Agricultural non-point source pollution early warning method and system and electronic equipment

Abstract

The invention relates to the field of agricultural environment monitoring, and provides an agricultural non-point source pollution early warning method and system and electronic equipment, and the method comprises the steps: obtaining a first pollution monitoring parameter of a to-be-monitored region; predicting a second pollution monitoring parameter of the to-be-monitored area in a set time period; when the second pollution monitoring parameter meets the first early warning condition, first-level early warning is carried out, and when the second pollution monitoring parameter does not meet the first early warning condition but meets the second early warning condition, second-level early warning is carried out; and calculating the pollution contribution rate of the to-be-monitored area to the downstream area through the graph neural network, and if the pollution contribution rate meets a third early warning condition, performing third-level early warning. The invention discloses an agricultural non-point source pollution monitoring method and device, aims to overcome the defects that pollution monitoring means in the prior art are difficult to accurately identify and early warn pollution and cannot position pollution sources, can perform different levels of early warning on agricultural non-point source pollution through a graded early warning mechanism, and can realize real-time monitoring, dynamic early warning and accurate treatment of regional pollution.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural environment monitoring, and in particular to an agricultural non-point source pollution early warning method and system, and electronic equipment. Background Art

[0002] Agricultural non-point source pollution refers to the pollution of the ecological environment caused by nutrients such as nitrogen, phosphorus, and organic matter produced in the agricultural production process due to the irrational use of chemical inputs such as fertilizers, pesticides, and ground film, as well as the untimely or improper treatment of livestock and aquaculture waste, crop straw, etc., which, driven by rainfall and topography, use surface and underground runoff and soil erosion as carriers, accumulate excessively in the soil or enter the receiving water bodies.

[0003] Rice fields in plain river network areas are in a state of flooding for a long time. After fertilizing the rice fields, fertilizer elements such as nitrogen and phosphorus are easily discharged into the ditch river network, causing agricultural non-point source pollution. Since the water systems in the river network areas where the rice fields are located are often complex, the pollution monitoring methods in the existing technology are difficult to accurately identify and warn of pollution, and it is impossible to locate the source of pollution, resulting in serious environmental pollution caused by agricultural non-point source pollution. Summary of the invention

[0004] The present invention provides an agricultural non-point source pollution early warning method and system, and electronic equipment to solve the defects that the pollution monitoring means in the prior art are difficult to accurately identify and warn of pollution, and cannot locate the pollution source. The solution of the present application can provide early warnings at different levels for agricultural non-point source pollution through a graded early warning mechanism, and can realize real-time monitoring of regional pollution, dynamic early warning and precise control.

[0005] The present invention provides an agricultural non-point source pollution early warning method, comprising: Acquire a first pollution monitoring parameter of the area to be monitored; Based on the first pollution monitoring parameter, predicting the second pollution monitoring parameter of the area to be monitored within a set time period through a pre-built prediction model; When the second pollution monitoring parameter meets the first warning condition, a first-level warning is issued; when the second pollution monitoring parameter does not meet the first warning condition but meets the second warning condition, a second-level warning is issued; When the second pollution monitoring parameter does not meet the first warning condition and the second warning condition, the pollution contribution rate of the monitored area to the downstream area is calculated through the graph neural network. If the pollution contribution rate meets the third warning condition, a third-level warning is issued.

[0006] According to the agricultural non-point source pollution early warning method provided by the present invention, the prediction model is a coupling model of a soil and water resources assessment model and a long short-term memory network model.

[0007] According to the agricultural non-point source pollution early warning method provided by the present invention, the construction process of the prediction model includes: Collecting training data, wherein the training data includes historical pollution monitoring parameters of the area to be monitored; Based on the training data, the soil and water resources assessment model and the long short-term memory network model are trained respectively; The output of the soil and water resources assessment model is used as the input of the long short-term memory network model, and the long short-term memory network model is iteratively optimized. Based on the output of the long short-term memory network model, the soil and water resources assessment model is iteratively optimized.

[0008] According to the agricultural non-point source pollution early warning method provided by the present invention, the second pollution monitoring parameter includes soil nitrogen content and the rainfall probability of the monitored area within a set time period, and the first early warning condition includes: The soil nitrogen content exceeds a nitrogen content threshold, and the rainfall probability in the monitored area within a set time period exceeds a rainfall probability threshold.

[0009] According to the agricultural non-point source pollution early warning method provided by the present invention, the second pollution monitoring parameter includes the phosphorus content of the drainage outlet of the monitored area and the downstream flow rate of the monitored area, and the second early warning condition includes: The phosphorus content at the drainage outlet of the area to be monitored exceeds the phosphorus content threshold, and the downstream flow velocity of the area to be monitored is lower than the flow velocity threshold.

[0010] According to the agricultural non-point source pollution early warning method provided by the present invention, the downstream area of ​​the area to be monitored is determined by the following method: Constructing a virtual river network model of the area to be monitored, wherein the virtual river network model at least reflects the geographic spatial data and hydrological data of the area to be monitored; Based on the virtual river network model, determining the pollutant diffusion path of the area to be monitored; Based on the pollutant diffusion path, a downstream area of ​​the area to be monitored is determined.

