Water pollution data monitoring method and system based on multi-target analysis
By employing multi-objective analysis methods and non-negative matrix factorization algorithms, combined with the opening and closing status of dams and hydrological characteristics, the problem of pollution source analysis in complex dam-controlled rivers was solved, enabling accurate identification and management support of pollution sources.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 江苏省南京环境监测中心
- Filing Date
- 2026-05-27
- Publication Date
- 2026-06-26
AI Technical Summary
Existing pollution source apportionment technologies struggle to accurately identify pollution sources and their emission characteristics in complex, sluice-controlled rivers. Traditional methods fail under conditions of significant temporal and spatial variations and hydrodynamic instability, making it difficult to implement pipeline renovation and non-point source pollution control measures.
A multi-objective analysis method was adopted, which collected GIS maps, rainfall time series, flow time series and water quality monitoring time series, combined with the opening and closing status of dams to perform hydrological segmentation, and constructed a two-dimensional water quality observation matrix. The source analysis was performed using a non-negative matrix factorization algorithm with ratio constraints, and impulse constraints and delay constraints were introduced. The emission attributes of pollution sources were identified by cross-validation of chemical characteristics and hydrological response characteristics.
It enables precise identification of pollution sources in complex sluice-controlled rivers, improves the accuracy and precision of pollution source analysis, and supports differentiated pipeline repair and non-point source pollution control decisions.
Smart Images
Figure CN122286202A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of water pollution source tracing technology, specifically to a water pollution data monitoring method and system based on multi-objective analysis. Background Technology
[0002] In complex sluice-controlled rivers, the frequent opening and closing of sluice gates and dams causes dynamic changes in water flow, making it highly susceptible to pollutant accumulation during periods of stagnant flow and concentrated pollutant discharge during periods of free flow. This creates unique and complex hydrodynamic and water quality evolution patterns. These watersheds typically traverse urban-rural complexes, where urban sewage discharge and agricultural runoff are intertwined, resulting in extremely complex pollution causes. To achieve effective governance, accurately identifying pollution sources and their emission characteristics has become the primary task in overcoming the bottleneck of pollution reduction.
[0003] Existing pollution source apportionment technologies face numerous limitations in such complex scenarios. Traditional source inventory methods struggle to address emission characteristics with significant spatiotemporal variations, diffusion models are limited by their inability to accurately characterize the unsteady hydrodynamic conditions under dam control, and receptor models suffer from fundamental assumptions that are difficult to satisfy in the frequent water exchange of dam-controlled rivers. More critically, current management practices often rely on single water quality concentration thresholds for pollution identification, which is highly susceptible to failure in complex dam-controlled rivers. For example, during and after rainfall events, concentration thresholds alone are insufficient to distinguish between "pulse pollution from combined sewer overflows during rainfall," "delayed discharge from culverts after rainfall," and "non-point source pollution from agricultural runoff." These three types of pollution all exhibit exceedances in total phosphorus and ammonia nitrogen levels, and often occur in conjunction with hydrological disturbances, making it impossible to accurately pinpoint whether the problem stems from mixed pipe connections, culvert accumulation, or agricultural non-point source pollution based solely on concentration data. This ambiguity in source apportionment hinders the implementation of targeted pipe network upgrades and non-point source pollution control measures, severely restricting the systematic improvement of the regional water environment. Summary of the Invention
[0004] The purpose of this invention is to provide a water pollution data monitoring method and system based on multi-objective analysis to solve the problems mentioned in the background art.
[0005] To address the aforementioned technical problems, this invention provides a water pollution data monitoring method based on multi-objective analysis, comprising:
[0006] S100. Collect GIS maps of the target watershed, rainfall time series, flow time series of monitoring sections, and water quality monitoring time series. Among them, the water quality monitoring time series should include at least ammonia nitrogen concentration time series and total phosphorus concentration time series.
[0007] GIS maps are used to describe the spatial topology of river networks, dam locations, culvert sections, and drainage outlets within a target watershed.
[0008] Rainfall time series refers to the cumulative precipitation data collected by rain gauges within a target watershed at a set time step.
[0009] Flow time series refers to the flow rate data collected by the flow meter at the monitoring section at a set time step.
[0010] Water quality monitoring time series refers to the ammonia nitrogen concentration and total phosphorus concentration values collected by automatic water quality monitoring stations at a set time step.
[0011] The monitoring sections include the main trunk section set up on the main stream, the tributary section set up at the confluence of tributaries, and the upstream and downstream sections set up before and after the dam.
[0012] S200. Based on the opening and closing status of dams and sluices in the target watershed, the flow time series is hydrologically segmented to calculate the surface runoff component time series and the baseflow component time series. Specifically, this includes:
[0013] S201. Analyze the opening and closing status of sluice gates and dams within the target watershed. When the sluice gates and dams are fully closed, the current period is defined as the sluice gate-controlled flow retention period. When the sluice gates and dams are open, the current period is defined as the free-flow period.
[0014] S202, Adopt Digital filtering is used to segment the flow time series. During the gated flow stagnation period, filtering parameters are set. Substitute the first preset value into the formula to calculate the time series of the base current component.
[0015] During periods of smooth flow, set the filter parameters. Substitute the second preset value into the formula to calculate the time series of the base current component; the formula is:
[0016] ;
[0017] In the formula, for The base current component value at time; For the previous moment The base current component value; for Actual measured total flow rate at any given time; It is the baseflow index; it reflects the long-term proportion of baseflow in total flow.
[0018] The filtering parameters are used to control the smoothness of the changes in the base current between adjacent time steps.
[0019] During the sluice gate closure period, the closure of the gate and dam results in poor water flow and a slow decay of the baseflow recession curve. Therefore, filter parameters are set. A smaller first preset value is used to match the slow decay rate.
[0020] During the free-flow period, the water flow is strong when the dam is open, and the baseflow response is relatively fast. Therefore, filter parameters are set accordingly. A larger second preset value is used to match a faster decay rate.
[0021] use Digital filtering uses a recursive algorithm to smoothly separate the stable baseflow component, which is affected by groundwater recharge and soil infiltration, from the mixed total flow time series, thereby separating the rainfall-driven surface runoff component.
[0022] S203. Calculate the time series of surface runoff components, using the following formula: ;in, for Surface runoff component value at time.
[0023] S300. Construct a two-dimensional water quality observation matrix containing ammonia nitrogen and total phosphorus, and set impulse constraints and delay constraints by combining the time series of rainfall, surface runoff component and baseflow component.
