed area to capture diffusion changes, forming an initial distribution that includes node coordinates, corresponding terrain complexity scores, and estimated concentrations.

[0083] The pollution diffusion path modeling system starts from the initial pollution occurrence location and deploys initial path nodes along the initial diffusion axis according to the spacing rules corresponding to the terrain complexity score. For example, a node is deployed every 5 meters in complex terrain areas and every 10 meters in flat areas. Additional nodes are added before and after the obstructed area, such as 5 meters in front of and 5 meters behind the obstacle, to capture the diffusion changes of gas before and after the obstructed area. Each initial path node contains coordinate information, the terrain complexity score of the corresponding area, and the concentration value predicted based on the gas migration model, forming an initial distribution list.

[0084] Step 153: The pollution diffusion path modeling system starts the gas migration model to perform the inversion calculation of the current pollution diffusion path, simulates the gas diffusion process based on the initial distribution, and calculates the predicted gas concentration at each initial path node.

[0085] Among them, the pollution diffusion path modeling system starts the gas migration model. Based on optimized parameters, terrain feature data and meteorological condition data, the model simulates the initial distribution of the initial path nodes and simulates the diffusion process of gas from the initial pollution location along the initial diffusion axis. It calculates the predicted gas concentration value of each initial path node at different timestamps. For example, the predicted concentration value of node N1 at time T1 is C1 and at time T2 is C2, forming a node concentration prediction dataset.

[0086] Step 154: The pollution diffusion path modeling system retrieves the measured concentration values ​​from the original observation data at the corresponding locations, calculates the concentration error between the predicted and measured values ​​at each initial path node, and marks nodes with concentration errors exceeding the preset range as nodes to be adjusted.

[0087] The pollution diffusion path modeling system retrieves the original observation data corresponding to the initial path nodes. For example, the measured concentration value of the sensor at the location corresponding to node N1 at time T1 is C1'. The system calculates the difference between the predicted concentration value and the measured value, i.e., |C1-C1'|, to obtain the concentration error. The preset range is 0 to 0.1. If the concentration error exceeds 0.1, the node is marked as a node to be adjusted. For example, the concentration error of node N1 is 0.15, so it is marked as a node to be adjusted.

[0088] Step 155: The pollution diffusion path modeling system analyzes the causes of errors for the nodes to be adjusted by combining terrain feature data and meteorological condition data. Based on the causes of errors, it determines the adjustment direction and amount of the node positions, and at the same time corrects the correlation weights between nodes. The correlation weights reflect the degree of diffusion influence between nodes.

[0089] The pollution diffusion path modeling system first analyzes the terrain features of the nodes to be adjusted. For example, if node N1 is located near an obstacle, the complex terrain may lead to errors in concentration prediction. Then, it analyzes meteorological conditions, such as sudden changes in wind speed, which affect gas diffusion. Based on the above analysis, it determines the cause of the error, such as the error being caused by complex terrain. Then, it determines the adjustment direction of the node position, such as adjusting node N1 away from the obstacle by 2 meters. At the same time, it corrects the correlation weight between nodes. For example, the correlation weight between node N1 and its neighboring node N2, which was originally 0.5, is corrected to 0.6 due to the adjustment of the node position, reflecting the change in the degree of diffusion influence between nodes.

[0090] Step 1551: The pollution diffusion path modeling system extracts the coordinate information of the node to be adjusted and completes the positioning in the terrain digital model. Simultaneously, it retrieves the original observation data of the corresponding collection period of the node to be adjusted and the basic data used for inversion calculation to construct a node feature dataset.

[0091] The coordinates of node N1 to be adjusted are (X1, Y1). The pollution diffusion path modeling system locates these coordinates in the digital terrain model to determine the terrain area where the node is located. Then, it retrieves the original observation data for the corresponding collection period of the node, such as T1 to T5, including measured concentration values, wind speed and direction data, etc. At the same time, it retrieves the basic data used for inversion calculation, such as terrain feature data, meteorological condition data and model parameters. The above data are integrated to construct a node feature dataset, which includes node coordinates, original observation data, basic data and concentration prediction values.

[0092] Step 1552: The pollution diffusion path modeling system extracts topographic features, including elevation gradient, obstacle distribution and surface permeability, within a preset range around the node based on the node feature dataset, and extracts the meteorological feature change trend for the corresponding time period to form a topographic-meteorological feature correlation matrix.

[0093] The preset range is 10 meters around the node. The pollution diffusion path modeling system extracts the elevation gradient within this range, i.e., the elevation difference between adjacent points; the distribution pattern of obstacles, such as the shape and size of obstacles; the surface permeability, such as whether the surface is open or obstructed; and the meteorological feature change trend for the corresponding time period, such as the wind speed changing from 2 m / s to 3 m / s and the wind direction changing from east to southeast. The above features are arranged in rows and columns to form a topographic-meteorological feature correlation matrix. The rows of the matrix represent different topographic or meteorological features, the columns represent different time points, and the element values ​​are the specific values ​​of the features.

[0094] Step 1553: The pollution diffusion path modeling system matches the topographic-meteorological feature correlation matrix with the physical feature model of pollutant gas diffusion. It uses feature fit analysis to determine the main cause of error. If the fit between the topographic feature features and the diffusion model is lower than the preset threshold, then topographic shading is determined to be the main source of error. If the deviation between the trend of meteorological feature changes and the model input data exceeds the allowable range, then the fluctuation of meteorological conditions is determined to be the main source of error.

[0095] The physical feature model for pollutant gas diffusion includes the diffusion patterns of gases under different terrain and meteorological conditions. For example, the model specifies the diffusion coefficient under open terrain and stable wind speed. The pollution diffusion path modeling system matches the features in the terrain-meteorological feature correlation matrix with the corresponding features in the physical feature model and calculates the feature fit. The fit is the degree of similarity between the actual features and the model features. The preset threshold is 0.7. If the fit of the terrain feature is 0.6, which is lower than the threshold, then terrain shading is determined to be the main source of error. If the deviation between the trend of meteorological feature changes and the model input data is 0.2, which exceeds the allowable range of 0.1, then the fluctuation of meteorological conditions is determined to be the main source of error.

