A method and system for locating a pollution source

Abstract

This invention provides a method and system for tracing and locating pollution sources. In this invention, sky-ground coordinated data for a designated area is acquired; based on the sky-ground coordinated data, a pollution concentration identification model is used to identify the pollution concentration in the designated area, obtaining a spatiotemporal distribution matrix of pollution concentration; based on the acquired historical pollution event data of the designated area, a pollution source location model is used to locate the pollution source, obtaining candidate pollution source location information; based on the spatiotemporal distribution matrix of pollution concentration and the candidate pollution source location information, a particle swarm optimization algorithm is used to trace and locate the pollution source, obtaining the optimal pollution source location information in the designated area. This invention integrates multi-source information through sky-ground coordinated data, combined with intelligent models and optimization algorithms, to achieve rapid and accurate pollution source location, significantly shortening the tracking cycle, improving processing efficiency and accuracy, and effectively curbing pollution spread.

Description

Technical Field

[0001] This invention relates to the field of ecological and environmental monitoring technology, specifically to a method and system for tracing and locating pollution sources. Background Technology

[0002] Currently, ecological and environmental data is mainly collected by various sensors, and then inspectors complete the supervision of ecological and environmental protection and the tracking and location of pollution sources. The process of manually completing the supervision of ecological and environmental protection includes: "manual collection → on-site verification → report writing → pollution source location → problem feedback", with each step taking an average of 7-15 days.

[0003] The following problems exist in traditional pollution source tracing and location methods:

[0004] 1. Relying on various sensors to collect ecological and environmental data results in limited data and makes it difficult to accurately reflect changes in the ecological and environmental environment;

[0005] 2. The entire process involves manual intervention, and each step is time-consuming, leading to a decrease in tracking accuracy. Taking the water environment monitoring of a certain river basin as an example, it often takes more than two weeks from the discovery of pollutant exceeding the standard to the completion of source tracing, by which time the pollution may have already spread.

[0006] In summary, traditional pollution source tracing and location methods suffer from drawbacks such as slow processing speed, low efficiency, and poor accuracy. Summary of the Invention

[0007] To overcome the shortcomings of traditional pollution source tracing and location methods, such as slow processing speed, low efficiency, and poor accuracy, this invention provides a pollution source tracing and location method, comprising:

[0008] Acquire air-ground coordinated data for a designated area;

[0009] Based on the aforementioned sky-ground collaborative data, a pollution concentration identification model is used to identify the pollution concentration in the designated area, thereby obtaining a spatiotemporal distribution matrix of pollution concentration.

[0010] Based on the historical pollution event data of the designated area, the pollution source is located using a pollution source location model to obtain the location information of candidate pollution sources.

[0011] Based on the spatiotemporal distribution matrix of pollution concentration and the location information of the candidate pollution sources, the particle swarm optimization algorithm is used to track and locate the pollution sources to obtain the optimal pollution source location information in the set area.

[0012] Optionally, the pollution concentration identification model is a spatiotemporal graph convolutional network;

[0013] The step of identifying pollution concentration in the designated area based on the sky-ground collaborative data and using a pollution concentration identification model to obtain a spatiotemporal distribution matrix of pollution concentration includes:

[0014] Based on the aforementioned sky-ground collaborative data, the defined region is divided into multiple graph nodes using the spatiotemporal graph convolutional network.

[0015] Based on the multiple graph nodes, spatial correlation and temporal dynamic capture are performed using graph convolution and temporal convolution in the spatiotemporal graph convolutional network to obtain the spatiotemporal distribution matrix of pollution concentration.

[0016] Optionally, the step of using a particle swarm optimization algorithm to track and locate pollution sources based on the spatiotemporal distribution matrix of pollution concentration and the location information of candidate pollution sources, to obtain the optimal location information of pollution sources in the designated area, includes:

[0017] The candidate pollution source location information is used as the initial position of each particle in the particle swarm optimization algorithm, and the velocity of each particle is randomly generated.

[0018] Based on the initial position and velocity of each particle, a pollution diffusion simulation is performed to obtain the simulated diffusion distribution matrix of each particle. According to the simulated diffusion distribution matrix of each particle and the spatiotemporal distribution matrix of pollution concentration, the fitness function value of each particle is determined. Based on the fitness function value of each particle and preset constraints, the position and velocity of each particle are iteratively updated multiple times until the maximum number of iterations is reached. The position of the particle with the highest fitness function value at this time is taken as the optimal pollution source location information in the set area.

[0019] Optionally, determining the fitness function value of each particle based on the simulated diffusion distribution matrix of each particle and the spatiotemporal distribution matrix of the pollution concentration includes:

[0020] Based on each particle, the mean square error of the simulated diffusion distribution matrix of the particle and the spatiotemporal distribution matrix of the pollution concentration is determined as the fitness function value; or

[0021] Based on each particle, the spatial correlation index between the simulated diffusion distribution matrix of the particle and the spatiotemporal distribution matrix of the pollution concentration is determined as the fitness function value.

[0022] Optionally, after obtaining the optimal pollution source location information in the designated area by using a particle swarm optimization algorithm to track and locate pollution sources based on the spatiotemporal distribution matrix of pollution concentration and the location information of candidate pollution sources, the method further includes:

[0023] The environmental observation data used to obtain the optimal pollution source location information includes real-time wind speed, topography, and pollution concentration.

