A multi-parameter water quality synchronous detection and pollution source analysis and tracing method and system
By filtering source signals and analyzing features of the multimodal in-situ spectral matrix, and combining constrained optimization reconstruction with spatial continuity and ratio invariance constraints, the problem of quantitative separation of pollution sources in existing technologies has been solved, and accurate quantification of pollution source contribution weights and accurate representation of spatial distribution have been achieved.
Patent Information
- Application Number
- CN202610655221.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-13
- Publication Date
- 2026-06-23
Smart Images
Figure CN122259461A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of water quality monitoring technology, and in particular to a method and system for simultaneous multi-parameter water quality detection and pollution source analysis and tracing. Background Technology
[0002] In water environment quality monitoring and emergency response to sudden environmental incidents, simultaneous multi-parameter water quality monitoring and pollution source analysis of surface water basins are crucial for supporting precise pollution control decisions. Current technologies typically employ online monitoring or portable instruments to perform spectral scanning of water samples at multiple monitoring nodes in the target watershed, obtaining absorption and fluorescence spectra data for each node. Then, by comparing the response intensity of characteristic peaks in the mixed spectrum with the spectral similarity to standard spectra of known pollution sources, the direction of pollution sources is inferred from the concentration gradient changes along upstream and downstream nodes of the river channel.
[0003] However, when multiple potential pollution sources of different types exist within a watershed, the effects of torrential rain and other factors can lead to concentrated discharges from these sources. In such cases, the characteristic fluorescence peaks of various pollutants in the water body severely overlap, and their spectral response ranges highly intersect. Existing methods based on spectral similarity comparison can only qualitatively determine whether a signal profile of a known pollution source exists in the mixed spectrum. These methods cannot quantitatively separate the characteristic components of each potential pollution source from the overlapping composite pollution spectra, and even more so, they cannot determine the independent contribution of each pollution source to the pollution load of the same node water sample. As a result, source tracing analysis can only make directional inferences and cannot provide quantitative data support for accurately identifying the responsible party for pollution. Summary of the Invention
[0004] This invention provides a method and system for simultaneous multi-parameter water quality detection and pollution source analysis and tracing to solve the problems mentioned in the background art.
[0005] To achieve the above objectives, this invention provides a method for simultaneous multi-parameter water quality detection and pollution source apportionment, comprising: Source signals of the multimodal in-situ spectral matrix are screened, and the selected candidate source signals are analyzed to obtain the pollution feature fingerprint spectrum. The pollution feature fingerprint spectrum is reconstructed by constraint optimization on the source spectrum matrix. By minimizing the reconstruction error, the contribution weight of each potential pollution source to the water sample pollution is quantitatively solved. Based on the spatial index of nodes, the contribution weights of nodes are mapped to the geographic information system to generate a spatial distribution map of pollution characteristic components.
[0006] Preferably, the method for constructing the multimodal in-situ spectral matrix includes: At multiple monitoring nodes in the target water area, the same water sample is simultaneously scanned using both absorption and fluorescence spectra, and the spatial location and sampling time of the corresponding nodes are recorded. Using the Raman scattering peak intensity of pure water in the water sample as the internal standard for fluorescence spectroscopy and the characteristic absorbance of the water sample in the ultraviolet region as the internal standard for absorption spectroscopy, the intensity of the scanned absorption and fluorescence spectra was uniformly calibrated. The calibrated spectral data are correlated with spatial location and sampling time to generate a multimodal in-situ spectral matrix.
[0007] Preferably, the source signal filtering of the multimodal in-situ spectral matrix includes: Independent component analysis was performed on the multimodal in-situ spectral matrix to obtain all source signal components; Calculate the autocorrelation coefficient of the response intensity of each source signal component within the continuous time-series acquisition period at the same node, and remove source signal components with autocorrelation coefficients lower than a preset threshold; Apply nonnegativity constraints to the source signal components that are retained after elimination to filter out candidate source signals.
[0008] Preferably, the step of performing feature analysis on the screened candidate source signals to obtain a contamination feature fingerprint spectrum includes: Based on the spectral library of characteristic pollutants in the watershed, characteristic pollutant components are separated from each candidate source signal, and the spectral components of each characteristic pollutant component are extracted from the candidate source signal. Peak position identification and peak shape parameter extraction are performed for each spectral component, and the extracted peak position and peak shape parameters are used to construct a pollution feature fingerprint spectrum.
[0009] Preferably, the source spectrum matrix is constructed in the following ways: Identify the types of potential pollution sources in the target watershed and collect typical water samples from each type of pollution source; For each typical water sample, absorption spectral scanning and fluorescence spectral scanning were performed simultaneously to extract its characteristic fluorescent components, spectral ratio parameters, and characteristic absorbance ratio parameters; The pollution type identifiers, characteristic fluorescent components, spectral ratio parameters, and characteristic absorbance ratio parameters of each potential pollution source are organized into a source spectrum matrix for constrained optimization reconstruction.