[0096] Step 1554: The pollution diffusion path modeling system addresses the errors caused by terrain occlusion by analyzing the characteristic path of gas diffraction diffusion in conjunction with the spatial contour features of obstacles. The node to be adjusted is moved to the core area of ​​the diffraction path so that the node can capture the gas concentration characteristics after diffraction.

[0097] The spatial outline of the obstacle is rectangular, 10 meters long and 5 meters wide. The pollution diffusion path modeling system analyzes the characteristic path of gas diffraction diffusion. The diffraction path is the path by which gas diffuses to both sides after encountering the obstacle. The core area is the region with higher concentration in the diffraction path. Then, the node to be adjusted is moved to this core area. For example, node N1, which was originally located in front of the obstacle, is moved to the core area of ​​the diffraction path on the right side of the obstacle, so that the node can capture the gas concentration characteristics after diffraction and reduce the concentration error.

[0098] Step 1555: To address errors caused by fluctuations in meteorological conditions, the pollution diffusion path modeling system updates the dominant diffusion direction based on real-time meteorological characteristics and redeploys nodes along the new dominant direction, ensuring that the node distribution aligns with the actual gas diffusion trend.

[0099] Among them, real-time meteorological characteristics show that the prevailing wind direction has changed from east to southeast, and the pollution diffusion path modeling system updates the dominant diffusion direction to the southeast. Then, the nodes to be adjusted are redeployed along the new dominant direction. For example, node N1 was originally set along the east direction, but now it is adjusted to be set along the southeast direction, so that the node distribution is consistent with the actual gas diffusion trend and the accuracy of concentration prediction is improved.

[0100] Step 1556: Based on the adjusted spatial distribution relationship of the path nodes and the correlation of concentration characteristics of adjacent nodes, the pollution diffusion path modeling system constructs the diffusion influence logic between nodes and redefines the allocation rules of node association weights so that the association weights reflect the degree of diffusion influence between nodes.

[0101] The adjusted spatial distribution of path nodes is as follows: node N1 is located southeast of N2, at a distance of 5 meters; the concentration characteristics of adjacent nodes are correlated such that the concentration change of N1 is significantly affected by the concentration of N2; the pollution diffusion path modeling system constructs diffusion influence logic between nodes, for example, an increase in the concentration of N2 will cause the concentration of N1 to increase after a period of time; based on this logic, the allocation rules for node association weights are redefined, for example, the association weight between N1 and N2 is determined according to their distance and concentration correlation, the closer the distance and the stronger the correlation, the higher the weight, so that the association weight accurately reflects the degree of diffusion influence between nodes.

[0102] Step 156: The pollution diffusion path modeling system re-inputs the adjusted path nodes into the gas migration model for inversion calculation, jumps to the node concentration error calculation and position adjustment step, until the concentration error of all path nodes is within the preset range, and completes the iterative update of the path nodes in the current pollution diffusion path.

[0103] The pollution diffusion path modeling system re-inputs the adjusted path node coordinates, association weights, and other information into the gas migration model. The model performs inversion calculations based on the new node information to obtain new concentration prediction values. Then, it jumps to step 154 ​​to calculate the new concentration error. If the concentration error of any node still exceeds the preset range, steps 155 to 156 are repeated until the concentration error of all nodes is within the range of 0 to 0.1, thus completing the iterative update of the path nodes.

[0104] Step 157: The pollution diffusion path modeling system connects the iteratively updated path nodes sequentially according to the diffusion time sequence, labels the concentration information and corresponding timestamp of each iteratively updated path node, generates the pollution diffusion inversion path, and outputs it.

[0105] The pollution diffusion path modeling system first extracts the timestamps of the path nodes after iterative updates and sorts the nodes in chronological order. Then, it connects the nodes sequentially according to the diffusion time sequence, for example, connecting the node at time T1 to the node at time T2 to form a path. It labels the concentration information of each node, for example, the concentration of node N1 is C1 and the timestamp is T1. Finally, it generates the pollution diffusion inversion path, which is represented in the form of lines, labeled with node information and timestamps, and outputs it to a display device or storage device.

[0106] Step 1571: The pollution diffusion path modeling system extracts the timestamps and concentration feature data of all updated path nodes, constructs a node spatiotemporal feature matrix, and sorts the updated path nodes in an orderly manner based on the timestamps to determine the temporal evolution logic of pollution diffusion.

[0107] The updated path nodes have timestamps from T1 to T100, and the concentration feature data are the concentration values ​​of each node at the corresponding timestamp. The pollution diffusion path modeling system constructs a spatiotemporal feature matrix of nodes. The rows of the matrix represent nodes, the columns represent timestamps and concentration features, and the element values ​​are the concentration values ​​at the corresponding timestamps. Then, based on the timestamps, the nodes are sorted in chronological order. For example, the node at time T1 is placed first, and the node at time T100 is placed last, thus determining the temporal evolution logic of pollution diffusion, that is, the diffusion order of pollutant gases from morning to night.

[0108] Step 1572: The pollution diffusion path modeling system analyzes the rationality of concentration transitions between adjacent nodes based on the sorted node sequence and the continuity characteristics of gas diffusion. If there are abrupt concentration changes between adjacent nodes that do not conform to the physical laws of diffusion, a transition feature model is constructed based on the spatiotemporal characteristics between adjacent nodes to generate transition nodes that conform to the diffusion trend and fill the gaps in diffusion features between adjacent nodes.