[0024] The environmental observation data is input into the pollution source diffusion prediction model to simulate the diffusion path and obtain the pollution diffusion path.

[0025] Optionally, the pollution source diffusion prediction model is a diffusion model driven by deep reinforcement learning (DRL).

[0026] Optionally, the training process of the pollution source diffusion prediction model includes:

[0027] Acquire sample environmental observation data, which includes meteorological data and pollution concentration data;

[0028] Based on the meteorological data and the pollution concentration data, a dual-delay deep deterministic strategy gradient algorithm is used to obtain the agent's action quantities; the action quantities include the diffusion coefficient of the Gaussian plume model and the wind speed adjustment.

[0029] Based on the action quantity, the agent runs a Gaussian plume model to generate a simulated concentration field as the agent's state quantity.

[0030] Using a reward function, the reward value of the simulated concentration field and the actual observed concentration field corresponding to the pollution concentration data is calculated; based on the reward value, the DRL-driven diffusion model is iteratively trained to obtain the pollution source diffusion prediction model.

[0031] Optionally, the reward function satisfies the following formula:

[0032]

[0033] Wherein, r is the reward value, and MSE is the mean square error between the actual pollution path and the simulated pollution path.

[0034] Optionally, the policy gradient algorithm is either the Proximal Policy Optimization (PPO) algorithm or the Deep Deterministic Policy Gradient (DDPG) algorithm.

[0035] On the other hand, the present invention also provides a pollution source tracking and location system, comprising:

[0036] The acquisition module is used to acquire sky-ground coordinated data for a specified area;

[0037] The pollution concentration identification module is used to identify the pollution concentration in the designated area based on the sky-ground collaborative data and a pollution concentration identification model to obtain a spatiotemporal distribution matrix of pollution concentration.

[0038] The pollution source location module is used to locate pollution sources based on historical pollution event data of the designated area using a pollution source location model to obtain candidate pollution source location information; and to track and locate pollution sources using a particle swarm optimization algorithm based on the spatiotemporal distribution matrix of pollution concentration and the candidate pollution source location information to obtain the optimal pollution source location information in the designated area.

[0039] Optionally, the pollution concentration identification model is a spatiotemporal graph convolutional network;

[0040] The pollution concentration identification module is specifically used to divide the set area into multiple graph nodes based on the sky-ground collaborative data and the spatiotemporal graph convolutional network; based on the multiple graph nodes, spatial correlation and temporal dynamic capture are performed using graph convolution and temporal convolution in the spatiotemporal graph convolutional network to obtain the spatiotemporal distribution matrix of pollution concentration.

[0041] Optionally, the pollution source localization module is specifically used to use the candidate pollution source location information as the initial position of each particle in the particle swarm optimization algorithm, and randomly generate the velocity of each particle; perform pollution diffusion simulation based on the initial position and velocity of each particle to obtain the simulated diffusion distribution matrix of each particle; determine the fitness function value of each particle according to the simulated diffusion distribution matrix of each particle and the spatiotemporal distribution matrix of pollution concentration; and perform multiple iterative updates on the position and velocity of each particle based on the fitness function value of each particle and preset constraints until the maximum number of iterations is reached, and take the position of the particle with the highest fitness function value at this time as the optimal pollution source location information in the set area.

[0042] Optionally, the pollution source localization module is specifically used to determine the mean square error of the simulated diffusion distribution matrix of the particle and the spatiotemporal distribution matrix of the pollution concentration as the fitness function value based on each particle; or, based on each particle, to determine the spatial correlation index of the simulated diffusion distribution matrix of the particle and the spatiotemporal distribution matrix of the pollution concentration as the fitness function value.

[0043] Optionally, the pollution source tracking and location system further includes:

[0044] The pollution diffusion path prediction module is used to acquire environmental observation data of the optimal pollution source location information, including real-time wind speed, topography, and pollution concentration; the environmental observation data is input into the pollution source diffusion prediction model to simulate the diffusion path and obtain the pollution diffusion path.

[0045] Optionally, the pollution source diffusion prediction model is a diffusion model driven by deep reinforcement learning (DRL).

[0046] Optionally, the training process of the pollution source diffusion prediction model includes: acquiring sample environmental observation data, which includes meteorological data and pollution concentration data; based on the meteorological data and the pollution concentration data, using a dual-delay deep deterministic policy gradient algorithm to obtain the agent's action quantities; the action quantities include the diffusion coefficient and wind speed adjustment of the Gaussian plume model; based on the action quantities, using the agent to run the Gaussian plume model to generate a simulated concentration field as the agent's state quantity; using a reward function to calculate the reward value of the simulated concentration field and the actual observed concentration field corresponding to the pollution concentration data; and iteratively training the DRL-driven diffusion model based on the reward value to obtain the pollution source diffusion prediction model.

[0047] Optionally, the reward function satisfies the following formula:

[0048]

[0049] Wherein, r is the reward value, and MSE is the mean square error between the actual pollution path and the simulated pollution path.

[0050] Optionally, the policy gradient algorithm is either the Proximal Policy Optimization (PPO) algorithm or the Deep Deterministic Policy Gradient (DDPG) algorithm.

