A water eutrophication monitoring and pollution early warning system

Abstract

This application relates to the technical field of lake eutrophication detection, and discloses a water eutrophication monitoring and pollution early warning system, including a data acquisition module, a data processing module, a model training module, and a monitoring and early warning module. The data acquisition module collects eutrophication factors at each monitoring station, including nutrient concentrations and cyanobacteria-sensitive factors. The data processing module organizes the eutrophication factors at each monitoring station and extracts their correlation features. The model training module trains a pollution prediction model. The monitoring and early warning module predicts the cyanobacteria-sensitive factors at each monitoring station based on the correlation features between the pollution prediction model and the eutrophication factors, and issues pollution early warnings based on the predicted values ​​of the cyanobacteria-sensitive factors. This application achieves high-precision prediction of water eutrophication by modeling the spatiotemporal correlation between nutrient concentrations and cyanobacteria response factors in water bodies.

Description

Technical Field

[0001] This application relates to the technical field of lake eutrophication detection, specifically to a water eutrophication monitoring and pollution early warning system. Background Technology

[0002] With the continuous increase in industrial emissions, agricultural non-point source pollution, and urban runoff, the accumulation of nutrients such as total nitrogen and total phosphorus in surface water bodies such as lakes and reservoirs is becoming increasingly serious, leading to frequent eutrophication and cyanobacterial blooms that pose a significant threat to ecosystems and public health. Traditional methods for monitoring eutrophication mainly rely on manual water sampling and laboratory testing, which are time-consuming and slow to respond, making it difficult to meet the real-time early warning needs for sudden water quality events such as cyanobacterial blooms. In recent years, with the development of sensor technology and the Internet of Things (IoT), automated water quality monitoring systems based on remote sensors and wireless communication have emerged. These systems can collect key indicators such as total nitrogen, total phosphorus, and chlorophyll a in water bodies in real time, and upload the data to a central platform for analysis and processing via wireless network, improving monitoring efficiency and the timeliness of information transmission. However, existing eutrophication monitoring and prediction systems still have many limitations.

[0003] Existing systems often lack effective mechanisms for integrating and analyzing data from multiple monitoring stations during data processing. This prevents the uncovering of potential spatiotemporal causal relationships between different monitoring points, relying solely on single-point time-series trends. Consequently, the understanding of eutrophication diffusion mechanisms is incomplete, and early warning responses are delayed. Most predictive models are still trained based on historical data from single stations, neglecting the impact of nutrient load inputs from surrounding stations on local cyanobacteria growth, thus limiting the model's generalization ability to complex ecosystems. Furthermore, current systems typically lack the ability to model the lag and driving relationships between cyanobacteria response indicators such as chlorophyll a, transparency, and dissolved oxygen and nutrient concentrations, making it difficult to explain certain nonlinear burst phenomena.

[0004] For example, patent application CN114113530A discloses a remote sensing-based method for monitoring lake eutrophication, including the following steps: Step 1: First, multiple water bodies around the target location are sampled. After sampling, the composition of the water bodies at different locations is analyzed. After analysis, remote sensing eutrophication monitoring stations are inserted and set up at multiple locations within the designated lake. This method, by setting up remote sensing eutrophication monitoring stations and floating boxes at multiple locations on the lake, can utilize remote sensing cameras on multiple monitoring stations to monitor multiple locations within the lake. When problems are detected in multiple locations, a sampling drone inside the floating box automatically samples and brings back the sampled water for researchers to analyze, thus conducting remote sensing monitoring.

[0005] Patent application CN108614086A discloses a method for evaluating lake eutrophication, relating to water environment pollution assessment methods. It aims to address the impact of multiple water quality parameter measurements on the total trophic index (TSI) in existing calculation methods. The method involves: 1. Sampling the lake to obtain water quality samples; 2. Measuring pH, salinity, water temperature, SD, TN, TP, DOC, Chla, and aOACs on the lake water samples; 3. Measuring aOACs, including ap and aCDOM; 4. Calculating the TSI using the measured aOACs, with data points evenly distributed on both sides of the regression line, thereby classifying the lake's eutrophication level. This method only requires measuring the absorption coefficient of optically active substances in the lake to calculate the TSI value using this model, which can then be used to evaluate lake eutrophication.