[0010] Preferably, the step of quantitatively determining the contribution weight of each potential pollution source to water sample pollution by minimizing reconstruction error includes: Construct a spatial continuity constraint term to associate the contribution weights of each pollution source at the current node with the contribution weights of the corresponding pollution sources at the upstream adjacent nodes; Under the constraint of spatial continuity, the multimodal in-situ spectral matrix of each node is reconstructed by constraint optimization along the direction of water flow, and the spatial continuous sequence of pollution source contribution weights of each node is solved iteratively.
[0011] Preferably, the constraint optimization reconstruction along the water flow direction under spatial continuity constraints includes: Extract the fluorescence component ratio parameters of each potential pollution source from the source spectrum matrix and construct ratio invariance constraints; Under the combined constraints of spatial continuity and ratio invariance, the multimodal in-situ spectral matrix is reconstructed by constraint optimization along the water flow direction. The reconstruction residual is then iteratively minimized to obtain the contribution weights.
[0012] Preferably, the node-based spatial index maps the contribution weights of nodes to a geographic information system to generate a spatial distribution map of pollution feature components, including: Based on the spatial index of each node, the contribution weight of each pollution source is constructed into a set of spatial discrete points in the geographic information system, resulting in discrete contribution points carrying spatial coordinates. Spatial interpolation is performed on discrete contribution points along the water flow direction to transform the contribution weights of each pollution source into a continuous spatial grid.
[0013] Preferably, after spatially interpolating the discrete contribution points along the water flow direction to convert the contribution weights of each pollution source into a continuous spatial grid, the method further includes: For multiple consecutive sampling time frames, the contribution weights of each pollution source are spatially interpolated along the water flow direction into a continuous spatial grid to obtain the continuous spatial grid sequence corresponding to each time frame; The spatial migration vectors of each pollution source contribution weight are extracted by performing grid-by-grid temporal difference on consecutive spatial grids of adjacent time frames. Based on the spatial migration vector, the predicted spatial distribution of the contribution weights of each pollution source at the next moment is deduced along the direction of water flow. The predicted spatial distribution is then fused with the current spatial distribution map of pollution characteristic components to generate a spatial distribution map of pollution characteristic components.
[0014] To address the aforementioned problems, this invention also provides a multi-parameter water quality simultaneous detection and pollution source apportionment system, the system comprising: The spectral fingerprinting module is used to filter the source signals of the multimodal in-situ spectral matrix and perform feature analysis on the filtered candidate source signals to obtain the pollution feature fingerprint spectrum. The pollution source quantitative analysis module is used to perform constrained optimization reconstruction of the pollution feature fingerprint spectrum on the source spectrum matrix. By minimizing the reconstruction error, the contribution weight of each potential pollution source to the water sample pollution is quantitatively solved. The source map generation module is used for node-based spatial indexing, mapping the contribution weights of nodes to the geographic information system to generate a spatial distribution map of pollution characteristic components.
[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention uses constrained optimization reconstruction to directly solve for the independent contribution weight of each potential pollution source from the overlapping spectral signals of multiple pollution sources. This method can quantitatively separate different pollution source components in the composite pollution signal and obtain the specific contribution value of each pollution source to the water sample pollution. This improves the source tracing conclusion from directional inference to quantitative judgment and provides a direct calculation basis for accurately identifying the pollution responsible party.
[0016] 2. By jointly introducing spatial continuity constraints and ratio invariance constraints in the constrained optimization reconstruction, the contribution weights of each node are smoothly transitioned along the water flow direction. At the same time, the inherent stability of the ratio relationship of fluorescent components of the same pollution source in spatial migration is maintained. Moreover, the dual constraints eliminate the ambiguity of weight allocation, improve the uniqueness and spatial traceability of the contribution weight solution, and enable the spatial distribution map of pollution characteristic components to continuously show the complete diffusion gradient of pollution contribution along the river channel. Attached Figure Description
[0017] Figure 1 This is a flowchart of a method for simultaneous multi-parameter water quality detection and pollution source analysis and tracing in this invention; Figure 2 This is a flowchart of the contribution weight determination method in this invention; Figure 3 This is a module structure diagram of a multi-parameter water quality synchronous detection and pollution source analysis and tracing system according to the present invention. Detailed Implementation
[0018] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0019] Implementation 1, refer to Figures 1-2 As shown in this embodiment, a method for simultaneous multi-parameter water quality detection and pollution source analysis includes: Step 1: Select source signals from the multimodal in-situ spectral matrix and perform feature analysis on the selected candidate source signals to obtain the pollution feature fingerprint spectrum.
[0020] When conducting in-situ detection of the aquatic environment, multimodal spectroscopy is easily affected by environmental noise, water background interference, and overlapping and mixed signals from multiple pollution sources. Traditional methods lack unified light intensity calibration, which can easily cause signal distortion. Source signal screening can also easily result in the accidental deletion of effective pollution signals and the retention of short-term disturbances, making it difficult to accurately extract pollution characteristics.
