Watershed pollution control measures optimization method employing response surface methodology coupled with genetic algorithm

Abstract

A watershed pollution control measures optimization method employing response surface methodology coupled with a genetic algorithm, the method relating to the cross-disciplinary field of environmental engineering, environmental system simulation and prediction technology, and computer technology. The present invention solves the problem of low optimization efficiency in existing optimization methods. The invention comprises: by using known data, establishing a geographic database for an area; by using the geographic database, calculating a single load of each pollution source, and accordingly calculating the load reduction amount of each pollution source in a single control measure scenario; calculating the total nitrogen pollution load reduction rate and the total phosphorus pollution load reduction rate, and computing the total economic benefit in the single control measure scenario; setting the maximization of the economic benefit, total nitrogen load reduction rate, and total phosphorus load reduction rate as objectives, and establishing a regression function by adopting response surface methodology; and by adopting an NSGA-II genetic algorithm, performing multi-objective optimization on the regression function, and selecting an optimal management measure for a watershed to undergo pollution control. The present invention is suitable for optimizing pollution control measures for a watershed.

Description

Optimization method of watershed pollution control measures based on response surface coupled genetic algorithm Technical Field

[0001] The present invention belongs to the intersecting field of environmental engineering, environmental system simulation and prediction technology and computer technology. Background Art

[0002] With the acceleration of industrialization and urbanization, excessive emissions of nutrients such as nitrogen and phosphorus have led to increasingly prominent watershed pollution problems. Accurately understanding the transformation patterns of nitrogen and phosphorus pollutants and the load contributions of different pollution sources is a prerequisite for reducing watershed pollution loads and formulating pollution strategies. Furthermore, the formulation of watershed pollution control policies requires a coordinated process of environmental and economic benefits. Single watershed pollution control projects often face problems such as unstable pollution reduction effects and high operating costs, and are prone to conflicts between primary and secondary priorities.

[0003] Accurately quantifying watershed pollution presents the following challenges: First, pollution sources are complex, encompassing non-point sources such as rural life, livestock and poultry farming, and agricultural production, as well as point sources such as urban life and industrial sources. Second, pollutants exhibit significant diversity in location, emission intensity, and emission forms, with complex migration and transformation patterns. Relying solely on empirical coefficients makes it difficult to accurately estimate the nitrogen and phosphorus loads of various pollution sources, especially non-point source pollution. Mechanistic models, such as the distributed watershed hydrological model SWAT, are needed to infer the forms of nitrogen and phosphorus loads and simulate the migration and transformation processes of pollutants. Mechanistic models often suffer from a high degree of uncertainty in simulating non-point source pollution, making the rationality of the results difficult to guarantee.

[0004] Using intelligent multi-objective optimization algorithms to explore best management practices (BMPs) that simultaneously achieve economic and environmental benefits is beneficial for improving pollution models' ability to cope with uncertainty and risk analysis. The NSGA-II genetic optimization algorithm utilizes a non-dominated sorting-based approach, incorporating the concept of congestion and an elitist strategy. This effectively reduces computational complexity, fully accounts for the spatial heterogeneity of each unit within a watershed, and obtains optimal parameters for measures. However, the optimization process often faces the following challenges: First, it overemphasizes the spatial heterogeneity of the watershed, resulting in trivial analysis results that render policy ineffective. Second, the mechanism model coupled to the multi-objective optimization algorithm requires running the mechanism model for each simulation, while conventional optimization algorithms require hundreds of simulations. Furthermore, running the mechanism model itself is time-consuming, significantly reducing optimization efficiency.

[0005] Summary of the Invention

[0006] The present invention aims to solve the problem of low optimization efficiency in existing optimization methods and provides a watershed pollution control measure optimization method based on a response surface coupled genetic algorithm.

[0007] The method for optimizing watershed pollution control measures based on a response surface coupled genetic algorithm of the present invention comprises:

[0008] Step 1: Establish a geographic database of the area to be polluted based on the elevation data, land use data, soil type data, meteorological data, soil property data, and pollution statistics of the area to be polluted;

[0009] Step 2: Calculate the single load of all pollution sources in the area using the geographic database;

[0010] Step 3: Using the single load, calculate the load reduction of all pollution sources under a single control measure scenario;

[0011] Step 4: Using the load reductions from all pollution sources, calculate the total nitrogen pollution load reduction rate and the total phosphorus pollution load reduction rate, and simultaneously calculate the total economic benefits under a single control measure scenario;

[0012] Step 5: maximizing economic benefits, maximizing total nitrogen load reduction rate, and maximizing total phosphorus load reduction rate as goals, using response surface methodology to establish regression functions between the goals and each control measure;

[0013] Step 6: Use the NSGA-II genetic algorithm to perform multi-objective optimization on the regression function to screen out the optimal management measures for the watershed to be polluted.