[0006] All of the above patents suffer from the problem described in this background: the lack of an effective mechanism for integrating and analyzing data from multiple monitoring stations makes it impossible to uncover potential spatiotemporal causal relationships between different monitoring points.

[0007] The information disclosed in this background section is intended only to enhance the understanding of the overall background of this application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0008] The technical problem to be solved by this application is to overcome the defects of the prior art and provide a water eutrophication monitoring and pollution early warning system, which realizes high-precision prediction of water eutrophication by modeling the spatiotemporal correlation between nutrient concentration in water and cyanobacteria response factors.

[0009] To solve the above-mentioned technical problems, this application provides the following technical solution:

[0010] A water eutrophication monitoring and pollution early warning system includes a data acquisition module, a data processing module, a model training module, and a monitoring and early warning module; wherein:

[0011] The data acquisition module is used to collect eutrophication factors of the water body at each monitoring station; the eutrophication factors include the nutrient concentration of the water body and cyanobacteria-sensitive factors;

[0012] The data processing module is used to organize the eutrophication factors of each monitoring station and extract the correlation features of the eutrophication factors.

[0013] The model training module is used to collect historical data on eutrophication factors in water bodies and train pollution prediction models.

[0014] The monitoring and early warning module calculates the predicted value of eutrophication factors for each monitoring station based on the correlation characteristics between the pollution prediction model and eutrophication factors, and issues pollution early warnings based on the predicted values ​​of cyanobacteria-sensitive factors.

[0015] As a preferred embodiment of the eutrophication monitoring and pollution early warning system for water bodies described in this application, the nutrient concentrations include at least total nitrogen content, total phosphorus content, ammonia nitrogen content, nitrate nitrogen content, and phosphate content; the cyanobacteria-sensitive factors include at least the chlorophyll a concentration, transparency, dissolved oxygen, and pH value of the water body.

[0016] The data acquisition module includes a first monitoring unit and a second monitoring unit, wherein the first monitoring unit is used to collect the nutrient concentration of the water at each monitoring station; and the second monitoring unit is used to collect the cyanobacteria-sensitive factors of the water at each monitoring station.

[0017] As a preferred embodiment of the eutrophication monitoring and pollution early warning system for water bodies described in this application, the data processing module includes a data processing unit and a feature extraction unit. The data processing unit is used to clean the eutrophication factors at each monitoring station, number each monitoring station, and process the eutrophication factors at each station into time series. The time series includes the time series of nutrient concentrations in the water body at each monitoring station and the time series of cyanobacteria-sensitive factors in the water body. The dimension of any time series is N, where N is a positive integer.

[0018] The feature extraction unit is used to extract the correlation features of the eutrophication factors, including establishing a distance matrix of the eutrophication factors and a correlation matrix of the eutrophication factors at each monitoring station.

[0019] As a preferred embodiment of the eutrophication monitoring and pollution early warning system for water bodies described in this application, the method by which the feature extraction unit establishes the distance matrix of eutrophication factors is as follows:

[0020] Initialize an n×n distance matrix, where n is the number of monitoring stations; calculate the distance between any two monitoring stations and assign values ​​to the distance matrix; wherein the value of any element in the distance matrix is ​​the distance between the corresponding two monitoring stations.

[0021] As a preferred embodiment of the eutrophication monitoring and pollution early warning system for water bodies described in this application, the method by which the feature extraction unit establishes the correlation matrix of eutrophication factors for any monitoring station is as follows:

[0022] Initialize a p×q correlation matrix, where p is the number of nutrient types corresponding to the nutrient concentration; and q is the number of cyanobacterial sensitive factors.

[0023] Calculate the correlation influence of any nutrient concentration on any cyanobacterial sensitizer, and assign values ​​to the elements in the correlation matrix; wherein, the element in the i-th row and j-th column of the correlation matrix represents the correlation influence of the i-th nutrient concentration on the j-th cyanobacterial sensitizer, denoted as c. ij The range of values ​​for i is 0, 1, ..., p; the range of values ​​for j is 0, 1, ..., q.