[0024] A ratio-constrained nonnegative matrix factorization algorithm is used to perform source analysis on the two-dimensional water quality observation matrix, decomposing it into a source feature matrix and a source contribution time coefficient matrix. Specifically, this includes:
[0025] S301, Set the time window length to Time series of ammonia nitrogen concentration and total phosphorus concentration were extracted to construct a two-dimensional water quality observation matrix. The first row is the ammonia nitrogen concentration sequence, and the second row is the total phosphorus concentration sequence.
[0026] S302, Set the number of sources for source resolution to be... The two-dimensional water quality observation matrix is decomposed into source feature matrices. Source contribution time coefficient matrix ,satisfy .
[0027] Among them, the source feature matrix The row vectors correspond to the ammonia nitrogen and total phosphorus characteristic terms, and the column vectors correspond to... One source of pollution.
[0028] Source contribution time coefficient matrix The row vectors correspond to The pollution source and the first The row vector corresponds to the surface runoff response component, the first Row vectors correspond to implicit time-delay response components, and column vectors correspond to time series components. That moment.
[0029] S303. Constructing the objective function Its expression is: And solve .
[0030] in, for Norm; used to measure the magnitude of matrix reconstruction error. and These are regularization parameters; they control the penalty weights of the time constraint and ratio constraint in the overall objective function, respectively.
[0031] This is a time constraint term; used in the constraint matrix. The time dynamic characteristics. This is a ratio constraint term; used in the constraint matrix. The stoichiometric characteristics.
[0032] While ensuring that the matrix product after decomposition restores the original water quality observation matrix as much as possible, constraints based on temporal physical laws and water quality chemical characteristics are introduced to avoid solutions without physical meaning from pure mathematical decomposition, thus ensuring that the analyzed pollution sources have real geochemical and hydrological attributes.
[0033] Specifically, it includes:
[0034] S3031, Regarding the source contribution time coefficient matrix The first corresponding surface runoff response row vector Set pulse constraint terms .
[0035] in, It is a proportionality coefficient; used to match the dimensional and magnitude differences between the contribution coefficient and surface runoff.
[0036] The calculated values are for the impulse constraint terms corresponding to the surface runoff response; This represents the total number of moments in the time series. For the first Each pollution source is at all times The contribution coefficient.
[0037] Mandatory requirement The contribution time variation curves of pollution sources are highly synchronized with the variation curves of surface runoff components. When the sum of the squared differences between the two is minimized, it indicates that the source is emitted in a pulse-like, instantaneous manner with surface runoff during rainfall, thus identifying the combined sewer overflow pulse source.
[0038] S3032, Regarding the source contribution time coefficient matrix The corresponding hidden delay response in the middle row vector Set delay constraints .
[0039] in, For a moment The amount of rainfall; This is a delay kernel function; used to describe different delay durations. Weighting of the impact of rainfall on current sewage discharge.
[0040] The maximum delay time step; represents the longest lag effect time. This is the time delay step variable.
[0041] The calculated value is the delay constraint term corresponding to the implicit delay response; For the first Each pollution source is at all times The contribution coefficient.
[0042] Mandatory requirement The contribution variation curve of the pollution source is presented as the delayed convolution result of the previous rainfall. When the sum of the squared differences between the two is the minimum, it indicates that the emission of the source does not occur immediately, but is accumulated and slowly released after the rainfall enters the culvert, thus identifying the delayed discharge source of the culvert.
[0043] S3033, Combine the pulse constraint term and the delay constraint term to obtain the time constraint term. .
[0044] Surface runoff responds to rainfall instantaneously and impulsively. Therefore, the impulse constraint term forces the time coefficient of the surface runoff response component to be proportional to the time series of the surface runoff component.
[0045] Due to the pipeline storage and sediment release mechanism, the sewage in the culvert has a lag effect and a prolonged effect in response to rainfall. Therefore, the delay constraint term adopts the convolution form of rainfall amount and delay kernel function to simulate the delayed discharge process of rainfall after being stored in the culvert.
[0046] S3034. Extract the source feature matrix The Middle column vector ;in, For the first The ammonia nitrogen source characteristic values of each pollution source For the first The total phosphorus source characteristic value of each pollution source.
[0047] S3035, Setting the threshold range for domestic sewage ratio And agricultural non-source heat pump ratio threshold range .
[0048] in, and These are the lower and upper thresholds for the ratio of ammonia nitrogen to total phosphorus in domestic sewage, respectively. and These represent the lower and upper thresholds for the ratio of ammonia nitrogen to total phosphorus in agricultural non-point sources, respectively.
[0049] Domestic sewage is rich in human excrement, and its ammonia nitrogen to total phosphorus ratio is usually high.
[0050] The application of chemical fertilizers in agricultural non-point source fertilizers results in relatively high total phosphorus content, and the ratio of ammonia nitrogen to total phosphorus is usually low.
[0051] S3036. Constructing ratio constraint terms:
[0052] ;
[0053] In the formula, A constant set to prevent division by zero. and These are weighting coefficients; they respectively control the intensity of penalties for exceeding the upper limit of ammonia nitrogen standards and for falling below the lower limit.
[0054] Based on prior knowledge of water quality, the nitrogen-phosphorus ratio ranges for the two types of pollution sources were defined, and then... Function implementation of penalty mechanism:
[0055] When the ratio of the source features obtained by parsing deviates from the reasonable range, a huge penalty term is generated, which forces the decomposition algorithm to assign features with high and low ratios to different source factors, thereby effectively separating domestic sewage sources from agricultural non-point sources.
[0056] S304, Fixed Source Feature Matrix The source contribution time coefficient matrix is updated using a multiplication update rule. The formula is:
[0057] ;
[0058] In the formula, For assignment update operators; This is element-wise multiplication; corresponding elements of the matrix are multiplied one by one.
[0059] Source feature matrix The transpose of . and The time constraint terms in the objective function are respectively related to The negative and positive components in the gradient non-negative decomposition, where both the negative and positive components are non-negative matrices.
[0060] With a fixed source feature matrix Under the premise of using a multiplication update rule to iteratively solve the matrix. .
[0061] The multiplication rule naturally guarantees the updated matrix The non-negativity of all elements, combined with the gradient information of the added constraint terms, ensures that each iteration converges in a direction that satisfies the time constraints and minimizes the reconstruction error.
