A method for simulating pollutant concentration diffusion based on grid optimization and self-regulation

Abstract

The present application relates to a pollutant concentration diffusion simulation method based on grid optimization and self-regulation. The method includes: dividing the simulation area into basic grids to obtain basic grids; calculating the pollutant concentration field of the simulation area according to the spatial position of the monitoring station and the pollutant concentration value; dynamically adjusting the grid density of the basic grid after generating the concentration distribution matrix according to the pollutant concentration field of the simulation area; calculating the computational complexity of each grid area using the grid density of the adjusted basic grid, assigning pollutant concentration calculation tasks according to the cloud platform and weighted polling algorithm, and the pollutant concentration prediction results of each grid constitute the pollutant concentration distribution prediction results of the simulation area; dynamically adjusting the pollutant concentration distribution prediction results according to the Kalman filter and the acquired real-time environmental data to obtain the final pollutant concentration prediction value. The use of this method can improve the accuracy and real-time performance of pollution diffusion simulation.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a pollutant concentration diffusion simulation method based on grid optimization and self-regulation. Background Art

[0002] With the continuous acceleration of industrialization and urbanization, environmental pollution problems are becoming increasingly serious, especially the spread of biochemical pollutants such as chemical agents and biological pathogens, which pose a huge threat to environmental quality, ecosystems and human health. In practical applications, pollutant diffusion simulation technology has become a key tool for environmental governance and emergency response. In sudden environmental pollution incidents, the path and impact range of pollutant diffusion are often complex and difficult to predict. Therefore, rapid and accurate simulation of the diffusion process of pollutants has become a core requirement in the field of environmental management and emergency response. Traditional pollution diffusion simulation methods usually use fixed grid division technology and static calculation mode, which have many limitations.

[0003] Existing pollution diffusion models are usually calculated based on fixed grid division. This method can meet the simulation needs of small and medium-scale pollution scenes to a certain extent, but it is difficult to take into account both accuracy and efficiency in high-concentration pollution areas and low-concentration pollution areas. In high-concentration areas, the fixed grid resolution may be insufficient, resulting in the simulation results failing to meet actual needs, while in low-concentration areas, excessively high grid density will cause a large amount of unnecessary computing overhead, significantly reducing resource utilization. In addition, traditional simulation methods usually rely on single-machine computing and cannot flexibly expand computing resources, resulting in slow simulation speeds when facing large-scale or complex scenes, making it difficult to meet the real-time requirements of emergencies. These technical bottlenecks have formed major obstacles in terms of accuracy, efficiency and flexibility in pollutant diffusion simulation. In addition, another limitation of traditional models is their lack of responsiveness to dynamic changes in the environment. The pollutant diffusion process is significantly affected by meteorological conditions (such as wind speed, wind direction, temperature, humidity, etc.), which may change over time, and their dynamic characteristics are difficult to display in a timely manner. Existing methods usually use static model parameters, which are difficult to respond to changes in environmental data in real time, resulting in deviations between simulation results and actual conditions, weakening the accuracy of pollution prediction and the effectiveness of emergency response. Therefore, in order to deal with the problem of pollution diffusion in complex dynamic environments, a new technology is needed that can adjust the grid density in real time, flexibly schedule computing resources, and optimize the model based on dynamic environmental data to provide an accurate and efficient solution for the simulation of biochemical pollutant diffusion. Summary of the invention

[0004] Based on this, it is necessary to provide a pollutant concentration diffusion simulation method based on grid optimization and self-regulation that can improve the accuracy and real-time performance of pollution diffusion simulation in response to the above technical problems.

[0005] A pollutant concentration diffusion simulation method based on grid optimization and self-regulation, the method comprising:

[0006] Acquire the simulation area; divide the simulation area into basic grids to obtain basic grids; calculate the pollutant concentration field of the simulation area according to the spatial position of the monitoring station and the pollutant concentration value; generate a concentration distribution matrix according to the pollutant concentration field of the simulation area;

[0007] The preset concentration threshold is compared with the concentration distribution matrix, and the grid density of the basic grid is dynamically adjusted according to the comparison result; the computational complexity of each grid area is calculated using the adjusted grid density of the basic grid, and the computational complexity is normalized to obtain the weight value of the grid;

[0008] According to the cloud platform and the weighted pollutant concentration calculation algorithm, the pollutant concentration prediction result of each grid is obtained; the pollutant concentration prediction result of each grid constitutes the pollutant concentration distribution prediction result of the simulation area;

[0009] The pollutant concentration distribution prediction results are dynamically adjusted according to the Kalman filter and the real-time environmental data obtained, the observation equation and Kalman gain are calculated, and the final pollutant concentration prediction value is calculated using the observation equation and Kalman gain.