[0021] Based on this, the present invention provides a preferred embodiment, wherein the multimodal in-situ spectral matrix construction method includes: At multiple monitoring nodes in the target water area, the same water sample is simultaneously scanned using both absorption and fluorescence spectra, and the spatial location and sampling time of the corresponding nodes are recorded. Using the Raman scattering peak intensity of pure water in the water sample as the internal standard for fluorescence spectroscopy and the characteristic absorbance of the water sample in the ultraviolet region as the internal standard for absorption spectroscopy, the intensity of the scanned absorption and fluorescence spectra was uniformly calibrated. The calibrated spectral data are correlated with spatial location and sampling time to generate a multimodal in-situ spectral matrix.
[0022] In this embodiment of the invention, source signal filtering of the multimodal in-situ spectral matrix includes: Independent component analysis was performed on the multimodal in-situ spectral matrix to obtain all source signal components; Calculate the autocorrelation coefficient of the response intensity of each source signal component within the continuous time-series acquisition period at the same node, and remove source signal components with autocorrelation coefficients lower than a preset threshold; Apply nonnegativity constraints to the source signal components that are retained after elimination to filter out candidate source signals.
[0023] In this embodiment of the invention, feature analysis is performed on the screened candidate source signals to obtain a contamination feature fingerprint spectrum, including: Based on the spectral library of characteristic pollutants in the watershed, characteristic pollutant components are separated from each candidate source signal, and the spectral components of each characteristic pollutant component are extracted from the candidate source signal. Peak position identification and peak shape parameter extraction are performed for each spectral component, and the extracted peak position and peak shape parameters are used to construct a pollution feature fingerprint spectrum.
[0024] Specifically, multiple monitoring nodes can be set up along key sections of the main stream and tributaries in the target water area. For example, monitoring points can be set up at the upstream and downstream control sections of the watershed and at the confluence of tributaries. Water samples of the same depth can be selected at each monitoring node, and absorption spectral scanning and fluorescence spectral scanning can be carried out simultaneously. The spatial location of each node and the sampling time of the water sample can be recorded simultaneously.
[0025] Then, the Raman scattering peak intensity of pure water in the water sample was used as the internal standard for fluorescence spectroscopy, and the characteristic absorbance of the water sample in the ultraviolet region was used as the internal standard for absorption spectroscopy. The intensity of the absorption and fluorescence spectra obtained by scanning was uniformly calibrated by comparing the two standards. The calibrated spectral data and the corresponding spatial location sampling time were correlated and integrated to generate a multimodal in-situ spectral matrix.
[0026] Furthermore, the generated multimodal in-situ spectral matrix is subjected to independent component analysis and decomposition. By splitting the mixed spectral signals layer by layer, all independently existing source signal components are separated. For example, the characteristic signals corresponding to organic and inorganic pollution in water bodies are completely separated from the mixed spectrum.
[0027] It should be noted that, based on the historical signal samples accumulated from long-term continuous time-series spectral acquisition at each monitoring node in the target water area, the temporal variation differences between the stable background signal of the water body and the short-term disturbance interference signal from the outside world can be distinguished. In this way, a preset judgment threshold for the autocorrelation coefficient is defined, forming a fixed screening reference standard. The correlation degree of the response intensity of each source signal component within the continuous time-series acquisition period of the same node is calculated, thereby obtaining the response intensity autocorrelation coefficient. If the obtained autocorrelation coefficient is lower than the defined preset judgment threshold, the source signal component is removed.
[0028] Furthermore, considering the inherent characteristics of the physical response of the spectral source signal of water pollution, it can be known that the spectral response intensity corresponding to the pollutant only has a positive characterization state and will not have a negative characterization such as reverse attenuation.
[0029] Therefore, based on the aforementioned inherent physical mechanism, a non-negativity constraint screening rule for the source signal components can be pre-set. The source signal components remaining after elimination can be screened using this pre-set rule, and signals that conform to the constraint logic can be retained as candidate source signals.
[0030] Furthermore, a characteristic pollutant spectral library adapted to the watershed area is retrieved. The spectral library contains the standard spectral characteristics of various conventional pollutants in the watershed. Then, characteristic pollutant components are separated for each candidate source signal based on the spectral library, and the spectral components corresponding to each characteristic pollutant component are extracted from the candidate source signal.
[0031] Finally, peak position identification and peak shape related feature parameter extraction are performed on each of the split spectral components in sequence. All identified peak position information and extracted peak shape parameters are uniformly arranged and combined to form a pollution feature fingerprint spectrum.
[0032] In summary, this scheme simultaneously collects water sample absorption and fluorescence spectra at multiple monitoring nodes. Using the pure water Raman scattering peak and the characteristic absorbance in the ultraviolet region as dual internal standards, it completes unified light intensity calibration and correlates spatiotemporal information to construct a multimodal in-situ spectral matrix. Then, it performs independent component analysis decomposition on the matrix, eliminates low-correlation interference signals by combining the time-series autocorrelation coefficient, applies non-negativity constraints to screen out effective candidate source signals, and separates characteristic pollutant components based on the spectral library of characteristic pollutants in the watershed, extracting peak position and peak shape parameters to construct a pollution characteristic fingerprint spectrum.