[0062] S305, Fixed Source Contribution Time Coefficient Matrix The source feature matrix is updated using a multiplication update rule. The formula is:
[0063] ;
[0064] In the formula, Source contribution time coefficient matrix The transpose of .
[0065] and The ratio constraint terms in the objective function are respectively related to The negative and positive components in the gradient non-negative decomposition; and both the negative and positive components are non-negative matrices.
[0066] Contribution time coefficient matrix of fixed source Under the premise of using a multiplication update rule to iteratively solve the matrix. Similarly, the multiplication rule is used to maintain the non-negativity of the elements, and gradient correction with ratio constraints is incorporated to ensure that the decomposed source eigenvalues conform to the physical law of the nitrogen-phosphorus ratio between domestic sewage and agricultural non-point source pollution.
[0067] In nonnegative matrix factorization, it is essential to ensure that the matrix elements remain nonnegative throughout the update iteration process. By decomposing the gradient with constraints into positive and negative components and employing a fractional multiplication update rule, it can be ensured that in each iteration, the effect of subtracting the denominator from the numerator is achieved through multiplication proportionally. As long as the initial matrix is nonnegative, the updated matrix elements will not contain negative values.
[0068] If conventional additive gradient descent is used for updating, the addition of a penalty term can easily make it difficult to control the iteration step size, resulting in negative concentration values or contribution coefficients, which violates the physical meaning.
[0069] S306. Repeat steps S304 and S305 until the iterative change of the objective function is less than the set convergence threshold. .
[0070] S400. Based on the concentration ratio of ammonia nitrogen to total phosphorus in the source feature matrix, and combined with the correlation between the source contribution time coefficient matrix and the rainfall time series, identify the emission attributes of various pollution sources, and mark the spatial distribution location of the pollution sources corresponding to the emission attributes on the GIS map. Specifically, this includes:
[0071] S401, Calculate the source feature matrix The ratio of ammonia nitrogen to total phosphorus characteristic values for each column. .
[0072] S402, when Within the threshold range of domestic sewage ratio and corresponding Mid-row vector and rainfall sequence When the cross-correlation coefficient is less than or equal to the set correlation threshold, the column is identified as a source of mixed sewage from dry streams.
[0073] S403, when Within the threshold range of domestic sewage ratio and When the cross-correlation coefficient is greater than the set correlation threshold, it is identified as a combined sewer and stormwater pulse source.
[0074] S404, when Within the agricultural non-source heat pump ratio threshold range and corresponding Mid-row vector and surface runoff components When the cross-correlation coefficient is greater than or equal to the set correlation threshold, the column is identified as a non-point source of farmland runoff.
[0075] Although dry-flow sewage and combined sewage discharge both belong to domestic sewage and have the same high ammonia nitrogen / total phosphorus ratio, dry-flow sewage is less affected by rainfall, and its time coefficient has a low correlation with rainfall.
[0076] Combined stormwater and sewage discharge is caused by rainwater runoff and is highly correlated with rainfall.
[0077] Agricultural non-point sources are characterized by low ammonia nitrogen / total phosphorus ratios, and their excretion mainly occurs with surface runoff. Therefore, their time coefficient is highly correlated with surface runoff components.
[0078] Accurate source tracing is achieved by combining chemical fingerprinting with hydrologically driven two-dimensional cross-referencing.
[0079] S405, Calculation Time of the first Contribution weight of pollution sources ,formula: On the GIS map, different colors are rendered at the corresponding monitoring section locations to indicate the contribution weight of various pollution sources.
[0080] in, For the first The column vector length of each pollution source; No. Each pollution source is at all times The contribution coefficient; For the first The column vector length of each pollution source; For the first Each pollution source is at all times The contribution coefficient.
[0081] By calculating the normalized contribution weights, the magnitude interference caused by the difference in absolute flow rate at the monitoring sections is eliminated, making the pollution source structure of different sections horizontally comparable.
[0082] Dynamic rendering based on weights on GIS maps can intuitively and quantitatively display the real-time evolution patterns of dominant pollution sources at different spatial locations, helping managers to quickly identify key pollution discharge areas and critical pollution sources.
[0083] The present invention also provides a water pollution data monitoring system based on multi-objective analysis, including a data acquisition module, a hydrological segmentation module, a source apportionment module, and a visualization module.
[0084] The data acquisition module collects GIS maps of the target watershed, rainfall time series, flow time series of monitoring sections, and water quality monitoring time series, including ammonia nitrogen concentration time series and total phosphorus concentration time series.
[0085] The hydrological segmentation module performs hydrological segmentation on the flow time series based on the opening and closing status of the dam, and calculates the surface runoff component time series and the baseflow component time series.
[0086] The source apportionment module constructs a two-dimensional water quality observation matrix containing ammonia nitrogen and total phosphorus. Combining the time series of rainfall, surface runoff component and baseflow component with pulse constraints and delay constraints, a non-negative matrix factorization algorithm with ratio constraints is used to perform source apportionment on the two-dimensional water quality observation matrix, and the source feature matrix and source contribution time coefficient matrix are obtained.
[0087] The visualization module identifies pollution source emission attributes based on the concentration ratio of ammonia nitrogen to total phosphorus in the source feature matrix and the correlation between the source contribution time coefficient matrix and the rainfall sequence, and marks the spatial distribution location of pollution sources corresponding to the emission attributes on the GIS map.
[0088] Compared with the prior art, the beneficial effects achieved by the present invention are:
[0089] Existing technologies typically employ a single, fixed filter parameter for baseflow segmentation, failing to consider the drastic changes in river hydrodynamic conditions caused by dam or gate control. This results in significant discrepancies between the segmented surface runoff and baseflow. This invention, by dynamically adjusting the filter parameter to match the actual hydrological receding patterns under dam control, significantly improves the accuracy of baseflow segmentation in watersheds affected by dam control, providing accurate hydrological component inputs for subsequent pollution source analysis.
[0090] Traditional nonnegative matrix factorization (NMF) relies solely on mathematical statistical characteristics, and the resulting components often fail to correspond to real-world physical sources. This invention introduces impulse and delay constraints, injecting prior hydrophysical mechanisms into the matrix factorization optimization process. This forces the algorithm to decouple instantaneous response sources from delayed response sources, effectively overcoming the source aliasing phenomenon present in traditional NMF algorithms, and ensuring that the decomposed source contribution time coefficients have clear physical interpretability.