[0010] The above-mentioned pollutant concentration diffusion simulation method based on grid optimization and self-regulation first constructs a concentration distribution matrix by dividing the basic grid and combining it with the monitoring station data. Based on this, the grid density is dynamically adjusted by comparing it with the concentration threshold. The grid is refined in the high-concentration area to capture details, and the low-concentration area is reasonably simplified to avoid resource waste and precision loss. Secondly, the Kalman filter is used to integrate real-time environmental data to dynamically correct the prediction results. Since the pollutant diffusion is greatly affected by environmental factors such as meteorology, this can reduce the deviation from the actual situation. Furthermore, the grid calculation complexity is considered and normalized to obtain the weight value, and the calculation tasks are allocated based on the cloud platform and weighted polling algorithm to ensure the calculation accuracy of each grid and the rational use of resources. In terms of real-time improvement, dynamic grid density adjustment reduces the redundant calculation amount in the low-concentration area and accelerates the simulation process; the cloud platform and weighted polling algorithm give full play to the scalability and resource scheduling flexibility of the cloud platform to efficiently handle tasks; the Kalman filter responds to environmental changes in a timely manner and quickly updates the simulation results, so that the simulation can track the pollutant diffusion trend in real time, thereby effectively coping with the pollution diffusion problem in a complex dynamic environment, and providing an accurate and efficient solution for the simulation of biochemical pollutant diffusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 The figure is a flow chart of a pollutant concentration diffusion simulation method based on grid optimization and self-regulation in one embodiment. DETAILED DESCRIPTION

[0012] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0013] In one embodiment, Figure 1 As shown, a pollutant concentration diffusion simulation method based on grid optimization and self-regulation is provided, comprising the following steps:

[0014] Step 102, obtaining a simulation area; dividing the simulation area into basic grids to obtain basic grids; calculating the pollutant concentration field of the simulation area according to the spatial position of the monitoring station and the pollutant concentration value; generating a concentration distribution matrix according to the pollutant concentration field of the simulation area.

[0015] Step 104, compare the preset concentration threshold with the concentration distribution matrix, and dynamically adjust the grid density of the basic grid according to the comparison result; use the adjusted grid density of the basic grid to calculate the computational complexity of each grid area, normalize the computational complexity, and obtain the weight value of the grid.

[0016] Traditional pollution diffusion simulation methods usually use fixed grid division technology, which has obvious shortcomings when facing different concentration areas. For example, in high-concentration pollution areas, fixed grid resolution is often difficult to accurately depict the rapid changes in pollutant concentrations, resulting in the simulation results being unable to accurately present the actual distribution characteristics of pollutants, and many key details are lost, limiting the simulation accuracy.

[0017] This application first divides the simulation area into basic grids to obtain the basic grid, and then calculates the pollutant concentration field of the simulation area based on the spatial position of the monitoring station and the pollutant concentration value, and then generates a concentration distribution matrix. Then the pre-set concentration threshold is compared with the concentration distribution matrix, and the grid density of the basic grid is dynamically adjusted according to the comparison result. In this way, in high-concentration areas, the grid density can be increased in time according to the concentration distribution, and details such as slight changes in pollutant concentration and changes in concentration gradients can be captured more finely, just like using a finer "screen" to filter pollutant concentration information, so that the simulation results are more in line with the actual high-concentration pollution distribution, thereby improving the simulation accuracy.

[0018] For example, if a sudden pollution incident occurs in a chemical park and the pollutant concentration in the local area is extremely high, the method of the present application can refine the grid in the high-concentration area and accurately present the rapid changes in concentration of pollutants as they diffuse outward from the leakage source. The traditional fixed grid method may blur these details and fail to accurately reflect the actual concentration changes.