[0051] On the other hand, the present invention also provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus;

[0052] The memory is used to store one or more programs;

[0053] When the one or more programs are executed by the at least one processor, the pollution source tracing and location method as described in any of the above statements is implemented.

[0054] On the other hand, the present invention also provides a readable storage medium having an executable program stored thereon, which, when executed, implements the pollution source tracing and locating method as described in any one of the above.

[0055] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0056] This invention provides a method and system for tracing and locating pollution sources. The method includes: acquiring coordinated air-ground data of a designated area; identifying pollution concentration in the designated area using a pollution concentration identification model based on the coordinated air-ground data to obtain a spatiotemporal distribution matrix of pollution concentration; locating pollution sources using a pollution source location model based on the acquired historical pollution event data of the designated area to obtain candidate pollution source location information; and tracing and locating pollution sources using a particle swarm optimization algorithm based on the spatiotemporal distribution matrix of pollution concentration and the candidate pollution source location information to obtain the optimal pollution source location information in the designated area. This invention integrates multi-source information through coordinated air-ground data, combined with intelligent models and optimization algorithms, to achieve rapid and accurate pollution source location, significantly shortening the tracking cycle, improving processing efficiency and accuracy, and effectively curbing pollution spread. Attached Figure Description

[0057] Figure 1 This is a flowchart illustrating the pollution source tracing and location method of the present invention;

[0058] Figure 2 This is a schematic diagram of the pollution source tracing and positioning system of the present invention;

[0059] Figure 3 This is a schematic diagram of the electronic device of the present invention. Detailed Implementation

[0060] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0061] Example 1:

[0062] The present invention provides a method for tracing and locating pollution sources, the flowchart of which is shown below. Figure 1 As shown, it includes:

[0063] Step 101: Obtain sky-ground coordinated data for the designated area;

[0064] Step 102: Based on the aforementioned sky-ground collaborative data, use the pollution concentration identification model to identify the pollution concentration in the designated area and obtain the spatiotemporal distribution matrix of pollution concentration;

[0065] Step 103: Based on the historical pollution event data of the designated area, use the pollution source location model to locate the pollution source and obtain the location information of the candidate pollution source;

[0066] Step 104: Based on the spatiotemporal distribution matrix of pollution concentration and the location information of the candidate pollution sources, use the particle swarm optimization algorithm to track and locate the pollution sources to obtain the optimal pollution source location information in the set area.

[0067] The pollution source tracing and location method provided in this embodiment of the invention is applied to electronic devices, such as personal computers (PCs) and servers.

[0068] In this invention, an electronic device collects combined sky-ground data for a designated area and uses this combined sky-ground data to track and locate pollution sources. This combined sky-ground data includes space-based data, air-based data, and ground-based data. Space-based data includes, but is not limited to, satellite data, which may be hyperspectral data; air-based data includes, but is not limited to, drone data; and ground-based data includes, but is not limited to, sensor data.

[0069] For example, data is collected through space-based, airborne, and ground-based monitoring. First, low-Earth orbit (LEO) satellites with an altitude not exceeding 500 km are deployed for space-based monitoring. These LEO satellites carry a hyperspectral imager with a spectral band not exceeding 200 and a synthetic aperture radar (SAR). The hyperspectral imager acquires spectral data for a designated area, and combined with SAR's ability to penetrate clouds, a spatiotemporal adaptive compressed sensing algorithm is used to reconstruct images of obscured areas, yielding radar data. The designated area can be monitored at predetermined time intervals, not exceeding one hour.

[0070] The spatiotemporal adaptive compressed sensing algorithm can be implemented using the following formula:

[0071]

[0072] Where Y represents the observation data acquired by the satellite, A is the preset sensing matrix, X is the reconstructed image, λ is the preset regularization parameter, and ||X|| TV This represents the Total Variation (TV) norm of the reconstructed image X.

[0073] Aerial monitoring is achieved by deploying drones equipped with hydrogen fuel cells. These drones have a flight time of up to 10 hours and are equipped with lidar and gas chromatographs. They can autonomously fly along pollutant concentration gradients based on dynamic trajectory planning algorithms, collecting three-dimensional pollution distribution data in real time.

[0074] The dynamic trajectory planning algorithm can be implemented using the following formula:

[0075]

[0076] Where C(x,y,t) represents the pollutant concentration field collected by the UAV, v(t) represents the UAV speed, and P routeThe flight trajectory is represented by T, where T represents the total length of the time window considered when the UAV plans its trajectory, and t represents the point in time.

[0077] Ground-based monitoring is achieved by deploying multiple micro sensors. Data such as PM2.5, VOCs concentration, temperature, and humidity are transmitted between the sensors via the Long Range Wide Area Network (LoRaWAN) protocol. After noise is processed by adaptive Kalman filtering, a high-precision ground pollution map is formed.

[0078] In this invention, an electronic device is equipped with a pollution concentration identification model. After the electronic device acquires the combined air-ground data of a set area, it can input the combined air-ground data into the pollution concentration identification model, so that the pollution concentration identification model can identify the pollution concentration of the set area based on the combined air-ground data and obtain the spatiotemporal distribution matrix of pollution concentration.