[0011] According to the agricultural non-point source pollution early warning method provided by the present invention, the pollution contribution rate of the monitored area to the downstream area is calculated by using a graph neural network, including: Calculating the concentration of pollutants in the area to be monitored and the concentration of pollutants in the downstream area of ​​the area to be monitored by the graph neural network, wherein the pollutants include nitrogen and phosphorus; Based on the pollutant concentration of the area to be monitored and the pollutant concentration of the downstream area of ​​the area to be monitored, the pollution contribution rate of the area to be monitored to the downstream area is determined.

[0012] According to the agricultural non-point source pollution early warning method provided by the present invention, the graph neural network is constructed by the following method: The to-be-monitored areas are used as graph nodes, and the rivers between the to-be-monitored areas are used as edges to construct a graph structure; The graph structure is iteratively trained to optimize the weights of the edges in the graph structure to obtain the graph neural network.

[0013] The present invention also provides an agricultural non-point source pollution early warning system, comprising: A parameter acquisition module, used to acquire a first pollution monitoring parameter of the area to be monitored; A parameter prediction module, used to predict a second pollution monitoring parameter of the area to be monitored within a set time period based on the first pollution monitoring parameter through a pre-built prediction model; A first warning module, configured to issue a first-level warning when the second pollution monitoring parameter meets a first warning condition, and issue a second-level warning when the second pollution monitoring parameter does not meet the first warning condition but meets a second warning condition; The second early warning module is used to calculate the pollution contribution rate of the monitored area to the downstream area through a graph neural network when the second pollution monitoring parameter does not meet the first early warning condition and the second early warning condition, and to issue a third-level early warning if the pollution contribution rate meets the third early warning condition.

[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, any of the above-mentioned agricultural non-point source pollution early warning methods is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, any of the above-mentioned agricultural non-point source pollution early warning methods is implemented.

[0016] The present invention also provides a computer program product, including a computer program, which implements any of the above-mentioned agricultural non-point source pollution early warning methods when executed by a processor.

[0017] In the agricultural non-point source pollution early warning method provided by the present invention, the pollution situation in the future time period can be predicted based on the pre-constructed prediction model to achieve early warning of the future pollution situation, and strict warning conditions are set for each level of warning through a multi-level warning method. When the corresponding warning conditions are met, the corresponding warning is issued, which can help the staff to accurately judge the agricultural non-point source pollution in the monitored area. When both the first warning condition and the second warning condition are not met, the key monitoring area can be determined by calculating the pollution contribution rate through the graph neural network, and a third-level warning can be issued. In this way, the source of pollution can be traced, which is conducive to the subsequent rapid pollution control. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 This is one of the flow diagrams of the agricultural non-point source pollution early warning method provided by the embodiment of the present invention; Figure 2 This is the second flow chart of the agricultural non-point source pollution early warning method provided by the embodiment of the present invention; Figure 3 It is a structural schematic diagram of a graph neural network model provided by an embodiment of the present invention; Figure 4 This is one of the structural schematic diagrams of the agricultural non-point source pollution early warning system provided by an embodiment of the present invention; Figure 5 This is the second structural diagram of the agricultural non-point source pollution early warning system provided by the embodiment of the present invention; Figure 6 It is a schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Figure 1 It is a flow chart of the agricultural non-point source pollution early warning method provided by an embodiment of the present invention.

[0022] like Figure 1 As shown, this embodiment provides an agricultural non-point source pollution early warning method, including: Step 101, obtaining a first pollution monitoring parameter of the area to be monitored; The area to be monitored may be a plain river network area, and there may be multiple areas to be monitored. That is, in practical applications, the plain river network area may be divided into multiple areas to be monitored, and each area to be monitored may be monitored separately. For example, each rice field may be divided into a monitoring area, and each rice field may be monitored separately.

[0023] The first pollution monitoring parameter obtained can reflect the pollution situation of the area to be monitored. The first pollution monitoring parameter can be obtained from multiple channels. For example, sensors can be deployed in the area to be monitored, and the nitrogen content and phosphorus content in the soil can be collected by the sensors. The nitrogen and phosphorus sensors set in the soil can be evenly set in the rice fields, and more sensors can be set at the ridges of the rice fields. For example, double-density nitrogen and phosphorus sensors can be set. The nitrogen and phosphorus sensors can be used to monitor the migration of nitrogen and phosphorus elements in the rice fields under flooding conditions. During implementation, a multi-parameter water quality sensor can also be set at the drainage outlet of the rice field to collect the pH value, conductivity, ammonia nitrogen content, total phosphorus data of the water quality of the rice field in real time. A buoy-type monitoring station can also be deployed at the intersection of the plain river network. The monitoring station is integrated with a flow rate monitor and a water quality sensor for monitoring the flow rate.

[0024] In addition, data of the monitored area can be collected through aerial remote sensing platforms, such as drones equipped with hyperspectral and thermal infrared cameras. The camera's spectrum can be 400 to 1000 nanometers. Within three days after fertilizing the paddy field, the drone can fly daily along the ditches of the paddy field. In actual applications, the drone can scan the paddy field at a low altitude of 30 meters. The drone can invert the chlorophyll content through hyperspectral data to mark areas with uneven fertilization. Satellite data can also be integrated to extract the NDWI index to monitor the suspended matter content in the river network. For example, the satellite data can be Sentinel-2 satellite data.