[0019] At the same time, in areas of low pollution concentration, traditional fixed grid division cannot flexibly change the grid density, so a relatively high grid resolution will be set to cope with possible complex situations. However, this causes a lot of unnecessary computational overhead, which not only wastes computing resources, but also these redundant calculations will not substantially improve the overall simulation accuracy. It may even lead to unreasonable resource allocation due to excessive attention to the details of low-concentration areas (which have relatively little impact on the overall pollution situation), affecting the accuracy control of the simulation in key high-concentration areas.

[0020] This application dynamically adjusts the grid density, and can appropriately reduce the grid density in low-concentration areas. Under the premise of ensuring that the general distribution trend of pollutants can be reasonably reflected, excessive useless calculations can be avoided, so that computing resources can be more concentrated on high-concentration areas and key concentration change areas that have a greater impact on simulation accuracy. For example, in a large area of ​​low concentration formed by the dilution and diffusion of pollutants in a city, this application can use a relatively sparse grid to describe its approximate concentration, and leave more computing resources for high-concentration areas such as those near pollution sources that require more accurate simulation, thereby optimizing the overall simulation accuracy.

[0021] The computational complexity of each grid area is calculated using the grid density of the adjusted basic grid, and then normalized to obtain the grid weight value. Based on the characteristics of each grid area, such as the complexity of pollutant concentration changes within the grid, the size of the spatial range, and other factors, the computational complexity is comprehensively considered, providing a scientific quantitative basis for the subsequent reasonable allocation of computing resources, so that each grid can obtain computing resource investment that matches its actual needs, thereby ensuring the accuracy of simulation calculations in each area and improving the accuracy of the overall simulation.

[0022] Step 106, the pollutant concentration calculation task is allocated according to the cloud platform and the weighted polling algorithm to obtain the pollutant concentration prediction result of each grid; the pollutant concentration prediction result of each grid constitutes the pollutant concentration distribution prediction result of the simulation area.

[0023] The pollutant concentration calculation tasks are allocated based on the cloud platform and weighted polling algorithm. The cloud platform has strong computing power and scalability, and can meet the computing needs of large-scale complex simulation scenarios. The weighted polling algorithm allocates the computing tasks to each grid more reasonably based on the grid weight values ​​obtained earlier, ensuring that the pollutant concentration prediction tasks of each grid can be processed in a timely and accurate manner, avoiding simulation errors caused by unreasonable task allocation or limited computing resources, and helping to improve the accuracy of pollutant concentration diffusion simulation in the entire simulation area.

[0024] Step 108, dynamically adjust the pollutant concentration distribution prediction result according to the Kalman filter and the acquired real-time environmental data, calculate the observation equation and the Kalman gain, and use the observation equation and the Kalman gain to calculate the final pollutant concentration prediction value.

[0025] The Kalman filter is used to dynamically adjust the prediction results in real time in combination with real-time environmental data, so that the simulation can quickly respond to dynamic changes in the environment. When faced with real-time changes in environmental factors such as meteorological conditions, the simulation results can be updated in a timely manner, ensuring the real-time nature of the simulation and allowing the simulation results to more promptly reflect the actual diffusion of pollutants.

[0026] In the above-mentioned pollutant concentration diffusion simulation method based on grid optimization and self-regulation, firstly, the concentration distribution matrix is ​​constructed by dividing the basic grid and combining the monitoring station data. Based on this, the grid density is dynamically adjusted by comparing with the concentration threshold. The grid is refined in the high-concentration area to capture details, and the low-concentration area is reasonably simplified to avoid resource waste and precision loss. Secondly, the Kalman filter is used to integrate real-time environmental data to dynamically correct the prediction results. Since the pollutant diffusion is greatly affected by meteorological and other environmental factors, this can reduce the deviation from the actual situation. Furthermore, the grid calculation complexity is considered and normalized to obtain the weight value, and the calculation tasks are allocated based on the cloud platform and weighted polling algorithm to ensure the calculation accuracy of each grid and the rational use of resources. In terms of real-time improvement, dynamic grid density adjustment reduces the redundant calculation amount in the low-concentration area and accelerates the simulation process; the cloud platform and weighted polling algorithm give full play to the scalability and resource scheduling flexibility of the cloud platform to efficiently handle tasks; the Kalman filter responds to environmental changes in a timely manner and quickly updates the simulation results, so that the simulation can track the pollutant diffusion trend in real time, thereby effectively coping with the pollution diffusion problem in a complex dynamic environment, and providing an accurate and efficient solution for the simulation of biochemical pollutant diffusion.