[0079] For example, this pollution concentration identification model extracts and fuses features from sky-ground collaborative data, uses satellite pixels, UAV waypoints, and sensor locations as graph nodes, and dynamically calculates cross-platform correlation through an attention mechanism to generate a spatiotemporal distribution matrix of pollution concentration.

[0080] In addition, the electronic device is equipped with a pollution source location model. This model can input historical pollution event data for a designated area, allowing it to locate pollution sources based on this data and obtain candidate pollution source location information for the designated area. This historical pollution event data includes, but is not limited to, pollution type, pollution concentration, meteorological conditions, pollution extent, and geographical information.

[0081] Among them, the pollution source location model can output candidate pollution source location information by outputting probability heatmaps or discrete location points.

[0082] The electronic device inputs the spatiotemporal distribution matrix of pollution concentration and the location information of candidate pollution sources into the particle swarm optimization algorithm. The particle swarm optimization algorithm is initialized based on the location information of the candidate pollution sources. Specifically, the location information of the candidate pollution sources is used as the initial position of the particles. The pollution diffusion process is simulated iteratively, the mean square error between the simulation results and the spatiotemporal distribution matrix of pollution concentration is calculated, and the particle trajectory is optimized in combination with the constraints of geographical obstacles, thereby obtaining the optimal pollution source location information.

[0083] The electronic device will determine the optimal pollution source location information in the set area based on the optimal pollution source location information determined by the particle swarm optimization algorithm.

[0084] This invention obtains sky-ground collaborative data through multi-platform collaborative observation, eliminating blind spots of single data sources. Furthermore, it effectively captures the spatiotemporal evolution of pollution through dynamic correlation analysis of pollution source concentration identification models. Based on pollution source localization models and particle swarm optimization algorithms, under complex constraints, combined with the spatiotemporal distribution matrix of pollution concentration, it maintains efficient optimization capabilities, thereby accurately locating the optimal pollution source location information.

[0085] In order to accurately locate pollution sources and improve the efficiency of pollution source location, based on the above embodiments, in this invention, the pollution concentration identification model is a spatiotemporal graph convolutional network;

[0086] Based on combined sky-ground data, a pollution concentration identification model is used to identify the pollution concentration in a designated area, resulting in a spatiotemporal distribution matrix of pollution concentration, including:

[0087] Based on sky-ground collaborative data, a spatiotemporal graph convolutional network is used to divide a defined region into multiple graph nodes;

[0088] Based on multiple graph nodes, spatial correlation and temporal dynamic capture are performed using graph convolution and temporal convolution in a spatiotemporal graph convolutional network to obtain the spatiotemporal distribution matrix of pollution concentration.

[0089] In this invention, the pollution concentration identification model is a spatiotemporal graph convolutional network (ST-GCN).

[0090] Electronic devices input sky-ground coordinated data into the ST-GCN. After receiving the sky-ground coordinated data, the ST-GCN divides the designated area into multiple graph nodes. For example, the ST-GCN uses satellite pixels, UAV waypoints, and sensor locations of the designated area carried in the sky-ground coordinated data as graph nodes. Based on multiple graph nodes, the ST-GCN uses graph convolution and temporal convolution to perform spatial correlation and dynamic temporal capture to obtain the spatiotemporal distribution matrix of pollution concentration.

[0091] For example, ST-GCN employs a dual-branch structure, consisting of a spatial branch and a temporal branch. The spatial branch is used for spatial feature extraction, specifically by aggregating features from adjacent graph nodes through graph convolution operations. The temporal branch is used for temporal feature extraction, specifically by capturing the periodic changes in pollution diffusion through temporal convolution.

[0092] In the spatial feature extraction stage, ST-GCN calculates the correlation between each graph node through an attention mechanism, dynamically generating an adjacency matrix. This attention mechanism adaptively strengthens the weights of connection edges between nodes surrounding the pollution source, enabling cross-platform correlation. In the temporal feature extraction stage, the temporal convolutional layer employs a dilated convolution structure, effectively capturing long-term pollution evolution data while maintaining computational efficiency.

[0093] The attention mechanism can be implemented using the following formula:

[0094]

[0095] Among them, e ij h represents the weight of the edge connecting graph node i and graph node j. i h represents the feature corresponding to graph node i. j W represents the feature corresponding to graph node j. q and W k is a preset learnable parameter, and d is a preset parameter.

[0096] ST-GCN fuses the extracted spatial and temporal features to obtain the spatiotemporal distribution matrix of pollution concentration.

[0097] To accurately locate pollution sources and improve the efficiency of pollution source location, based on the above embodiments, in this invention, the above-mentioned method of using a particle swarm optimization algorithm to track and locate pollution sources based on the spatiotemporal distribution matrix of pollution concentration and candidate pollution source location information, to obtain the optimal pollution source location information in a set area, includes:

[0098] The candidate pollution source location information is used as the initial position of each particle in the particle swarm optimization algorithm, and the velocity of each particle is randomly generated.

[0099] Pollution diffusion simulation is performed based on the initial position and velocity of each particle to obtain the simulated diffusion distribution matrix of each particle. Based on the simulated diffusion distribution matrix of each particle and the spatiotemporal distribution matrix of pollution concentration, the fitness function value of each particle is determined. Based on the fitness function value of each particle and the preset constraints, the position and velocity of each particle are iteratively updated multiple times until the maximum number of iterations is reached. The position of the particle with the highest fitness function value at this time is taken as the optimal pollution source location information in the set area.

