Water quality influence related factor analysis method based on multi-source data

Abstract

The invention relates to the technical field of resources and environments, in particular to a multi-source data-based water quality influence related factor analysis method, which comprises the following steps: A, constructing a clustering analysis algorithm, B, constructing a principal component analysis algorithm, C, constructing a multiple linear stepwise regression algorithm, and D, constructing a clustering-principal component-multiple linear stepwise regression real-time online tool. The method comprehensively considers the multi-source data which has the most direct influence on the water quality, analyzes the water quality influence factors from the data perspective, is small in operation resource occupation, easy in data acquisition, simple and convenient in data processing and high in operation speed, and can realize online real-time analysis; by means of the method, qualitative and quantitative analysis can be rapidly conducted on main driving factors and driving force of section water quality types and characteristics and section water quality standard exceeding conditions, meanwhile, the three statistical analysis methods are used in series, the input data accuracy of each analysis method can be effectively guaranteed, interference factors are eliminated, and the analysis result is more accurate and reliable.

Description

Technical Field

[0001] The present invention relates to the field of resource and environmental technology, and in particular to a method for analyzing water quality influencing related factors based on multi-source data. Background Art

[0002] The factors affecting the surface water environment quality are complex, involving many aspects including natural factors (such as rainfall, basin topography characteristics, basin soil characteristics) and human factors (such as domestic source emissions, industrial source emissions, and agricultural non-point source pollution).

[0003] At present, the methods for analyzing the factors affecting water quality can be divided into two categories: ① Mathematical models based on the surface water environment dynamics and the laws of material migration and transformation. Commonly used water environment models include pollution load models, receiving water quality models, comprehensive watershed models integrating pollution load and water quality, water quality models and comprehensive watershed models, etc. This type of method can accurately simulate the hydrodynamic parameters such as flow velocity and flow direction of unit water body, the pollutant concentration of unit water body, the change of pollutants along the way, and the impact of pollution source emissions on water quality. However, it often requires the input of a large amount of detailed basic data, including basin topography, soil, land use, population, water body boundaries, bottom elevation, water quality along the way, pollution source emissions along the way, etc. Since basic data are difficult to obtain , and its calculation process has high requirements on computers, which leads to the fact that the actual application effect of data prediction is often unsatisfactory; ② Numerical analysis method based on statistical methods. This type of method mainly analyzes the correlation between water quality data and its potential influencing factors through statistical analysis. Commonly used statistical methods include: principal component analysis, multivariate linear regression analysis, correlation analysis, etc. The advantage of this type of method is that it only looks for the relationship and internal logic between data from the data perspective, and the calculation is simple and fast; the disadvantage is that the single application of this type of method usually cannot well explain the causal relationship and internal connection between data, and needs to be tested in actual application. In order to solve the above technical problems, we designed a method for analyzing water quality influencing related factors based on multi-source data. Summary of the invention

[0004] The purpose of the present invention is to provide a method for analyzing water quality influencing related factors based on multi-source data, which has the advantages of small computing resource occupancy, easy data acquisition, simple data processing, fast running speed and the ability to realize online real-time analysis, and the analysis results are more accurate and reliable. It solves the problem that basic data acquisition is difficult, the calculation process has high requirements on computers, resulting in poor actual application effect of data prediction and poor accuracy of data analysis results.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for analyzing water quality related factors based on multi-source data, comprising the following steps:

[0006] A: Cluster analysis algorithm construction;

[0007] B: Construction of principal component analysis algorithm;

[0008] C: Construction of multivariate linear stepwise regression algorithm;

[0009] D: Construction of real-time online tool for clustering-principal component-multivariate linear stepwise regression.