[0025] During implementation, the nitrogen and phosphorus sensors in the soil can upload the collected parameters once at a set time. For example, data can be uploaded every 15 minutes. After the edge node eliminates outliers, it can be used as the first pollution monitoring parameter. Outliers include nitrogen and phosphorus reading drift caused by moisture interference.

[0026] Step 102, based on the first pollution monitoring parameter, predicting the second pollution monitoring parameter of the area to be monitored within a set time period through a pre-built prediction model; During implementation, the prediction model can be a coupling model of a soil and water resources assessment model and a long short-term memory network model (Long Short-Term Memory, LSTM), wherein the Soil and Water Assessment Tool (SWAT) can be used to simulate the hydrological cycle of rice fields, and the LSTM model can be used to predict the second pollution monitoring parameters within a set time period in the future. In practical applications, it can be used to predict the second pollution monitoring parameters in the next three days, wherein the second pollution monitoring parameters may include pollution peaks.

[0027] In this step, by inputting the first pollution monitoring parameter obtained in step 101 into the prediction model, the nitrogen and phosphorus loss flux in kg / ha can be calculated through the SWAT model, and the pollutant concentration curve of the drainage outlet of the rice field in the next three days can be predicted through the LSTM model.

[0028] Step 103, when the second pollution monitoring parameter meets the first warning condition, a first-level warning is issued, and when the second pollution monitoring parameter does not meet the first warning condition but meets the second warning condition, a second-level warning is issued; In actual applications, the above-mentioned first-level warning and second-level warning can be issued in the staff's monitoring station. For example, the warning reminder can be issued by an indicator light or a buzzer. For example, the first-level warning can be issued by a yellow indicator light, and the second-level warning can be issued by a red indicator light.

[0029] Step 104, when the second pollution monitoring parameter does not meet the first warning condition and the second warning condition, the pollution contribution rate of the monitored area to the downstream area is calculated by the graph neural network. If the pollution contribution rate meets the third warning condition, a third-level warning is issued.

[0030] The Graph Neural Network (GNN) in this step can be used to analyze the pollution diffusion path and locate high-risk fields whose pollution contribution rate meets the third warning conditions. High-risk fields may bring great pollution hazards to downstream areas, and early positioning can help with subsequent timely treatment.

[0031] In an exemplary embodiment, the prediction model is a coupling model of a soil and water resources assessment model and a long short-term memory network model.

[0032] In an exemplary embodiment, the process of constructing the prediction model includes: Collecting training data, wherein the training data includes historical pollution monitoring parameters of the area to be monitored; Based on the training data, the soil and water resources assessment model and the long short-term memory network model are trained respectively; The output of the soil and water resources assessment model is used as the input of the long short-term memory network model, and the long short-term memory network model is iteratively optimized. Based on the output of the long short-term memory network model, the soil and water resources assessment model is iteratively optimized.

[0033] The following is a detailed introduction to the construction process of the prediction model, which includes the following steps.

[0034] 1. Data collection and preprocessing, First, the coupled model requires the collection and organization of multiple types of data, including but not limited to: Meteorological data: including precipitation, temperature, humidity, wind speed, etc., usually from weather stations or remote sensing data.

[0035] Soil data: including soil type, soil fertility, soil moisture, etc.

[0036] Agricultural management data: such as data on agricultural activities such as fertilization, pesticide use, and irrigation.

[0037] Hydrological data: including watershed hydrological characteristics, such as runoff, groundwater level, soil moisture, etc.

[0038] Pollutant data: Concentration data of different types of agricultural pollutants (such as nitrogen, phosphorus, pesticides, etc.) in water or soil.

[0039] After that, the data preprocessing process is carried out, including but not limited to: Missing value handling: Use interpolation, mean filling, or previous value filling methods to handle missing values ​​in the data.

[0040] Data standardization: Standardize or normalize input data (such as meteorological data, pollutant concentrations, etc.) so that all types of data have the same scale so that the LSTM model can be better trained.

[0041] Time alignment: Align all data according to a unified time step (such as day, month, etc.) to ensure consistency between different data sources.

[0042] 2. Construction of SWAT model The core function of the SWAT model is to simulate the dynamic behavior of hydrological processes and pollutants in soil and water. The construction of the SWAT model usually includes the following steps: 2.1 Model Initialization Input data: including meteorological data (precipitation, temperature, etc.), land use data, soil data, watershed boundary data, etc.

[0043] Determine the model area: Divide the study area into watersheds, sub-watersheds, land use types, etc., and establish the basic spatial unit of the SWAT model.

[0044] 2.2 Hydrological process simulation Precipitation and Evaporation: Simulates precipitation infiltration, evaporation and plant transpiration.

[0045] Runoff and Soil Moisture: Calculate hydrological processes such as surface runoff and groundwater infiltration to obtain data such as soil moisture and runoff.

[0046] Pollutant simulation: The loss and migration of agricultural non-point source pollutants (such as nitrogen, phosphorus, pesticides, etc.) are calculated through the model to obtain the changes in pollutant concentrations.