[0091] Existing technologies typically rely on a single water quality concentration or a single hydrological indicator for source identification, failing to distinguish between pollution sources with similar concentration characteristics but different driving mechanisms (such as dry-flow baseflow and combined sewer overflow). This invention, through cross-validation of chemical characteristics and hydrological response characteristics, accurately separates dry-flow baseflow from rainwater runoff, both belonging to domestic sewage, achieving more refined identification of pollution source emission attributes and providing differentiated decision support for pipeline network repair and non-point source pollution control.
[0092] The absolute concentrations of different pollution sources fluctuate significantly over time, making it difficult to reliably distinguish source types based solely on absolute concentration. This invention utilizes the inherent stoichiometry (nitrogen-phosphorus ratio) of pollution sources as a priori constraint to anchor the chemical fingerprint of pollution sources from a characteristic dimension. This avoids the contradictory nitrogen-phosphorus inversion phenomenon in the source characteristic values obtained from source apportionment, thereby improving the chemical rationality and classification accuracy of the source apportionment results. Attached Figure Description
[0093] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:
[0094] Figure 1 This is a flowchart illustrating the water pollution data monitoring method based on multi-objective analysis of the present invention. Detailed Implementation
[0095] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0096] Please see Figure 1 This invention provides a water pollution data monitoring method based on multi-objective analysis, comprising:
[0097] S100. Collect GIS maps of the target watershed, rainfall time series, flow time series of monitoring sections, and water quality monitoring time series. Among them, the water quality monitoring time series should include at least ammonia nitrogen concentration time series and total phosphorus concentration time series.
[0098] GIS maps are used to describe the spatial topology of river networks, dam locations, culvert sections, and drainage outlets within a target watershed.
[0099] Rainfall time series refers to the cumulative precipitation data collected by rain gauges within a target watershed at a set time step.
[0100] Flow time series refers to the flow rate data collected by the flow meter at the monitoring section at a set time step.
[0101] Water quality monitoring time series refers to the ammonia nitrogen concentration and total phosphorus concentration values collected by automatic water quality monitoring stations at a set time step.
[0102] The monitoring sections include the main trunk section set up on the main stream, the tributary section set up at the confluence of tributaries, and the upstream and downstream sections set up before and after the dam.
[0103] S200. Based on the opening and closing status of dams and sluices in the target watershed, the flow time series is hydrologically segmented to calculate the surface runoff component time series and the baseflow component time series. Specifically, this includes:
[0104] S201. Analyze the opening and closing status of sluice gates and dams within the target watershed. When the sluice gates and dams are fully closed, the current period is defined as the sluice gate-controlled flow retention period. When the sluice gates and dams are open, the current period is defined as the free-flow period.
[0105] S202, Adopt Digital filtering is used to segment the flow time series. During the gated flow stagnation period, filtering parameters are set. Substitute the first preset value into the formula to calculate the time series of the base current component.
[0106] During periods of smooth flow, set the filter parameters. Substitute the second preset value into the formula to calculate the time series of the base current component; the formula is:
[0107] ;
[0108] In the formula, for The base current component value at time; For the previous moment The base current component value; for Actual measured total flow rate at any given time; It is the baseflow index; it reflects the long-term proportion of baseflow in total flow.
[0109] The filtering parameters are used to control the smoothness of the changes in the base current between adjacent time steps.
[0110] During the sluice gate closure period, the closure of the gate and dam results in poor water flow and a slow decay of the baseflow recession curve. Therefore, filter parameters are set. A smaller first preset value is used to match the slow decay rate.
[0111] During the free-flow period, the water flow is strong when the dam is open, and the baseflow response is relatively fast. Therefore, filter parameters are set accordingly. A larger second preset value is used to match a faster decay rate.
[0112] use Digital filtering uses a recursive algorithm to smoothly separate the stable baseflow component, which is affected by groundwater recharge and soil infiltration, from the mixed total flow time series, thereby separating the rainfall-driven surface runoff component.
[0113] S203. Calculate the time series of surface runoff components, using the following formula: ;in, for Surface runoff component value at time.
[0114] S300. Construct a two-dimensional water quality observation matrix containing ammonia nitrogen and total phosphorus, and set impulse constraints and delay constraints by combining the time series of rainfall, surface runoff component and baseflow component.
[0115] A ratio-constrained nonnegative matrix factorization algorithm is used to perform source analysis on the two-dimensional water quality observation matrix, decomposing it into a source feature matrix and a source contribution time coefficient matrix. Specifically, this includes:
[0116] S301, Set the time window length to Time series of ammonia nitrogen concentration and total phosphorus concentration were extracted to construct a two-dimensional water quality observation matrix. The first row is the ammonia nitrogen concentration sequence, and the second row is the total phosphorus concentration sequence.
[0117] S302, Set the number of sources for source resolution to be... The two-dimensional water quality observation matrix is decomposed into source feature matrices. Source contribution time coefficient matrix ,satisfy .
[0118] Among them, the source feature matrix The row vectors correspond to the ammonia nitrogen and total phosphorus characteristic terms, and the column vectors correspond to... One source of pollution.
[0119] Source contribution time coefficient matrix The row vectors correspond to The pollution source and the first The row vector corresponds to the surface runoff response component, the first Row vectors correspond to implicit time-delay response components, and column vectors correspond to time series components. That moment.
[0120] S303. Constructing the objective function Its expression is: And solve .
[0121] in, for Norm; used to measure the magnitude of matrix reconstruction error. and These are regularization parameters; they control the penalty weights of the time constraint and ratio constraint in the overall objective function, respectively.
[0122] This is a time constraint term; used in the constraint matrix. The time dynamic characteristics. This is a ratio constraint term; used in the constraint matrix. The stoichiometric characteristics.
[0123] While ensuring that the product of the matrices after decomposition restores the original water quality observation matrix as much as possible (reconstructing the error term on the left), constraints based on the laws of time physics and the chemical characteristics of water quality are introduced to avoid solutions with no physical meaning (such as source confusion) that may occur in pure mathematical decomposition, thus ensuring that the analyzed pollution sources have real geochemical and hydrological attributes.
[0124] Specifically, it includes:
[0125] S3031, Regarding the source contribution time coefficient matrix The first corresponding surface runoff response row vector Set pulse constraint terms .
[0126] in, It is a proportionality coefficient; used to match the dimensional and magnitude differences between the contribution coefficient and surface runoff.