[0010] Preferably, in step A, cluster analysis is a kind of unsupervised learning, which is often used to divide samples of unknown categories. Its principle is to divide these unknown samples into several clusters according to certain rules, cluster related samples in the same cluster, and divide dissimilar samples into different clusters, so as to analyze the intrinsic properties between samples and the laws of connection between each other. The partition-based method is the simplest and most widely used clustering method. It clusters objects by dividing them into mutually exclusive clusters, and each object belongs to and only belongs to one cluster; the partition result aims to make the similarity between clusters low and the similarity within clusters high. K-means clustering is the most widely used clustering method in the partition-based clustering algorithm. By calculating the distance between the sample point and the centroid of the cluster, the sample points close to the centroid of the cluster are divided into the same cluster. K-means measures the similarity between them by the distance between the samples. The farther the distance between the two samples, the lower the similarity, otherwise the higher the similarity. The K-means algorithm is as follows:

[0011] a. Select K initial centroids (K needs to be specified by the user). The initial centroids can be randomly selected, and each centroid is a class.

[0012] b. For each remaining sample point, calculate the Euclidean distance from each centroid, and assign it to the cluster with the centroid with the smallest distance between them, and calculate the centroid of each new cluster;

[0013] c. After all sample points are divided, the centroid position of each cluster is recalculated according to the division situation, and then the distance from each sample point to the centroid of each cluster is iteratively calculated to re-divide all sample points;

[0014] d. Repeat b and c until the centroid no longer changes or the maximum number of iterations is reached.

[0015] Preferably, in step B, the principal component analysis method is the most widely used data dimensionality reduction algorithm. The main idea of ​​PCA is to map n-dimensional features to k-dimensional features. This k-dimensional feature is a new orthogonal feature also called principal component. It is a k-dimensional feature reconstructed on the basis of the original n-dimensional feature. There are two main methods for implementing the principal component analysis algorithm:

[0016] (1) Implement the PCA algorithm based on the eigenvalue decomposition covariance matrix:

[0017] ① De-averaging (i.e. decentralization), that is, subtracting the mean value of each feature from its own.

[0018] ② Calculate the covariance matrix. Note: Whether or not to divide the sample size n or n-1 has no effect on the obtained eigenvector.

[0019] ③ Use the eigenvalue decomposition method to find the eigenvalues ​​and eigenvectors of the covariance matrix.

[0020] ④ Sort the eigenvalues ​​from large to small, select the largest k among them, and then use their corresponding k eigenvectors as row vectors to form the eigenvector matrix P.

[0021] ⑤ Convert the data into a new space constructed by k eigenvectors, that is, Y=PX.

[0022] (2) Implement the PCA algorithm based on the singular value decomposition (SVD) covariance matrix:

[0023] ① Remove the mean value, that is, subtract the mean value of each feature.

[0024] ②Calculate the covariance matrix.

[0025] ③ Calculate the eigenvalues ​​and eigenvectors of the covariance matrix through SVD.

[0026] ④ Sort the eigenvalues ​​from large to small, select the largest k among them, and then use their corresponding k eigenvectors as column vectors to form an eigenvector matrix.

[0027] ⑤Convert the data into a new space constructed by k feature vectors.

[0028] Preferably, in step C, multivariate linear stepwise regression is a method based on linear regression, which mainly solves the problem of multivariate collinearity, that is, the relationship is not linearly independent. It is a regression method for feature extraction based on variable explanatory power. The idea is to introduce variables one by one, and after introducing a new variable, test the old variables that have been selected for the regression model one by one, and delete the variables that are considered meaningless, until no new variables are introduced and no old variables are deleted, so as to ensure that every variable in the regression model is meaningful.

[0029] When a variable is introduced, we first check whether this variable causes significant changes in the model (F test). If there is a significant change, a t test is performed on all variables. When the originally introduced variable no longer changes significantly due to the introduction of the later added variable, this variable is eliminated to ensure that the regression equation only contains significant variables before each introduction of a new variable, until no significant explanatory variables are selected into the regression equation and no insignificant explanatory variables are eliminated from the regression equation, and finally an optimal set of variables is obtained. This method adopts the multivariate linear stepwise regression analysis method for the characteristics of each pollution type and seasonal changes in the section, and constructs a correlation model between the main pollutants in the section and rainfall, flow, temperature, and emissions from upstream sewage treatment plants, industrial enterprises and other pollution sources, in order to quantitatively analyze the main driving force (or contribution) of section pollution.