[0014] Furthermore, in the present invention, in step 2, all pollution sources include: urban domestic sewage pollution, industrial wastewater pollution, rural domestic sewage pollution, livestock and aquaculture wastewater pollution, agricultural fertilizer application pollution, atmospheric nitrogen deposition pollution and regional soil nutrient loss pollution.

[0015] Furthermore, in the present invention, the pollution load of urban domestic sewage is: P3 = E3bN3c (1-W t ) / 1000000

[0016] Where P3 is the pollution load of urban domestic sewage discharge in the watershed to be polluted, E3 is the per capita comprehensive domestic water consumption in the watershed to be polluted, b is the pollution reduction coefficient of the watershed to be polluted, N3 is the urban population in the watershed to be polluted, c is the pollutant concentration in the watershed to be polluted, and W t For the efficiency of urban domestic sewage treatment;

[0017] The pollution load of industrial wastewater is:

[0018] The amount of industrial wastewater and various pollutants generated is estimated by the output value of the secondary industry in the basin to be polluted, and the pollution load is calculated based on the amount of industrial wastewater and the amount of various pollutants generated.

[0019] Furthermore, in the present invention, the pollution load of rural domestic sewage is: P1=E1N1[(1-k)+k(1-W c )] / 1000

[0020] Where P1 is the pollution load of rural domestic sewage discharge in the watershed to be polluted, E1 is the pollution quota of rural domestic sewage in the watershed to be polluted, N1 is the number of rural population in the watershed to be polluted, k is the proportion of administrative villages that treat domestic sewage in the watershed to be polluted; W c Efficiency of rural domestic sewage treatment in watersheds to be polluted;

[0021] The pollution load of livestock and aquaculture wastewater is: P2=(∑E i M i +E s M s ) / 365

[0022] Where P2 is the pollution load of livestock and aquatic wastewater discharge in the watershed to be polluted, E i M is the pollution production quota of livestock and poultry i in the watershed to be polluted. i is the number of livestock and poultry i in the watershed to be polluted, E s The pollution quota for aquatic products in the watershed to be polluted is M s The amount of aquaculture in the watershed where pollution control is to be carried out;

[0023] The pollution load of agricultural fertilizer application is:

[0024] The amount of nitrogen and phosphorus fertilizers to be applied is determined through statistical yearbooks. Based on the actual agricultural practices in the basin, 60%, 30%, and 10% of the annual fertilizer application amount are set as base fertilizer, topdressing, and ear fertilizer, respectively. The base fertilizer is organic fertilizer, and the rest is inorganic fertilizer. The pollution load when all fertilizers are applied to the ground surface is obtained.

[0025] The pollution load of atmospheric nitrogen deposition is:

[0026] According to the provinces and areas of the sub-basins within the watershed to be polluted, the amount of atmospheric nitrogen deposition is determined by interpolation and extrapolation, and then converted into inorganic fertilizer to obtain the pollution load of atmospheric nitrogen deposition.

[0027] Furthermore, in the present invention, the regional soil nutrient loss pollution load is:

[0028] The SWAT model is used to simulate the pollution load of soil nutrient loss by taking the pollution load of urban domestic sewage, industrial wastewater, rural domestic sewage, livestock and aquaculture wastewater, agricultural fertilizer application pollution, atmospheric nitrogen deposition pollution and meteorological data as inputs.

[0029] Furthermore, in the present invention, the process of establishing the SWAT model is as follows:

[0030] Based on elevation data, the hydrological analysis module is called to extract river network data within the watershed to be polluted and to divide the watershed into sub-basins. Hydrological response units (HRUs) are generated based on land use type, soil type, and slope. The HRUs are used as the basic calculation units of the model, and the boundary condition of the model is: the data area threshold is 5%.

[0031] Furthermore, in the present invention, in step three, the management measures include: fertilizer reduction measures, no-till measures, contour planting measures, vegetation buffer zone measures, grass-planted waterway measures, terrace engineering measures and returning farmland to forest measures.

[0032] Furthermore, in the present invention, in step 3, the method for calculating the load reduction of different pollution sources under a single control measure scenario is:

[0033] Using the formula:

[0034] Calculate the pollution load reduction rate R under a single control measure scenario P , P pre is the average annual pollution load of the entire basin from a single pollution source, P aft It is the average annual pollution load of the entire river basin after the implementation of a single measure.

