Agricultural non-point source pollution load calculation method based on trend pre-judgment

Abstract

The invention relates to the technical field of farmland non-point source pollution control, in particular to an agricultural non-point source pollution load calculation method based on trend pre-judgment. The invention discloses an agricultural non-point source pollution load calculation method based on trend pre-judgment. The method comprises the steps of S1, data collection and preprocessing; s2, trend modeling and future prediction; s3, future pollution load calculation; s4, calculating a future pollution load weight; s5, a preliminary integration formula of non-point source pollution load calculation; s6, performing iterative optimization and convergence judgment; and S7, result calculation. Through fusion of future trend prediction, entropy weight method weight distribution and a dynamic correction function, the problems of insufficient foresight, weight staticization and environmental response lag in the prior art are solved. According to the method, the current pollution load calculation result can be dynamically corrected, and a scientific basis is provided for agricultural non-point source pollution prevention and control.

Description

Technical Field

[0001] The present invention relates to the technical field of farmland non-point source pollution prevention and control, and specifically relates to a calculation method for agricultural non-point source pollution load based on trend prediction. Background Art

[0002] Agricultural non-point source pollution is one of the main sources causing water eutrophication, soil degradation and ecosystem imbalance. Its main pollution sources include chemical fertilizers, pesticides, livestock and poultry breeding waste, and microplastics, etc., which have the characteristics of dispersion, concealment and randomness. Accurately calculating the agricultural non-point source pollution load is the key basis for formulating pollution prevention and control policies, optimizing resource allocation and assessing ecological risks.

[0003] The current mainstream calculation methods for agricultural non-point source pollution load mainly include the following three categories: (1) Statistical models: regression analysis or correlation models based on historical data, but without considering the temporal characteristics of pollution source changes, unable to predict future trends, resulting in results lagging behind actual environmental changes. (2) Mechanistic models (such as SWAT, HSPF): relying on physical mechanisms to simulate the pollution migration process, but requiring a large number of parameters and complex calculations, lacking a real-time correction mechanism for dynamic environmental factors (such as rainfall, soil type), and being difficult to adapt to dynamic environmental changes. (3) Data-driven models (such as machine learning): although they can capture non-linear relationships, the weight allocation depends on experience or static data, unable to reflect the contribution degree changes of future pollution sources, lacking the ability to predict future trends, and lacking a mechanism for dynamically correcting weights.

[0004] Therefore, there is an urgent need for a calculation method for agricultural non-point source pollution load that can combine future trend prediction and dynamic correction, so as to improve the accuracy and practicality of calculation results and provide a scientific basis for agricultural non-point source pollution prevention and control. Summary of the Invention

[0005] (1) The problems to be solved by the present invention are: the problems of not considering future trends, insufficient dynamic correction and static weight allocation in the prior art.

[0006] (2) Technical Solution

[0007] A calculation method for agricultural non-point source pollution load based on trend prediction includes the following steps:

[0008] S1: Data collection and preprocessing: Collect historical-year related data on farmland non-point source pollution from official statistical data, research reports, and field monitoring data, including: Pollution source data: perfluorinated compound quantity Q PFOS , pesticide usage Q pesticide , microplastic content C microplastic ; Environmental data: farmland area A, rainfall R, soil type, temperature, etc.; Data cleaning: Remove outliers and fill in missing values, and organize the data by year;

[0009] S2: Trend Modeling and Future Prediction: Based on historical data, establish trend models for various pollution sources: number; ∈ t : white noise error term, t: the t-th year;

[0010] Use the trained trend models to predict the data of various pollution sources for the next N years:

[0011]

[0012] where, Y t-p is the pollution data for the (t - p)-th year;

[0013] S3: Future Pollution Load Calculation: Based on the predicted values for the next N years, calculate the pollution loads of various pollution sources;

[0014] S4: Future Pollution Load Weight Calculation:

[0015] Construct the data matrix X = [x ij m×n ;