[0030] Preferably, in step D, a cluster analysis tool is first used to cluster the cross-section water quality based on the time scale to identify the main types and pollution characteristics of water pollution in each river section. The input data is the long-term series water quality data of a single section. It is assumed that five water quality indicators, namely TP, TN, DO, NH3-N and CODMn, are considered in particular, and there are a total of n time series data.

[0031] After cluster analysis, k groups of classified data can be obtained. Assume that there are x time series data in the (1)th pollution type, y time series data in the (2)th pollution type, and z time series data in the (k)th pollution type.

[0032] Pollution type characteristic analysis: characteristic analysis of the output water pollution types of various sections, including ① identifying the main water quality characteristics and main pollution factors of each pollution type according to the overall distribution of factors in each pollution type; ② the temporal changes and temporal distribution characteristics of pollution types; ③ the high incidence seasons of various pollution types, and their temporal correlations with the dry season / flood season / normal water season; ④ the occurrence frequency of various pollution types, and determining the dominant pollution type of the section (the pollution type with the highest occurrence frequency).

[0033] For the main water pollution types identified by cluster analysis, the principal component analysis method was used to analyze their driving factors, and qualitatively analyze the most important influencing factors for this type of water pollution. It was assumed that the driving factors corresponding to type (1) pollution were mainly rainfall, flow, ammonia nitrogen and total nitrogen concentrations of upstream pollution source A, and ammonia nitrogen and total nitrogen concentrations of upstream pollution source B (the specific driving factors of each section were optimized based on actual conditions).

[0034] After using cluster analysis and principal component analysis to analyze the main pollution types of the section, the main pollution factors of each type and their main driving factors, the multivariate linear regression analysis tool was further used to analyze the driving factors and driving forces behind each main pollution factor. By constructing a correlation model between the main pollution factors and the main component driving factors such as rainfall, flow, temperature, and emissions from water-related pollution sources identified by principal component analysis, it is used to qualitatively and quantitatively analyze the main driving forces (or contributions) that lead to section pollution.

[0035] The output data can be used to obtain the correlation coefficient between each variable and the dependent variable, thereby qualitatively analyzing the driving force of each driving factor.

[0036] Analysis of main driving factors and driving forces: Based on the correlation coefficient of each driving factor, analyze ① the positive and negative correlation between the driving factor and the main pollutants in the section; ② quantitatively analyze the driving force of each driving factor on the main pollutants in the section to determine the strong correlation factors of the main pollutants in each pollution type.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The clustering-principal component-multivariate linear stepwise regression tool constructed by the present invention comprehensively considers the multi-source data that has the most direct impact on water quality, including real-time data such as meteorology, hydrology, and large-scale online monitoring of pollution sources, and analyzes the factors affecting water quality from a data perspective. It occupies small computing resources, is easy to acquire data, is simple to process data, has a fast running speed, and can realize online real-time analysis. It can quickly conduct qualitative and quantitative analysis of the cross-section water quality type and its characteristics, and the main driving factors and driving forces of the cross-section water quality exceeding the standard. At the same time, the series use of the three statistical analysis methods can effectively ensure the accuracy of the input data of each analysis method, eliminate interference factors, and make the analysis results more accurate and reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the framework of the multivariate linear regression analysis algorithm in the present invention;

[0040] Figure 2 This is a schematic diagram of a single set of input data in the present invention;

[0041] Figure 3 A schematic diagram of multiple groups of classified output data in the present invention;

[0042] Figure 4 Schematic diagram of input data for principal component analysis in the present invention;

[0043] Figure 5 This is a schematic diagram of the PCA dimension reduction result data output in the present invention;

[0044] Figure 6It is a data schematic diagram of the pollution factors and driving factors in the present invention. DETAILED DESCRIPTION

[0045] See also Figure 1-Figure 6 , a method for analyzing water quality related factors based on multi-source data, comprising the following steps:

[0046] A: Cluster analysis algorithm construction;

[0047] B: Construction of principal component analysis algorithm;

[0048] C: Construction of multivariate linear stepwise regression algorithm;

[0049] D: Construction of real-time online tool for clustering-principal component-multivariate linear stepwise regression.