[0033] As a preferred implementation method, this solution can eliminate spectral deviations caused by instruments and the environment, ensure data consistency and accuracy, accurately remove mixed pollution signals, eliminate invalid interference without losing effective features, and quickly generate exclusive pollution feature fingerprint spectra, providing accurate and stable feature support for subsequent quantitative analysis of pollution sources and spatial source tracing.
[0034] Step 2: Reconstruct the pollution feature fingerprint spectrum on the source spectrum matrix using constrained optimization. By minimizing the reconstruction error, quantitatively solve the contribution weight of each potential pollution source to the water sample pollution.
[0035] It should be noted that traditional water pollution source apportionment lacks spatial continuity constraints, and the contribution weights of pollution sources are prone to abrupt distortion along the water flow. Based on this, the present invention provides a preferred embodiment, wherein the source spectrum matrix is constructed in the following manner: Identify the types of potential pollution sources in the target watershed and collect typical water samples from each type of pollution source; For each typical water sample, absorption spectral scanning and fluorescence spectral scanning were performed simultaneously to extract its characteristic fluorescent components, spectral ratio parameters, and characteristic absorbance ratio parameters; The pollution type identifiers, characteristic fluorescent components, spectral ratio parameters, and characteristic absorbance ratio parameters of each potential pollution source are organized into a source spectrum matrix for constrained optimization reconstruction.
[0036] In this embodiment of the invention, the contribution weight of each potential pollution source to water sample pollution is quantitatively determined by minimizing the reconstruction error, including: Construct a spatial continuity constraint term to associate the contribution weights of each pollution source at the current node with the contribution weights of the corresponding pollution sources at the upstream adjacent nodes; Under the constraint of spatial continuity, the multimodal in-situ spectral matrix of each node is reconstructed by constraint optimization along the direction of water flow, and the spatial continuous sequence of pollution source contribution weights of each node is solved iteratively.
[0037] In this embodiment of the invention, under spatial continuity constraints, constraint optimization reconstruction along the water flow direction includes: Extract the fluorescence component ratio parameters of each potential pollution source from the source spectrum matrix and construct ratio invariance constraints; Under the combined constraints of spatial continuity and ratio invariance, the multimodal in-situ spectral matrix is reconstructed by constraint optimization along the water flow direction. The reconstruction residual is then iteratively minimized to obtain the contribution weights.
[0038] In practice, specific pollution source categories can be defined for plain river network basins. For example, sewage discharge from industrial parks along the river, surface runoff from contiguous farmland, and centralized discharge of urban domestic sewage can be identified as potential pollution source types for the target basin. Water samples can be taken at designated points such as the end of industrial wastewater discharge outlets, the confluence of farmland ditches, and the overflow points of domestic sewage to collect water bodies that can fully represent the background characteristics of this type of pollution discharge as typical water samples.
[0039] Then, each typical water sample collected is placed in a light-proof, static container to avoid stray light interference. Simultaneously, absorption and fluorescence spectral scanning are performed. For example, the water sample static time is uniformly set before spectral acquisition is started. Characteristic fluorescent components specific to various pollutants are separated from the continuous spectrum. The spectral ratio parameter is calculated by comparing the response amplitude of different spectral bands. Then, the characteristic absorbance ratio parameter is extracted based on the difference in absorbance intensity in the ultraviolet region.
[0040] Then, the pollution type identifier corresponding to each potential pollution source is matched with the characteristic fluorescent components, spectral ratio parameters and characteristic absorbance ratio parameters extracted from the spectrum of typical water samples, and then systematically arranged and entered according to category to form a source spectrum matrix that can fully carry the spectral characteristics of various pollution sources and is dedicated to constrained optimization and reconstruction.
[0041] Furthermore, spatial continuity constraints can be established based on the connectivity of upstream and downstream water bodies in natural river channels and the downstream migration of pollutants. For example, two adjacent monitoring nodes in the main stream can be set as related objects, and the change in the contribution weight of various pollution sources at the current node can be linked with the change in the contribution weight of the same type of pollution sources at the upstream adjacent node.
[0042] When the natural flow direction of the river is used as the order of calculation, the multimodal in-situ spectral matrix corresponding to each monitoring node is constrained and optimized and reconstructed one by one from the upstream starting section to the downstream control section. The contribution ratio of various types of pollution is continuously adjusted, and the spatial continuous sequence of the pollution source contribution weight of each node along the water area is obtained by iterative deduction and solution.
[0043] Furthermore, the inherent fluorescence component ratios of various pollution sources can be retrieved from the established source spectrum matrix. For example, the fixed fluorescence component ratios of industrial pollution and agricultural pollution can be selected as references. Based on the characteristic that the component ratios of similar pollution sources will not change with water flow migration, a ratio invariance constraint can be constructed.
[0044] Under the combined constraints of spatial continuity and ratio invariance, the multimodal in-situ spectral matrix can be reconstructed node by node along the direction of river flow. For example, the deviation between the measured spectrum and the reconstructed spectrum can be compared section by section, the weight allocation ratio can be finely adjusted, and the contribution weights of each potential pollution source can be stably obtained after iteratively narrowing the range of reconstruction residuals.
