Water pollution monitoring and processing system based on multi-dimensional data analysis

Abstract

The present application relates to the technical field of ecological environment monitoring, in particular to a water pollution monitoring and processing system based on multi-dimensional data analysis, comprising a dynamic data cube construction module, a pollution event slice extraction module, a multi-dimensional feature fingerprint coupling module, a pollution source reverse tracing module and a self-adaptive early warning triggering module. Through receiving water quality parameters of Internet of Things sensors, meteorological data of meteorological and hydrological stations and satellite remote sensing spectral data, a dynamic data cube with unified space-time coordinates and characteristic axes is constructed. Entropy change analysis and adaptive threshold method are used to identify pollution event slices. Feature coupling analysis is performed on pollution events, and pollution feature fingerprints are generated by fusing spectral fluctuation patterns, physicochemical gradients and toxicity responses. Dynamic time warping algorithm is used to realize pollution source positioning and contribution weight estimation, and based on early warning level driving, corresponding disposal equipment is started to realize intelligent closed-loop control of pollution identification, tracing and treatment.

Description

Technical Field

[0001] This invention relates to the field of ecological environment monitoring technology, and in particular to a water pollution monitoring and treatment system based on multidimensional data analysis. Background Technology

[0002] In scenarios such as watershed environmental protection, aquatic ecological restoration, and pollution emergency response, accurate and efficient automatic monitoring and intelligent treatment of water pollution have become core requirements for current water environment management. With the development of IoT sensors, remote sensing observation, and meteorological and hydrological technologies, water environment data sources are characterized by multi-source heterogeneity, spatiotemporal multi-scale, and rapid dynamic changes.

[0003] Existing technologies for pollution monitoring suffer from the following main problems: First, multi-source water environment data lacks a unified, structured integration framework, resulting in significant differences in spatiotemporal scales and low information utilization efficiency. Second, pollution identification generally relies on single or limited water quality indicators, making it difficult to characterize the comprehensive features and dynamic evolution patterns of pollution events. Third, in the pollution source tracing and response phases, there is a lack of precise source tracing methods that combine time-series behavior with spatial characteristics, as well as a lack of early warning triggering mechanisms that can automatically link and respond to pollution levels. These shortcomings limit the intelligence and automation level of water pollution monitoring systems, making it difficult to support the application requirements of precise governance and efficient scheduling. Summary of the Invention

[0004] To achieve the above objectives, this invention provides a water pollution monitoring and treatment system based on multidimensional data analysis. It provides a pollution monitoring and treatment system that integrates multi-source data, has spatial resolution capabilities and an intelligent response mechanism, so as to improve the accuracy of pollution identification and the timeliness of governance decisions.

[0005] A water pollution monitoring and treatment system based on multidimensional data analysis includes a dynamic data cube construction module, a pollution event slice extraction module, a multidimensional feature fingerprint coupling module, a pollution source reverse tracing module, and an adaptive early warning triggering module.

[0006] The dynamic data cube construction module receives water quality parameters, meteorological data from meteorological and hydrological stations, and satellite remote sensing spectral data uploaded in real time by the IoT sensor array deployed in the monitored water area, and generates a dynamic data cube by aligning them with spatiotemporal coordinates.

[0007] The pollution event slice extraction module performs a spatiotemporal sliding window scan on the dynamic data cube. If the entropy change value of the water quality parameter in the window exceeds the adaptive threshold, the data in that window is extracted as a pollution event slice.

[0008] The multidimensional feature fingerprint coupling module inputs the pollution event slices into the feature coupling engine, and simultaneously extracts the spectral reflectance fluctuation pattern, physicochemical parameter gradient distribution and biotoxicity response curve, and fuses them to generate pollution feature fingerprints.

[0009] The pollution source reverse tracing module performs dynamic time-normalization matching between the pollution feature fingerprint and the pollution source feature database, and outputs the pollution source location coordinates and pollution contribution weight.

[0010] The adaptive early warning triggering module generates graded early warning commands based on the pollution source location coordinates and pollution contribution weight, driving the aeration equipment or adsorption device in the target water area to start.

[0011] Optionally, the dynamic data cube building module includes:

[0012] Data standardization processing: Receive water quality parameters, meteorological data from meteorological and hydrological stations, and satellite remote sensing spectral data uploaded in real time by IoT sensor arrays deployed in the monitored water area. Normalize the three types of data respectively and output standardized water quality parameters, standardized meteorological data, and standardized satellite remote sensing spectral data.