[0127] The calculated values are for the impulse constraint terms corresponding to the surface runoff response; This represents the total number of moments in the time series. For the first Each pollution source is at all times The contribution coefficient.
[0128] Mandatory requirement The contribution time variation curves of pollution sources are highly synchronized with the variation curves of surface runoff components. When the sum of the squared differences between the two is minimized, it indicates that the source is emitted in a pulse-like, instantaneous manner with surface runoff during rainfall, thus identifying the combined sewer overflow pulse source.
[0129] S3032, Regarding the source contribution time coefficient matrix The corresponding hidden delay response in the middle row vector Set delay constraints .
[0130] in, For a moment The amount of rainfall; This is a delay kernel function; used to describe different delay durations. Weighting of the impact of rainfall on current sewage discharge.
[0131] The maximum delay time step; represents the longest lag effect time. This is the time delay step variable.
[0132] The calculated value is the delay constraint term corresponding to the implicit delay response; For the first Each pollution source is at all times The contribution coefficient.
[0133] Mandatory requirement The contribution variation curve of the pollution source is presented as the delayed convolution result of the previous rainfall. When the sum of the squared differences between the two is the minimum, it indicates that the emission of the source does not occur immediately, but is accumulated and slowly released after the rainfall enters the culvert, thus identifying the delayed discharge source of the culvert.
[0134] S3033, Combine the pulse constraint term and the delay constraint term to obtain the time constraint term. .
[0135] Surface runoff responds to rainfall instantaneously and impulsively. Therefore, the impulse constraint term forces the time coefficient of the surface runoff response component to be proportional to the time series of the surface runoff component.
[0136] Due to the pipeline storage and sediment release mechanism, the sewage in the culvert has a lag effect and a prolonged effect in response to rainfall. Therefore, the delay constraint term adopts the convolution form of rainfall amount and delay kernel function to simulate the delayed discharge process of rainfall after being stored in the culvert.
[0137] S3034. Extract the source feature matrix The Middle column vector ;in, For the first The ammonia nitrogen source characteristic values of each pollution source For the first The total phosphorus source characteristic value of each pollution source.
[0138] S3035, Setting the threshold range for domestic sewage ratio And agricultural non-source heat pump ratio threshold range .
[0139] in, and These are the lower and upper thresholds for the ratio of ammonia nitrogen to total phosphorus in domestic sewage, respectively. and These represent the lower and upper thresholds for the ratio of ammonia nitrogen to total phosphorus in agricultural non-point sources, respectively.
[0140] Domestic sewage is rich in human excrement, and its ammonia nitrogen to total phosphorus ratio is usually high (i.e., (Higher range).
[0141] Agricultural non-point source fertilizer application leads to relatively high total phosphorus content, and its ammonia nitrogen to total phosphorus ratio is usually low (i.e., (Lower range).
[0142] S3036. Constructing ratio constraint terms:
[0143] ;
[0144] In the formula, A constant set to prevent division by zero. and These are weighting coefficients; they respectively control the intensity of penalties for exceeding the upper limit of ammonia nitrogen standards and for falling below the lower limit.
[0145] Based on prior knowledge of water quality, the nitrogen-phosphorus ratio ranges for the two types of pollution sources were defined, and then... Function implementation of penalty mechanism:
[0146] When the ratio of the source features deviates from the reasonable range (the nitrogen-phosphorus ratio of domestic sewage is too high and the nitrogen-phosphorus ratio of agricultural non-point sources is too low), a huge penalty term is generated, which forces the decomposition algorithm to assign the features with high and low ratios to different source factors, thereby effectively separating domestic sewage sources from agricultural non-point sources.
[0147] S304, Fixed Source Feature Matrix The source contribution time coefficient matrix is updated using a multiplication update rule. The formula is:
[0148] ;
[0149] In the formula, For assignment update operators; This is element-wise multiplication; corresponding elements of the matrix are multiplied one by one (Hadamard product).
[0150] Source feature matrix The transpose of . and The time constraint terms in the objective function are respectively related to The negative and positive components in the gradient non-negative decomposition, where both the negative and positive components are non-negative matrices.
[0151] With a fixed source feature matrix Under the premise of using a multiplication update rule to iteratively solve the matrix. .
[0152] The multiplication rule naturally guarantees the updated matrix The non-negativity of all elements, combined with the gradient information of the added constraint terms, ensures that each iteration converges in a direction that satisfies the time constraints and minimizes the reconstruction error.
[0153] S305, Fixed Source Contribution Time Coefficient Matrix The source feature matrix is updated using a multiplication update rule. The formula is:
[0154] ;
[0155] In the formula, Source contribution time coefficient matrix The transpose of .
[0156] and The ratio constraint terms in the objective function are respectively related to The negative and positive components in the gradient non-negative decomposition; and both the negative and positive components are non-negative matrices.
[0157] Contribution time coefficient matrix of fixed source Under the premise of using a multiplication update rule to iteratively solve the matrix. Similarly, the multiplication rule is used to maintain the non-negativity of the elements, and gradient correction with ratio constraints is incorporated to ensure that the decomposed source eigenvalues conform to the physical law of the nitrogen-phosphorus ratio between domestic sewage and agricultural non-point source pollution.
[0158] In nonnegative matrix factorization, it is essential to ensure that the matrix elements remain nonnegative throughout the update iteration process. The gradient with constraints is decomposed into positive and negative components (i.e.,...). (and both are non-negative), and adopts the fractional multiplication update rule, which can ensure that in each iteration, the effect of subtracting the denominator from the numerator is achieved through multiplication proportion. As long as the initial matrix is non-negative, the updated matrix elements will not have negative values.
[0159] If conventional additive gradient descent is used for updating, the addition of a penalty term can easily make it difficult to control the iteration step size, resulting in negative concentration values or contribution coefficients, which violates the physical meaning.
[0160] S306. Repeat steps S304 and S305 until the iterative change of the objective function is less than the set convergence threshold. .
[0161] S400. Based on the concentration ratio of ammonia nitrogen to total phosphorus in the source feature matrix, and combined with the correlation between the source contribution time coefficient matrix and the rainfall time series, identify the emission attributes of various pollution sources, and mark the spatial distribution location of the pollution sources corresponding to the emission attributes on the GIS map. Specifically, this includes:
[0162] S401, Calculate the source feature matrix The ratio of ammonia nitrogen to total phosphorus characteristic values for each column. .