[0045] It should be noted that, as a preferred implementation method, constrained optimization reconstruction can complete the overall calculation and deduction according to the given objective function. This function integrates three error dimensions: spectral reconstruction residual, spatial weight difference, and fluorescence component ratio deviation. The smaller the overall value of the function, the higher the matching degree between the reconstructed spectral features and the actual in-situ spectrum of the water area.
[0046] The objective function formula is as follows:
[0047] in The multimodal in-situ spectral matrix generated after the current monitoring node is calibrated and spatiotemporally correlated is matched with the normalized spectral intensity characterization content of different wavelength channels or excitation and emission wavelength combinations, and the original in-situ spectral information of a single node is completely preserved.
[0048] in The number of dimensions is consistent with the number of potential pollution source types included in the source spectrum matrix. Each item in the vector corresponds to the pollution contribution weight of a type of pollution source in the water sample at the current node. All content follows the laws of spectral physical response to maintain a non-negative state, avoiding inverse numerical representations that have no practical meaning.
[0049] in It is constructed by extracting feature information from the source spectrum matrix that is suitable for the pollution background of the watershed at the current node. Each row of the matrix corresponds to the unique spectral change profile of a potential pollution source, and fully includes the characteristic fluorescence component and characteristic absorbance ratio parameter of the corresponding pollution source.
[0050] Indicates the current node The spectral reconstruction residual is used to characterize the overall difference between the original multimodal in-situ spectral matrix of the current node and the predicted spectrum reconstructed by the contribution weight and source spectral feature matrix. The smaller the difference, the more ideal the reconstruction spectrum is in terms of the restoration effect of the real water body spectrum.
[0051] The square of the Frobenius norm is calculated by squaring each element in the matrix and summing the results. This method quantifies the overall cumulative scale of various error terms and intuitively reflects the overall degree of reconstruction error.
[0052] As a penalty coefficient for the influence of spatial continuity constraints on the objective function, it is preset according to the characteristics of the annual water flow rate of the target watershed and the density of the monitoring nodes. If the river flow is rapid and the water replacement speed is fast or the monitoring nodes are densely arranged, the constraint effect of this coefficient will be weakened accordingly.
[0053] in We contribute weight vectors to pollution sources that have completed iterative convergence at adjacent monitoring nodes upstream of the water flow. In the process of solving node by node along the water flow, we prioritize the completion of the calculation and directly substitute them into the objective function of the current node to participate in error accounting.
[0054] This represents a spatial continuity constraint term, which achieves its constraint effect by calculating the overall difference between the contribution weight vectors of the current node and its upstream neighboring nodes. It can limit abrupt changes in the contribution weights of various pollution sources along the water flow direction, ensuring that the weight distribution presents a smooth transition.
[0055] As a penalty coefficient representing the weight of the constraint on the invariance of the regulation ratio in the objective function, it is preset based on the fluctuation pattern of the component proportion accumulated from long-term spectral monitoring of various pollution sources in the watershed. If the annual fluctuation range of the fluorescent component proportion of a certain type of pollution source is weak, the constraint of this coefficient will be increased accordingly.
[0056] in It is composed of the weight ratios of fluorescent components of various pollution sources in the current node contribution weight vector, which intuitively reflects the relative composition relationship between fluorescent components of different pollution sources under the current reconstruction result.
[0057] The intrinsic fixed fluorescence component ratio parameters of each pollution source are directly extracted from the source spectrum matrix and integrated into a reference set, which serves as a unified benchmark for verifying whether the component ratios of the reconstruction results conform to the inherent characteristics of the pollution sources.
[0058] This represents the ratio invariance constraint term, which measures the deviation between the current reconstructed solution fluorescence component ratio and the intrinsic ratio of the pollution source recorded in the source spectrum matrix. It can constrain the same pollution source to maintain a constant ratio of its characteristic fluorescence components throughout the entire spatial migration process in the water area.
[0059] It should be noted that when performing iterative solutions node by node along the direction of water flow, if the calculation object is the first monitoring node at the source of the watershed, then... The system is set to an unconstrained state, with the optimization objectives being only to reduce the spectral reconstruction residual and adhere to the ratio invariance constraint. If the computational object is the other monitoring nodes downstream of the first node, the contribution weight vector that has already been solved upstream is directly introduced into the objective function computation.
[0060] The contribution weight of each node is obtained by iteratively minimizing the constrained objective function using the alternating least squares algorithm. Here, the alternating least squares algorithm is carried out in a step-by-step fixed alternating optimization loop mode. It can first fix the existing feature arrangement of the source spectrum feature matrix, then iteratively update the pollution source contribution weight vector of the current node, then fix the updated weight vector, and reversely correct the matching contour of the source spectrum feature matrix. This loop is repeated multiple times until the overall calculation result tends to be stable.
[0061] In summary, this scheme predefines the convergence threshold based on the numerical fluctuation patterns of the objective function accumulated through long-term, extensive iterative calculations in the watershed, and considering the characteristic that the numerical values no longer fluctuate significantly after weight convergence. When the change in the objective function after a single iteration falls below this preset threshold, the iteration process is terminated, and the final contribution weight of the current node is output. .