[0047] The output of the SWAT model includes hydrological data and pollutant data. The hydrological data includes runoff, evaporation, soil moisture, etc., and the pollutant data includes nitrogen and phosphorus loss, pesticide concentration, etc. These output data will serve as input for the subsequent LSTM model.

[0048] 3. Construction and training of LSTM model The LSTM (Long Short-Term Memory) model is used to process time series data and capture long-term dependencies. In the coupled SWAT-LSTM model, the LSTM model is used to further process the output of the SWAT model and other related time series data (such as meteorological data, agricultural management data, etc.), thereby improving the accuracy of pollutant concentration prediction.

[0049] The inputs of the LSTM model include: pollutant concentrations output by the SWAT model, time series data, and seasonal characteristics. The pollutant concentrations output by the SWAT model include nitrogen and phosphorus concentrations, runoff, soil moisture, etc. The time series data include meteorological data, fertilizer application, irrigation amount, etc. The seasonal characteristics include months, quarters, etc. These data help LSTM capture seasonal fluctuations.

[0050] 3.1 LSTM model architecture design Input layer: The input layer accepts a variety of input features, mainly including historical pollutant concentrations, hydrological data output by the SWAT model, meteorological data, and agricultural management data.

[0051] LSTM hidden layer: The LSTM layer is used to learn long-term dependencies in time series data. Multiple LSTM layers can be set, and the number of neurons in each layer is determined according to the complexity of the problem.

[0052] Fully connected layer: The output of the LSTM layer will be passed to the fully connected layer (Dense Layer), and the final prediction result will be generated through the fully connected layer.

[0053] Output layer: The output layer is used to predict the pollutant concentration at a future time step.

[0054] 3.2 LSTM model training Loss function: The mean square error (MSE) is usually used as the loss function, and the optimization goal is to minimize the difference between the predicted value and the actual value.

[0055] Optimization algorithm: Use algorithms such as Adam optimizer or RMSprop to optimize the training process of the LSTM model.

[0056] Hyperparameter tuning: Adjust the hyperparameters of LSTM (such as learning rate, batch size, number of hidden layer neurons, etc.) through grid search or random search to improve model performance.

[0057] 3.3 LSTM model prediction After the model training is completed, the LSTM model can be used to predict the future pollutant concentration. Based on historical data, the LSTM model can predict the trend of pollutant concentration changes at a certain moment in the future.

[0058] 4. Integration of SWAT-LSTM coupling model The coupling of the SWAT-LSTM model is to effectively combine the outputs of the SWAT model and the LSTM model to improve the prediction ability of pollutant concentrations.

[0059] 4.1 Data Flow and Coupling Methods Phase I (SWAT simulation): First, the SWAT model is used to simulate the hydrological process and output preliminary concentration data of pollutants (such as runoff, nitrogen and phosphorus losses, etc.).

[0060] Phase 2 (LSTM prediction): The data output by the SWAT model and other time series data are input into the LSTM model for time series prediction. By learning from historical data, the LSTM model can predict the changing trend of pollutant concentrations in the future.

[0061] 4.2 Coupling process SWAT model operation: First, the SWAT model simulation is performed to obtain data such as pollutant concentration, surface runoff, and soil moisture.

[0062] LSTM model training: Use historical data (including pollutant concentrations and hydrological and meteorological data output by SWAT) to train the LSTM model and learn patterns and long-term dependencies in time series.

[0063] Joint optimization: Through joint training and feedback mechanism, the output of the SWAT model can provide more accurate input for the LSTM model, and the output of the LSTM model can also provide a reference for subsequent simulations of the SWAT model.

[0064] After the model training is completed, the coupling model can also be used to predict agricultural non-point source pollution and monitor changes in pollutant concentrations in real time.

[0065] Combining the advantages of SWAT and LSTM models, the changing trend of pollutant concentrations can be predicted at multiple time steps, so that corresponding environmental protection measures can be taken in advance.

[0066] You can also use mean square error (MSE), root mean square error (RMSE) and R² value as evaluation indicators, compare with actual monitoring data, evaluate the prediction accuracy of the model, and further optimize the model based on the evaluation results.

[0067] In an exemplary embodiment, the second pollution monitoring parameter includes soil nitrogen content and the probability of rainfall in the monitored area within a set time period, and the first warning condition includes: The soil nitrogen content exceeds a nitrogen content threshold, and the rainfall probability in the monitored area within a set time period exceeds a rainfall probability threshold.

[0068] In implementation, the nitrogen content threshold can be 150 mg / kg and the rainfall probability threshold can be 60%.

[0069] In an exemplary embodiment, the second pollution monitoring parameter includes the phosphorus content at the drain outlet of the monitored area and the downstream flow rate of the monitored area, and the second early warning condition includes: The phosphorus content at the drainage outlet of the area to be monitored exceeds the phosphorus content threshold, and the downstream flow velocity of the area to be monitored is lower than the flow velocity threshold.

[0070] In implementation, the phosphorus content threshold may be 0.4 mg / L, and the flow velocity threshold of the downstream flow velocity may be 0.1 m / s.