[0109] The continuity characteristic of gas diffusion requires that the node concentration changes smoothly between adjacent time stamps. For example, if the node concentration is C1 at time T1, it should be C1±0.05 at time T2. If the concentration changes abruptly between adjacent nodes, such as C1 at time T1 and C1+0.2 at time T2, it does not conform to the physical laws of diffusion. The pollution diffusion path modeling system constructs a transition feature model based on the spatiotemporal characteristics of adjacent nodes, such as the concentration values, time intervals, and topographic and meteorological conditions at times T1 and T2. The model simulates the transition process of concentration from C1 to C1+0.2 and generates transition nodes, such as inserting a node at time T1.5 with a concentration of C1+0.1 between T1 and T2 to fill the gaps in the diffusion characteristics.

[0110] Step 1573: The pollution diffusion path modeling system constructs multi-dimensional feature labels for each updated path node and supplementary transition node. The content of the multi-dimensional feature labels covers node coordinates, concentration characteristics, and topographic and meteorological background information for the corresponding time period.

[0111] The multidimensional feature label is a structure containing multiple fields, including node coordinates (X, Y), concentration features (concentration value, concentration change rate), terrain information for the corresponding time period (elevation, obstacle distribution), and meteorological background information (wind speed, wind direction, temperature, humidity). For example, the multidimensional feature label of node N1 is {(X1, Y1), (C1, ΔC1), (H1, O1), (V1, D1, T1, M1)}, where H1 is elevation, O1 is obstacle distribution, V1 is wind speed, D1 is wind direction, T1 is temperature, and M1 is humidity.

[0112] Step 1574: The pollution diffusion path modeling system uses the updated temporal order of path nodes as the axis and combines the strength differences of node concentration characteristics to set path visualization rules, linking concentration characteristics with path presentation form, so that the path reflects the differences in pollution levels in different areas.

[0113] Among them, the path visualization rules stipulate that the higher the concentration value, the thicker the path line and the darker the color; for example, the path line corresponding to the node with a concentration value of C1 is 1 pixel wide and blue, and the path line corresponding to the node with a concentration value of C2 (C2>C1) is 2 pixels wide and red. Based on this rule, the pollution diffusion path modeling system associates the node concentration characteristics with the path presentation form, so that the generated pollution diffusion inversion path of the XXX Basin can intuitively reflect the differences in the degree of pollution in different areas.

[0114] Step 1575: The pollution diffusion path modeling system overlays and merges the sorted node sequence with the digital terrain model of the pollution monitoring area to establish the spatial relationship between path nodes and terrain elements, and obtain the merged path data.

[0115] The pollution diffusion path modeling system first acquires a digital terrain model of the pollution monitoring area, which includes information such as terrain undulations and obstacle locations within the area. Then, it matches the coordinates of the sorted node sequence with the coordinates of the terrain model and overlays the nodes onto the terrain model. It establishes spatial relationships between path nodes and terrain features. For example, if node N1 is located at an elevation point in the terrain model, it associates the elevation value of that point with obstacle information. Finally, it obtains the fused path data, which includes node coordinates, terrain information, and concentration characteristics.

[0116] Step 1576: The pollution diffusion path modeling system constructs a pollution diffusion inversion path based on the fused path data and multi-dimensional feature labels, and generates a complete output result based on the pollution diffusion inversion path, including a path time series evolution diagram, spatial distribution visualization, and node feature lookup table.

[0117] The pollution diffusion path modeling system first connects nodes in chronological order based on the fused path data to form a pollution diffusion inversion path; then it generates a path time-series evolution map, which shows the distribution of path nodes and concentration changes at different time stamps; it generates a spatial distribution visualization, overlaying the path onto the terrain model to intuitively display the spatial location of the path; it generates a node feature lookup table, which contains multi-dimensional feature label information for each node, and the corresponding feature can be queried by node number; and it integrates the above results into a complete output result, which is then output to the user terminal or storage system.

[0118] Optionally, the method further includes: Step 210: The pollution diffusion path modeling system receives real-time data collected by multi-source sensors within the pollution monitoring area, compares the real-time data with the historical observation data corresponding to the inversion path at the same time period, and extracts the differences between the real-time data and the historical data in terms of gas concentration, spatial distribution and changing trends to form a difference feature set. The difference feature set reflects the degree of deviation between the current pollution diffusion state and the historical inversion path.

[0119] The pollution diffusion path modeling system receives real-time data from multiple sources, such as gas concentration data and wind speed data at time T101; retrieves historical data from the same time period corresponding to the inversion path, such as observation data from time T1 to T10; performs feature comparison, calculates the difference in gas concentration, spatial distribution, and trend between real-time and historical data; extracts the above difference features, such as real-time concentration being 0.2 higher than historical concentration, more concentrated spatial distribution, and faster trend of change, to form a difference feature set; the larger the element value of the difference feature set, the greater the deviation between the current pollution diffusion state and the historical inversion path.

[0120] Step 220: The pollution diffusion path modeling system inputs the differential feature set into the gas migration model, adjusts the input weights of the model in combination with the determined optimization parameters, and generates a predicted path of real-time pollution diffusion trend. The predicted path includes path node evolution information within a preset future time period.

[0121] The pollution diffusion path modeling system inputs a set of differential features into a gas migration model. The model combines the determined optimization parameters and adjusts the input weights, such as increasing the weight of concentration difference features in the set of differential features. Based on the adjusted weights, the model predicts the pollution diffusion trend in a future preset time period, such as the next hour. It generates a predicted path, which contains the evolution information of path nodes in the future time period, such as the location and concentration prediction value of node N101 at time T102, forming a predicted path dataset.

[0122] Step 230: The pollution diffusion path modeling system associates the predicted path with the functional zoning data of the pollution monitoring area, identifies the protected areas involved in the predicted path, analyzes the concentration evolution characteristics of the path nodes within the protected areas, and outputs the regional pollution risk identification results.

[0123] The functional zoning data of the pollution monitoring area includes residential areas, industrial areas, ecological protection areas, etc. The pollution diffusion path modeling system spatially correlates the predicted path with the functional zoning data, identifies the protected areas through which the predicted path passes, such as ecological protection areas, analyzes the concentration evolution characteristics of path nodes within the protected area, such as the trend of node concentration changes over time, and outputs regional pollution risk identification results based on the concentration evolution characteristics, such as indicating that the ecological protection area has a high pollution risk in the next hour.