[0035] Furthermore, in the present invention, in step 4, the total economic benefit under a single control measure scenario is calculated:

[0036] Use the formula: B=YQ B -Q M

[0037] Calculate the total economic benefit B under a single control measure scenario, where Y is the additional economic benefit obtained by the beneficiary and the government under a single measure scenario, and Q is B is the construction cost under a single measure scenario, Q M It is the operation and maintenance cost under a single measure scenario.

[0038] Furthermore, in the present invention, in step six, the population size of the NSGA-II genetic algorithm is 100, the number of iterations is 300, the number of targets is 3, the number of variables is 3, and the parameter values ​​of the crossover probability is 0.9, the simulated binary crossover distribution parameter is 20, the mutation probability is 0.1, and the polynomial mutation distribution parameter is 20 are adopted.

[0039] The method described in this paper uses the SWAT model (Distributed Soil and Hydrological Simulation Software) to calculate and simulate nitrogen and phosphorus pollution loads in a watershed. Based on this, a BMP (Best Management Practice) strategy for watershed pollution control is proposed. The response surface methodology is used to fit the discrete simulation results of the hydrological model into a continuous objective function. The powerful search and optimization capabilities of the genetic algorithm are then leveraged to determine the optimal parameters for watershed control measures. This effectively improves optimization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] FIG1 is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only 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 making creative work are within the scope of protection of the present invention. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other in the absence of conflict.

[0042] Specific embodiment 1: This embodiment is described in detail with reference to FIG1 . The method for optimizing watershed pollution control measures based on a response surface coupled genetic algorithm described in this embodiment includes:

[0043] Step 1: Establish a geographic database of the area to be polluted based on the elevation data, land use data, soil type data, meteorological data, soil property data and pollution statistics of the area to be polluted;

[0044] Step 2: Calculate the single load of all pollution sources in the area using the geographic database;

[0045] Step 3: Using the single load, calculate the load reduction of all pollution sources under a single control measure scenario;

[0046] Step 4: Using the load reductions from all pollution sources, calculate the total nitrogen pollution load reduction rate and the total phosphorus pollution load reduction rate, and simultaneously calculate the total economic benefits under a single control measure scenario;

[0047] Step 5: maximizing economic benefits, maximizing total nitrogen load reduction rate, and maximizing total phosphorus load reduction rate as goals, using response surface methodology to establish regression functions between the goals and each control measure;

[0048] Step 6: Use the NSGA-II genetic algorithm to perform multi-objective optimization on the regression function to screen out the optimal management measures for the watershed to be polluted.

[0049] The present invention establishes the objective function through the response surface method and uses the genetic algorithm to perform multi-objective optimization to obtain the ideal factor level combination. First, based on the environmental and economic benefits of the several BMP measures used individually, three more prominent management measures are selected to participate in the construction of the response surface. The three-factor three-level Box-

[0050] BenhnkenDesign combined design experiment. Taking total nitrogen reduction rate, total phosphorus reduction rate and economic net benefit as responses, further design experiments and determine the response values ​​through simulation and calculation.

[0051] Furthermore, in this embodiment, in step 2, all pollution sources include: urban domestic sewage pollution, industrial wastewater pollution, rural domestic sewage pollution, livestock and aquaculture wastewater pollution, agricultural fertilizer application pollution, atmospheric nitrogen deposition pollution and regional soil nutrient loss pollution.

[0052] Furthermore, in this embodiment, the pollution load of rural domestic sewage is: P1=E1N1[(1-k)+k(1-W c )] / 1000

[0053] Where P1 is the pollution load of rural domestic sewage discharge in the watershed to be polluted, E1 is the pollution quota of rural domestic sewage in the watershed to be polluted, N1 is the number of rural population in the watershed to be polluted, k is the proportion of administrative villages that treat domestic sewage in the watershed to be polluted; W c Efficiency of rural domestic sewage treatment in watersheds to be polluted;

[0054] The pollution load of livestock and aquaculture wastewater is: P2=(∑E i M i +E s M s ) / 365

[0055] Where P2 is the pollution load of livestock and aquatic wastewater discharge in the watershed to be polluted, E i M is the pollution production quota of livestock and poultry i in the watershed to be polluted. i is the number of livestock and poultry i in the watershed to be polluted, E sThe pollution quota for aquatic products in the watershed to be polluted is M s The amount of aquaculture in the watershed where pollution control is to be carried out;

[0056] The pollution load of agricultural fertilizer application is:

[0057] The amount of nitrogen and phosphorus fertilizers to be applied is determined through statistical yearbooks. Based on the actual agricultural practices in the basin, 60%, 30%, and 10% of the annual fertilizer application amount are set as base fertilizer, topdressing, and ear fertilizer, respectively. The base fertilizer is organic fertilizer, and the rest is inorganic fertilizer. The pollution load when all fertilizers are applied to the ground surface is obtained.