[0024] As a preferred embodiment of the water eutrophication monitoring and pollution early warning system described in this application, the model training module includes a historical data unit and a model training unit. The historical data unit is used to collect historical eutrophication factors for each monitoring station and transmit the historical eutrophication factors to the data processing module. The data processing module establishes the time series of historical eutrophication factors, the distance matrix of historical eutrophication factors, and the correlation matrix of historical eutrophication factors for each monitoring station and transmits them to the model training unit.

[0025] The model training unit divides the training set and validation set based on the time series of historical eutrophication factors, the distance matrix of historical eutrophication factors, and the correlation matrix of historical eutrophication factors for each monitoring station, and trains, validates, and optimizes the pollution prediction model. The trained pollution prediction model is then saved and deployed to the monitoring and early warning module.

[0026] As a preferred embodiment of the eutrophication monitoring and pollution early warning system for water bodies described in this application, the feature extraction unit is configured with a correlation evaluation strategy to calculate the correlation influence of any nutrient concentration at each monitoring station on any cyanobacterial sensitive factor; the correlation evaluation strategy specifically includes:

[0027] For any monitoring station, obtain the time series of historical eutrophication factors, including the time series of each nutrient concentration constructed based on historical eutrophication factors and the time series of cyanobacterial sensitive factors;

[0028] Based on the time series of the historical eutrophication factors, the correlation index between the concentration of the i-th nutrient salt and the j-th cyanobacterial sensitive factor at different lag steps τ is calculated, and the correlation index with the largest absolute value is denoted as R. ijmax Where τ is an integer, ranging from -N+1 to N-1; the feature extraction unit is also configured with an association threshold r; based on R ijmax The correlation influence is assigned a value based on the correlation threshold r, as follows:

[0029] If R ijmax The absolute value of R is greater than the correlation threshold r, and R ijmax If the corresponding lag step is a positive integer, then the correlation influence degree c ij The value is assigned to τ;

[0030] If R ijmax The absolute value of R is less than or equal to the correlation threshold r, or R ijmax The corresponding lag step size is a negative integer, and the correlation influence degree c ij The value is assigned to 0.

[0031] As a preferred embodiment of the eutrophication monitoring and pollution early warning system for water bodies described in this application, the formula for calculating the correlation index between any nutrient concentration and any cyanobacterial sensitive factor at any monitoring station is as follows:

[0032]

[0033] Among them, R ij (τ) represents the correlation index between the concentration of the i-th nutrient and the j-th cyanobacterial sensitive factor at a lag of τ time steps; τ represents the lag step size; N represents the time series dimension of the cyanobacterial sensitive factor in the water body; x t Let represent the t-th element in the time series of the concentration of the i-th nutrient, where t ranges from 1, 2, ..., N; x0 represents the mean element of the time series of the concentration of the i-th nutrient; y t+τ y0 represents the (t+τ)th element in the time series of the j-th cyanobacterial sensitizer; y0 represents the mean element of the time series of the j-th cyanobacterial sensitizer.

[0034] As a preferred embodiment of the water eutrophication monitoring and pollution early warning system described in this application, the monitoring and early warning module includes a prediction model unit and an early warning unit; wherein the prediction model unit is equipped with a trained pollution prediction model for calculating the predicted value of cyanobacteria-sensitive factors at each monitoring station.

[0035] The pollution prediction model takes into account the time series of nutrient concentrations in the water at each monitoring station, the time series of cyanobacterial sensitive factors, the distance matrix of eutrophication factors, and the correlation matrix of eutrophication factors. The output is the predicted values ​​of nutrient concentrations and cyanobacterial sensitive factors in the water at each monitoring station over the next M time steps, where M is a positive integer.

[0036] The early warning unit is equipped with a maximum threshold for the concentration of each nutrient and a maximum threshold for each cyanobacterial sensitive factor. When the predicted value of the concentration of any nutrient in the water at any monitoring station at any future time step is higher than the corresponding maximum threshold, or the predicted value of any cyanobacterial sensitive factor is higher than the corresponding maximum threshold, the early warning unit issues a pollution early warning message.