[0124] Step 240: The pollution diffusion path modeling system generates pollution diffusion control suggestions based on the regional pollution risk identification results and predicted paths. The pollution diffusion control suggestions include the optimization direction of sensor deployment and the priority of diffusion blocking.

[0125] The pollution diffusion path modeling system determines the optimization direction of sensor deployment based on regional pollution risk identification results, such as high-pollution-risk areas, for example, by increasing the number of sensors in the area; it determines the diffusion blocking priority based on the diffusion direction of the predicted path, for example, by prioritizing the installation of blocking facilities at the front end of the diffusion path; and it generates pollution diffusion control suggestions, which are presented in text form, such as "add 5 gas concentration sensors in the ecological protection area and prioritize the installation of activated carbon adsorption devices at the front end of the predicted path".

[0126] Step 250: The pollution diffusion path modeling system integrates the predicted path, regional pollution risk identification results, and pollution diffusion control recommendations into a dynamic monitoring report and outputs it.

[0127] The pollution diffusion path modeling system integrates the visualization of predicted paths, textual explanations of regional pollution risk identification results, and pollution diffusion control recommendations into a dynamic monitoring report. The report includes a cover, table of contents, predicted path analysis, risk identification results, and control recommendations. The report is output in PDF or web page format and sent to the management platform or displayed on the monitoring platform.

[0128] Through the above embodiments, a dynamic monitoring and control system for pollution diffusion was constructed, realizing a full-chain response from real-time data comparison to risk warning and control recommendations. By receiving real-time data collected by multi-source sensors and comparing features with historical inversion path observation data, the deviation between the current pollution diffusion state and historical paths is accurately captured. The difference feature set is input into the gas migration model and the input weights are adjusted. The generated real-time pollution diffusion trend prediction path can effectively reflect the future direction of pollution evolution. By associating the prediction path with functional zoning data to identify protected areas and analyzing concentration evolution characteristics, the output regional pollution risk identification results can accurately locate high-risk areas. Control recommendations generated based on risk results and prediction paths provide practical guidance for prevention and control from two aspects: sensor deployment optimization and diffusion blocking priority. Finally, the integrated dynamic monitoring report provides comprehensive and timely decision-making basis, thereby improving the initiative and accuracy of pollution prevention and control.

[0129] Optionally, the method further includes: Step 310: The pollution diffusion path modeling system acquires geographic information data and pollution source registration data of the pollution monitoring area, associates and calibrates the geographic information data with the spatial coordinates of the inverted path, and establishes the spatial correspondence between path nodes and geographic elements.

[0130] The geographic information data includes the location coordinates of geographic elements such as roads, rivers, and buildings in the pollution monitoring area; the pollution source registration data includes the location, type, and emission status of known pollution sources in the area; the pollution diffusion path modeling system associates and calibrates the coordinate system of the geographic information data with the spatial coordinates of the inverted path to ensure that the coordinates of the two are consistent; then, the spatial correspondence between path nodes and geographic elements is established, for example, if node N1 is located next to a road, it is associated with the name and location information of that road.

[0131] Step 320: The pollution diffusion path modeling system extracts the geographical regions corresponding to the starting point of the inverted path and the high-concentration nodes along the path based on spatial correspondence; combined with the emission characteristics and spatial location information of various pollution sources in the pollution source registration data, it analyzes the matching relationship between the spatiotemporal characteristics of the path nodes and the characteristics of potential pollution sources, and constructs the association feature matrix between nodes and potential pollution sources.

[0132] The pollution diffusion path modeling system extracts the geographical area corresponding to the starting point of the inverted path, such as the vicinity of a factory; it also extracts the geographical areas corresponding to high-concentration nodes along the path, such as the area around a river; then, combined with pollution source registration data, it analyzes the emission characteristics of pollution sources in the above areas, such as the type and amount of pollutants emitted by the factory, and spatial location information; it analyzes the spatiotemporal characteristics of the path nodes, such as concentration change trends and time distribution, and their matching relationship with the characteristics of potential pollution sources, such as whether the node concentration change trend is consistent with the factory emission period; and it constructs an association feature matrix, where the rows of the matrix represent path nodes, the columns represent potential pollution sources, and the element values ​​are the degree of matching.

[0133] Step 330: The pollution diffusion path modeling system inputs the correlation feature matrix into the gas migration model, locates the potential source areas of pollution diffusion through the model's back-inference function, and outputs the feature description information of the source areas, which includes the regional range and emission feature matching degree.

[0134] The pollution diffusion path modeling system inputs the associated feature matrix into the gas migration model. Based on the determined optimization parameters, the model uses the back-inference function to infer the location of the pollution source from the pollution diffusion path; it locates the potential source area of ​​pollution diffusion, such as the area where a factory is located; and outputs the feature description information of the source area, which is within 100 meters of the factory. The emission feature matching degree is 0.85, indicating that the pollution source and path node features in this area have a high degree of matching.

[0135] Step 340: The pollution diffusion path modeling system retrieves historical sensor data from the vicinity of potential source areas, verifies the correlation with the initial node data of the inversion path, corrects the matching parameters in the correlation feature matrix, and obtains optimized pollution source location results.

[0136] The pollution diffusion path modeling system retrieves historical data from sensors around potential source areas, such as concentration data from sensors around factories at times T1 to T10; it verifies the correlation with the initial node data of the inversion path, such as the concentration data of the initial node at time T1, and calculates the correlation coefficient between the two; if the correlation coefficient is low, it corrects the matching parameters in the correlation feature matrix, such as adjusting the weight of pollution source emission characteristics; and it re-inputs the data into the model for back-dive to obtain optimized pollution source location results, such as identifying the factory as the main pollution source.