[0100] In this invention, by combining the particle swarm optimization algorithm with the spatiotemporal distribution matrix of pollution concentration, the optimal pollution source location information can be selected from the candidate pollution source location information.

[0101] Specifically, the location information of candidate pollution sources is used as the initial position of each particle in the particle swarm optimization algorithm, and the velocity of each particle is randomly generated. Based on the initial position and velocity of each particle, pollution diffusion simulation is performed to obtain the simulated diffusion distribution matrix of each particle.

[0102] The electronic device determines the fitness function value of each particle based on the simulated diffusion distribution matrix and the spatiotemporal distribution matrix of pollution concentration. Then, based on the fitness function value and preset constraints, it iteratively updates the position and velocity of each particle multiple times until the maximum number of iterations is reached. The position of the particle with the highest fitness function value at this point is taken as the optimal pollution source location information in the designated area.

[0103] For example, constraints include, but are not limited to, geographical obstacle constraints and pollution source release time constraints.

[0104] In this invention, during the iterative optimization process based on the particle swarm optimization algorithm, each particle can simulate Gaussian plume diffusion based on its current position to generate a simulated diffusion distribution matrix.

[0105] The termination condition for the particle swarm optimization algorithm can be set to the maximum number of iterations or the fitness function value being less than a threshold for 10 consecutive generations.

[0106] To accurately locate pollution sources and improve the efficiency of pollution source location, based on the above embodiments, in this invention, determining the fitness function value of each particle based on the simulated diffusion distribution matrix and the spatiotemporal distribution matrix of pollution concentration includes:

[0107] For each particle, the mean square error of the simulated diffusion distribution matrix of the particle and the spatiotemporal distribution matrix of the pollution concentration is determined as the fitness function value; or

[0108] For each particle, the spatial correlation coefficient between the simulated diffusion distribution matrix of the particle and the spatiotemporal distribution matrix of the pollution concentration is determined as the fitness function value.

[0109] In this invention, the electronic device can determine the fitness function value in a variety of ways, including but not limited to based on mean square error or based on spatial correlation coefficient.

[0110] In one example, for each particle, the electronic device determines the mean square error of the simulated diffusion distribution matrix and the spatiotemporal distribution matrix of the pollution concentration for that particle, and uses this mean square error as the fitness function value.

[0111] In another example, for each particle, the electronic device determines the spatial correlation coefficient between the particle's simulated diffusion distribution matrix and the spatiotemporal distribution matrix of pollution concentration, and uses this spatial correlation coefficient as the fitness function value.

[0112] To accurately locate pollution sources and improve the efficiency of pollution source location, based on the above embodiments, this invention, after obtaining the optimal pollution source location information in a set area by using a particle swarm optimization algorithm based on the spatiotemporal distribution matrix of pollution concentration and candidate pollution source location information, further includes:

[0113] Environmental observation data is used to obtain optimal pollution source location information. The environmental observation data includes real-time wind speed, topography, and pollution concentration.

[0114] Environmental observation data is input into a pollution source diffusion prediction model to simulate the diffusion path and obtain the pollution diffusion path.

[0115] In this invention, after determining the optimal pollution source location information, the electronic device can also monitor and simulate the pollution at the optimal pollution source location information to obtain the pollution diffusion path.

[0116] Specifically, the electronic device acquires environmental observation data containing optimal pollution source location information. This environmental observation data includes, but is not limited to, real-time wind speed, topography, and pollution concentration. The electronic device is equipped with a pollution source diffusion prediction model. After acquiring the environmental observation data, the electronic device inputs this data into the pollution source diffusion prediction model to simulate the diffusion path and obtain the pollution diffusion path.

[0117] Specifically, the electronic device can normalize the real-time wind speed, topography, and pollution concentration of the optimal pollution source location information, and then concatenate the normalization results into a spatiotemporal tensor. The electronic device then inputs this spatiotemporal tensor into the pollution source diffusion prediction model.

[0118] In order to accurately locate pollution sources and improve the efficiency of pollution source location, based on the above embodiments, in this invention, the pollution source diffusion prediction model is a diffusion model driven by deep reinforcement learning (DRL).

[0119] To accurately locate pollution sources and improve the efficiency of pollution source location, based on the above embodiments, the training process of the pollution source diffusion prediction model in this invention includes:

[0120] Obtain environmental observation data for the sample, including meteorological data and pollution concentration data;

[0121] Based on meteorological and pollution concentration data, a dual-delay deep deterministic policy gradient algorithm is used to obtain the agent's action quantities; the action quantities include the diffusion coefficient of the Gaussian plume model and the wind speed adjustment.

[0122] Based on the action quantity, the agent runs a Gaussian plume model to generate a simulated concentration field as the agent's state quantity.

[0123] Using a reward function, the reward value of the simulated concentration field and the actual observed concentration field corresponding to the pollution concentration data are calculated; based on the reward value, the DRL-driven diffusion model is iteratively trained to obtain the pollution source diffusion prediction model.

[0124] In this invention, the pollution source diffusion prediction model is a diffusion model driven by deep reinforcement learning (DRL).