[0037] As a preferred embodiment of the eutrophication monitoring and pollution early warning system for water bodies described in this application, the pollution prediction model includes an input layer, an LSTM layer, an attention layer, a fusion layer, and an output layer; wherein:

[0038] The input layer is used to receive the time series of eutrophication factors, distance matrix, and correlation matrix of each monitoring station as input;

[0039] LSTM layers are used to extract the temporal features of time series of eutrophication factors;

[0040] The attention layer is used to extract the spatial features of the distance matrix and the correlation matrix;

[0041] The fusion layer is used to generate the fused features of the temporal and spatial features;

[0042] Based on the fusion features, the output layer calculates and outputs the predicted value of the eutrophication factor for each monitoring station at the next M time steps.

[0043] Compared with the prior art, the beneficial effects achieved by this application are as follows:

[0044] This invention constructs a time series of nutrient concentration data in water bodies, combines it with a distance matrix between monitoring stations, and further establishes a correlation matrix that reflects the spatiotemporal driving relationship between nutrient input and cyanobacterial response. This not only helps to identify the diffusion path and scope of influence in the process of eutrophication, but also reveals the role of different monitoring stations in the cyanobacterial growth response, enhancing the system's understanding of the evolution mechanism of cyanobacterial blooms.

[0045] The eutrophication prediction model trained in this invention introduces an LSTM structure to capture the temporal evolution characteristics of nutrient concentration and cyanobacterial response factors. At the same time, it uses an attention mechanism to process the distance matrix and the cyanobacterial sensitivity causal matrix, extracts the spatial influence weights between different monitoring stations, and achieves spatiotemporal feature fusion, thereby significantly improving the prediction accuracy and early warning for cyanobacterial outbreaks. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0047] Figure 1 This is a schematic diagram of the water eutrophication monitoring and pollution early warning system provided by the present invention.

[0048] Figure 2 This is a schematic diagram of the working mode of the water eutrophication monitoring and pollution early warning system provided by the present invention. Detailed Implementation

[0049] The technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solution of this application, rather than limitations thereof. In the absence of conflict, the embodiments and technical features in the embodiments can be combined with each other.

[0050] This embodiment introduces a water eutrophication monitoring and pollution early warning system, referring to... Figure 1 The system includes a data acquisition module, a data processing module, a model training module, a monitoring and early warning module, and a feedback and optimization module; the functions of each module are as follows: Figure 2 As shown, where:

[0051] The data acquisition module is used to collect eutrophication factors at each monitoring station; the eutrophication factors include the nutrient concentration of the water body and cyanobacteria-sensitive factors.

[0052] The nutrient concentrations include total nitrogen, total phosphorus, ammonia nitrogen, nitrate nitrogen, and phosphate content. Nutrient concentration reflects the concentration of the main nutrients in the water body that support the large-scale reproduction of algae and other phytoplankton. For example, total nitrogen is the total amount of various nitrogen elements in the water, such as ammonia nitrogen, nitrate nitrogen, and organic nitrogen, and is an important source for protein synthesis by cyanobacteria. Phosphate is the soluble form of phosphorus, which is the preferred form for direct absorption by cyanobacteria. Long-term accumulation or short-term large-scale input of these nutrients can induce eutrophication, leading to cyanobacterial blooms and water pollution. The cyanobacteria-sensitive factors include the chlorophyll a concentration, transparency, dissolved oxygen, and pH value of the water body. These cyanobacteria-sensitive factors are physicochemical parameters used to characterize the water body's response to cyanobacterial growth. For example, chlorophyll a represents the photosynthetic pigment content within cyanobacteria and is a direct indicator of algal biomass; an increase in water pH is often associated with active cyanobacterial photosynthesis. The temporal fluctuations of cyanobacterial sensitive factors are a response to eutrophication. Therefore, this scheme improves the predictive model's ability to calculate the timing of algal blooms by monitoring the changing trends of these indicators and establishing their correlation with nutrient concentrations.

[0053] The data acquisition module includes a first monitoring unit and a second monitoring unit. The first monitoring unit is used to collect the nutrient concentration of the water at each monitoring station; the second monitoring unit is used to collect the cyanobacteria-sensitive factors of the water at each monitoring station. Both the first and second monitoring units are equipped with multi-parameter water quality analysis instruments, including automatic sampling systems, and can be mounted on buoys, platforms, or unmanned monitoring vessels.