[0071] In an exemplary embodiment, after a first-level warning is issued, water pollution control can also be carried out on rice fields, including pushing a "suspend drainage" command and starting a drone re-inspection. After a second-level warning is issued, the water conservancy facilities can be linked to close the gates, and a variable fertilization map can be generated. The fertilizer machine can be controlled through the agricultural machinery Internet of Things to reduce the application amount around the ridge by 20%, that is, to reduce fertilization.

[0072] In an exemplary embodiment, the downstream area of ​​the area to be monitored is determined by the following method: Constructing a virtual river network model of the area to be monitored, wherein the virtual river network model at least reflects the geographic spatial data and hydrological data of the area to be monitored; Based on the virtual river network model, determining the pollutant diffusion path of the area to be monitored; Based on the pollutant diffusion path, a downstream area of ​​the area to be monitored is determined.

[0073] The virtual river network model reflects the path of pollutant diffusion by simulating river systems, watersheds and their hydrological processes. The following types of data are usually required to generate a virtual river network model: a. Geospatial Data Digital Elevation Model (DEM): Provides height information of the earth's surface and is used to generate topographic features of watersheds and river networks. DEM data helps determine the direction of water flow and the connectivity of rivers.

[0074] Geographic location of rivers and water bodies: including the coordinates, length, and watershed boundaries of rivers, lakes, and wetlands. River grid data in GIS (geographic information system) datasets (such as hydrological zoning data) can be used.

[0075] Land use / land cover data: These data reflect the land use of different areas in the basin (such as agriculture, forest, urban, wetland, etc.), which has a direct impact on the distribution of pollution sources and the initial concentration of pollutants.

[0076] b. Hydrological data Precipitation data: Historical and real-time precipitation data, especially precipitation data under heavy rain scenarios, directly affects water flow and pollutant dispersion within the basin.

[0077] Flow data: The rate and volume of water flowing in a river, used to model the transport of pollutants. Water flows in different river sections within a watershed can be estimated based on historical hydrological data or real-time monitoring data.

[0078] c. Pollutant data Pollutant concentration data: including the concentration data of pollutants such as nitrogen, phosphorus, and heavy metals in water bodies. These data can be collected in real time through soil sensors, water quality monitoring points and other equipment.

[0079] Pollution source distribution data: Determine the specific location of pollution sources (such as farmland, industrial emission outlets, etc.) and understand the intensity of pollution sources.

[0080] d. Hydrological models and meteorological data SWAT model or similar hydrological models: These models can help calculate the water flow distribution within the basin, the amount of pollutants lost and the hydrological response of the basin.

[0081] Meteorological data: Factors such as temperature, wind speed, and humidity have an impact on hydrological processes such as precipitation and evaporation, and can affect the dynamics of water flow after heavy rain.

[0082] e. River attributes River channel geometry parameters: Parameters such as river channel width, depth, slope, etc. can be used to estimate the flow rate and flow capacity of the river.

[0083] Hydraulic data: including flow velocity, flow rate, etc., used to calculate the diffusion of pollutants in water bodies.

[0084] The process of simulating pollutant diffusion paths under heavy rain scenarios using a virtual river network model involves the following steps: a. Construction of virtual river network Terrain modeling: Using DEM data, the location of each node (monitoring point or geographic unit) in the basin and its adjacent nodes are determined through water flow direction analysis. A virtual river network is established based on the precipitation and river flow characteristics of the basin.

[0085] Generation of network structure: Construct a river network diagram based on the actual topography and hydrological data of the basin (such as basin boundaries, river paths, etc.). Nodes represent river sections or water units, and edges represent the direction and intensity of water flow.

[0086] b. Simulation of precipitation and water flow Precipitation input: In a rainstorm scenario, precipitation increases dramatically, so it is necessary to dynamically input precipitation based on rainstorm forecast data (such as precipitation intensity and precipitation distribution).

[0087] Water flow simulation: Use hydrological models (such as SWAT, HEC-HMS, etc.) to simulate the flow of surface water caused by precipitation and determine the propagation path of pollutants in the water flow. Calculate how pollutants spread from upstream to downstream based on parameters such as flow rate, river slope and water depth.

[0088] c. Pollutant diffusion simulation Initial concentration of pollutants: Set the initial pollutant concentration based on the concentration data of the pollution source (such as agricultural fertilization, industrial emissions, etc.). Pollutants usually spread from the pollution source node to the surrounding water or soil.

[0089] Pollutant transport process: In the virtual river network model, pollutants will spread with the movement of water. The diffusion of pollutants is related to factors such as the speed of water flow, the geometry of the river channel, and the exchange rate of water bodies.

[0090] River spread: The spread of pollutants in water bodies follows the physical processes of diffusion, diffusion and convection. In areas with higher water flow rates, pollutants spread faster, and vice versa.

[0091] Pollutant concentration calculation: During the simulation, the pollutant concentration at each node (such as a river section or a water quality monitoring point) changes over time. The mass conservation equation and flow velocity equation can be used to calculate the concentration change of the pollutant in each time step.

[0092] d. Specific simulation steps for heavy rain scenarios Precipitation input: During a rainstorm, the amount of precipitation increases significantly, leading to increased surface runoff. Therefore, the amount and intensity of precipitation during the rainstorm are input into the model.