[0050] In step A, cluster analysis is a kind of unsupervised learning, which is often used to divide samples of unknown categories. Its principle is to divide these unknown samples into several clusters according to certain rules, cluster related samples in the same cluster, and divide dissimilar samples into different clusters, so as to analyze the intrinsic properties and the relationship between samples. The partition-based method is the simplest and most widely used clustering method. It clusters objects by dividing them into mutually exclusive clusters. Each object belongs to and only belongs to one cluster; the partition result aims to make the similarity between clusters low and the similarity within clusters high. K-means clustering is the most widely used clustering method among the partition-based clustering algorithms. By calculating the distance between the sample point and the cluster centroid, the sample points close to the cluster centroid are divided into the same cluster. K-means measures the similarity between them by the distance between the samples. The farther the distance between the two samples, the lower the similarity, otherwise the higher the similarity. The K-means algorithm is as follows:

[0051] a. Select K initial centroids (K needs to be specified by the user). The initial centroids can be randomly selected, and each centroid is a class.

[0052] b. For each remaining sample point, calculate the Euclidean distance from each centroid, and assign it to the cluster with the centroid with the smallest distance between them, and calculate the centroid of each new cluster;

[0053] c. After all sample points are divided, the centroid position of each cluster is recalculated according to the division situation, and then the distance from each sample point to the centroid of each cluster is iteratively calculated to re-divide all sample points;

[0054] d. Repeat b and c until the centroid no longer changes or the maximum number of iterations is reached.

[0055] In step B, the principal component analysis method is the most widely used data dimensionality reduction algorithm. The main idea of ​​PCA is to map n-dimensional features to k-dimensional features. This k-dimensional feature is a new orthogonal feature also called the principal component. It is a k-dimensional feature reconstructed on the basis of the original n-dimensional feature. There are two main methods to implement the principal component analysis algorithm:

[0056] (1) Implement the PCA algorithm based on the eigenvalue decomposition covariance matrix:

[0057] Input: Dataset X = {x1, x2, …, xn}, which needs to be reduced to k dimensions.

[0058] ① De-averaging (i.e. decentralization), that is, subtracting the mean value of each feature from its own.

[0059] ② Calculate the covariance matrix. Note: Whether or not to divide the sample size n or n-1 has no effect on the obtained eigenvector.

[0060] ③ Use the eigenvalue decomposition method to find the eigenvalues ​​and eigenvectors of the covariance matrix.

[0061] ④ Sort the eigenvalues ​​from large to small, select the largest k among them, and then use their corresponding k eigenvectors as row vectors to form the eigenvector matrix P.

[0062] ⑤ Convert the data into a new space constructed by k eigenvectors, that is, Y=PX.

[0063] (2) Implement the PCA algorithm based on the singular value decomposition (SVD) covariance matrix:

[0064] Input: Dataset X = {x1, x2, …, xn}, which needs to be reduced to k dimensions.

[0065] ① Remove the mean value, that is, subtract the mean value of each feature.

[0066] ②Calculate the covariance matrix.

[0067] ③ Calculate the eigenvalues ​​and eigenvectors of the covariance matrix through SVD.

[0068] ④ Sort the eigenvalues ​​from large to small, select the largest k among them, and then use their corresponding k eigenvectors as column vectors to form an eigenvector matrix.

[0069] ⑤Convert the data into a new space constructed by k feature vectors.