[0163] S402, when Within the threshold range of domestic sewage ratio and corresponding Mid-row vector and rainfall sequence When the cross-correlation coefficient is less than or equal to the set correlation threshold, the column is identified as a source of mixed sewage from dry streams.
[0164] S403, when Within the threshold range of domestic sewage ratio and When the cross-correlation coefficient is greater than the set correlation threshold, it is identified as a combined sewer and stormwater pulse source.
[0165] S404, when Within the agricultural non-source heat pump ratio threshold range and corresponding Mid-row vector and surface runoff components When the cross-correlation coefficient is greater than or equal to the set correlation threshold, the column is identified as a non-point source of farmland runoff.
[0166] Although dry-flow sewage and combined sewage discharge both belong to domestic sewage and have the same high ammonia nitrogen / total phosphorus ratio, dry-flow sewage is less affected by rainfall, and its time coefficient has a low correlation with rainfall.
[0167] Combined stormwater and sewage discharge is caused by rainwater runoff and is highly correlated with rainfall.
[0168] Agricultural non-point sources are characterized by low ammonia nitrogen / total phosphorus ratios, and their excretion mainly occurs with surface runoff. Therefore, their time coefficient is highly correlated with surface runoff components.
[0169] Accurate source tracing is achieved by combining chemical fingerprints (ratios) with hydrological-driven (correlation) two-dimensional cross-determination.
[0170] S405, Calculation Time of the first Contribution weight of pollution sources ,formula: On the GIS map, different colors are rendered at the corresponding monitoring section locations to indicate the contribution weight of various pollution sources.
[0171] in, For the first The column vector length of each pollution source; No. Each pollution source is at all times The contribution coefficient; For the first The column vector length of each pollution source; For the first Each pollution source is at all times The contribution coefficient.
[0172] By calculating the normalized contribution weights, the magnitude interference caused by the difference in absolute flow rate at the monitoring sections is eliminated, making the pollution source structure of different sections horizontally comparable.
[0173] Dynamic rendering based on weights on GIS maps can intuitively and quantitatively display the real-time evolution patterns of dominant pollution sources at different spatial locations, helping managers to quickly identify key pollution discharge areas and critical pollution sources.
[0174] The present invention also provides a water pollution data monitoring system based on multi-objective analysis, including a data acquisition module, a hydrological segmentation module, a source apportionment module, and a visualization module.
[0175] The data acquisition module collects GIS maps of the target watershed, rainfall time series, flow time series of monitoring sections, and water quality monitoring time series, including ammonia nitrogen concentration time series and total phosphorus concentration time series.
[0176] The hydrological segmentation module performs hydrological segmentation on the flow time series based on the opening and closing status of the dam, and calculates the surface runoff component time series and the baseflow component time series.
[0177] The source apportionment module constructs a two-dimensional water quality observation matrix containing ammonia nitrogen and total phosphorus. Combining the time series of rainfall, surface runoff component and baseflow component with pulse constraints and delay constraints, a non-negative matrix factorization algorithm with ratio constraints is used to perform source apportionment on the two-dimensional water quality observation matrix, and the source feature matrix and source contribution time coefficient matrix are obtained.
[0178] The visualization module identifies pollution source emission attributes based on the concentration ratio of ammonia nitrogen to total phosphorus in the source feature matrix and the correlation between the source contribution time coefficient matrix and the rainfall sequence, and marks the spatial distribution location of pollution sources corresponding to the emission attributes on the GIS map.
[0179] Example 1: A typical urban-rural complex watershed with gate control in East China was selected as the experimental object. The total area of the watershed is 238 km², the main stream is 42 km long, and there is a medium-sized gate, three main tributaries, 12 municipal drainage outlets and an 8.6 km culvert section in the watershed.
[0180] The upper reaches of the basin are mainly agricultural planting areas, while the middle and lower reaches pass through urban built-up areas. There are multiple pollution problems such as mixed discharge of rainwater and sewage, sewage accumulation in culverts and farmland runoff. Frequent opening and closing of sluice gates and dams leads to complex hydrodynamic conditions.
[0181] Four key monitoring sections were set up within the basin, forming a three-dimensional monitoring network covering the main stream, tributaries, and upstream and downstream of sluice gates and dams.
[0182] Main section S1: Located at the watershed outlet, used to monitor the overall pollution output load of the watershed;
[0183] Tributary section S2: Located 500m upstream of the confluence of the largest tributary, it is used to monitor agricultural non-point source pollution and sewage discharge along the tributary.
[0184] S3 section upstream of the gate: located 300m upstream of the gate, used to monitor the accumulation of pollutants upstream of the gate;
[0185] S4 section downstream of the dam: located 500m downstream of the dam, used to monitor the diffusion of pollutants after the dam discharges.
[0186] The data collection period is from June 1, 2025 to June 30, 2025 (30 days, 720 hours). The time step for all monitoring devices is uniformly set to 1 hour. The data collection parameters are as follows:
[0187] GIS map data: Obtain a 1:5000 high-precision electronic map of the watershed, and mark the spatial topology of the river network, dam locations, culvert sections, and all drainage outlets.
[0188] Rainfall data: Hourly cumulative precipitation data collected by three automatic rain gauges within the basin were used, and the arithmetic mean was taken as the basin surface rainfall.
[0189] Flow data: The flow rate of water is collected in real time at each monitoring section using an acoustic Doppler velocity profiler.
[0190] Water quality data: Ammonia nitrogen (NH3-N) and total phosphorus (TP) concentrations were collected at each monitoring section using online automatic water quality monitoring stations, with monitoring accuracies of ±0.01 mg / L and ±0.001 mg / L, respectively.
[0191] The statistical characteristics of the collected raw monitoring data are shown in Table 1 below. All data meet the requirements of the "Technical Specification for Monitoring Surface Water and Wastewater" (HJ / T91-2002).
[0192] Table 1
[0193]
[0194] Based on the gate operation records, the 30-day monitoring period is divided into:
[0195] The gate-controlled congestion period will be from 0:00 on June 3 to 23:00 on June 6, from 0:00 on June 15 to 23:00 on June 18, and from 0:00 on June 25 to 23:00 on June 28, for a total of 12 days.
[0196] Smooth flow period: the remaining 18 days.