[0093] Changes in water flow dynamics: As precipitation increases, water flow rates also increase, affecting the paths and rates at which pollutants are transported.

[0094] Pollutant diffusion path tracking: Based on the dynamic changes of water flow, simulate how pollutants spread to different river sections and water bodies with water flow. In some cases, pollutants may enter tributaries or downstream lakes and other water bodies.

[0095] Pollutant Concentration Prediction: Use a time-step-based model to predict changes in pollutant concentrations at each node to identify hotspots of pollutant spread.

[0096] e. Early warning and response Based on the transmission path and concentration of pollutants, an early warning system can be set up at key nodes (such as water quality monitoring points, drainage outlets, etc.) to monitor the spread of pollutants in real time and generate emergency response measures (such as closing drainage outlets, initiating ecological restoration, etc.) based on the pollution diffusion path and concentration.

[0097] In this embodiment, determining the pollutant diffusion path has the following beneficial effects: 1. Identify pollution sources and pollution spread pathways Pollution source location: By simulating the diffusion path of pollutants, the source and diffusion direction of pollutants can be accurately identified, helping to determine the location of pollution sources (such as farmland, industrial emission outlets, etc.).

[0098] Pollution propagation pathway analysis: By simulating the diffusion paths of pollutants under different scenarios, it is possible to reveal how pollutants spread from the source to other areas (such as rivers, lakes, groundwater, etc.) and predict the flow and diffusion trends of pollutants.

[0099] 2. Predict the distribution of pollutants in space and time Spatial distribution: Simulating the diffusion path of pollutants can help predict the concentration changes of pollutants in different geographical locations (such as river sections, water bodies, soil, etc.), thereby understanding the spatial distribution of pollutants in the basin.

[0100] Temporal changes: By simulating the changes of pollutants over time, it is possible to predict the dynamic changes of pollutant concentrations in the future, providing a time reference for pollution prevention, control and governance.

[0101] 3. Assess the impact of pollutants on the environment and ecosystems Impact assessment: Simulating pollutant diffusion pathways helps assess the potential impacts of pollutants on the environment, ecosystems, and water quality. For example, by predicting whether pollutants will enter drinking water sources, wetlands, or ecologically sensitive areas, the threat to biodiversity, water quality, and ecosystem functions can be assessed.

[0102] Risk assessment: By evaluating the diffusion path of pollutants, the pollution risk of pollutants to downstream areas, water bodies and ecosystems can be quantified, providing a scientific basis for management departments to take corresponding prevention and control measures.

[0103] 4. Optimize water resource management and pollution prevention Early warning and emergency response: By simulating the diffusion path of pollutants, early warning information can be provided for pollution events. Especially under extreme weather conditions such as heavy rain, the prediction of the diffusion path of pollutants can help take emergency response measures in advance, such as closing drains, strengthening water quality monitoring or starting water treatment systems.

[0104] Governance strategy formulation: Simulating the diffusion path of pollutants helps design and optimize pollution prevention measures. For example, understanding the propagation path of pollutants from the source to the target area can help formulate reasonable water purification and pollutant control strategies (such as ecological restoration, sewage treatment, source control, etc.).

[0105] 5. Support decision-making and policy implementation Decision support: Simulating the diffusion path of pollutants can provide decision support for governments, environmental management agencies, scientific researchers, etc. By understanding the path and trend of pollutant diffusion, relevant departments can make reasonable environmental protection policies and measures to promote pollution control.

[0106] Monitoring and regulation: By simulating the diffusion paths of pollutants, governments can better monitor emissions from pollution sources and ensure that relevant laws and policies are effectively implemented.

[0107] 6. Improve pollutant traceability Pollution source tracing: In some pollution accidents or environmental pollution incidents, the source and propagation path of pollutants can be traced by simulating the diffusion path of pollutants. This is very important for accountability, environmental restoration and the formulation of remedial measures.

[0108] In an exemplary embodiment, the step of calculating the pollution contribution rate of the monitored area to the downstream area by using a graph neural network includes: Calculating the concentration of pollutants in the area to be monitored and the concentration of pollutants in the downstream area of ​​the area to be monitored by the graph neural network, wherein the pollutants include nitrogen and phosphorus; Based on the pollutant concentration of the area to be monitored and the pollutant concentration of the downstream area of ​​the area to be monitored, the pollution contribution rate of the area to be monitored to the downstream area is determined.

[0109] In an exemplary embodiment, the graph neural network is constructed by the following method: The to-be-monitored areas are used as graph nodes, and the rivers between the to-be-monitored areas are used as edges to construct a graph structure; The graph structure is iteratively trained to optimize the weights of the edges in the graph structure to obtain the graph neural network.

[0110] Figure 3 It is a structural schematic diagram of the graph neural network model provided by an embodiment of the present invention.

[0111] like Figure 3 Specifically, when establishing the graph structure, each monitoring point can be defined as a node , each hydrological connectivity is an edge , the edge weight Indicates that the pollutant is in the slave node Flow to Node process, the strength of the hydrological flow or the ability of pollution to spread.

[0112] After the graph structure is constructed, the edge weights can be adjusted using the training mechanism of the graph neural network (such as gradient descent). Optimization is performed so that the model can learn the appropriate pollution propagation pattern and accurately calculate the pollution contribution rate.