[0027] In one embodiment, the pollutant concentration field includes concentration estimates of a plurality of monitoring stations; and calculating the pollutant concentration field of the simulation area according to the spatial locations of the monitoring stations and the pollutant concentration values ​​includes:

[0028] The estimated concentration of the monitoring station in the simulation area is calculated based on the spatial location of the monitoring station and the pollutant concentration value.

[0029] ;

[0030] in, It's location The estimated concentration at For location The mean value of is the weight coefficient of the known pollution source, Known pollution source The concentration value of For pollution source The mean of .

[0031] In a specific embodiment, pollutant data (such as PM2.5, NOx, SO2, etc.) and meteorological data (such as wind speed, wind direction, temperature, humidity, etc.) from multiple environmental monitoring stations (such as air quality monitoring stations) are read in. These data are used to estimate the pollutant concentration at each grid point, thereby providing accurate input for pollution diffusion simulation.

[0032] For example, suppose there are data from two monitoring stations, monitoring station 1 is at location (30.0°N, 120.0°E), with a PM2.5 concentration of 120 µg / m³, and monitoring station 2 is at location (30.1°N, 120.1°E), with a PM2.5 concentration of 50 µg / m³. Through Kriging interpolation, the system calculates the pollutant concentration field of the entire area based on the spatial location of the monitoring stations and the pollutant concentration values, as follows:

[0033] ;

[0034] in, It's location The estimated concentration at is the mean value at that location, is the weight coefficient of the known pollution source, is the concentration value of the known pollution source, is the mean value of the pollution source location. This step expands the concentration data of the monitoring point to the entire area, providing the necessary pollutant concentration distribution for dynamic adjustment of the grid density.

[0035] In one embodiment, dynamically adjusting the mesh density of the base mesh according to the comparison result includes:

[0036] If the concentration distribution matrix , refine the basic grid, then the grid density of the adjusted basic grid is:

[0037] ;

[0038] like , refine the basic grid, then the grid density of the adjusted basic grid is:

[0039] ;

[0040] in, is the initial mesh density, is the grid adjustment scale factor.

[0041] In one embodiment, the computational complexity of each grid area is calculated using the grid density of the adjusted basic grid, including:

[0042] The computational complexity of each grid area is calculated using the grid density of the adjusted basic grid:

[0043] ;

[0044] in, is the grid area, is the mesh density of the adjusted base mesh.

[0045] In a specific embodiment, the complexity of all grids is calculated, and the results are normalized to convert the calculation complexity into corresponding weight values. ,Right now:

[0046] ;

[0047] in Indicates The weight of the grid, is the total number of grids.

[0048] In one embodiment, pollutant concentration calculation tasks are allocated based on a cloud platform and a weighted round-robin algorithm to obtain pollutant concentration prediction results for each grid, including:

[0049] Initialize the computing node set according to the resource situation of the cloud platform , each node With initial weight , all pollutant concentration calculation tasks are used as task pools Enter the weighted polling algorithm, the remaining processing capacity of each node is calculated by its initial weight Indicates that the weight of the task is calculated based on the pollutant concentration and The relative size of pollutant concentrations with high weights is assigned to nodes with high weights first;

[0050] Each assigned task After that, update the remaining weight of the node:

[0051] ;

[0052] If the remaining weight of a node is not enough to continue receiving tasks, the node is skipped and assigned to the next node;

[0053] In the nodes assigned with the pollutant concentration calculation task, the pollutant concentration is calculated according to the pollution diffusion model to obtain the pollutant concentration prediction result of each grid.

[0054] In a specific embodiment, it is assumed that there are three computing nodes , and their initial weights are , , There are five grid tasks , and their weights are . Task allocation steps according to the weighted round-robin algorithm:

[0055] Round 1: Mission Assigned to the node with the highest weight (Remaining weight 0). Task Assigned to (Remaining weight 0). Task Assigned to (Remaining weight 0).

[0056] Second round: After all node weights are reset and updated, the remaining tasks and Assigned to and .

[0057] Final result: All tasks are assigned, and tasks are assigned to high-performance nodes first. The entire process completes the dynamic allocation of resources and load balancing.