[0016] where, m: number of samples; n: number of variables; i: sample index; j: variable index; x ij : the data value of the j-th pollution source in the i-th sample;

[0017] Calculate the weights of various pollution sources for the next N years:

[0018]

[0019]

[0020] where, Ej: information entropy of the j-th type of pollution; p ij : x ij standardized; wj: weight of the j-th type of pollution;

[0021] S5: Preliminary Integration Formula for Non-Point Source Pollution Load Calculation: Establish a preliminary integration formula for calculating the non-point source pollution load in current agriculture:

[0022]

[0023] where, w i : weight of the i-th pollution source for the next N years; L i : current pollution load of the i-th pollution source; R: future rainfall; R0: reference rainfall; K: soil erosion factor; K0: reference soil erosion factor; w1, w2, w3: weight coefficients obtained by fitting historical data; k: exponential coefficient reflecting the non-linear impact of rainfall on pollution load; ​

[0024] S6: Iterative Optimization and Convergence Judgment:

[0025] Calculate and optimize the weight w through the AI algorithm i and the exponential coefficient k to minimize the prediction error of the integration formula;

[0026] S7: Result Calculation:

[0027] When the convergence condition is reached, the final calculation result L of the farmland non-point source pollution load can be calculated and output current .

[0028] Preferably, the perfluorinated compounds include perfluoro and polyfluoroalkyl compounds; the pesticides include herbicides, insecticides, and fungicides.

[0029] Preferably, in S1, the method for filling missing values is the mean filling method:

[0030]

[0031] where x missing : the missing value of the pollution source;

[0032] Preferably, the pollution loads of various pollution sources in S3 include:

[0033] Perfluorinated Compound Pollution Load:

[0034] L PFAS = Q PFAS × α × R

[0035] where Q PFAS : the amount of perfluorinated compounds; α: the loss coefficient (related to soil type); R: the rainfall erosion factor;

[0036] Pesticide Pollution Load:

[0037] L pesticide = Q pesticide × β × K oc

[0038] where Q pesticide : the amount of pesticide used; β: the migration coefficient; K oc : the soil adsorption coefficient;

[0039] Microplastic Pollution Load:

[0040] L microplastic = C microplastic × A × γ

[0041] where C microplastic : the microplastic content; A: the farmland area; γ: the surface runoff carrying rate;

[0042] Preferably, the AI algorithm described in S6 is the particle swarm optimization algorithm: the optimization weight w to be optimized i and the exponential coefficient k are encoded as particles of the AI algorithm, and N particles are randomly generated. Each particle represents a set of parameter values, and the particle velocity v is iteratively updated i and the position x i :

[0043]

[0044] where: ω is the inertia weight, c1, c2 are the learning factors, r1, r2 are random numbers from 0 to 1, p best is the individual historical optimal position, and g best is the global optimal position;

[0045] The parameter constraints of the particle swarm optimization algorithm include: w1 + w2 + w3 = 1 and k > 0;

[0046] Preferably, the convergence condition described in S7 is:

[0047]

[0048] where, L t is the measured pollution load, is the model prediction value.

[0049] Advantages of the present invention:

[0050] (1) Prospective: Through future trend prediction, the future pollution source change trend is used to reverse-correct the current load calculation, improving the scientificity of long-term prevention and control strategies. (2) Dynamic: Introduce a dynamic correction function and a comprehensive correction function to respond in real time to the dynamic changes of environmental factors such as rainfall and soil type. (3) Expandable: New pollution sources or new parameters can be incorporated to expand the application scope of the method. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0052] Figure 1 is the method flow chart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] The technical solution of the present invention will be clearly and completely described below in conjunction with embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0054] A method for calculating agricultural non-point source pollution load based on trend prediction includes the following steps:

[0055] S1: Data collection and preprocessing: Collect data related to farmland non-point source pollution in historical years from official statistical data, research reports, and field monitoring data, including: Pollution source data: perfluorinated compound quantity Q PFOS , pesticide usage Q pesticide , microplastic content C microplastic ; Environmental data: farmland area A, rainfall R, soil type, temperature, etc.; Data cleaning: Remove outliers and fill in missing values, and organize the data by year; The perfluorinated compounds include perfluoroalkyl and polyfluoroalkyl compounds; Pesticides include herbicides, insecticides, and fungicides.