[0058] The pollution load of atmospheric nitrogen deposition is:

[0059] According to the provinces and areas of the sub-basins within the watershed to be polluted, the amount of atmospheric nitrogen deposition is determined by interpolation and extrapolation, and then converted into inorganic fertilizer to obtain the pollution load of atmospheric nitrogen deposition.

[0060] Furthermore, in this embodiment, the regional soil nutrient loss pollution load is:

[0061] The SWAT model is used to simulate the pollution load of soil nutrient loss by taking the pollution load of urban domestic sewage, industrial wastewater, rural domestic sewage, livestock and aquaculture wastewater, agricultural fertilizer application pollution, atmospheric nitrogen deposition pollution and meteorological data as inputs.

[0062] In the present invention, the pollution discharge coefficients of livestock, poultry and aquatic products in each city are determined with reference to the "Handbook of Pollution Emission Calculation Methods and Coefficients for Emission Source Statistical Surveys" and the "National Large-Scale Livestock and Poultry Farming Pollution Survey and Prevention Countermeasures". After calculating the pollution load of rural domestic sewage and livestock, poultry and aquatic breeding wastewater according to the above method, it is converted into organic fertilizer and added to the surface soil of all farmland once a day using the continuous fertilization option in the SWAT management operation. The nitrogen and phosphorus conversion coefficients of common nitrogen fertilizers containing 46% nitrogen and phosphorus fertilizers containing 17% phosphorus are used, and the fertilizer database is edited accordingly, and four types of fertilizers, namely inorganic nitrogen fertilizer, inorganic phosphorus fertilizer, organic nitrogen fertilizer and organic phosphorus fertilizer, are set respectively so that they can be selected according to different needs.

[0063] Furthermore, in this embodiment, the process of establishing the SWAT model is as follows:

[0064] Based on elevation data, the hydrological analysis module is called to extract river network data within the watershed to be polluted and to divide the watershed into sub-basins. Hydrological response units (HRUs) are generated based on land use type, soil type, and slope. The HRUs are used as the basic calculation units of the model, and the boundary condition of the model is: the data area threshold is 5%.

[0065] In the present invention, the SWAT (Distributed Soil and Hydrological Simulation Software) model is used to create an HRU (small watershed unit), and meteorological data, point source emission input, and fertilizer data are written in sequence. The sub-basin database is edited, and the management input data (.mgt) is edited according to the agricultural management process. After the setting is completed, the database is rewritten. After the input is complete, the total running time of the SWAT model is set (the first three years are the warm-up period), the model is run, and the simulation results are saved after completion. After obtaining the simulation results, the model is calibrated and verified using the SWAT-CUP automatic calibration software. The evaluation of the calibration effect mainly examines the Nash efficiency coefficient (NSE) and the coefficient of determination (R2).

[0066] Furthermore, in this embodiment, in step three, the management measures include: fertilizer reduction measures, no-till measures, contour planting measures, vegetation buffer zone measures, grassed waterway measures, terrace engineering measures and returning farmland to forest measures.

[0067] Furthermore, in this embodiment, in step 3, the method for calculating the load reduction amount of different pollution sources under a single control measure scenario is as follows:

[0068] Using the formula:

[0069] Calculate the pollution load reduction rate R under a single control measure scenario P , P pre is the average annual pollution load of the entire basin from a single pollution source, P aft It is the average annual pollution load of the entire river basin after the implementation of a single measure.

[0070] Furthermore, in this embodiment, in step 4, the total economic benefits under a single control measure scenario are calculated:

[0071] Use the formula: B=YQ B -Q M

[0072] Calculate the total economic benefit B under a single control measure scenario, where Y is the additional economic benefit obtained by the beneficiary and the government under a single measure scenario, and Q is B is the construction cost under a single measure scenario, Q M It is the operation and maintenance cost under a single measure scenario.

[0073] Furthermore, in this embodiment, in step six, the population size of the NSGA-II genetic algorithm is 100, the number of iterations is 300, the number of targets is 3, the number of variables is 3, and the parameter values ​​of the crossover probability is 0.9, the simulated binary crossover distribution parameter is 20, the mutation probability is 0.1, and the polynomial mutation distribution parameter is 20 are adopted.

[0074] In this implementation, the value range of each variable of the NSGA-II genetic algorithm is the same as the range during the Expert Design parameter optimization, and the final Pareto optimal surface is obtained. The optimal management measures for the watershed are obtained through analysis.