[0137] Step 350: The pollution diffusion path modeling system combines the pollution source location results and the inversion path to generate a pollution source tracing report. The pollution source tracing report includes source location information, diffusion path verification basis, and key node analysis results of pollution transmission.

[0138] The pollution diffusion path modeling system first organizes the source location, type, and emission characteristics from the pollution source location results; then it organizes the diffusion path verification basis, such as the correlation between historical sensor data and path node data; it analyzes the key nodes of pollution transmission, such as the node with the highest concentration and the node where the diffusion direction changes; and integrates the above content into a pollution source tracing report, which is presented in text and chart form and output for pollution control decision-making.

[0139] As can be seen, steps 310 and 350 form a complete pollution source tracing technology chain, effectively solving the problem of difficulty in locating pollution sources: acquiring geographic information data and pollution source registration data and correlating and calibrating the spatial coordinates of the inversion path, establishing a spatial correspondence between path nodes and geographic elements; extracting the geographic areas of the starting point of the inversion path and high-concentration nodes, and constructing a correlation feature matrix in conjunction with pollution source registration data, realizing feature matching between path nodes and potential pollution sources; inputting the correlation matrix into the model to locate potential source areas through reverse inference, and the output feature description information narrows down the scope of source investigation; retrieving historical data from sensors around the source area for correlation verification and correcting matching parameters, further optimizing the pollution source location results; finally, combining the location results and the pollution source tracing report generated by the inversion path, clearly presenting source information, verification basis, and key node analysis, realizing pollution responsibility identification and source control, and improving the efficiency and accuracy of pollution source tracing.

[0140] This application embodiment synchronously collects raw observation data using multi-source sensors deployed in the pollution monitoring area. A data association index is established by combining the spatial deployment location of the multi-source sensors and the data acquisition timestamps, integrating multi-dimensional and multi-temporal observation data. This solves the problems of scattered data and weak spatiotemporal correlation in traditional pollution monitoring. Furthermore, by combining the data association index with the physical diffusion characteristics of pollutant gases, spatiotemporal features characterizing gas migration are extracted, ensuring the relevance and accuracy of feature extraction. Based on these spatiotemporal features, a basic model structure for the gas migration model is established. The correlation weights between input and output are determined through feature response analysis, and a parameter self-learning strategy is generated. The design achieves dynamic optimization of model parameters, improving the model's adaptability to complex pollution diffusion scenarios. A parameter self-learning strategy is initiated, and parameters are iteratively updated using a particle swarm optimization space. Fitness values ​​drive target parameter optimization, ensuring optimal model parameters and enhancing prediction accuracy. Finally, the optimized parameters are imported into a gas migration model for inversion calculations. The initial distribution of path nodes is determined by combining terrain features and meteorological conditions, and nodes are iteratively updated through error feedback, outputting accurate pollution diffusion inversion paths. This achieves efficient and accurate inversion of pollution diffusion paths during the target monitoring period, providing a scientific basis for pollution source tracing and risk prevention. This design improves the accuracy, efficiency, and adaptability of pollution diffusion path modeling.

[0141] Based on the same inventive concept, embodiments of this application also provide a pollution diffusion path modeling system. See also... Figure 2 As shown, it is a schematic diagram of a possible pollution diffusion path modeling system provided in an embodiment of this application. Figure 2 In the pollution diffusion path modeling system 200, there are a processor 210 and a memory 220. The processor 210 and the memory 220 are connected to each other via a communication bus. The memory 220 stores computer programs that can be executed by the processor 210. By executing the instructions stored in the memory 220, the processor 210 can perform the steps of the above-mentioned multi-sensor-based pollution diffusion path modeling method.

[0142] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium including a computer program. When the computer program is run on a pollution diffusion path modeling system, it causes the pollution diffusion path modeling system to perform the steps of the aforementioned multi-sensor-based pollution diffusion path modeling method. In some possible implementations, various aspects of the multi-sensor-based pollution diffusion path modeling method provided in this application can also be implemented as a program product, including a computer program. When the program product is run on a pollution diffusion path modeling system, the computer program causes the pollution diffusion path modeling system to perform the steps of the aforementioned multi-sensor-based pollution diffusion path modeling method. For example, the pollution diffusion path modeling system can perform actions such as... Figure 1 The steps are shown in the diagram. The computer-readable storage medium includes volatile or non-volatile or a combination thereof, and may be removable or non-removable. Examples of computer-readable storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium.

[0143] like Figure 3 The diagram shown is a functional block diagram of the pollution diffusion path modeling system provided in this embodiment of the application. The pollution diffusion path modeling system includes a pollution diffusion path modeling device, which includes: The data acquisition and association module is used to synchronously acquire raw observation data through multi-source sensors deployed in the pollution monitoring area, and to establish a data association index based on the spatial deployment location of the multi-source sensors and the data acquisition timestamp. The data association index is used to establish the spatiotemporal correspondence between different raw observation data. The spatiotemporal feature extraction module is used to extract spatiotemporal features characterizing gas migration from the original observation data by combining the data association index and the physical diffusion characteristics of polluting gases. The spatiotemporal features include the spatial distribution gradient and temporal variation law of gas concentration. The learning strategy generation module is used to establish the basic model structure of the gas migration model based on the spatiotemporal features, associate the mapping relationship between the original observation data and the spatiotemporal features through the basic model structure, and determine the correlation weight between the input and output of the gas migration model by combining feature responsivity analysis, and generate the parameter self-learning strategy of the gas migration model. The model parameter update module is used to start the parameter self-learning strategy and initialize the particle swarm search space. The spatiotemporal features are input into the gas migration model to obtain the initial parameter combination. The parameter fitness value is calculated by iteratively updating the position of the particle swarm and the target parameters of the gas migration model are updated based on the parameter fitness value. After multiple iterations, the optimal parameter combination is obtained as the optimization parameters. The diffusion path inversion module is used to import the optimized parameters into the gas migration model to drive the inversion calculation of the current pollution diffusion path. It determines the initial distribution of path nodes by combining the topographic features and meteorological conditions of the pollution monitoring area, adjusts the node positions and correlations through node concentration error feedback to iteratively update the path nodes in the current pollution diffusion path, and outputs the pollution diffusion inversion path based on the updated path node information. The current pollution diffusion path and the pollution diffusion inversion path both correspond to the target monitoring period.