[0056] The mean filling method is used to fill in the missing values:

[0057]

[0058] where x missing : the missing value of the pollution source;

[0059] S2: Trend modeling and future prediction: Based on historical data, establish trend models for various pollution sources: number; ∈ t : white noise error term, t: the t-th year;

[0060] Use the trained trend model to predict the data of various pollution sources in the next N years:

[0061]

[0062] where Y t-p is the pollution data in the (t - p)-th year;

[0063] S3: Future pollution load calculation: Based on the predicted values in the next N years, calculate the pollution loads of various pollution sources:

[0064] Perfluorinated compound pollution load:

[0065] L PFAS = Q PFAS ×α×R

[0066] where Q PFAS: Perfluorinated compound content; α: Loss coefficient, related to soil type; R: Rainfall erosion factor, calculated by the RUSLE model;

[0067] Pesticide pollution load:

[0068] L pesticide = Q pesticide ×β×K oc

[0069] Among them, Q pesticide : Pesticide application amount; β: Migration coefficient; K oc : Soil adsorption coefficient;

[0070] Microplastic pollution load:

[0071] L microplastic = C microplastic ×A×γ

[0072] Among them, C microplastic : Microplastic content; A: Farmland area; γ: Surface runoff carrying rate;

[0073] S4: Future pollution load weight calculation:

[0074] Construct the data matrix X = [x ij m×n ;

[0075] Among them, m: Number of samples; n: Number of variables; i: Sample index; j: Variable index; x ij : Data value of the jth pollution source in the ith sample;

[0076] Calculate the weights of various pollution sources in the next N years:

[0077]

[0078] Among them, Ej: Information entropy of the jth type of pollution; p ij : x ij Normalized; wj: Weight of the jth type of pollution;

[0079] S5: Preliminary integration formula for calculating non-point source pollution load: Establish a preliminary integration formula for calculating the current agricultural non-point source pollution load:

[0080]

[0081] Among them, w i : Weight of the ith pollution source in the next N years; L i ​:Pollution load of the i-th type of pollution source today; R: Future rainfall; R0: Benchmark rainfall (obtained by calculating the historical mean); K: Soil erosion factor (related to soil type); K0: Benchmark soil erosion factor; w1, w2, w3: Weight coefficients, obtained by fitting historical data; k: Exponential coefficient, reflecting the non-linear impact of rainfall on pollution load;

[0082] S6: Iterative optimization and convergence judgment:

[0083] Calculate and optimize the weight w through the AI algorithm i and the exponential coefficient k to minimize the prediction error of the integration formula. When the prediction error is less than the preset threshold, the model converges;

[0084] The AI algorithm mentioned above is the particle swarm optimization algorithm: The weight w i to be optimized and the exponential coefficient k are encoded as particles of the AI algorithm. Randomly generate N particles, and each particle represents a set of parameter values. Iteratively update the particle velocity v i and position x i :

[0085]

[0086] where: ω is the inertia weight, c1, c2 are learning factors, r1, r2 are random numbers from 0 to 1, p best is the individual historical optimal position, g best is the global optimal position;

[0087] The parameter constraints of the particle swarm optimization algorithm include: w1 + w2 + w3 = 1 and k > 0;

[0088] The convergence condition mentioned above is:

[0089]

[0090] where, L t is the measured pollution load, is the model prediction value;

[0091] S7: Result calculation:

[0092] When the convergence condition is reached, the final calculation result L of the farmland non-point source pollution load can be calculated and output current .