[0054] The data processing module is used to organize the eutrophication factors of each monitoring station and extract the correlation features of the eutrophication factors.

[0055] The data processing module includes a data processing unit and a feature extraction unit. The data processing unit is used to clean the eutrophication factors of each monitoring station, number each monitoring station, and process the eutrophication factors of each monitoring station into time series. The time series includes the time series of nutrient concentrations in the water body of each monitoring station and the time series of cyanobacteria-sensitive factors. The dimension of any time series is N, where N is a positive integer.

[0056] The feature extraction unit is used to extract the correlation features of the eutrophication factors, including establishing a distance matrix of the eutrophication factors and a correlation matrix of the eutrophication factors at each monitoring station; wherein, the method for establishing the distance matrix of the eutrophication factors is as follows:

[0057] Initialize an n×n distance matrix, where n is the number of monitoring stations; calculate the distance between any two monitoring stations and assign values ​​to the distance matrix; wherein the value of any element in the distance matrix is ​​the distance between the corresponding two monitoring stations;

[0058] The method for establishing the correlation matrix of eutrophication factors at any monitoring station is as follows:

[0059] Initialize a p×q correlation matrix, where p is the number of nutrient types corresponding to the nutrient concentration; and q is the number of cyanobacterial sensitive factors.

[0060] Calculate the correlation influence of any nutrient concentration on any cyanobacterial sensitizer, and assign values ​​to the elements in the correlation matrix; wherein, the element in the i-th row and j-th column of the correlation matrix represents the correlation influence of the i-th nutrient concentration on the j-th cyanobacterial sensitizer, denoted as c. ij The range of values ​​for i is 0, 1, ..., p; the range of values ​​for j is 0, 1, ..., q.

[0061] The feature extraction unit is configured with an association evaluation strategy to calculate the correlation influence of any nutrient concentration at each monitoring station on any cyanobacterial sensitive factor; the association evaluation strategy specifically includes:

[0062] For any monitoring station, obtain the time series of historical eutrophication factors, including the time series of each nutrient concentration constructed based on historical eutrophication factors and the time series of cyanobacterial sensitive factors;

[0063] Based on the time series of the historical eutrophication factors, the correlation index between the concentration of the i-th nutrient salt and the j-th cyanobacterial sensitive factor at different lag steps τ is calculated, and the correlation index with the largest absolute value is denoted as R. ijmaxWhere τ is an integer, ranging from -N+1 to N-1; the feature extraction unit is also configured with an association threshold r; based on R ijmax The correlation influence is assigned a value based on the correlation threshold r, as follows:

[0064] If R ijmax The absolute value of R is greater than the correlation threshold r, and R ijmax If the corresponding lag step is a positive integer, then the correlation influence degree c ij The value is assigned to τ;

[0065] If R ijmax The absolute value of R is less than or equal to the correlation threshold r, or R ijmax The corresponding lag step size is a negative integer, and the correlation influence degree c ij The value is assigned to 0.

[0066] For any monitoring station, the formula for calculating the correlation index between the concentration of any nutrient and any cyanobacterial sensitive factor is as follows:

[0067]

[0068] Among them, R ij (τ) represents the correlation index between the concentration of the i-th nutrient and the j-th cyanobacterial sensitive factor at a lag of τ time steps; τ represents the lag step size; N represents the time series dimension of the cyanobacterial sensitive factor in the water body; x t Let represent the t-th element in the time series of the concentration of the i-th nutrient, where t ranges from 1, 2, ..., N; x0 represents the mean element of the time series of the concentration of the i-th nutrient; y t+τ y0 represents the (t+τ)th element in the time series of the j-th cyanobacterial sensitizer; y0 represents the mean element of the time series of the j-th cyanobacterial sensitizer.

[0069] This application introduces an evaluation strategy for the lagged correlation between eutrophication factors and cyanobacterial response factors; based on the correlation between nutrient concentration factors and cyanobacterial sensitivity factors at different time lag steps, a correlation matrix is ​​constructed to characterize the delayed driving process of nutrient input changes on the ecological response of aquatic bodies.