[0013] Time dimension alignment: Based on a unified timestamp, the standardized water quality parameters, standardized meteorological data and standardized satellite remote sensing spectral data are resampled to generate time-aligned water quality parameters, time-aligned meteorological data and time-aligned satellite remote sensing spectral data with consistent time resolution;

[0014] Spatial dimension alignment: Using the spatial interpolation algorithm of the geographic information system, the time-aligned water quality parameters, time-aligned meteorological data and time-aligned satellite remote sensing spectral data are mapped to a unified grid coordinate system to generate spatially aligned water quality parameters, spatially aligned meteorological data and spatially aligned satellite remote sensing spectral data with spatial coordinate matching;

[0015] Three-dimensional data integration: The spatially aligned water quality parameters, meteorological data, and satellite remote sensing spectral data are integrated along the time axis, horizontal spatial axis, and data feature axis to generate a dynamic data cube containing spatiotemporal attributes and multidimensional features. Its data structure satisfies the following:

[0016] Each grid cell stores a triplet of longitude, latitude, and timestamp;

[0017] The data feature axis integrates subsets of water quality parameters, meteorological data, and spectral data.

[0018] Optionally, the pollution event slice extraction module includes:

[0019] Spatiotemporal sliding window initialization: Set the initial size of the spatiotemporal sliding window, and perform a sliding scan along the time axis and horizontal spatial axis of the dynamic data cube. Each slide extracts the window dataset within the current window.

[0020] Water quality parameter entropy change calculation: Perform multivariate information entropy analysis on a subset of water quality parameters in each window dataset, calculate the entropy difference between the current window and the previous time window, and output the water quality parameter entropy change value;

[0021] Adaptive threshold generation: Based on the fluctuation range of water quality parameter entropy values ​​in the dynamic data cube of the same region over the past 30 days, the adaptive threshold is dynamically calculated using the quantile statistical method.

[0022] Pollution event slice determination and extraction: If the entropy change value of the water quality parameter exceeds the adaptive threshold corresponding to the current window, the slice extraction mechanism will be triggered to completely capture the spatiotemporal coordinate range and all data features of the dataset in the window and generate a pollution event slice.

[0023] If the threshold is not exceeded, the sliding window moves to the next position and repeats the spatiotemporal sliding window initialization, water quality parameter entropy change calculation, and adaptive threshold generation.

[0024] Optionally, the multidimensional feature fingerprint coupling module includes:

[0025] Pollution event slice analysis: Receive the pollution event slice generated by the pollution event slice extraction module, input it into the feature coupling engine, and separate the satellite remote sensing spectral data and water quality parameters from it;

[0026] Spectral reflectance fluctuation pattern extraction: In the feature coupling engine, continuous wavelet transform is performed on satellite remote sensing spectral data to extract frequency band energy distribution features and generate spectral reflectance fluctuation patterns;

[0027] Physicochemical parameter gradient distribution extraction: In the feature coupling engine, spatial kriging interpolation is performed on the water quality parameters, and the rate of change vector field of each parameter in three-dimensional space is calculated to generate the physicochemical parameter gradient distribution.

[0028] Optionally, the multidimensional feature fingerprint coupling module further includes:

[0029] Biotoxicity response curve generation: In the feature coupling engine, based on the concentration data of toxic substances in the water quality parameters, the biological inhibition rate is calculated through the dose-response function to generate a biotoxicity response curve;

[0030] Feature fusion and fingerprint generation: The spectral reflectance fluctuation mode, physicochemical parameter gradient distribution and biotoxicity response curve are input into the fusion unit of the feature coupling engine. The feature dimensionality reduction and fusion are performed by principal component analysis to output the pollution feature fingerprint.

[0031] Optionally, the pollution source reverse tracing module includes:

[0032] Contamination feature fingerprint reception: Obtain the contamination feature fingerprint output by the multidimensional feature fingerprint coupling module and convert its format into a time series matrix;

[0033] Pollution source feature database query: Access the pre-built pollution source feature database, retrieve the feature fingerprint time series of all potential pollution sources, and generate a candidate pollution source feature set;

[0034] Dynamic time warping matching execution: Perform dynamic time warping matching on the pollution feature fingerprint and each feature sequence in the candidate pollution source feature set, calculate the optimal curved path distance between sequences, and output the matching distance matrix.