[0070] In step C, multivariate linear stepwise regression is a method based on linear regression. It mainly solves the problem of multivariate collinearity, that is, the relationship is not linearly independent. It is a regression method based on variable explanatory power to extract features. The idea is to introduce variables one by one, and after introducing a new variable, test the old variables that have been selected for the regression model one by one, and delete the variables that are considered meaningless until no new variables are introduced and no old variables are deleted, so as to ensure that every variable in the regression model is meaningful;

[0071] When a variable is introduced, we first check whether this variable causes significant changes in the model (F test). If there is a significant change, a t test is performed on all variables. When the originally introduced variable no longer changes significantly due to the introduction of the later added variable, this variable is eliminated to ensure that the regression equation only contains significant variables before each introduction of a new variable, until no significant explanatory variables are selected into the regression equation and no insignificant explanatory variables are eliminated from the regression equation, and finally an optimal set of variables is obtained. This method adopts the multivariate linear stepwise regression analysis method for the characteristics of each pollution type and seasonal changes in the section, and constructs a correlation model between the main pollutants in the section and rainfall, flow, temperature, and emissions from upstream sewage treatment plants, industrial enterprises and other pollution sources, in order to quantitatively analyze the main driving force (or contribution) of section pollution.

[0072] In step C, cluster analysis tools are first used to cluster the water quality of sections based on the time scale to identify the main types and pollution characteristics of water pollution in each river basin section. The input data is the long-term water quality data of a single section. It is assumed that five water quality indicators, including TP, TN, DO, NH3-N and CODMn, are considered, and there are n time series data in total, such as Figure 2 As shown:

[0073] After cluster analysis, k groups of classified data can be obtained. Assume that there are x time series data in the (1)th pollution type, y time series data in the (2)th pollution type, and z time series data in the (k)th pollution type. For example, Figure 3 As shown:

[0074] Pollution type characteristic analysis: characteristic analysis of the output water pollution types of various sections, including ① identifying the main water quality characteristics and main pollution factors of each pollution type according to the overall distribution of factors in each pollution type; ② the temporal changes and temporal distribution characteristics of pollution types; ③ the high incidence seasons of various pollution types, and their temporal correlations with the dry season / flood season / normal water season; ④ the occurrence frequency of various pollution types, and determining the dominant pollution type of the section (the pollution type with the highest occurrence frequency).

[0075] For the main water pollution types identified by cluster analysis, the principal component analysis method is used to analyze their driving factors, and qualitatively analyze the most important influencing factors of this type of water pollution. It is assumed that the driving factors corresponding to type (1) pollution are mainly rainfall, flow, ammonia nitrogen and total nitrogen concentrations of upstream pollution source A, and ammonia nitrogen and total nitrogen concentrations of upstream pollution source B (the specific driving factors of each section are optimized according to the actual situation), such as Figure 4 As shown:

[0076] Where Q is flow, Pre is rainfall, ANH3-N is NH3-N emission from pollution source A, ATN is TN emission from pollution source A, BNH3-N is NH3-N emission from pollution source B, BTN is TN emission from pollution source B. The output data is the main driving factors causing NH3-N pollution, the main pollutant component, in the first category of water pollution, such as Figure 5 As shown:

[0077] That is, the main component driving factors of the (1) type of water pollution are rainfall, flow, NH3-N emissions from pollution source A, and NH3-N emissions from pollution source B. Similarly, the main component driving factors of the (k) type of pollution can be obtained.

[0078] After using cluster analysis and principal component analysis to analyze the main pollution types of the section, the main pollution factors of each type and their main driving factors, the multivariate linear regression analysis tool was further used to analyze the driving factors and driving forces behind each main pollution factor. By constructing a correlation model between the main pollution factors and the main component driving factors such as rainfall, flow, temperature, and emissions from water-related pollution sources identified by principal component analysis, it is used to qualitatively and quantitatively analyze the main driving forces (or contributions) that lead to section pollution.

[0079] If the main pollutant of type (1) is NH3-N, and the main component driving factors are rainfall, flow, NH3-N emissions from pollution source A, and NH3-N emissions from pollution source B, then NH3-N is taken as the dependent variable, and other driving data are taken as independent variables, such as Figure 6 As shown:

[0080] The output data can be used to obtain the correlation coefficient between each variable and the dependent variable, so as to qualitatively analyze the driving force of each driving factor. For example, the output correlation coefficient can be expressed as:

[0081]

[0082] Analysis of main driving factors and driving forces: Based on the correlation coefficient of each driving factor, analyze ① the positive and negative correlation between the driving factor and the main pollutants in the section; ② quantitatively analyze the driving force of each driving factor on the main pollutants in the section to determine the strong correlation factors of the main pollutants in each pollution type.