[0075] The present invention obtains the rural domestic sewage production quota and rural domestic sewage treatment efficiency based on the pollution source census bulletin, obtains the rural population based on the statistical yearbook of the area to be polluted, and obtains the proportion of administrative villages based on literature data, and jointly calculates the rural domestic sewage load; obtains the livestock and aquatic product pollution quota based on the pollution source census bulletin, obtains the livestock and aquatic product breeding volume based on the statistical yearbook of the area to be polluted, and jointly calculates the livestock and aquatic breeding load; determines the nitrogen fertilizer and phosphate fertilizer application amount based on the statistical yearbook, and calculates the agricultural fertilizer load; determines the atmospheric nitrogen deposition amount in the area to be polluted during the simulation period based on literature research, and calculates the atmospheric nitrogen deposition load using interpolation and extrapolation methods; the above four types of non-point source loads are combined Converted into fertilizer, it was input into the farmland surface soil module of the SWAT model using the continuous fertilization option tool. The per capita comprehensive domestic water consumption and urban domestic sewage treatment efficiency were obtained based on the pollution source census bulletin. The pollution reduction coefficient and pollutant concentration were determined based on the environmental statistics bulletin. The urban population was obtained based on the statistical yearbook of the area to be polluted and controlled, and the urban domestic sewage load was calculated together. The amount of industrial wastewater and various pollutants generated was obtained based on the environmental statistics bulletin. The output value of the secondary industry was obtained based on the statistical yearbook of each province, and the industrial load was calculated using the output value deduction method. The point source emissions and meteorological data of the above two types of point source pollution were input into the SWAT model to simulate the soil nutrient loss load.

[0076] The core effects of the present invention are mainly reflected in the following four points:

[0077] This paper uses the SWAT model to simulate the load contribution and migration patterns of watershed pollution sources, and proposes calculation methods for point and non-point source pollution loads, including urban domestic sewage, industrial wastewater, rural domestic sewage, livestock and aquaculture wastewater, agricultural fertilizer application, and atmospheric nitrogen deposition, providing a solid data foundation for watershed pollution assessment and control.

[0078] The present invention is based on a method of initially enumerating scenarios and uniformly arranging simulations across the entire watershed. It uses the response surface methodology to obtain the objective function required by the optimization algorithm and couples it with the NSGA-II genetic algorithm for multi-objective optimization. This achieves continuity of the enumerated discrete parameters while significantly reducing the computational complexity of the optimization algorithm, making efficient parameter optimization a reality and the resulting watershed pollution control measures more practical.

[0079] The present invention is data-driven and suitable for nitrogen and phosphorus pollution control in different river basins. It has certain universality and transferability, and provides a universal solution for river basin pollution control. It can effectively promote the promotion of my country's river basin ecological environment protection work to the grassroots level and popularize it within the national river basin.

[0080] Specific embodiment:

[0081] This invention has been successfully applied in the analysis of nitrogen and phosphorus pollution in Basin A and the optimization of agricultural management measures. Basin A is the most important tributary of the Yangtze River. Its point source and non-point source pollution are complex. The main contribution to the basin pollution is agricultural non-point source pollution, and the pollutants are total nitrogen and total phosphorus. In order to effectively control agricultural non-point source pollution in Basin A, it is urgent to scientifically evaluate the current status of agricultural non-point source pollution in Basin A and the implementation effect of agricultural management measures in Basin A, so as to determine practical and feasible parameters for agricultural non-point source management measures and achieve synergy between environmental and economic benefits. The specific implementation process is as follows:

[0082] 1. Construction of geographic database;

[0083] The construction of the SWAT model primarily requires DEM digital elevation data, land use data, soil raster data, and meteorological data. The calculation of the model's pollution inputs requires statistical data on population, livestock production, and fertilizer application rates. Model calibration requires measured hydrological and water quality data within the watershed. The calculation of the economic benefits of agricultural management measures requires data on commodity prices and construction costs. Specific data sources and details are provided in Table 1.

[0084] The DEM data was obtained from the USGS website of the United States Geological Survey. The downloaded SRTM1Arc-Second Global dataset was extracted, projected, and preprocessed to obtain a usable watershed ADEM layer.

[0085] Land use data were obtained through the website of Resources and Environmental Science and Data Center. The downloaded 2020 1km land cover data were reclassified and other preprocessed to obtain the land use layer of Basin A.