[0058] Assuming that after a high-pollution area is refined, the computational complexity reaches 4 times of the initial level, while that of a low-pollution area is only 0.5 times of the initial level, the computational nodes in the high-pollution area can be increased by 3 times, while the low-pollution area can be reduced to 50% of the original level. This not only optimizes computational efficiency, but also avoids the problem of uneven resource allocation.

[0059] During the running of computing tasks, the cloud platform monitors the load (CPU, memory usage, etc.) and execution progress of each computing node in real time. These monitoring data are collected in real time through distributed monitoring tools (such as Prometheus or Zabbix) and aggregated to a central scheduling system for subsequent load balancing decisions. For example, in a 10-node computing cluster, the CPU usage of a node exceeds 80%, while the usage of another node is only 30%. At this time, the load will be marked as uneven. If a node is overloaded due to too many tasks assigned to a high-density grid, the load balancing mechanism is started to migrate some tasks of the node to nodes with lighter loads. The basis for migration is the remaining available resources of the node and the priority of the task. For example, for a node with an occupancy rate of more than 80%, the system can choose to migrate 20% of its load to a node with an occupancy rate of less than 50%.

[0060] Task migration is a key link in load balancing. In order to ensure that the task migration process is efficient and does not affect the calculation progress, the system adopts the following measures: Before migrating the task, the system will save the intermediate state of the task being executed (such as grid data, concentration distribution matrix, etc.) to a shared storage system (such as HDFS or distributed database). After the migration is completed, the new node resumes the task from the save point to avoid recalculating the completed part.

[0061] Take a sudden pollution incident in an industrial area as an example. The system needs to simulate the high-concentration area around the pollution source with high precision. In this case, the mesh of the high-concentration area is refined, resulting in a surge in the computational complexity of some nodes. Without load balancing, these nodes may be overloaded, delaying the entire simulation process.

[0062] Through the load balancing mechanism: the system monitors that the resource utilization rate of computing nodes in high-concentration areas exceeds 90%, and immediately triggers the task migration mechanism; migrates some low-priority tasks (such as calculations in low-concentration areas far away from pollution sources) to other nodes with lighter loads; dynamically allocates more resources to computing nodes in high-concentration areas, including additional memory and CPU resources. After load balancing, the diffusion simulation that was originally expected to take 2 hours to complete was actually completed within 1 hour, and the simulation accuracy of the high-concentration area was guaranteed.

[0063] In the process of pollution concentration calculation, the pollution diffusion model can be CALPUFF.

[0064] In one embodiment, calculating the observation equation and the Kalman gain includes:

[0065] The observation equation is calculated as

[0066] ;

[0067] in, is the pollutant concentration observed in real time, is the observation matrix, mapping the system state to the observation value, is the system state, representing the pollutant concentration predicted by the model, is the observation noise, which represents the sensor measurement error.

[0068] In one embodiment, calculating the observation equation and the Kalman gain further includes:

[0069] The Kalman gain is calculated as

[0070] ;

[0071] in, is the forecast error covariance matrix, which represents the An estimate of the uncertainty or error in the forecast at the previous moment; is the covariance matrix of the observation noise, which reflects the error or uncertainty of the observation data itself, and usually depends on the accuracy of the sensor and the observation environment; is the observation matrix, which converts the predicted results of pollutant concentration distribution into the actual monitored pollutant concentration.

[0072] In one embodiment, the updated pollutant concentration prediction value is obtained by using the observation equation and the Kalman gain calculation, including:

[0073] The updated pollutant concentration prediction value is obtained by using the observation equation and Kalman gain calculation:

[0074] ;

[0075] in, is the pollutant concentration observed in real time, is the predicted result of pollutant concentration distribution, is the Kalman gain, is the estimated pollutant concentration after filter update.

[0076] In a specific embodiment, the pollutant concentration distribution prediction results and the real-time pollutant concentration observation values ​​are obtained (the real-time monitoring data includes the pollutant concentrations of each environmental monitoring station (such as PM2.5, PM10, etc.), etc.) and the above two types of data are input into the Kalman filter to self-regulate the pollutant concentration.

[0077] The Kalman filter can be used to predict and update pollutant concentrations. First, the prediction step of the pollution diffusion model is used to obtain an estimate of the pollutant concentration. . Establish the observation equation:

[0078] ;

[0079] in, is the observed value, indicating the actual pollutant concentration, is the observation matrix, mapping the system state to the observation value, is the system state, representing the pollutant concentration predicted by the model, is the observation noise, which represents the sensor measurement error.