[0083] As can be concluded from the above table: in the first type of pollution, the driving forces corresponding to the main pollutant NH3-N are rainfall, flow, NH3-N emissions from pollution source A and ammonia nitrogen emissions from pollution source B, from strong to weak, and they are all positively correlated. Rainfall is the main driving factor of NH3-N pollution.

[0084] Similarly, the main pollutants, main component driving factors and driving forces of the (k)th pollution type can be obtained.

[0085] In summary: this method for analyzing water quality-related factors based on multi-source data, through the coordinated use of the above steps, solves the problems of difficulty in obtaining basic data and high requirements for computers in the calculation process, resulting in poor actual application effects of data prediction and poor accuracy of data analysis results.

Claims

1. A method for analyzing water quality related factors based on multi-source data, comprising the following steps: A: Cluster analysis algorithm construction; B: Construction of principal component analysis algorithm; C: Construction of multivariate linear stepwise regression algorithm; D: Construction of real-time online tool for clustering-principal component-multivariate linear stepwise regression.

2. The method for analyzing water quality related factors based on multi-source data according to claim 1, characterized in that: In step A, cluster analysis is a kind of unsupervised learning, which is often used to divide samples of unknown categories. Its principle is to divide these unknown samples into several clusters according to certain rules, cluster related samples in the same cluster, and divide dissimilar samples into different clusters, so as to analyze the intrinsic properties between samples and the laws of their connection. The partition-based method is the simplest and most widely used clustering method. It clusters objects by dividing them into mutually exclusive clusters, and each object belongs to and only belongs to one cluster; the partition result aims to make the similarity between clusters low and the similarity within clusters high. K-means clustering is the most widely used clustering method among the partition-based clustering algorithms. By calculating the distance between the sample point and the cluster centroid, the sample points close to the cluster centroid are divided into the same cluster. K-means measures the similarity between them by the distance between the samples. The farther the distance between the two samples, the lower the similarity, otherwise the higher the similarity. The K-means algorithm is as follows: a. Select K initial centroids (K needs to be specified by the user). The initial centroids can be randomly selected, and each centroid is a class. b. For each remaining sample point, calculate the Euclidean distance from each centroid, and assign it to the cluster with the centroid with the smallest distance between them, and calculate the centroid of each new cluster; c. After all sample points are divided, the centroid position of each cluster is recalculated according to the division situation, and then the distance from each sample point to the centroid of each cluster is iteratively calculated to re-divide all sample points; d. Repeat b and c until the centroid no longer changes or the maximum number of iterations is reached.

3. The method for analyzing water quality related factors based on multi-source data according to claim 1 is characterized by: In step B, the principal component analysis method is the most widely used data dimensionality reduction algorithm. The main idea of ​​PCA is to map n-dimensional features to k-dimensional features. This k-dimensional feature is a new orthogonal feature also called the principal component. It is a k-dimensional feature reconstructed on the basis of the original n-dimensional feature. There are two main methods to implement the principal component analysis algorithm: (1) Implement the PCA algorithm based on the eigenvalue decomposition covariance matrix: ① De-averaging (i.e. decentralization), that is, subtracting the mean value of each feature from its own. ② Calculate the covariance matrix. Note: Whether or not to divide the sample size n or n-1 has no effect on the obtained eigenvector. ③ Use the eigenvalue decomposition method to find the eigenvalues ​​and eigenvectors of the covariance matrix. ④ Sort the eigenvalues ​​from large to small, select the largest k among them, and then use their corresponding k eigenvectors as row vectors to form the eigenvector matrix P. ⑤ Convert the data into a new space constructed by k eigenvectors, that is, Y=PX. (2) Implement the PCA algorithm based on the singular value decomposition (SVD) covariance matrix: ① Remove the mean value, that is, subtract the mean value of each feature. ②Calculate the covariance matrix. ③ Calculate the eigenvalues ​​and eigenvectors of the covariance matrix through SVD. ④ Sort the eigenvalues ​​from large to small, select the largest k among them, and then use their corresponding k eigenvectors as column vectors to form an eigenvector matrix. ⑤Convert the data into a new space constructed by k feature vectors.