[0086] Soil type data were obtained from the website of the National Tibetan Plateau Science Data Center. The China Soil Dataset (v1.2) based on the High-Wide Soil Database (HWSD) was downloaded. The preprocessing was basically the same as that of the land use layer, but there were slight differences in the reclassification. The soil type with the largest area in the same soil group was selected as the representative of the group to simplify the classification. Ultimately, the soil types were reclassified into 21 types.

[0087] Meteorological data were obtained from the National Oceanic and Atmospheric Administration (NOAA) website. A meteorological database was constructed based on the collected daily maximum and minimum temperatures and rainfall data from 18 weather stations in and around Basin A from 2013 to 2022. Data for missing dates, as well as wind speed, relative humidity, and solar radiation, were automatically generated using a weather generator based on the World Weather Database (CFSR_World). SWAT has strict requirements for the format of meteorological databases; all data must be complete for every day during the study period. The downloaded raw data were organized using Python, with unit conversion, data deduplication and correction, and missing values ​​imputed (using a placeholder of -99). The files were then split by station and data type. Finally, a weather station index file was created, indicating the station number and location information, to generate a spatial distribution map of each weather station.

[0088] Other statistical data used for the calculation, including data used for pollution input calculation, measured hydrological and water quality data, and data used for economic benefit calculation, were downloaded and obtained through the Statistical Yearbooks of Provinces M and N (urban and rural population, secondary industry output value, livestock and aquaculture output, and nitrogen and phosphorus fertilizer application in each county and city in the two provinces from 2013 to 2022), the Environmental Statistical Bulletins of Provinces M and N (industrial wastewater and various pollutants generated in the two provinces from 2013 to 2022), the National Environmental Monitoring Network (average monthly flow data of each hydrological station from 2020 to 2021, and monthly measured water quality data of each monitoring section from November 2019 to December 2020), market research, and the China Agricultural Product Price Survey Yearbook (average selling price of rice and wheat, and average selling price of various types of nitrogen and phosphorus fertilizers).

[0089] 2. Construction of watershed process model and calculation of nitrogen and phosphorus loads;

[0090] The SWAT model was established in Basin A, which was divided into 1552 sub-basins and 13035 HRUs. The results of calibrating 20 water quantity and water quality parameters in the SWAT model based on the measured values ​​of the monitoring stations in Basin A showed that the NSE of each parameter was greater than 0.7, and R 2 The results are all above 0.8, indicating a good calibration effect. Based on this, the parameters were modified and the model was re-run to verify the model. It was found that the verification results still met the NSE greater than 0.7 and R 2 When it is greater than 0.8, the overall fitting effect is good.

[0091] Table 1 Calibration and verification results of hydrological stations and monitoring sections in Basin A

[0092] At the same time, the correlation analysis between precipitation and runoff, nitrogen and phosphorus pollution loads in Basin A was conducted, and it was found that precipitation and runoff showed a strong positive correlation, and its R 2 reached 0.740; compared with the correlation between nitrogen load and precipitation (R2 =0.476), the correlation between phosphorus load and precipitation (R 2 =0.588), and the changes in precipitation and phosphorus load were almost synchronized, while the changes in nitrogen load lagged behind. This phenomenon was caused by the fact that the migration and transformation speed of phosphorus was slower than that of nitrogen and its migration process was more dependent on precipitation flushing.

[0093] 3. Optimal design of watershed pollution control measures based on response surface coupled genetic algorithm;

[0094] First, seven single agricultural management measures were established. Through SWAT simulation and benefit calculations, the pollution reduction rates and net benefits of each measure were determined. Engineered measures, such as vegetation buffers and terraces, achieved nitrogen and phosphorus reduction rates exceeding 15%, significantly outperforming non-engineering measures such as fertilizer reduction and no-tillage (which both achieved nitrogen and phosphorus reduction rates below 6%). The seven measures were ranked from highest to lowest in terms of reduction rate: returning farmland to forest > vegetation buffers > terraces > grassed waterways > fertilizer reduction > contour planting > no-tillage. The returning farmland to forest measure, which had the highest reduction rate, achieved a 46.71% reduction in total nitrogen and a 47.00% reduction in total phosphorus, while the no-till measure, which had the lowest reduction rate, only achieved a 2.42% reduction in total nitrogen and a 2.59% reduction in total phosphorus. Non-engineering measures, due to their lack of construction costs, generally have positive net benefits, such as fertilization, which has a net benefit of 513 million yuan per year. Engineered measures, due to their higher construction and operating costs, generally have lower net benefits, such as terraces, which have a net benefit of -1.273 billion yuan per year. From the perspective of comprehensive economic and environmental benefits, the comprehensive effect of engineering measures is still better. For example, the net benefit of vegetation buffer zones is -150 million yuan / a, slightly lower than non-engineering measures, but it can achieve a reduction rate of 23.74% in total nitrogen and 26.83% in total phosphorus, which is significantly higher than non-engineering measures.