[0080] Calculate Kalman gain: Kalman gain represents the relative weight between the model prediction results and the observed data. The larger the gain, the higher the weight of the observed data.

[0081] The Kalman gain can be expressed as :

[0082] ;

[0083] in, is the forecast error covariance matrix, which represents the An estimate of the uncertainty or error in the forecast at the previous moment; is the covariance matrix of the observation noise, which reflects the error or uncertainty of the observation data itself, and usually depends on the accuracy of the sensor and the observation environment; is the observation matrix, which converts the state estimate (i.e., the predicted value of the pollutant concentration distribution) into the actual observation value (such as the actual monitored pollutant concentration).

[0084] Forecast error covariance matrix The uncertainty of the model's prediction of pollutant concentrations is quantified. In the Kalman filter, the error covariance matrix changes continuously with each time step update. Its update formula is:

[0085] ;

[0086] Where 𝐼 is the identity matrix. At each time step, the difference between the predicted value and the observed value will affect the subsequent calculation of the error covariance matrix, which in turn affects the calculation of the Kalman gain.

[0087] At the initial moment, it is assumed that there is a rough error covariance matrix (which can be determined based on experience or prior information), and then the error covariance is continuously modified with each prediction and update.

[0088] The covariance matrix 𝑅 of the observation noise reflects the noise level of the data collected from the monitoring equipment. For each pollutant monitoring instrument, the measurement error may come from multiple factors (such as sensor accuracy, environmental changes, external interference, etc.). Typically, 𝑅 is a diagonal matrix in which the diagonal elements represent the noise variance of each observed variable (pollutant concentration) at a certain time step. Assuming that we have estimated the standard deviation of the observation noise through error analysis or through experimental data (for example, the error when the monitoring station measures the pollutant concentration is ±5 µg / m³), then we can use the variance of this value (that is, the square of the error) as an element in 𝑅. For example, if the observation error of the pollutant concentration is normally distributed, the observation noise can be represented by the standard deviation, Each element of can be expressed as:

[0089] ;

[0090] in, The pollutant concentration is The standard deviation of the measurement error of each type.

[0091] Kalman Gain Controls the weighted ratio of predicted values ​​to observed values. The larger the gain, the greater the influence of the observed values ​​and the smaller the influence of the predicted values; conversely, the smaller the gain, the greater the prediction influence of the model.

[0092] Therefore, the final calculation of the Kalman gain is determined by the error covariance, observation noise, and the difference in predicted values. A higher error covariance and lower observation noise will result in a larger gain, making the model rely more on the observed data; conversely, more on the model predictions.

[0093] Next, update the predicted pollutant concentration based on the actual observed value:

[0094] ;

[0095] in, is the pollutant concentration observed in real time, is the pollutant concentration predicted by the CALPUFF model, is the Kalman gain, is the estimated pollutant concentration after filter update.

[0096] Is the difference between the observed value and the predicted value, i.e. the incremental (Innovation) or residual. This residual is used to correct the prediction results. By calculating this difference, the Kalman filter can identify the error between the predicted value and the actual observed value and weight it when updating the estimate.

[0097] For example, in pollutant concentration prediction, if the Kalman filter predicts a pollution concentration of 50 µg / m³, but the real-time observed pollution concentration is 45 µg / m³, the residual is:

[0098] ;

[0099] This error value will adjust the final pollutant concentration estimate through a weighted combination of the Kalman gain and the predicted results. Update the error covariance: Update the error covariance matrix to reflect the uncertainty changes brought by the new observation data;

[0100] ;

[0101] in, is the updated error covariance matrix, is the identity matrix.

[0102] Through the above steps, the model can reflect environmental changes in real time and adjust the predicted value of pollutant concentration. For example, suppose that after filtering, the predicted pollutant concentration is adjusted from the initial 200 µg / m³ to 250 µg / m³. This process quickly adjusts the concentration and distribution of pollutants according to the observed values ​​of the environment, thereby improving the accuracy of the prediction.