[0125] The training process of the DRL-driven diffusion model includes: acquiring sample environmental observation data, which includes meteorological data and pollution concentration data; based on the meteorological data and pollution concentration data, using the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm to obtain the agent's action variables; the action variables include the diffusion coefficient of the Gaussian plume model and the wind speed adjustment; based on the action variables, using the agent to run the Gaussian plume model to generate a simulated concentration field as the agent's state variable; using a reward function, calculating the reward value of the simulated concentration field and the actual observed concentration field corresponding to the pollution concentration data; and iteratively training the DRL-driven diffusion model based on the reward value to obtain the pollution source diffusion prediction model.

[0126] Specifically, the TD3 algorithm is implemented using the following formula:

[0127]

[0128] Where Q(s,a) represents the expected reward value of the current state variable s and action variable a, where s is the state variable, a is the action variable, E represents the operation of calculating the expected value, and r and γ are preset parameters. The output of the objective Q-function is used to stabilize the training process and reduce overestimation problems, π. φ’ (s') represents the action output by the target policy network based on the next state s', θ' i φ' are parameters of the target policy network, and i may represent one of the two Q networks, i = 1, 2.

[0129] In this invention, the action quantities of the intelligent agent include the diffusion coefficient and wind speed adjustment of the Gaussian plume model. The diffusion coefficient is used to control the horizontal diffusion range of pollutants, and the wind speed adjustment is used to adjust the advection term in the Gaussian plume model to match the actual wind field.

[0130] In order to accurately locate pollution sources and improve the efficiency of pollution source location, based on the above embodiments, in this invention, the reward function satisfies the following formula:

[0131]

[0132] Where r is the reward value, and MSE is the mean square error between the actual pollution path and the simulated pollution path.

[0133] In this invention, the reward function can be the reciprocal of the mean square error between the simulated concentration field and the actual observed concentration field, in order to maximize alignment.

[0134] For example, the reward function can be implemented using the following formula:

[0135]

[0136] Where r is the reward value, and MSE is the mean square error between the actual pollution path and the simulated pollution path.

[0137] In this invention, the convergence of the pollution source diffusion prediction model can be determined based on the reward function. If the model converges, the action quantity at that point is used as the network parameter of the model. If the model fails to converge, the action quantity is optimized using a policy gradient algorithm until convergence occurs.

[0138] In order to accurately locate pollution sources and improve the efficiency of pollution source location, based on the above embodiments, in this invention, the above policy gradient algorithm is either the Proximal Policy Optimization (PPO) algorithm or the Deep Deterministic Policy Gradient (DDPG) algorithm.

[0139] In this invention, the policy gradient algorithm is either the Proximal Policy Optimization (PPO) algorithm or the Deep Deterministic Policy Gradient (DDPG) algorithm.

[0140] Based on the disclosure of this invention, a lightweight AI chip can be deployed at the sensor node. By using the model distillation compression algorithm, the algorithm and model provided by this invention can be distilled and compressed to obtain the ST-GCN micro model, thereby realizing the detection and early warning of local pollution concentration anomalies.

[0141] The present invention will now be described with reference to a specific embodiment, which includes the following steps:

[0142] (1) Obtain sky-ground coordinated data for a designated area;

[0143] (2) Based on the sky-ground collaborative data, the pollution concentration identification model is used to identify the pollution concentration in the designated area and obtain the spatiotemporal distribution matrix of pollution concentration.

[0144] (3) Based on the historical pollution event data of the designated area, the pollution source is located using the pollution source location model to obtain the location information of the candidate pollution source;

[0145] (4) Based on the spatiotemporal distribution matrix of pollution concentration and the location information of candidate pollution sources, the particle swarm optimization algorithm is used to track and locate pollution sources and obtain the optimal pollution source location information in the set area.

[0146] (5) Obtain environmental observation data for the optimal pollution source location information, including real-time wind speed, topography and pollution concentration;

[0147] (6) Input environmental observation data into the pollution source diffusion prediction model to simulate the diffusion path and obtain the pollution diffusion path.

[0148] This invention integrates multi-source information through sky-ground collaborative data, combined with intelligent models and optimization algorithms, to achieve rapid and accurate location of pollution sources, significantly shorten the tracking cycle, improve processing efficiency and accuracy, and effectively curb the spread of pollution.

[0149] This invention relates to an integrated air-ground ecological environment monitoring system and pollution source tracing method. Based on a three-dimensional monitoring network constructed using satellite remote sensing, low-altitude drone surveillance, and ground sensors, it achieves comprehensive perception and precise source tracing of air, water, and soil pollution. The invention includes a data acquisition module, an automatic pollution source identification module, a dynamic data visualization platform, and a pollution diffusion simulation module. Through multi-source data fusion and machine learning algorithms, it can quickly locate pollution sources and simulate their diffusion paths, providing a scientific basis for pollution control. This invention has the advantages of wide monitoring coverage and high positioning accuracy, and is widely applicable to the field of comprehensive ecological environment monitoring.

[0150] Example 2:

[0151] Based on the same inventive concept, this invention also provides a pollution source tracking and positioning system, the structural schematic diagram of which is shown below. Figure 2 As shown, it includes:

[0152] Module 201 is used to acquire sky-ground coordinated data for a designated area;

[0153] The pollution concentration identification module 202 is used to identify the pollution concentration in a set area based on sky-ground collaborative data and a pollution concentration identification model to obtain a spatiotemporal distribution matrix of pollution concentration.