[0095] Table 3 Economic and environmental benefits of agricultural management measures scenarios

[0096] The response surface optimization results showed that among the groups of measures involved in constructing the response surface, the one with the best comprehensive effect could achieve a reduction rate of 38.72% for total nitrogen and 38.10% for total phosphorus, with a net cost of -910 million yuan / a. Compared with a single measure, there was a significant improvement in the reduction rate, and the reduction in net benefit was still acceptable. The R values ​​of the regression equations for the reduction rates of total nitrogen, total phosphorus, and net benefit obtained were 2 The pre values ​​were all higher than 0.97, and the significance of each model was extremely high, indicating a good fitting effect. The results of variance analysis showed that the most significant interactive factors affecting the reduction rates of total nitrogen and total phosphorus were the width of the vegetation buffer zone and the width of the terraced field.

[0097] Table 4 Pareto optimal solution set of multi-objective optimization model

[0098] The results of the combined BMP optimization design based on response surface and NSGA-II show that the obtained Pareto optimal solution set contains a total of 30 solutions, which can basically achieve a reduction of nitrogen and phosphorus pollution load of more than 35% in the entire watershed, with a net benefit between -300 million yuan / a and -3 billion yuan / a. According to the reduction rate and net benefit, it can be divided into high-benefit zone, high-reduction zone and compromise zone; the best agricultural management measure is determined to be the middle solution 12 in the compromise zone, which can achieve a reduction of 40.04% in total nitrogen and 39.22% in total phosphorus, with a net cost of -1.134 billion yuan / a. Compared with the results of multiple groups and automatic optimization in the response surface experiment, there is a significant improvement in the reduction of nitrogen and phosphorus loads and net benefits in the watershed.

[0099] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely illustrative of the principles and applications of the invention. It should be understood that many modifications may be made to the illustrative embodiments, and that other arrangements may be devised, without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in ways other than those described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be used in conjunction with other described embodiments.

Claims

1. An optimization method for basin pollution control measures based on response surface coupled genetic algorithm, characterized in that, it includes: Step 1: Based on the elevation data, land use data, soil type data, meteorological data, soil property data and pollution statistics data of the area to be polluted controlled, establish a geographical database for this area; Step 2: Using the geographical database, calculate the single load of all pollution sources in this area; Step 3: Using the single load, calculate the load reduction amount of all pollution sources under the scenario of a single control measure; Step 4: Using the load reduction amounts of all pollution sources, calculate the total nitrogen pollution load reduction rate and the total phosphorus pollution load reduction rate, and at the same time calculate the total economic benefit under the scenario of a single control measure; Step 5: Simultaneously aiming at maximizing economic benefits, maximizing the total nitrogen load reduction rate and maximizing the total phosphorus load reduction rate, adopt the response surface method to establish a regression function between the above objectives and each control measure; Step 6: Adopt the NSGA-II genetic algorithm to perform multi-objective optimization on the regression function, and screen out the optimal management measures for the basin to be polluted controlled.

2. The optimization method for basin pollution control measures based on response surface coupled genetic algorithm according to claim 1, characterized in that, in Step 2, the above all pollution sources include: urban domestic sewage pollution, industrial wastewater pollution, rural domestic sewage pollution, livestock and poultry aquaculture wastewater pollution, agricultural chemical fertilizer application pollution, atmospheric nitrogen deposition pollution and regional soil nutrient loss pollution.

3. The optimization method for basin pollution control measures based on response surface coupled genetic algorithm according to claim 2, characterized in that, The pollution load of urban domestic sewage is: P 3 = E 3 bN 3 c(1 - W t ) / 1000000 Wherein, P 3 is the total nitrogen pollution load of domestic sewage discharge in urban areas of the basin to be subject to pollution control, E 3 is the per capita comprehensive domestic water consumption in the basin to be subject to pollution control, b is the pollution coefficient of the basin to be subject to pollution control, N 3 is the urban population in the basin to be subject to pollution control, c is the pollutant concentration in the basin to be subject to pollution control, W t is the treatment efficiency of domestic sewage in urban areas; the pollution load of industrial wastewater is: Estimate the industrial wastewater and the generation amounts of various pollutants through the output value of the secondary industry in the basin to be polluted controlled, and calculate the pollution load according to the amount of industrial wastewater and the generation amounts of various pollutants.