[0197] Using dynamic parameters The digital filtering method is used for base current segmentation, with the following parameter settings:
[0198] Baseline index (Based on statistical analysis of long-term hydrological observation data of the basin);
[0199] Gate control lag period filter parameters (First preset value, matching a slow water receding rate);
[0200] Flow period filter parameters (Second preset value, matching the rapid water receding rate);
[0201] Basic current segmentation calculation formula: ;
[0202] Formula for calculating surface runoff components: .
[0203] Set time window length (Covering the entire 30-day monitoring period), time series of ammonia nitrogen and total phosphorus concentrations at each monitoring section were extracted to construct a two-dimensional water quality observation matrix. The first row is the ammonia nitrogen concentration sequence, and the second row is the total phosphorus concentration sequence.
[0204] Preset number of pollution sources (Corresponding to dry-flow sewage inflow source, rainwater and sewage mixed discharge pulse source, and farmland runoff non-point source).
[0205] Regularization parameters ;
[0206] Convergence threshold ;
[0207] Threshold range of ammonia nitrogen / total phosphorus ratio in domestic sewage ;
[0208] Agricultural non-point source ammonia nitrogen / total phosphorus ratio threshold range ;
[0209] Maximum delay time step Hour;
[0210] The delay kernel function uses an exponential decay function. ;
[0211] Construct an objective function that includes a reconstruction error term, a time constraint term, and a ratio constraint term:
[0212] ;
[0213] Among them, time constraint item It includes pulse constraints and delay constraints.
[0214] Ratio constraint This method is used to constrain the nitrogen-phosphorus ratio of pollution sources within a reasonable range; iterative solutions are obtained using an alternating multiplication update rule. and The matrix iterative process is as follows:
[0215] 1. Initialize the nonnegative matrix and ;
[0216] 2. Fixed ,renew : ;
[0217] 3. Fix ,renew : ;
[0218] Repeat steps 2-3 until... Finally, after 128 iterations, the convergence condition was met.
[0219] Taking tributary section S2 as an example, the source feature matrix obtained by decomposition is... Source contribution time coefficient matrix The key results are as follows:
[0220] Source feature matrix (Ammonia nitrogen, total phosphorus):
[0221] Source 1: (7.82, 1.02), ammonia nitrogen / total phosphorus ratio = 7.67;
[0222] Source 2: (6.95, 0.91), ammonia nitrogen / total phosphorus ratio = 7.64;
[0223] Source 3: (2.18, 1.05), ammonia nitrogen / total phosphorus ratio = 2.08;
[0224] Correlation analysis of source contribution time coefficients:
[0225] The Pearson correlation coefficient between Source 1 and rainfall was 0.12.
[0226] The Pearson correlation coefficient between Source 2 and rainfall was 0.87.
[0227] The Pearson correlation coefficient between source 3 and surface runoff is 0.91.
[0228] Two-dimensional cross-determination is performed based on the nitrogen-to-phosphorus ratio of source characteristics and hydrological correlation:
[0229] Source 1: Nitrogen-to-phosphorus ratio 7.67 (within the range of domestic sewage), correlation coefficient with rainfall 0.12 (≤0.3) → dry-flow sewage mixed into the source;
[0230] Source 2: Nitrogen-to-phosphorus ratio 7.64 (within the range of domestic sewage), correlation coefficient with rainfall 0.87 (>0.3) → combined sewer pulse source;
[0231] Source 3: Nitrogen-to-phosphorus ratio 2.08 (within the agricultural non-point source range), correlation coefficient with surface runoff 0.91 (≥0.3) → farmland runoff non-point source;
[0232] Calculate the normalized contribution weights of various pollution sources at each time point: The average contribution weights of various pollution sources at tributary section S2 were obtained during the 30-day monitoring period:
[0233] Sources of sewage mixed with dry water: 38.2%; Pulse sources of combined sewer overflows: 31.5%; Non-point sources of farmland runoff: 30.3%.
[0234] GIS spatial visualization overlays the pollution source contribution weight data of each monitoring section with the watershed GIS map, and uses color gradient to render the pollution source structure of different sections, generating a watershed water pollution source tracing diagram.
[0235] Spatial analysis revealed that farmland runoff from upstream agricultural areas accounted for 62.7% of the total water volume, combined sewer overflows near urban built-up areas accounted for 58.3%, and sewage accumulation in the stagnant flow area in front of the sluice gate accounted for 47.9%.
[0236] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0237] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A water pollution data monitoring method based on multi-objective analysis, characterized in that: The method includes: S100. Collect GIS maps of the target watershed, rainfall time series, flow time series of monitoring sections, and water quality monitoring time series; among which, the water quality monitoring time series shall include at least ammonia nitrogen concentration time series and total phosphorus concentration time series; S200. Based on the opening and closing status of the dams in the target watershed, the flow time series is hydrologically segmented to calculate the surface runoff component time series and the baseflow component time series. S300. Construct a two-dimensional water quality observation matrix containing ammonia nitrogen and total phosphorus. Combine the time series of rainfall, surface runoff, and baseflow to set impulse and delay constraints. Use a non-negative matrix factorization algorithm with ratio constraints to perform source analysis on the two-dimensional water quality observation matrix, and decompose it to obtain the source feature matrix and the source contribution time coefficient matrix. S400. Based on the concentration ratio of ammonia nitrogen to total phosphorus in the source feature matrix, combined with the correlation between the source contribution time coefficient matrix and the rainfall time series, identify the emission attributes of various pollution sources, and mark the spatial distribution location of pollution sources corresponding to the emission attributes on the GIS map.
2. The water pollution data monitoring method based on multi-objective analysis according to claim 1, characterized in that: In S100, the GIS map is used to describe the spatial topology of the river network, dam locations, culvert sections, and drainage outlets within the target watershed. Rainfall time series refers to the cumulative precipitation data collected by rain gauges within a target watershed at a set time step; Flow time series refers to the flow rate data of water collected by the flow meter at the monitoring section at a set time step; Water quality monitoring time series refers to the ammonia nitrogen concentration and total phosphorus concentration values collected by automatic water quality monitoring stations at a set time step; The monitoring sections include the main trunk section set up on the main stream, the tributary section set up at the confluence of tributaries, and the upstream and downstream sections set up before and after the dam.