[0062] In summary, this scheme first identifies the types of potential pollution sources in the watershed, collects typical water samples and performs simultaneous spectral scanning, extracts characteristic parameters to construct a dedicated source spectrum matrix; then it constructs a spatial continuity constraint term to associate the weights of upstream and downstream nodes, and simultaneously extracts the ratio parameters of the fluorescent components of pollution sources to construct ratio invariance constraints; finally, under dual constraints, it performs constraint optimization reconstruction of the spectral matrix along the water flow direction to minimize the reconstruction residual and iteratively solve for the contribution weights.
[0063] As a preferred embodiment of the present invention, the weights can be constrained to transition smoothly along the water flow, conforming to the downstream migration law of pollutants, avoiding abrupt distortion of weights, while locking in the inherent component characteristics of pollution sources, preventing reconstruction deviation, significantly reducing spectral reconstruction errors, and accurately quantifying the contribution weights of each pollution source, providing stable and reliable quantitative support for subsequent spatial mapping and source tracing of pollution.
[0064] Step 3: Based on the spatial index of nodes, map the contribution weights of nodes to the geographic information system to generate a spatial distribution map of pollution feature components.
[0065] In this embodiment of the invention, based on the spatial index of nodes, the contribution weights of nodes are mapped to a geographic information system to generate a spatial distribution map of pollution feature components, including: Based on the spatial index of each node, the contribution weight of each pollution source is constructed into a set of spatial discrete points in the geographic information system, resulting in discrete contribution points carrying spatial coordinates. Spatial interpolation is performed on discrete contribution points along the water flow direction to transform the contribution weights of each pollution source into a continuous spatial grid.
[0066] This process, which involves spatial interpolation of discrete contribution points along the water flow direction to convert the contribution weights of each pollution source into a continuous spatial grid, also includes: For multiple consecutive sampling time frames, the contribution weights of each pollution source are spatially interpolated along the water flow direction into a continuous spatial grid to obtain the continuous spatial grid sequence corresponding to each time frame; The spatial migration vectors of each pollution source contribution weight are extracted by performing grid-by-grid temporal difference on consecutive spatial grids of adjacent time frames. Based on the spatial migration vector, the predicted spatial distribution of the contribution weights of each pollution source at the next moment is deduced along the direction of water flow. The predicted spatial distribution is then fused with the current spatial distribution map of pollution characteristic components to generate a spatial distribution map of pollution characteristic components.
[0067] In practice, the spatial index information of each monitoring node is first clarified. The spatial index is associated with the actual geographical location of each node. For example, if the suburban river basin is selected as the target area, monitoring nodes are set up near the upstream source, the confluence of the tributaries in the middle reaches, and the sewage outlets in the downstream. The spatial index of each node corresponds to unique geographical latitude and longitude information. The contribution weights of the three types of pollution sources, namely industrial pollution, agricultural pollution and domestic pollution, obtained by solving each node are bound to the latitude and longitude information of the corresponding node one by one.
[0068] Specifically, the bound weights and coordinates can be entered one by one in the geographic information carrying interface to construct a spatial discrete point set, in which each discrete point carries clear spatial coordinates and the corresponding pollution source contribution weight, and finally the discrete contribution points carrying spatial coordinates are obtained.
[0069] Furthermore, spatial interpolation is performed on the constructed discrete contribution points along the natural extension direction of the river flow from upstream to downstream. For example, two adjacent discrete contribution points are selected as interpolation benchmarks, and the flow distance and weight difference between the two points are calculated. The weight values of the missing positions in the middle are gradually supplemented according to the distance ratio of the flow direction, so as to ensure that the supplemented weight values have a smooth transition with the weights of upstream and downstream nodes and avoid abrupt changes. Through this point-by-point interpolation and completion method, the originally scattered point contribution weight data is transformed into a continuous spatial grid covering the entire target watershed.
[0070] Then, a fixed sampling interval is selected to set a continuous sampling time frame, for example, every 2 hours is a sampling time frame. Water samples are collected continuously for 8 time frames for detection and analysis. The contribution weight of each pollution source in each time frame is processed according to the interpolation method along the water flow direction mentioned above. The discrete contribution points of each time frame are converted into corresponding continuous spatial grids. The continuous spatial grids of the 8 time frames are arranged in the order of sampling to obtain the continuous spatial grid sequence corresponding to each time frame.
[0071] Next, select the continuous spatial grids corresponding to two adjacent time frames in the continuous spatial grid sequence, such as the grids of the second and third time frames. For each grid cell whose position is completely corresponding in the two grids, calculate the difference in contribution weight between the grid cells corresponding to the grid cells in the later time frame and the previous time frame. Record the magnitude and direction of the weight change of each grid cell. Integrate the difference and direction information of all grid cells to extract the spatial migration vector of the contribution weight of each pollution source changing with time.
[0072] Furthermore, based on the extracted spatial migration vectors, the migration patterns of each pollution source are clarified. For example, by using the spatial migration vectors, it is discovered that the contribution weight of industrial pollution migrates downstream along the water flow direction at a fixed amplitude. Combining the current water flow velocity and migration amplitude, the distribution of the contribution weight of each pollution source at the next moment is deduced grid by grid along the water flow extension direction, thus obtaining the predicted spatial distribution at the next moment.