[0070] For any monitoring station, the time series of its historical nutrient concentration factors and the corresponding time series of cyanobacterial sensitivity factors are extracted, and the correlation index of each pair of indicators is calculated within a set lag window. If the absolute value of the maximum correlation index corresponding to any lag step is greater than the set correlation threshold, and its lag step is a positive integer, it indicates that the nutrient factor has a significant positive impact on the corresponding cyanobacterial sensitivity factor. The influence weight of this pair of factors in the causal matrix is ​​then assigned to the corresponding lag step. If the absolute value of the maximum correlation does not exceed the threshold or the lag step is negative, it is determined that there is no significant causal relationship between the two. This strategy and method can not only identify whether there is a significant time-lag association between water nutrient input and ecological response, but also achieve quantitative extraction of influence paths, providing support for interpretable prediction of cyanobacterial risk.

[0071] The model training module is used to collect historical data on eutrophication factors in water bodies and train pollution prediction models.

[0072] The model training module includes a historical data unit and a model training unit. The historical data unit is used to collect historical eutrophication factors for each monitoring station and transmit the historical eutrophication factors to the data processing module. The data processing module establishes the time series of historical eutrophication factors, the distance matrix of historical eutrophication factors, and the correlation matrix of historical eutrophication factors for each monitoring station and transmits them to the model training unit.

[0073] The model training unit divides the training set and validation set based on the time series of historical eutrophication factors, the distance matrix of historical eutrophication factors, and the correlation matrix of historical eutrophication factors for each monitoring station, and trains, validates, and optimizes the pollution prediction model. The trained pollution prediction model is then saved and deployed to the monitoring and early warning module.

[0074] The monitoring and early warning module calculates the predicted value of eutrophication factors for each monitoring station based on the correlation characteristics between the pollution prediction model and eutrophication factors, and issues pollution early warnings based on the predicted values ​​of cyanobacteria-sensitive factors.

[0075] The monitoring and early warning module includes a prediction model unit and an early warning unit; wherein, the prediction model unit is equipped with a trained pollution prediction model, which is used to calculate the predicted value of the cyanobacteria sensitive factor for each monitoring station.

[0076] The pollution prediction model takes into account the time series of nutrient concentrations in the water at each monitoring station, the time series of cyanobacterial sensitive factors, the distance matrix of eutrophication factors, and the correlation matrix of eutrophication factors. The output is the predicted values ​​of nutrient concentrations and cyanobacterial sensitive factors in the water at each monitoring station over the next M time steps, where M is a positive integer.

[0077] The pollution prediction model includes an input layer, an LSTM layer, an attention layer, a fusion layer, and an output layer; wherein:

[0078] The input layer is used to receive the time series of eutrophication factors, distance matrix, and correlation matrix of each monitoring station as input;

[0079] LSTM layers are used to extract the temporal features of the time series of eutrophication factors; the time series of eutrophication factors for each monitoring station are encoded by one or more LSTM units to capture long-term dependencies in the time series.

[0080] An attention layer is used to extract the spatial features of the distance and correlation matrices; a self-attention mechanism is used to process the distance and correlation matrices. This application employs a self-attention mechanism to process the distance and correlation matrices. The attention mechanism dynamically assigns spatial weights to the eutrophication factor time series of each monitoring station based on the geographical distance between monitoring stations and the causal response relationship between nutrient factors and cyanobacterial response factors, thereby more effectively modeling the ecological impacts and response paths between different stations. In the attention layer, the eutrophication time series data of each station are weighted and fused with the time series data of other stations; by learning the weight allocation process, the model can automatically identify other key stations that have a greater impact on the cyanobacterial response changes of the target station. This mechanism improves the model's spatial sensitivity and explanatory power for the risk of regional cyanobacterial outbreaks in lakes, and is particularly suitable for complex aquatic environments affected by multiple pollution sources or exhibiting spatial heterogeneity in eutrophication processes.

[0081] The fusion layer is used to generate the fused features of the temporal and spatial features. First, a multi-head attention mechanism is used to comprehensively consider the interaction between time and space, and the output of the LSTM layer is combined with the output of the attention layer. Then, one or more fully connected layers are used to perform feature fusion, further fusing feature information from different dimensions.