[0113] In the GNN model, information is transmitted through neighboring nodes of the graph. Each node updates its own state through the information of its neighboring nodes, and the information update rule is usually expressed as:

[0114] Is a node The hidden state after the jth iteration (which can be understood as the feature representation of the node), It is the edge The weight of (i.e., hydrological connectivity), Is a node The set of neighbor nodes of is the activation function (such as ReLU), and b is the bias term.

[0115] Initial contamination Initialization can be performed based on actual pollution concentration data.

[0116] At each time t, the transmission of pollution is based on the pollution of neighboring nodes and edge weights. The specific pollution propagation formula is:

[0117] in, Is a node external sources of contamination.

[0118] In practical applications, in order to calculate the pollution contribution rate, the contribution of each monitoring point can be reflected based on the final concentration of the pollutant. Assume that the target pollution point is node , each node can be calculated by the following steps For Node Pollution contribution: Pollution contribution rate Can be defined as a node For Node The contribution ratio of the dye concentration is calculated as follows:

[0119] in: Is a node The pollution concentration at the final time T is, The target pollution point Pollution concentration at the final time T.

[0120] The agricultural non-point source pollution early warning system provided by the present invention is described below. The agricultural non-point source pollution early warning system described below and the agricultural non-point source pollution early warning method described above can be referenced to each other.

[0121] Figure 4 This is one of the structural schematic diagrams of the agricultural non-point source pollution early warning system provided by the embodiment of the present invention.

[0122] Figure 5 This is the second structural diagram of the agricultural non-point source pollution early warning system provided by the embodiment of the present invention.

[0123] like Figure 4 and Figure 5 As shown, the agricultural non-point source pollution early warning system provided in this embodiment includes: The parameter acquisition module 401 is used to acquire the first pollution monitoring parameter of the area to be monitored; A parameter prediction module 402 is used to predict a second pollution monitoring parameter of the monitored area within a set time period based on the first pollution monitoring parameter by using a pre-built prediction model; A first warning module 403, configured to issue a first-level warning when the second pollution monitoring parameter meets a first warning condition, and issue a second-level warning when the second pollution monitoring parameter does not meet the first warning condition but meets a second warning condition; The second warning module 404 is used to calculate the pollution contribution rate of the monitored area to the downstream area through a graph neural network when the second pollution monitoring parameter does not meet the first warning condition and the second warning condition, and to issue a third-level warning if the pollution contribution rate meets the third warning condition.

[0124] like Figure 5 As shown, the agricultural non-point source pollution early warning system provided by the present embodiment can collect the required first pollution monitoring parameters through the data collection layer, wherein specific parameters can be collected through the aerial remote sensing platform and the ground sensor network and sent to the data collection layer, and the data of the data collection layer is then sent to the cloud platform layer, and multi-source data fusion is performed on all data in the cloud platform layer, and then data analysis and early warning are performed in the data analysis module, and the data analysis module may include the above-mentioned parameter acquisition module 401, parameter prediction module 402, first early warning module 403 and second early warning module 404. After the data processing of the above-mentioned data analysis module, the early warning result can be obtained, and the various early warning results obtained can be sent to the decision support interface of the early warning system. The decision support interface can be displayed on a display screen. On the one hand, the early warning signals of the first early warning, the second early warning and the third early warning can be displayed on the decision support interface. In addition, a pollution heat map can be generated according to the first pollution monitoring parameter collected above. The pollution heat map can more intuitively reflect the severity of pollution in each area. Not only that, after the early warning, certain governance measures can be provided according to the severity of the pollution, for example, the amount of fertilizer can be adjusted, that is, a variable fertilization prescription map can be provided.

[0125] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other via the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the agricultural non-point source pollution early warning method, which includes: Acquire a first pollution monitoring parameter of the area to be monitored; Based on the first pollution monitoring parameter, predicting the second pollution monitoring parameter of the area to be monitored within a set time period through a pre-built prediction model; When the second pollution monitoring parameter meets the first warning condition, a first-level warning is issued; when the second pollution monitoring parameter does not meet the first warning condition but meets the second warning condition, a second-level warning is issued; When the second pollution monitoring parameter does not meet the first warning condition and the second warning condition, the pollution contribution rate of the monitored area to the downstream area is calculated through the graph neural network. If the pollution contribution rate meets the third warning condition, a third-level warning is issued.

[0126] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0127] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the agricultural non-point source pollution early warning method provided by the above methods, and the method includes: Acquire a first pollution monitoring parameter of the area to be monitored; Based on the first pollution monitoring parameter, predicting the second pollution monitoring parameter of the area to be monitored within a set time period through a pre-built prediction model; When the second pollution monitoring parameter meets the first warning condition, a first-level warning is issued; when the second pollution monitoring parameter does not meet the first warning condition but meets the second warning condition, a second-level warning is issued; When the second pollution monitoring parameter does not meet the first warning condition and the second warning condition, the pollution contribution rate of the monitored area to the downstream area is calculated through the graph neural network. If the pollution contribution rate meets the third warning condition, a third-level warning is issued.