[0073] Then, the predicted spatial distribution is overlaid and fused with the currently generated spatial distribution map of pollution characteristic components. The real-time weight information of the current distribution map is retained, and the trend information of the predicted distribution is supplemented. Finally, a spatial distribution map of pollution characteristic components that combines real-time distribution and future trend is generated.
[0074] If the difference in grid cell weights between adjacent time frames is positive, it indicates that the contribution weight of the pollution source at that location is increasing. If the difference is negative, it indicates that the weight is decreasing. Combining this with the direction of water flow, the specific direction of weight migration can be determined, ensuring that the extraction of spatial migration vectors accurately corresponds to the actual migration state of the pollution source.
[0075] It should be noted that the interpolation process is always carried out along the direction of water flow to avoid arbitrary interpolation across the direction of water flow. For example, the interpolation between upstream and downstream nodes strictly follows the direction of water flow from upstream to downstream to ensure that the continuous spatial grid can truly reflect the distribution gradient of pollution sources along the water flow and conform to the actual laws of water pollution migration.
[0076] Specifically, when extrapolating and predicting spatial distribution, the magnitude and direction of the changes in the spatial migration vector are strictly followed, and the migration trend is not arbitrarily adjusted. For example, if the migration vector of a certain type of pollution source shows that it migrates downstream by one grid cell per hour, the distribution at the next moment is extrapolated according to this magnitude, so as to ensure that the prediction results are consistent with the actual migration patterns.
[0077] Finally, when integrating the predicted spatial distribution with the current distribution map, it is crucial to clearly distinguish between real-time data and predicted data, without confusing their representational meanings. This ensures that the generated spatial distribution map of pollution characteristic components can reflect both the current pollution distribution status and accurately present future pollution migration trends, thus meeting the practical needs of water pollution source tracing and prediction.
[0078] In summary, this scheme maps the pollution source contribution weights to a spatial discrete point set in a geographic information system based on the spatial index of monitoring nodes, performs spatial interpolation along the water flow direction to convert the discrete weights into a continuous spatial grid, generates a continuous spatial grid sequence for multiple consecutive sampling time frames, extracts the spatial migration vector of the pollution contribution weights by grid-by-grid time difference, and predicts the weight distribution for the next time step based on the vectors and merges it with the current distribution map.
[0079] This solution can transform point monitoring data into a continuous spatial distribution covering the entire watershed, strictly following the laws of water flow migration to avoid spatial distortion. At the same time, it enables spatiotemporal dynamic analysis and trend prediction of pollution, generating a spatial distribution map of pollution characteristic components that combines real-time distribution and future migration trends. This significantly improves the completeness, dynamism, and predictability of pollution spatial visualization, providing intuitive and reliable spatial data support for accurate source tracing and control decisions of watershed pollution sources.
[0080] Example 2, as Figure 3 The diagram shown is a modular structure diagram of a multi-parameter water quality synchronous detection and pollution source apportionment system provided by the present invention, which includes: The spectral fingerprinting module is used to filter the source signals of the multimodal in-situ spectral matrix and perform feature analysis on the filtered candidate source signals to obtain the pollution feature fingerprint spectrum. The pollution source quantitative analysis module is used to perform constrained optimization reconstruction of the pollution feature fingerprint spectrum on the source spectrum matrix. By minimizing the reconstruction error, the contribution weight of each potential pollution source to the water sample pollution is quantitatively solved. The source map generation module is used for node-based spatial indexing, mapping the contribution weights of nodes to the geographic information system to generate a spatial distribution map of pollution characteristic components.
[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for simultaneous multi-parameter water quality detection and pollution source apportionment, characterized in that, The method includes: Source signals of the multimodal in-situ spectral matrix are screened, and the selected candidate source signals are analyzed to obtain the pollution feature fingerprint spectrum. The pollution feature fingerprint spectrum is reconstructed by constraint optimization on the source spectrum matrix. By minimizing the reconstruction error, the contribution weight of each potential pollution source to the water sample pollution is quantitatively solved. Based on the spatial index of nodes, the contribution weights of nodes are mapped to the geographic information system to generate a spatial distribution map of pollution characteristic components.
2. The method for simultaneous multi-parameter water quality detection and pollution source analysis as described in claim 1, characterized in that, The method for constructing the multimodal in-situ spectral matrix includes: At multiple monitoring nodes in the target water area, the same water sample is simultaneously scanned using both absorption and fluorescence spectra, and the spatial location and sampling time of the corresponding nodes are recorded. Using the Raman scattering peak intensity of pure water in the water sample as the internal standard for fluorescence spectroscopy and the characteristic absorbance of the water sample in the ultraviolet region as the internal standard for absorption spectroscopy, the intensity of the scanned absorption and fluorescence spectra was uniformly calibrated. The calibrated spectral data are correlated with spatial location and sampling time to generate a multimodal in-situ spectral matrix.