[0082] Based on the fused features, the output layer calculates and outputs the predicted values ​​of eutrophication factors for each monitoring station over the next M time steps. The output layer contains 2×M×N neurons, and through an appropriate activation function, such as a linear activation function, it maps the fused features to the predicted values ​​of continuous nutrient concentrations and cyanobacteria-sensitive factors.

[0083] This model structure effectively utilizes time-series data, distance information, and the correlation between nutrient concentration and cyanobacteria-sensitive factors to predict future changes in eutrophication factors in lakes, thus ensuring the accuracy and reliability of the prediction.

[0084] The early warning unit is configured with maximum threshold values ​​for each nutrient concentration and for each cyanobacterial sensitive factor. When the predicted value of any nutrient concentration or any cyanobacterial sensitive factor in the water body at any monitoring station at any future time step exceeds the corresponding maximum threshold, the early warning unit issues a pollution warning. The pollution warning information includes the monitoring station at risk of pollution and the predicted time when the nutrient concentration or cyanobacterial sensitive factor exceeds the standard.

[0085] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application 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.

[0086] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of protection of this application, and these forms are all within the protection scope of this application.

Claims

1. A water eutrophication monitoring and pollution early warning system, characterized in that: It includes a data acquisition module, a data processing module, a model training module, and a monitoring and early warning module; among which: The data acquisition module is used to collect eutrophication factors of the water body at each monitoring station; the eutrophication factors include the nutrient concentration of the water body and cyanobacteria-sensitive factors; The nutrient concentrations include at least total nitrogen, total phosphorus, ammonia nitrogen, nitrate nitrogen, and phosphate content; the cyanobacteria-sensitive factors include at least the chlorophyll a concentration, transparency, dissolved oxygen, and pH value of the water body. The data processing module is used to organize the eutrophication factors of each monitoring station and extract the correlation features of the eutrophication factors. The data processing module includes a data processing unit and a feature extraction unit. The data processing unit is used to clean the eutrophication factors of each monitoring station, number each monitoring station, and process the eutrophication factors of each monitoring station into time series. The time series includes the time series of nutrient concentrations in the water body at each monitoring station and the time series of cyanobacteria-sensitive factors in the water body. The dimension of any time series is N, where N is a positive integer. The feature extraction unit is used to extract the correlation features of the eutrophication factors, including establishing a distance matrix of the eutrophication factors and a correlation matrix of the eutrophication factors at each monitoring station. The method by which the feature extraction unit establishes the distance matrix of eutrophication factors is as follows: Initialize an n×n distance matrix, where n is the number of monitoring stations; calculate the distance between any two monitoring stations and assign values ​​to the distance matrix; wherein the value of any element in the distance matrix is ​​the distance between the corresponding two monitoring stations; The method by which the feature extraction unit establishes the correlation matrix of eutrophication factors for any monitoring site is as follows: Initialize a p×q correlation matrix, where p is the number of nutrient types corresponding to the nutrient concentration; and q is the number of cyanobacterial sensitivity factors. Calculate the correlation influence of any nutrient concentration on any cyanobacterial sensitizer, and assign values ​​to the elements in the correlation matrix; wherein, the element in the i-th row and j-th column of the correlation matrix represents the correlation influence of the i-th nutrient concentration on the j-th cyanobacterial sensitizer, denoted as . The value range of i is 0, 1, ..., p; the value range of j is 0, 1, ..., q; The feature extraction unit is configured with an association evaluation strategy to calculate the correlation influence of any nutrient concentration at each monitoring station on any cyanobacterial sensitive factor; the association evaluation strategy specifically includes: For any monitoring station, obtain the time series of historical eutrophication factors, including the time series of each nutrient concentration constructed based on historical eutrophication factors and the time series of cyanobacterial sensitive factors; Based on the time series of the historical eutrophication factors, the lag step lengths for the concentrations of the i-th nutrient and the j-th cyanobacteria-sensitive factor were calculated. The correlation index is set below, and the correlation index with the largest absolute value is marked as the correlation index. ;in, The integer value ranges from -N+1 to N-1; the feature extraction unit is also configured with an association threshold r; based on The correlation influence is assigned a value based on the correlation threshold r, as follows: like The absolute value is greater than the correlation threshold r, and If the corresponding lag step is a positive integer, then the correlation influence is... Assigned value ; like The absolute value is less than or equal to the correlation threshold r, or The corresponding lag step size is a negative integer, indicating the degree of correlation influence. The value is assigned to 0; For any monitoring station, the formula for calculating the correlation index between the concentration of any nutrient and any cyanobacterial sensitive factor is as follows: ； in, This indicates that the concentration of the i-th nutrient salt and the j-th cyanobacterial sensitive factor are in lag. The correlation index at each time step; Indicates the lag step size; N represents the time series dimension of the cyanobacteria-sensitive factor in the water body; Let t represent the t-th element in the time series of the concentration of the i-th nutrient, where t takes values ​​from 1, 2, ..., N; The mean element representing the time series of the concentration of the i-th nutrient; This represents the time series of the j-th cyanobacterial sensitivity factor. One element; The mean element represents the time series of the j-th cyanobacterial sensitivity factor; The model training module is used to collect historical data on eutrophication factors in water bodies and train pollution prediction models. The pollution prediction model takes into account the time series of nutrient concentrations in the water at each monitoring station, the time series of cyanobacterial sensitive factors, the distance matrix of eutrophication factors, and the correlation matrix of eutrophication factors. The output is the predicted values ​​of nutrient concentrations and cyanobacterial sensitive factors in the water at each monitoring station over the next M time steps; M is a positive integer. The pollution prediction model includes an input layer, an LSTM layer, an attention layer, a fusion layer, and an output layer; wherein: The input layer is used to receive the time series of eutrophication factors, distance matrix, and correlation matrix of each monitoring station as input; LSTM layers are used to extract the temporal features of time series of eutrophication factors; The attention layer is used to extract the spatial features of the distance matrix and the correlation matrix; The fusion layer is used to generate the fused features of the temporal and spatial features; Based on the fusion features, the output layer calculates and outputs the predicted value of the eutrophication factor for each monitoring station at the next M time steps; The monitoring and early warning module calculates the predicted value of eutrophication factors for each monitoring station based on the correlation characteristics between the pollution prediction model and eutrophication factors, and issues pollution early warnings based on the predicted values ​​of cyanobacteria-sensitive factors.