3. The water pollution data monitoring method based on multi-objective analysis according to claim 2, characterized in that: S200 includes: S201. Analyze the opening and closing status of sluice gates and dams within the target watershed. When the sluice gates and dams are fully closed, the current period is defined as the sluice gate-controlled flow retention period; when the sluice gates and dams are open, the current period is defined as the free flow period. S202, Adopt Digital filtering is used to segment the flow time series; during the gated flow stagnation period, filtering parameters are set. Substitute the first preset value into the formula to calculate the time series of the base current component; During periods of smooth flow, set the filter parameters. Substitute the second preset value into the formula to calculate the time series of the base current component; the formula is: ； In the formula, for The base current component value at time; For the previous moment The base current component value; for Actual measured total flow rate at any given time; The base current index; S203. Calculate the time series of surface runoff components, using the following formula: ;in, for Surface runoff component value at time.
4. The water pollution data monitoring method based on multi-objective analysis according to claim 3, characterized in that: The S300 includes: S301, Set the time window length to Time series of ammonia nitrogen concentration and total phosphorus concentration were extracted to construct a two-dimensional water quality observation matrix. The first row contains the ammonia nitrogen concentration sequence, and the second row contains the total phosphorus concentration sequence. S302, Set the number of sources for source resolution to be... The two-dimensional water quality observation matrix is decomposed into source feature matrices. Source contribution time coefficient matrix ,satisfy ; Among them, the source feature matrix The row vectors correspond to the ammonia nitrogen and total phosphorus characteristic terms, and the column vectors correspond to... One source of pollution; Source contribution time coefficient matrix The row vectors correspond to The pollution source and the first The row vector corresponds to the surface runoff response component, the first Row vectors correspond to implicit time-delay response components, and column vectors correspond to time series components. At that moment; S303. Constructing the objective function Its expression is: And solve ; in, for Norm; and For regularization parameters; For time constraints; This is a ratio constraint term.
5. The water pollution data monitoring method based on multi-objective analysis according to claim 4, characterized in that: In S303, the time constraint item The construction steps are as follows: S3031, Regarding the source contribution time coefficient matrix The first corresponding surface runoff response row vector Set pulse constraint terms ;in, This is the proportionality coefficient; S3032, Regarding the source contribution time coefficient matrix The corresponding hidden delay response in the middle row vector Set delay constraints ; in, For a moment The amount of rainfall; For delay kernel functions; This is the maximum delay time step; The time delay step size is a variable; S3033, Combine the pulse constraint term and the delay constraint term to obtain the time constraint term. .
6. The water pollution data monitoring method based on multi-objective analysis according to claim 4, characterized in that: In S303, the ratio constraint term The construction steps are as follows: S3034. Extract the source feature matrix The Middle column vector ;in, For the first The characteristic values of ammonia nitrogen source for each pollution source For the first Total phosphorus source characteristic value of each pollution source; S3035, Setting the threshold range for domestic sewage ratio And agricultural non-source heat pump ratio threshold range ; in, and These are the lower and upper thresholds for the ratio of ammonia nitrogen to total phosphorus in domestic sewage, respectively. and These are the lower and upper limits of the threshold for the ratio of ammonia nitrogen to total phosphorus in agricultural non-point sources, respectively. S3036. Constructing ratio constraint terms: ； In the formula, A constant set to prevent division by zero. and These are the weighting coefficients.
7. The water pollution data monitoring method based on multi-objective analysis according to claim 4, characterized in that: S300 also includes an updated solution step: S304, Fixed Source Feature Matrix The source contribution time coefficient matrix is updated using a multiplication update rule. The formula is: ； In the formula, This is element-wise multiplication; and The time constraint terms in the objective function are respectively related to The negative and positive components in the gradient non-negative decomposition, where both the negative and positive components are non-negative matrices; S305, Fixed Source Contribution Time Coefficient Matrix The source feature matrix is updated using a multiplication update rule. The formula is: ； In the formula, and The ratio constraint terms in the objective function are respectively related to The negative and positive components in the gradient non-negative decomposition; and both the negative and positive components are non-negative matrices; S306. Repeat steps S304 and S305 until the iterative change of the objective function is less than the set convergence threshold. .
8. The water pollution data monitoring method based on multi-objective analysis according to claim 6, characterized in that: The S400 includes: S401, Calculate the source feature matrix The ratio of ammonia nitrogen to total phosphorus characteristic values for each column. ; S402, when Within the threshold range of domestic sewage ratio and corresponding Mid-row vector and rainfall sequence When the cross-correlation coefficient is less than or equal to the set correlation threshold, the column is identified as a source of contamination from dry-flow sewage. S403, when Within the threshold range of domestic sewage ratio and When the cross-correlation coefficient is greater than the set correlation threshold, it is identified as a combined sewer and stormwater pulse source; S404, when Within the agricultural non-source heat pump ratio threshold range and corresponding Mid-row vector and surface runoff components When the cross-correlation coefficient is greater than or equal to the set correlation threshold, the column is identified as a farmland runoff non-source. S405, Calculation Time of the first Contribution weight of pollution sources ,formula: On the GIS map, different colors are rendered at the corresponding monitoring section locations to indicate the contribution weight of various pollution sources; in, For the first The column vector magnitude of each pollution source; No. Each pollution source is at all times The contribution coefficient; For the first The column vector magnitude of each pollution source; For the first Each pollution source is at all times The contribution coefficient.
9. A water pollution data monitoring system based on multi-objective analysis, applied to the water pollution data monitoring method based on multi-objective analysis as described in claim 1, characterized in that: The system includes a data acquisition module, a hydrological segmentation module, a source analysis module, and a visualization module; The data acquisition module collects GIS maps of the target watershed, rainfall time series, flow time series of monitoring sections, and water quality monitoring time series, including ammonia nitrogen concentration time series and total phosphorus concentration time series. The hydrological segmentation module performs hydrological segmentation on the flow time series based on the opening and closing status of the dam, and calculates the surface runoff component time series and the baseflow component time series; The source apportionment module constructs a two-dimensional water quality observation matrix containing ammonia nitrogen and total phosphorus. Combining the pulse constraints and delay constraints set by the rainfall time series, surface runoff component and baseflow component, the two-dimensional water quality observation matrix is decomposed into a source feature matrix and a source contribution time coefficient matrix. The visualization module identifies pollution source emission attributes based on the concentration ratio of ammonia nitrogen to total phosphorus in the source feature matrix and the correlation between the source contribution time coefficient matrix and the rainfall sequence, and marks the spatial distribution location of pollution sources corresponding to the emission attributes on the GIS map.