[0103] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0104] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0105] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0106] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A pollutant concentration diffusion simulation method based on grid optimization and self-regulation, characterized in that: The method comprises: Acquire a simulation area; divide the simulation area into basic grids to obtain basic grids; calculate the pollutant concentration field of the simulation area according to the spatial position of the monitoring station and the pollutant concentration value; generate a concentration distribution matrix according to the pollutant concentration field of the simulation area; Compare the preset concentration threshold with the concentration distribution matrix, and dynamically adjust the grid density of the basic grid according to the comparison result; calculate the computational complexity of each grid area using the adjusted grid density of the basic grid, normalize the computational complexity, and obtain the weight value of the grid; According to the cloud platform and the weighted pollutant concentration calculation algorithm, the pollutant concentration calculation task of the grid with a high weight value is assigned to the high-weight node to obtain the pollutant concentration prediction result of each grid; the pollutant concentration prediction result of each grid constitutes the pollutant concentration distribution prediction result of the simulation area; Dynamically adjust the pollutant concentration distribution prediction result according to Kalman filtering and the acquired real-time environmental data, calculate the observation equation and Kalman gain, and use the observation equation and Kalman gain to calculate the final pollutant concentration prediction value; Dynamically adjust the mesh density of the base mesh based on the comparison results, including: If the concentration distribution matrix , refine the base grid, then the grid density of the adjusted base grid is: ； like , refine the basic grid, then the grid density of the adjusted basic grid is: ； in, is the initial mesh density, is the grid adjustment scale factor.

2. The method according to claim 1, characterized in that The pollutant concentration field includes concentration estimation values ​​of a plurality of monitoring stations; the pollutant concentration field of the simulation area is calculated according to the spatial positions of the monitoring stations and the pollutant concentration values, including: The estimated concentration of the monitoring station in the simulation area is calculated based on the spatial location of the monitoring station and the pollutant concentration value. in, It's location The estimated concentration at For location The mean pollutant concentration at is the weight coefficient of the known pollution source, Known pollution source The concentration value of the pollutant concentration at For pollution source The mean pollutant concentration.

3. The method according to claim 1, characterized in that The computational complexity of each grid area is calculated using the grid density of the adjusted base grid, including: The computational complexity of each grid area is calculated using the grid density of the adjusted basic grid: in, is the grid area, is the mesh density of the adjusted base mesh.

4. The method according to claim 1, characterized in that According to the cloud platform and weighted pollutant concentration calculation algorithm, the pollutant concentration calculation tasks with high grid weight values ​​are assigned to high-weight nodes to obtain the pollutant concentration prediction results for each grid, including: Initialize the computing node set according to the resource situation of the cloud platform , each node With initial weight , all pollutant concentration calculation tasks are used as task pools Enter the weighted polling algorithm, the remaining processing capacity of each node is calculated by its initial weight Indicates that the weight value of the grid is calculated according to the pollutant concentration and The relative size of the grid weights is used to assign pollutant concentration calculation tasks with high grid weights to high-weight nodes first; Each assigned task After that, update the remaining weight of the node: If the remaining weight of a node is not enough to continue receiving tasks, the node is skipped and assigned to the next node; In the nodes assigned with the pollutant concentration calculation task, the pollutant concentration is calculated according to the pollution diffusion model to obtain the pollutant concentration prediction result of each grid.

5. The method according to claim 1, characterized in that Calculate the observation equation and Kalman gain, including: The observation equation is calculated as in, is the pollutant concentration observed in real time, is the observation matrix, mapping the system state to the observation value, is the system state, representing the pollutant concentration predicted by the model, is the observation noise, which represents the sensor measurement error.

6. The method according to claim 5, characterized in that Calculate the observation equation and Kalman gain, including: The Kalman gain is calculated as in, is the forecast error covariance matrix, which represents the An estimate of the uncertainty or error in the forecast at the previous moment; is the covariance matrix of the observation noise, which reflects the error or uncertainty of the observation data itself, depending on the accuracy of the sensor and the observation environment; is the observation matrix, which converts the predicted results of pollutant concentration distribution into the actual monitored pollutant concentration.

7. The method according to claim 1, characterized in that The updated pollutant concentration prediction value is obtained by using the observation equation and Kalman gain calculation, including: The updated pollutant concentration prediction value is obtained by using the observation equation and Kalman gain calculation: in, is the pollutant concentration observed in real time, is the predicted result of pollutant concentration distribution, is the Kalman gain, is the estimated pollutant concentration after filter update, represents the observation matrix.

Images