[0144] Accordingly, this application also provides a computer program product, which includes a computer program or instructions that, when executed by a processor, cause the processor to implement the steps in the above method embodiments. It should be understood that each step or combination of steps in the above method flow can be implemented by the computer program or instructions. Furthermore, these computer programs or instructions can be applied to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device, enabling the processor of the general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to function as an apparatus for implementing the corresponding functions in the above method embodiments.

[0145] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0146] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0147] Finally, it should be noted that the above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A pollution diffusion path modeling method based on multiple sensors, characterized in that, The method includes: Raw observation data is collected synchronously by multi-source sensors deployed in the pollution monitoring area. A data association index is established based on the spatial deployment location of the multi-source sensors and the data acquisition timestamp. The data association index is used to establish the spatiotemporal correspondence between different raw observation data. By combining the data association index and the physical diffusion characteristics of polluting gases, spatiotemporal features characterizing gas migration are extracted from the original observation data. These spatiotemporal features include the spatial distribution gradient and temporal variation law of gas concentration. Based on the aforementioned spatiotemporal features, a basic model structure for the gas migration model is established. The mapping relationship between the original observation data and the spatiotemporal features is associated through the basic model structure. The correlation weight between the input and output of the gas migration model is determined by combining feature responsivity analysis, and a parameter self-learning strategy for the gas migration model is generated. The parameter self-learning strategy is initiated and the particle swarm search space is initialized. The spatiotemporal features are input into the gas migration model to obtain the initial parameter combination. The parameter fitness value is calculated by iteratively updating the position of the particle swarm and the target parameters of the gas migration model are updated based on the parameter fitness value. After multiple iterations, the optimal parameter combination is obtained as the optimization parameters. The optimized parameters are imported into the gas migration model to drive the inversion calculation of the current pollution diffusion path. The initial distribution of path nodes is determined by combining the topographic features and meteorological conditions of the pollution monitoring area. The node positions and correlations are adjusted by node concentration error feedback to iteratively update the path nodes in the current pollution diffusion path. The pollution diffusion inversion path is output based on the updated path node information. The current pollution diffusion path and the pollution diffusion inversion path both correspond to the target monitoring period.

2. The pollution diffusion path modeling method based on multiple sensors as described in claim 1, characterized in that, The basic model structure for establishing a gas migration model based on the spatiotemporal features, through the basic model structure, associates the mapping relationship between the original observation data and the spatiotemporal features, and determines the correlation weights between the input and output of the gas migration model by combining feature responsivity analysis, generating a parameter self-learning strategy for the gas migration model, including: The spatiotemporal characteristics are decomposed into spatial distribution characteristics and temporal variation characteristics. The spatial distribution characteristics reflect the spatial gradient and distribution pattern of the concentration of polluting gas in the pollution monitoring area, and the temporal variation characteristics reflect the fluctuation pattern and trend of gas concentration at the calibration location over time. The raw observation data are classified according to the acquisition time period and sensor type, and a two-way mapping relationship is established between the classified raw observation data and spatial distribution characteristics and temporal variation characteristics. The basic model structure for constructing a gas migration model is based on the bidirectional mapping relationship. The basic model structure includes a feature input layer, a parameter operation layer, and a result output layer. The feature input layer is used to receive the spatiotemporal features, the parameter operation layer is used to perform model calculations, and the result output layer is used to output pollution diffusion-related parameters. By adjusting the values ​​of the spatiotemporal features as model inputs one by one using the controlled variable method, the changes in the predicted values ​​of pollutant gas concentration distribution output by the gas migration model are monitored. The degree of change in the predicted values ​​caused by adjusting each feature individually is normalized to obtain the normalized influence coefficients corresponding to each spatiotemporal feature. Based on the normalized influence coefficients, the correlation weights between each spatiotemporal feature and the model output are determined. The parameter adjustment rules are generated by combining the correlation weights and the bidirectional mapping relationship, and the initial parameter self-learning strategy of the gas migration model is generated by combining the parameter adjustment rules. Using some of the original observation data and corresponding spatiotemporal features as validation data, the gas migration model is input and the initial parameter self-learning strategy is activated. Based on the deviation between the model output after strategy execution and the actual monitoring results, the amplitude coefficient in the parameter adjustment rule is corrected to generate the parameter self-learning strategy.

3. The pollution diffusion path modeling method based on multiple sensors as described in claim 1, characterized in that, The process involves initiating the parameter self-learning strategy and initializing the particle swarm search space. The spatiotemporal features are input into the gas migration model to obtain an initial parameter combination. The parameter fitness value is calculated through iterative updates of the particle swarm's position, and the target parameters of the gas migration model are updated based on this fitness value. After multiple iterations, the optimal parameter combination is obtained as the optimization parameters, including: The parameter self-learning strategy is activated, the parameter adjustment rules and feature association weights contained in the parameter self-learning strategy are read, and the particle swarm search space is initialized based on the parameter value range of the gas migration model. Each particle corresponds to a set of model parameter combinations, and the position coordinates of the particle are mapped to the parameter values. The spatiotemporal features are input into the gas migration model, and the initial parameter combination is called to perform calculations to obtain the initial pollution diffusion simulation results. The concentration distribution data and time series change data in the initial pollution diffusion simulation results are extracted as the initial output. Based on the original observation data, the errors in spatial concentration distribution and temporal variation patterns between the initial output and the original observation data are calculated respectively; the spatial concentration distribution error and the temporal variation pattern error are normalized to obtain spatial error index and temporal error index respectively; based on the spatial error index, the temporal error index and their respective weights determined by the feature response degree analysis, a comprehensive fitness function is constructed and the initial fitness value is calculated; Based on the adjustment rules in the parameter self-learning strategy, and combined with the difference between the initial fitness value and the preset standard, the iteration direction and step size of the particle swarm are determined, and the position coordinates of each particle are updated to obtain a new parameter combination. Substitute the new parameter combination into the gas migration model and recalculate to obtain new pollution diffusion simulation results and calculate the corresponding fitness value. Record the optimal fitness value and corresponding parameter combination in each iteration. When the change in the optimal fitness value after multiple iterations is less than the preset change range, or when the number of iterations reaches the set upper limit, the iteration stops, and the final recorded optimal parameter combination is determined as the optimization parameter.