[0128] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the agricultural non-point source pollution early warning method provided by the above methods, the method comprising: Acquire a first pollution monitoring parameter of the area to be monitored; Based on the first pollution monitoring parameter, predicting the second pollution monitoring parameter of the area to be monitored within a set time period through a pre-built prediction model; When the second pollution monitoring parameter meets the first warning condition, a first-level warning is issued; when the second pollution monitoring parameter does not meet the first warning condition but meets the second warning condition, a second-level warning is issued; When the second pollution monitoring parameter does not meet the first warning condition and the second warning condition, the pollution contribution rate of the monitored area to the downstream area is calculated through the graph neural network. If the pollution contribution rate meets the third warning condition, a third-level warning is issued.

[0129] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0130] Through the description of the above implementation modes, those skilled in the art can clearly understand that each implementation mode can be implemented by means of software plus a necessary general hardware platform, or of course by hardware. Based on such an understanding, the above technical solution can essentially or in other words be embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0131] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. Agricultural non-point source pollution early warning method, characterized in that: include: Acquire a first pollution monitoring parameter of the area to be monitored; Based on the first pollution monitoring parameter, predicting the second pollution monitoring parameter of the area to be monitored within a set time period through a pre-built prediction model; When the second pollution monitoring parameter meets the first warning condition, a first-level warning is issued; when the second pollution monitoring parameter does not meet the first warning condition but meets the second warning condition, a second-level warning is issued; When the second pollution monitoring parameter does not meet the first warning condition and the second warning condition, the pollution contribution rate of the monitored area to the downstream area is calculated through the graph neural network. If the pollution contribution rate meets the third warning condition, a third-level warning is issued.

2. The agricultural non-point source pollution early warning method according to claim 1, characterized in that: The prediction model is a coupling model of a soil and water resources assessment model and a long short-term memory network model.

3. The agricultural non-point source pollution early warning method according to claim 2, characterized in that: The process of constructing the prediction model includes: Collecting training data, wherein the training data includes historical pollution monitoring parameters of the area to be monitored; Based on the training data, the soil and water resources assessment model and the long short-term memory network model are trained respectively; The output of the soil and water resources assessment model is used as the input of the long short-term memory network model, and the long short-term memory network model is iteratively optimized. Based on the output of the long short-term memory network model, the soil and water resources assessment model is iteratively optimized.

4. The agricultural non-point source pollution early warning method according to claim 1, characterized in that: The second pollution monitoring parameter includes soil nitrogen content and the probability of rainfall in the monitored area within a set time period, and the first warning condition includes: The soil nitrogen content exceeds a nitrogen content threshold, and the rainfall probability in the monitored area within a set time period exceeds a rainfall probability threshold.

5. The agricultural non-point source pollution early warning method according to claim 1, characterized in that: The second pollution monitoring parameter includes the phosphorus content at the outlet of the monitored area and the downstream flow rate of the monitored area, and the second early warning condition includes: The phosphorus content at the drainage outlet of the area to be monitored exceeds the phosphorus content threshold, and the downstream flow velocity of the area to be monitored is lower than the flow velocity threshold.

6. The agricultural non-point source pollution early warning method according to claim 1, characterized in that: The downstream area of ​​the area to be monitored is determined by the following method: Constructing a virtual river network model of the area to be monitored, wherein the virtual river network model at least reflects the geographic spatial data and hydrological data of the area to be monitored; Based on the virtual river network model, determining the pollutant diffusion path of the area to be monitored; Based on the pollutant diffusion path, a downstream area of ​​the area to be monitored is determined.

7. The agricultural non-point source pollution early warning method according to claim 1, characterized in that: The calculating the pollution contribution rate of the monitored area to the downstream area by using the graph neural network includes: Calculating the concentration of pollutants in the area to be monitored and the concentration of pollutants in the downstream area of ​​the area to be monitored by the graph neural network, wherein the pollutants include nitrogen and phosphorus; Based on the pollutant concentration of the area to be monitored and the pollutant concentration of the downstream area of ​​the area to be monitored, the pollution contribution rate of the area to be monitored to the downstream area is determined.

8. The agricultural non-point source pollution early warning method according to claim 7, characterized in that: The graph neural network is constructed by the following method: The to-be-monitored areas are used as graph nodes, and the rivers between the to-be-monitored areas are used as edges to construct a graph structure; The graph structure is iteratively trained to optimize the weights of the edges in the graph structure to obtain the graph neural network.

9. Agricultural non-point source pollution early warning system, characterized by: include: A parameter acquisition module, used to acquire a first pollution monitoring parameter of the area to be monitored; A parameter prediction module, used to predict a second pollution monitoring parameter of the area to be monitored within a set time period based on the first pollution monitoring parameter by using a pre-built prediction model; A first warning module, configured to issue a first-level warning when the second pollution monitoring parameter meets a first warning condition, and issue a second-level warning when the second pollution monitoring parameter does not meet the first warning condition but meets a second warning condition; The second early warning module is used to calculate the pollution contribution rate of the monitored area to the downstream area through a graph neural network when the second pollution monitoring parameter does not meet the first early warning condition and the second early warning condition, and to issue a third-level early warning if the pollution contribution rate meets the third early warning condition.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the agricultural non-point source pollution early warning method as described in any one of claims 1-8 is implemented.

Images