3. The method for simultaneous multi-parameter water quality detection and pollution source analysis as described in claim 2, characterized in that, The source signal filtering of the multimodal in-situ spectral matrix includes: Independent component analysis was performed on the multimodal in-situ spectral matrix to obtain all source signal components; Calculate the autocorrelation coefficient of the response intensity of each source signal component within the continuous time-series acquisition period at the same node, and remove source signal components with autocorrelation coefficients lower than a preset threshold; Apply nonnegativity constraints to the source signal components that are retained after elimination to filter out candidate source signals.
4. The method for simultaneous multi-parameter water quality detection and pollution source analysis as described in claim 3, characterized in that, The step of performing feature analysis on the screened candidate source signals to obtain the contamination feature fingerprint spectrum includes: Based on the spectral library of characteristic pollutants in the watershed, characteristic pollutant components are separated from each candidate source signal, and the spectral components of each characteristic pollutant component are extracted from the candidate source signal. Peak position identification and peak shape parameter extraction are performed for each spectral component, and the extracted peak position and peak shape parameters are used to construct a pollution feature fingerprint spectrum.
5. The method for simultaneous multi-parameter water quality detection and pollution source apportionment as described in claim 1, characterized in that, The source spectrum matrix is constructed in the following ways: Identify the types of potential pollution sources in the target watershed and collect typical water samples from each type of pollution source; For each typical water sample, absorption spectral scanning and fluorescence spectral scanning were performed simultaneously to extract its characteristic fluorescent components, spectral ratio parameters, and characteristic absorbance ratio parameters; The pollution type identifiers, characteristic fluorescent components, spectral ratio parameters, and characteristic absorbance ratio parameters of each potential pollution source are organized into a source spectrum matrix for constrained optimization reconstruction.
6. The method for simultaneous multi-parameter water quality detection and pollution source apportionment as described in claim 5, characterized in that, The method of quantitatively determining the contribution weight of each potential pollution source to water sample pollution by minimizing reconstruction error includes: Construct a spatial continuity constraint term to associate the contribution weights of each pollution source at the current node with the contribution weights of the corresponding pollution sources at the upstream adjacent nodes; Under the constraint of spatial continuity, the multimodal in-situ spectral matrix of each node is reconstructed by constraint optimization along the direction of water flow, and the spatial continuous sequence of pollution source contribution weights of each node is solved iteratively.
7. The method for simultaneous multi-parameter water quality detection and pollution source apportionment as described in claim 6, characterized in that, The constraint optimization reconstruction along the water flow direction under spatial continuity constraints includes: Extract the fluorescence component ratio parameters of each potential pollution source from the source spectrum matrix and construct ratio invariance constraints; Under the combined constraints of spatial continuity and ratio invariance, the multimodal in-situ spectral matrix is reconstructed by constraint optimization along the water flow direction. The reconstruction residual is then iteratively minimized to obtain the contribution weights.
8. The method for simultaneous multi-parameter water quality detection and pollution source analysis as described in claim 1, characterized in that, The node-based spatial index maps the contribution weights of nodes to a geographic information system, generating a spatial distribution map of pollution feature components, including: Based on the spatial index of each node, the contribution weight of each pollution source is constructed into a set of spatial discrete points in the geographic information system, resulting in discrete contribution points carrying spatial coordinates. Spatial interpolation is performed on discrete contribution points along the water flow direction to transform the contribution weights of each pollution source into a continuous spatial grid.
9. The method for simultaneous multi-parameter water quality detection and pollution source analysis as described in claim 8, characterized in that, The step of spatially interpolating discrete contribution points along the water flow direction to convert the contribution weights of each pollution source into a continuous spatial grid also includes: For multiple consecutive sampling time frames, the contribution weights of each pollution source are spatially interpolated along the water flow direction into a continuous spatial grid to obtain the continuous spatial grid sequence corresponding to each time frame; The spatial migration vectors of each pollution source contribution weight are extracted by performing grid-by-grid temporal difference on consecutive spatial grids of adjacent time frames. Based on the spatial migration vector, the predicted spatial distribution of the contribution weights of each pollution source at the next moment is deduced along the direction of water flow. The predicted spatial distribution is then fused with the current spatial distribution map of pollution characteristic components to generate a spatial distribution map of pollution characteristic components.
10. A multi-parameter water quality simultaneous detection and pollution source apportionment system, used to implement the multi-parameter water quality simultaneous detection and pollution source apportionment method according to any one of claims 1-9, characterized in that, The system includes: The spectral fingerprinting module is used to filter the source signals of the multimodal in-situ spectral matrix and perform feature analysis on the filtered candidate source signals to obtain the pollution feature fingerprint spectrum. The pollution source quantitative analysis module is used to perform constrained optimization reconstruction of the pollution feature fingerprint spectrum on the source spectrum matrix. By minimizing the reconstruction error, the contribution weight of each potential pollution source to the water sample pollution is quantitatively solved. The source map generation module is used for node-based spatial indexing, mapping the contribution weights of nodes to the geographic information system to generate a spatial distribution map of pollution characteristic components.