[0093] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for calculating agricultural non-point source pollution load based on trend prediction, characterized in that: The following steps are involved: S1: Data collection and preprocessing: Collect historical farmland non-point source pollution data from official statistics, research reports, and field monitoring data, including: Pollution source data: Perfluorinated compound amount Q PFOS , Pesticide usage Q pesticide , Microplastic Content C microplastic ; Environmental data: farmland area A, rainfall R, soil type, temperature, etc.; Data cleaning: remove outliers and fill in missing values, and organize data by year; S2: Trend modeling and future prediction: Based on historical data, establish trend models for various pollution sources: Among them, Δ d : difference order; φ i : autoregression coefficient; θ j : Moving average coefficient; ∈ t : white noise error term, t: the tth year; Use the trained trend model to predict various pollution source data for the next N years: Among them, Y t-p is the pollution data for year tp; S3: Calculation of future pollution load: Calculate the pollution load of various pollution sources based on the predicted values ​​for the next N years; S4: Calculation of future pollution load weights: Construct the data matrix X = [x ij ] m×n ; Where m: number of samples; n: number of variables; i: sample index; j: variable index; x ij : The data value of the jth pollution source in the i-th sample; Calculate the weights of various pollution sources in the next N years: Where Ej: information entropy of j-type pollution; p ij :x ij Standardization; wj: weight of type j pollution; S5: Preliminary integrated formula for calculating non-point source pollution load: Establish a preliminary integrated formula for calculating current agricultural non-point source pollution load: Among them, w i : The weight of the i-th type of pollution source in the next N years; L i : The pollution load of the current i-th type of pollution source; R: future rainfall; R0: baseline rainfall; K: soil erosion factor; K0: baseline soil erosion factor; w1, w2, w3: weight coefficients, obtained by fitting historical data; k: exponential coefficient, reflecting the nonlinear effect of rainfall on pollution load; S6: Iterative optimization and convergence judgment: Calculate the optimization weight w through AI algorithm i and exponential coefficient k, which minimizes the prediction error of the integrated formula; S7: Result calculation: When the convergence condition is reached, the final calculation result of farmland non-point source pollution load L can be calculated and output. current .

2. The method for calculating agricultural non-point source pollution load based on trend prediction according to claim 1 is characterized in that The perfluorinated compounds described in S1 include perfluorinated and polyfluoroalkyl compounds; the pesticides include herbicides, insecticides, and fungicides.

3. The method for calculating agricultural non-point source pollution load based on trend prediction according to claim 1 is characterized in that The missing values ​​described in S1 were imputed using the mean imputation method: Among them, x missing : Missing values ​​of pollution sources.

4. The method for calculating agricultural non-point source pollution load based on trend prediction according to claim 1 is characterized in that The pollution loads of various pollution sources described in S3 include: PFC pollution load: L PFAS =Q PFAS ×α×R Among them, Q PFAS : amount of perfluorinated compounds; α: loss coefficient; R: rainfall erosion factor; Pesticide pollution load: L pesticide =Q pesticide ×β×K oc Among them, Q pesticide : Pesticide usage; β: Migration coefficient; K oc : soil adsorption coefficient; Microplastic pollution load: L microplastic =C microplastic ×A×γ Among them, C microplastic : microplastic content; A: farmland area; γ: surface runoff carrying rate.

5. The method for calculating agricultural non-point source pollution load based on trend prediction according to claim 1, characterized in that the AI ​​algorithm described in S6 is a particle swarm optimization algorithm: the optimization weight to be optimized w i The exponential coefficient k is encoded as a particle of the AI ​​algorithm, N particles are randomly generated, each particle represents a set of parameter values, and the particle speed v is iteratively updated i and position x i : in: ω is the inertia weight, c1, c2 are learning factors, r1, r2 are random numbers between 0 and 1, p best is the individual's best historical position, g best is the global optimal position; The parameter constraints of the particle swarm optimization algorithm include: w1+w2+w3=1 and k>0.

6. The method for calculating agricultural non-point source pollution load based on trend prediction according to claim 1 is characterized in that The convergence conditions stated in S7 are: in, L t To measure the pollution load, is the model's predicted value.

Images