4. The method for analyzing water quality related factors based on multi-source data according to claim 1 is characterized by: In step C, multivariate linear stepwise regression is a method based on linear regression, which mainly solves the problem of multivariate collinearity, that is, the relationship is not linearly independent. It is a regression method for feature extraction based on variable explanatory power. The idea is to introduce variables one by one, and after introducing a new variable, test the old variables that have been selected for the regression model one by one, and delete the variables that are considered meaningless, until no new variables are introduced and no old variables are deleted, so as to ensure that every variable in the regression model is meaningful. When a variable is introduced, we first check whether this variable causes significant changes in the model (F test). If there is a significant change, a t test is performed on all variables. When the originally introduced variable no longer changes significantly due to the introduction of the later added variable, this variable is eliminated to ensure that the regression equation only contains significant variables before each introduction of a new variable, until no significant explanatory variables are selected into the regression equation and no insignificant explanatory variables are eliminated from the regression equation, and finally an optimal set of variables is obtained. This method adopts the multivariate linear stepwise regression analysis method for the characteristics of each pollution type and seasonal changes in the section, and constructs a correlation model between the main pollutants in the section and rainfall, flow, temperature, and emissions from upstream sewage treatment plants, industrial enterprises and other pollution sources, in order to quantitatively analyze the main driving force (or contribution) of section pollution.

5. The method for analyzing water quality related factors based on multi-source data according to claim 1, characterized in that: In step D, a cluster analysis tool is first used to cluster the cross-section water quality based on the time scale to identify the main types and pollution characteristics of water pollution in each river basin section. The input data is the long-term series water quality data of a single section. It is assumed that five water quality indicators, namely TP, TN, DO, NH3-N and CODMn, are considered in particular, and there are a total of n time series data. After cluster analysis, k groups of classified data can be obtained. Assume that there are x time series data in the (1)th pollution type, y time series data in the (2)th pollution type, and z time series data in the (k)th pollution type. Pollution type characteristic analysis: characteristic analysis of the output water pollution types of various sections, including ① identifying the main water quality characteristics and main pollution factors of each pollution type according to the overall distribution of factors in each pollution type; ② the temporal changes and temporal distribution characteristics of pollution types; ③ the high incidence seasons of various pollution types, and their temporal correlations with the dry season / flood season / normal water season; ④ the occurrence frequency of various pollution types, and determining the dominant pollution type of the section (the pollution type with the highest occurrence frequency). For the main water pollution types identified by cluster analysis, the principal component analysis method was used to analyze their driving factors, and qualitatively analyze the most important influencing factors for this type of water pollution. It was assumed that the driving factors corresponding to type (1) pollution were mainly rainfall, flow, ammonia nitrogen and total nitrogen concentrations of upstream pollution source A, and ammonia nitrogen and total nitrogen concentrations of upstream pollution source B (the specific driving factors of each section were optimized based on actual conditions). After using cluster analysis and principal component analysis to analyze the main pollution types of the section, the main pollution factors of each type and their main driving factors, the multivariate linear regression analysis tool was further used to analyze the driving factors and driving forces behind each main pollution factor. By constructing a correlation model between the main pollution factors and the main component driving factors such as rainfall, flow, temperature, and emissions from water-related pollution sources identified by principal component analysis, it is used to qualitatively and quantitatively analyze the main driving forces (or contributions) that lead to section pollution. The output data can be used to obtain the correlation coefficient between each variable and the dependent variable, thereby qualitatively analyzing the driving force of each driving factor. Analysis of main driving factors and driving forces: Based on the correlation coefficient of each driving factor, analyze ① the positive and negative correlation between the driving factor and the main pollutants in the section; ② quantitatively analyze the driving force of each driving factor on the main pollutants in the section to determine the strong correlation factors of the main pollutants in each pollution type.

Images