4. The pollution diffusion path modeling method based on multiple sensors as described in claim 3, characterized in that, The step of determining the iteration direction and step size of the particle swarm based on the adjustment rules in the parameter self-learning strategy, combined with the difference between the initial fitness value and the preset standard, and updating the position coordinates of each particle to obtain a new parameter combination includes: The adjustment rules in the parameter self-learning strategy are analyzed, and the parameter adjustment direction, adjustment ratio and constraint conditions corresponding to different fitness value ranges are extracted to establish a matching table between fitness values ​​and parameter adjustment rules. Calculate the difference between the initial fitness value and the preset standard, match the corresponding parameter adjustment direction and adjustment ratio from the matching table according to the interval where the difference is located, and correct the adjustment ratio in combination with the parameter constraints of the gas migration model to generate the target adjustment amount for each parameter. The inertia weight formula based on the particle swarm optimization algorithm is used to calculate the inertia weight of the current iteration by combining the initial fitness value. The inertia weight reflects the degree to which a particle retains its own historical position information. Based on the inertia weight, iteration direction, and target adjustment amount, combined with the particle's own historical best position and the global best position of the particle swarm, the position coordinates of each particle are updated, and the updated particle position coordinates are converted into corresponding parameter values ​​to obtain a new parameter combination. When the change in the optimal fitness value over multiple consecutive iterations is less than a preset change range, or when the number of iterations reaches a set upper limit, the iteration stops, and the final recorded optimal parameter combination is determined as the optimization parameters, including: The iteration termination is determined by two conditions: the first condition is that the change in the optimal fitness value for a consecutive preset number of iterations is less than the preset change range; the second condition is that the number of iterations reaches the preset maximum number of iterations. After each iteration, record the optimal fitness value and the corresponding iteration number of the current iteration, and calculate the absolute difference between the current optimal fitness value and the optimal fitness value of the previous iteration. The absolute difference is the magnitude of the change in fitness value. The system counts the number of consecutive instances where the fitness value changes by a smaller than a preset range. If this number is reached, the first termination condition is triggered, and the iteration is marked as ready to stop. Simultaneously, the system checks whether the current iteration count has reached the maximum iteration count. If it has, the second termination condition is triggered, and the iteration is marked as ready to stop. When any termination condition is triggered, the iteration process of the particle swarm is stopped, all optimal fitness values ​​and corresponding parameter combinations recorded during the iteration process are retrieved, and the parameter combination with the highest fitness value is selected. The selected parameter combinations are substituted into the gas migration model for verification calculation. Spatiotemporal features are input to obtain the verification simulation results. The final error between the verification simulation results and the original observation data is calculated. If the final error is within the preset tolerance range, the parameter combination is determined as the optimized parameter; if the final error exceeds the preset tolerance range, the termination condition is adjusted, the number of iterations is increased, and the iteration process is re-executed until a parameter combination that meets the preset tolerance range is obtained as the optimized parameter.

5. The pollution diffusion path modeling method based on multiple sensors as described in claim 1, characterized in that, The process involves importing the optimized parameters into the gas migration model to drive the inversion calculation of the current pollution diffusion path. This includes determining the initial distribution of path nodes based on the topographic features and meteorological conditions of the pollution monitoring area, adjusting node positions and relationships through node concentration error feedback to iteratively update the path nodes in the current pollution diffusion path, and outputting the pollution diffusion inversion path based on the updated path node information. This includes: The optimized parameters are imported into the gas migration model and the parameter configuration is completed. The terrain feature data and real-time meteorological condition data of the pollution monitoring area are read. The terrain feature data includes the elevation distribution, obstacle distribution and surface roughness of the pollution monitoring area. The meteorological condition data includes meteorological conditions and temperature and humidity information. Combining the terrain feature data and the meteorological condition data, the potential diffusion direction and obstruction area of ​​pollutants are analyzed, and initial path nodes are set in the dominant diffusion direction to form the initial distribution of path nodes; wherein, the spacing of the initial path nodes is adjusted according to the terrain complexity. The gas migration model is activated to perform the inversion calculation of the current pollution diffusion path. Based on the initial distribution, the gas diffusion process is simulated, and the predicted gas concentration at each initial path node is calculated. Retrieve the measured concentration values ​​from the original observation data at the corresponding location, calculate the concentration error between the predicted and measured concentration values ​​at each initial path node, and mark the nodes whose concentration errors exceed a preset range as nodes to be adjusted. For the nodes to be adjusted, the causes of errors are analyzed by combining the terrain feature data and the meteorological condition data. Based on the causes of errors, the adjustment direction and amount of the node position are determined. At the same time, the correlation weights between nodes are corrected. The correlation weights reflect the degree of diffusion influence between nodes. The adjusted path nodes are re-input into the gas migration model for inversion calculation, and the process jumps to the node concentration error calculation and position adjustment step until the concentration error of all path nodes is within the preset range, thus completing the iterative update of the path nodes in the current pollution diffusion path. Connect the iteratively updated path nodes sequentially according to the diffusion time sequence, label the concentration information and corresponding timestamp of each iteratively updated path node, generate the pollution diffusion inversion path and output it.