[0154] The pollution source location module 203 is used to locate pollution sources based on historical pollution event data of the set area and a pollution source location model to obtain candidate pollution source location information; and to track and locate pollution sources based on the spatiotemporal distribution matrix of pollution concentration and the candidate pollution source location information using a particle swarm optimization algorithm to obtain the optimal pollution source location information in the set area.

[0155] In one specific implementation, the pollution concentration identification model is a spatiotemporal graph convolutional network;

[0156] The pollution concentration identification module 202 is specifically used to divide a set area into multiple graph nodes based on sky-ground collaborative data and using a spatiotemporal graph convolutional network; based on the multiple graph nodes, it uses graph convolution and temporal convolution in the spatiotemporal graph convolutional network to perform spatial correlation and temporal dynamic capture to obtain the spatiotemporal distribution matrix of pollution concentration.

[0157] In one specific implementation, the pollution source localization module 203 is specifically used to use the candidate pollution source location information as the initial position of each particle in the particle swarm optimization algorithm, and randomly generate the velocity of each particle; perform pollution diffusion simulation based on the initial position and velocity of each particle to obtain the simulated diffusion distribution matrix of each particle; determine the fitness function value of each particle according to the simulated diffusion distribution matrix and the spatiotemporal distribution matrix of pollution concentration; and perform multiple iterative updates on the position and velocity of each particle based on the fitness function value of each particle and preset constraints until the maximum number of iterations is reached, and take the position of the particle with the highest fitness function value at this time as the optimal pollution source location information in the set area.

[0158] In one specific implementation, the pollution source localization module 203 is specifically used to determine the mean square error of the simulated diffusion distribution matrix of the particle and the spatiotemporal distribution matrix of pollution concentration as the fitness function value based on each particle; or, based on each particle, to determine the spatial correlation index of the simulated diffusion distribution matrix of the particle and the spatiotemporal distribution matrix of pollution concentration as the fitness function value.

[0159] In one specific implementation, the pollution source tracing and location system also includes:

[0160] The pollution diffusion path prediction module 204 is used to obtain environmental observation data of optimal pollution source location information. The environmental observation data includes real-time wind speed, topography and pollution concentration. The environmental observation data is input into the pollution source diffusion prediction model to simulate the diffusion path and obtain the pollution diffusion path.

[0161] In one specific implementation, the pollution source diffusion prediction model is a diffusion model driven by deep reinforcement learning (DRL).

[0162] In one specific implementation, the training process of the pollution source diffusion prediction model includes: acquiring sample environmental observation data, which includes meteorological data and pollution concentration data; based on the meteorological data and pollution concentration data, using a dual-delay deep deterministic policy gradient algorithm to obtain the agent's action quantities; the action quantities include the diffusion coefficient of the Gaussian plume model and the wind speed adjustment; based on the action quantities, using the agent to run the Gaussian plume model to generate a simulated concentration field as the agent's state quantity; using a reward function to calculate the reward value of the simulated concentration field and the actual observed concentration field corresponding to the pollution concentration data; and iteratively training the DRL-driven diffusion model based on the reward value to obtain the pollution source diffusion prediction model.

[0163] In one specific implementation, the reward function satisfies the following formula:

[0164]

[0165] Where r is the reward value, and MSE is the mean square error between the actual pollution path and the simulated pollution path.

[0166] In one specific implementation, the policy gradient algorithm is either the Proximal Policy Optimization (PPO) algorithm or the Deep Deterministic Policy Gradient (DDPG) algorithm.

[0167] Example 3:

[0168] like Figure 3 As shown, the present invention also provides an electronic device, which may be a computer device, a microcontroller device, a smart mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, processor, and transceiver component are connected via a bus; the memory can be used to store executable programs, and an exemplary executable program may include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, which can be accessed and / or modified when instructions are executed.

[0169] The processor may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and it is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of a pollution source tracking and positioning method in the above embodiments.

[0170] Example 4:

[0171] Based on the same inventive concept, this invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory). An electronic device readable storage medium is a memory device within an electronic device used to store programs and data. It is understood that the storage medium here can include both the built-in storage medium of the electronic device and extended storage media supported by the electronic device. The storage medium provides storage space, which stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more executable programs (including program code). It should be noted that the storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. Loading and executing one or more instructions stored in the storage medium by the processor can implement the steps of a pollution source tracing and location method in the above embodiments.

[0172] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0173] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0174] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0175] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0176] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the application, but these changes, modifications or equivalent substitutions are all within the scope of protection of the claims pending approval.