2. The eutrophication monitoring and pollution early warning system for water bodies as described in claim 1, characterized in that: The data acquisition module includes a first monitoring unit and a second monitoring unit, wherein the first monitoring unit is used to collect the nutrient concentration of the water at each monitoring station; and the second monitoring unit is used to collect the cyanobacteria-sensitive factors of the water at each monitoring station.

3. The eutrophication monitoring and pollution early warning system for water bodies as described in claim 2, characterized in that: The model training module includes a historical data unit and a model training unit. The historical data unit is used to collect historical eutrophication factors for each monitoring station and transmit the historical eutrophication factors to the data processing module. The data processing module establishes the time series of historical eutrophication factors, the distance matrix of historical eutrophication factors, and the correlation matrix of historical eutrophication factors for each monitoring station and transmits them to the model training unit. The model training unit divides the training set and validation set based on the time series of historical eutrophication factors, the distance matrix of historical eutrophication factors, and the correlation matrix of historical eutrophication factors for each monitoring station, and trains, validates, and optimizes the pollution prediction model. The trained pollution prediction model is then saved and deployed to the monitoring and early warning module.

4. The eutrophication monitoring and pollution early warning system for water bodies as described in claim 3, characterized in that: The monitoring and early warning module includes a prediction model unit and an early warning unit; wherein, the prediction model unit is equipped with a trained pollution prediction model, which is used to calculate the predicted value of the cyanobacteria sensitive factor for each monitoring station. The early warning unit is equipped with a maximum threshold for the concentration of each nutrient and a maximum threshold for each cyanobacterial sensitive factor. When the predicted value of the concentration of any nutrient in the water at any monitoring station at any future time step is higher than the corresponding maximum threshold, or the predicted value of any cyanobacterial sensitive factor is higher than the corresponding maximum threshold, the early warning unit issues a pollution early warning message.

Images