6. The pollution diffusion path modeling method based on multiple sensors as described in claim 5, characterized in that, The process of combining the terrain feature data and meteorological condition data to analyze the potential diffusion direction and obstruction area of ​​pollutants, and setting initial path nodes along the dominant diffusion direction to form the initial distribution of path nodes, includes: The terrain feature data is digitally processed to extract elevation change curves, obstacle outline coordinates, and surface roughness coefficients, and a digital terrain model is constructed, in which terrain elements affecting gas diffusion are identified. A time-series stability analysis was performed on the meteorological condition data to determine the prevailing wind direction and average wind speed. Based on the prevailing wind direction, the dominant diffusion direction of the pollutant gas was determined. Starting from the initial location of pollution occurrence, an initial diffusion axis was drawn along the prevailing wind direction. The initial diffusion axis is the core path direction of pollution diffusion. The terrain digitization model is superimposed on the initial diffusion axis to identify obstructed areas on and around the initial diffusion axis. The obstructed areas are areas that hinder gas diffusion, including areas where the elevation changes abruptly or where obstacles are located. Based on the terrain complexity evaluation index, which includes elevation change rate, obstacle density, and surface roughness coefficient, the terrain complexity of the area along the initial diffusion axis is scored. The spacing rules of the initial path nodes are determined according to the terrain complexity score. For complex terrain areas where the terrain complexity score reaches the first preset value, the node spacing is set as the first spacing; for terrain areas where the terrain complexity score is between the first and second preset values, the node spacing is set as the second spacing; for flat areas where the terrain complexity score is lower than the second preset value, the node spacing is set as the third spacing, and the first spacing is smaller than the second spacing, and the second spacing is smaller than the third spacing. Along the initial diffusion axis, initial path nodes are sequentially deployed according to the set spacing rules. Additional nodes are added before and after the obstructed area to capture diffusion changes, forming an initial distribution that includes node coordinates, corresponding terrain complexity scores, and estimated concentrations.

7. The pollution diffusion path modeling method based on multiple sensors as described in claim 5, characterized in that, For the nodes to be adjusted, the causes of errors are analyzed by combining the terrain feature data and the meteorological condition data. Based on the causes of errors, the adjustment direction and amount of the node positions are determined, and the correlation weights between nodes are corrected, including: Extract the coordinate information of the node to be adjusted and locate it in the terrain digitization model. Simultaneously retrieve the original observation data and basic data used for inversion calculation for the corresponding collection period of the node to be adjusted to construct a node feature dataset. Based on the node feature dataset, topographic features including elevation gradient, obstacle distribution and surface permeability are extracted within a preset range around the node, and the meteorological feature change trend of the corresponding time period is extracted to form a topographic-meteorological feature correlation matrix. The topographic-meteorological feature correlation matrix is ​​matched with the physical feature model of pollutant gas diffusion. The main cause of error is determined by feature fit analysis. If the fit between the topographic feature features and the diffusion model is lower than the preset threshold, the topographic shading is determined to be the main source of error. If the deviation between the trend of meteorological feature changes and the model input data exceeds the allowable range, the fluctuation of meteorological conditions is determined to be the main source of error. To address errors caused by terrain occlusion, the characteristic path of gas diffraction and diffusion is analyzed by combining the spatial contour features of obstacles. The node to be adjusted is moved to the core area of ​​the diffraction path so that the node can capture the gas concentration characteristics after diffraction. To address errors caused by fluctuating meteorological conditions, the dominant diffusion direction is updated based on real-time meteorological characteristics, and the nodes are redeployed along the new dominant direction to ensure that the node distribution is consistent with the actual gas diffusion trend. Based on the spatial distribution relationship of the adjusted nodes and combined with the concentration characteristics of adjacent nodes, a diffusion influence logic between nodes is constructed, and the allocation rules of node association weights are redefined so that the association weights reflect the diffusion transmission relationship between nodes.

8. The pollution diffusion path modeling method based on multiple sensors as described in claim 5, characterized in that, The updated path nodes are sequentially connected according to the diffusion time sequence, and the concentration information and corresponding timestamp of each node are labeled. This process is then integrated to form a complete pollution diffusion inversion path and output, including: Extract the timestamps and concentration features of all updated path nodes, construct a spatiotemporal feature matrix of nodes, and sort the updated path nodes in an orderly manner based on the timestamps to determine the temporal evolution logic of pollution diffusion. Based on the sorted node sequence, the rationality of the concentration transition between adjacent nodes is analyzed in conjunction with the continuity characteristics of gas diffusion. If there is a sudden change in concentration between adjacent nodes and it does not conform to the physical laws of diffusion, a transition feature model is constructed based on the spatiotemporal characteristics between adjacent nodes to generate transition nodes that conform to the diffusion trend and fill the gaps in diffusion features between adjacent nodes. Multidimensional feature labels are constructed for each updated path node and supplementary transition node. The content of the multidimensional feature labels covers node coordinates, concentration characteristics, and terrain and meteorological background information for the corresponding time period. Using the updated temporal order of path nodes as the axis, and combining the strength differences of node concentration characteristics, path visualization rules are set to associate concentration characteristics with path presentation form, so that the path reflects the differences in pollution levels in different areas. The sorted node sequence is overlaid and fused with the digital terrain model of the pollution monitoring area to establish the spatial relationship between path nodes and terrain features, and the fused path data is obtained. Based on the fused path data and multi-dimensional feature labels, a pollution diffusion inversion path is constructed. Based on the pollution diffusion inversion path, a complete output result including a path time series evolution diagram, spatial distribution visualization, and node feature lookup table is generated.

9. A pollution diffusion path modeling system, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the multi-sensor-based pollution diffusion path modeling method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a computer program that, when run on the pollution diffusion path modeling system, causes the pollution diffusion path modeling system to perform the steps of the multi-sensor-based pollution diffusion path modeling method according to any one of claims 1 to 8.

Images