Claims

1. A method of locating a source of pollution, characterized in that, include: Acquire air-ground coordinated data for a designated area; Based on the aforementioned sky-ground collaborative data, a pollution concentration identification model is used to identify the pollution concentration in the designated area, thereby obtaining a spatiotemporal distribution matrix of pollution concentration. Based on the historical pollution event data of the designated area, the pollution source is located using a pollution source location model to obtain the location information of candidate pollution sources. Based on the spatiotemporal distribution matrix of pollution concentration and the location information of the candidate pollution sources, the particle swarm optimization algorithm is used to track and locate the pollution sources to obtain the optimal pollution source location information in the set area. The environmental observation data used to obtain the optimal pollution source location information includes real-time wind speed, topography, and pollution concentration. The environmental observation data is input into the pollution source diffusion prediction model to simulate the diffusion path and obtain the pollution diffusion path. The pollution source diffusion prediction model is a diffusion model driven by deep reinforcement learning (DRL). The training process of the pollution source diffusion prediction model includes: Acquire sample environmental observation data, which includes meteorological data and pollution concentration data; Based on the meteorological data and the pollution concentration data, a dual-delay deep deterministic strategy gradient algorithm is used to obtain the agent's action quantities; the action quantities include the diffusion coefficient of the Gaussian plume model and the wind speed adjustment. Based on the action quantity, the agent runs a Gaussian plume model to generate a simulated concentration field as the state quantity of the agent. Using a reward function, the reward value of the simulated concentration field and the actual observed concentration field corresponding to the pollution concentration data is calculated; based on the reward value, the DRL-driven diffusion model is iteratively trained to obtain the pollution source diffusion prediction model.

2. The method of claim 1, wherein, The pollution concentration identification model is a spatiotemporal graph convolutional network; The step of identifying pollution concentration in the designated area based on the sky-ground collaborative data and using a pollution concentration identification model to obtain a spatiotemporal distribution matrix of pollution concentration includes: Based on the aforementioned sky-ground collaborative data, the defined region is divided into multiple graph nodes using the spatiotemporal graph convolutional network. Based on the multiple graph nodes, spatial correlation and temporal dynamic capture are performed using graph convolution and temporal convolution in the spatiotemporal graph convolutional network to obtain the spatiotemporal distribution matrix of pollution concentration.

3. The method of claim 1 or 2, wherein, The step of using the particle swarm optimization algorithm to track and locate pollution sources based on the spatiotemporal distribution matrix of pollution concentration and the location information of candidate pollution sources, to obtain the optimal pollution source location information in the designated area, includes: The candidate pollution source location information is used as the initial position of each particle in the particle swarm optimization algorithm, and the velocity of each particle is randomly generated. The pollution diffusion simulation is performed based on the initial position and speed of each particle to obtain a simulation diffusion distribution matrix of each particle; the fitness function value of each particle is determined according to the simulation diffusion distribution matrix of each particle and the pollution concentration spatiotemporal distribution matrix; the position and speed of each particle are iteratively updated based on the fitness function value of each particle and a preset constraint condition until a maximum iteration number is reached, and the position of the particle with the highest fitness function value at this time is taken as the optimal pollution source position information in the set region.

4. The method of claim 3, wherein, The fitness function value of each particle is determined according to the simulation diffusion distribution matrix of each particle and the pollution concentration spatiotemporal distribution matrix, including: The mean square error of the simulation diffusion distribution matrix of each particle and the pollution concentration spatiotemporal distribution matrix is determined as the fitness function value based on each particle; or The spatial correlation coefficient of the simulation diffusion distribution matrix of each particle and the pollution concentration spatiotemporal distribution matrix is determined as the fitness function value based on each particle.

5. The method of claim 1, wherein, The reward function satisfies the following formula: Wherein, the r is a reward value, and the MSE is the mean square error of the actual pollution path and the simulation pollution path.

6. The method of claim 1, wherein, The policy gradient algorithm is a proximal policy optimization (PPO) algorithm or a deep deterministic policy gradient (DDPG) algorithm.

7. A pollution source tracking and locating system characterized by, It includes: An acquisition module configured to acquire sky-ground collaborative data of a set region; A pollution concentration identification module configured to identify a pollution concentration of the set region based on the sky-ground collaborative data and using a pollution concentration identification model to obtain a pollution concentration spatiotemporal distribution matrix; A pollution source positioning module configured to position a pollution source based on historical pollution event data of the set region and using a pollution source positioning model to obtain candidate pollution source position information; A pollution source tracking and positioning module configured to track and position a pollution source based on the pollution concentration spatiotemporal distribution matrix and the candidate pollution source position information and using a particle swarm optimization algorithm to obtain optimal pollution source position information in the set region; A pollution diffusion path prediction module configured to acquire environmental observation data of the optimal pollution source position information, the environmental observation data including real-time wind speed, terrain, and pollution concentration; input the environmental observation data into the pollution source diffusion prediction model for diffusion path simulation to obtain a pollution diffusion path; The pollution source diffusion prediction model is a diffusion model driven by deep reinforcement learning (DRL); The training process of the pollution source diffusion prediction model includes: acquiring sample environmental observation data, the sample environmental observation data including meteorological data and pollution concentration data; Based on the meteorological data and the pollution concentration data, a double-delay deep deterministic policy gradient algorithm is used to obtain an action amount of an agent; the action amount includes a diffusion coefficient of a Gaussian plume model and a wind speed adjustment amount; based on the action amount, the agent runs the Gaussian plume model to generate a simulated concentration field as a state amount of the agent; a reward function is used to calculate a reward value of the simulated concentration field and an actual observed concentration field corresponding to the pollution concentration data; and based on the reward value, the DRL-driven diffusion model is iteratively trained to obtain a pollution source diffusion prediction model.

Images