4. The optimization method for basin pollution control measures based on response surface coupled genetic algorithm according to claim 3, characterized in that, The pollution load of rural domestic sewage is: P 1 = E 1 N 1 [(1 - k) + k(1 - W c )] / 1000 where P 1 is the pollution load of rural domestic sewage discharge in the basin to be controlled for pollution, E 1 is the pollution generation quota of rural domestic sewage in the basin to be controlled for pollution, N 1 is the rural population in the basin to be controlled for pollution, k is the proportion of administrative villages in the basin to be controlled for pollution that treat domestic sewage; W c is the treatment efficiency of rural domestic sewage in the basin to be controlled for pollution rate; the pollution load of livestock and poultry aquaculture wastewater is: P 2 = (∑E i M i + E s M s ) / 365 Wherein, P 2 is the pollution load of livestock and poultry and aquaculture sewage discharge in the basin to be controlled for pollution, E i is the pollution generation quota when livestock and poultry i are scattered in the basin to be controlled for pollution, M i is the number of livestock and poultry i in the basin to be controlled for pollution, E s is the pollution generation quota of aquaculture in the basin to be controlled for pollution, M s is the aquaculture volume in the basin to be controlled for pollution; the pollution load of agricultural chemical fertilizer application is: Determine the application amounts of nitrogen fertilizer and phosphorus fertilizer through the statistical yearbook. According to the actual agricultural cultivation rules in the basin, set 60%, 30% and 10% of the annual application amounts as base fertilizer, top dressing and ear fertilizer respectively, where the base fertilizer is organic fertilizer and the rest are inorganic fertilizers, and obtain the pollution load when the fertilizers are evenly applied on the ground surface; the pollution load of atmospheric nitrogen deposition is: According to the province to which the sub-basin in the basin to be polluted controlled belongs and the area of the sub-basin, determine the atmospheric nitrogen deposition amount through interpolation and extrapolation methods, and then convert it into inorganic fertilizer to obtain the pollution load of atmospheric nitrogen deposition.

5. The optimization method for basin pollution control measures based on response surface coupled genetic algorithm according to claim 4, characterized in that, the pollution load of regional soil nutrient loss is: Adopt the SWAT model, take the pollution load of urban domestic sewage, the pollution load of industrial wastewater, the pollution load of rural domestic sewage, the pollution load of livestock and poultry aquaculture wastewater, the pollution load of agricultural chemical fertilizer application pollution, the pollution load of atmospheric nitrogen deposition pollution and meteorological data as the inputs of the model, and simulate the pollution load of soil nutrient loss pollution.

6. The optimization method for basin pollution control measures based on response surface coupled genetic algorithm according to claim 5, characterized in that, the establishment process of the SWAT model is as follows: Based on elevation data, the river network data within the basin to be controlled for pollution is extracted by invoking the hydrological analysis module for sub-basin division; hydrological response units (HRUs) are generated according to land use type, soil type and slope; the hydrological response units (HRUs) are used as the basic calculation units of the model, and the boundary condition of the model is: the data area threshold is 5%.

7. The optimization method for basin pollution control measures based on response surface coupled genetic algorithm according to claim 1 or 6, characterized in that, in step three, the management measures include: fertilizer application reduction measures, no-till measures, contour planting measures, vegetative buffer strip measures, grassed waterway measures, terrace engineering measures and returning farmland to forest measures.

8. The optimization method for basin pollution control measures based on response surface coupled genetic algorithm according to claim 7, characterized in that, in step three, the calculation method for calculating the load reduction amount of different pollution sources under a single control measure scenario is: Using the formula: Calculate the pollution load reduction rate R under the scenario of a single control measure P , P pre is the annual average pollution load of the entire basin of a single pollution source, and P aft is the annual average pollution load of the entire basin after implementing a single measure.

9. The optimization method for basin pollution control measures based on response surface coupled genetic algorithm according to claim 1 or 7, characterized in that, in step four, the total economic benefit under a single control measure scenario is accounted: using the formula: B = Y - Q B -Q M Calculate the total economic benefit B in the scenario of a single control measure. In the formula, Y is the additional economic benefit obtained by the beneficiaries and the government in the single measure scenario, and Q B is the construction cost in the single measure scenario, and Q M is the operation and maintenance cost in the single measure scenario.

10. The optimization method for basin pollution control measures based on response surface coupled genetic algorithm according to claim 1, characterized in that, in step six, the population size of the NSGA-II genetic algorithm is 100, the number of iterations is 300, the number of objectives is 3, the number of variables is 3, and the parameter values of the crossover probability of 0.9, the simulated binary crossover distribution parameter of 20, the mutation probability of 0.1 and the polynomial mutation distribution parameter of 20 are adopted.

Images