[0035] Optionally, the pollution source reverse tracing module further includes:

[0036] Determining the location coordinates of pollution sources: Select the K candidate pollution sources with the smallest distance values ​​in the matching distance matrix, and calculate the location coordinates of the pollution sources based on their weighted average geographical coordinates, with the weights being inversely proportional to the matching distances;

[0037] Pollution contribution weight calculation: Based on the distance values ​​of the matching distance matrix, the pollution contribution weight of each pollution source to the current pollution event is generated by normalization using the softmax function.

[0038] Optionally, the adaptive early warning triggering module includes:

[0039] Pollution source data reception: Obtain the pollution source location coordinates and pollution contribution weights output by the pollution source reverse tracing module;

[0040] Warning level determination: Query the preset warning level mapping table based on the pollution contribution weight:

[0041] If the pollution contribution weight is greater than 0.6, a red-level warning instruction will be generated;

[0042] If 0.3 ≤ pollution contribution weight ≤ 0.6, an orange-level warning instruction will be generated;

[0043] If the pollution contribution weight is less than 0.3, a yellow-level warning instruction will be generated.

[0044] Optionally, the adaptive early warning triggering module further includes:

[0045] Equipment matching: Based on the pollution source location coordinates and the graded early warning command level, equipment matching is performed.

[0046] Red Order: Activate all aeration equipment and adsorption devices in the target water area;

[0047] Orange command: Activate adsorption devices within a 500-meter radius of the pollution source's location coordinates;

[0048] Yellow instruction: Activate only the aeration equipment;

[0049] Equipment drive signal generation: Convert equipment activation commands into IoT control protocol signals, wherein the aeration equipment receives oxygenation intensity parameters and the adsorption device receives adsorbent dosage parameters;

[0050] Emergency equipment activation: The drive signal is transmitted to the aeration equipment and / or adsorption device in the target area through the water IoT, triggering them to start operation according to the command parameters.

[0051] The beneficial effects of this invention are:

[0052] This invention, through a dynamic data cube construction module, integrates water quality parameters from IoT sensors, meteorological and hydrological station data, and satellite remote sensing spectral data. It employs steps such as data standardization, temporal dimension alignment, spatial dimension alignment, and three-dimensional data integration to construct a dynamic data cube with unified latitude and longitude indexes, timestamp indexes, and feature dimension indexes. This structure enables structured representation of cross-source data, overcoming the problems of inconsistent spatiotemporal scales in data acquisition and difficulties in data fusion in existing technologies, thus laying a high spatiotemporal resolution data foundation for pollution identification and tracking.

[0053] This invention introduces dynamically changing information entropy values ​​as a criterion for pollution outbreaks through a pollution event slice extraction module, and combines this with an adaptive threshold mechanism to achieve precise spatiotemporal localization of pollution events. Subsequently, a multi-dimensional feature fingerprint coupling module performs continuous wavelet transform, Kriging interpolation, and toxicity response modeling on the pollution event slices, extracting and fusing three core features: spectral reflectance fluctuation patterns, physicochemical parameter gradient distributions, and biotoxicity response curves. This ultimately constructs a pollution feature fingerprint, enabling in-depth analysis and traceable modeling of pollution events. This feature fusion method overcomes the problem of excessive reliance on single water quality indicators in traditional methods, improving the accuracy and stability of pollution event identification.

[0054] This invention utilizes a dynamic time warping algorithm to perform time-series matching of pollution feature fingerprints with historical samples in a pollution source feature database. It calculates the optimal curved path distance and outputs the pollution source location coordinates and pollution contribution weight accordingly, significantly improving the timeliness and accuracy of source tracing. Based on this, an adaptive early warning triggering module combines the pollution contribution weight and location coordinates to automatically determine the early warning level. It then matches aeration equipment and adsorption devices in the corresponding water area based on a tiered strategy. Through IoT control signals, it precisely drives the emergency response device, achieving fully automated closed-loop control of water pollution from identification and diagnosis to treatment, greatly improving environmental response capabilities and pollution control efficiency. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a schematic diagram of the system flow according to an embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of the pollution event slice extraction module in an embodiment of the present invention. Detailed Implementation

[0058] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0059] It should be noted that the use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when a specific feature, structure, or characteristic is described in connection with an embodiment, implementing such a feature, structure, or characteristic in conjunction with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the art.

[0060] Generally, terms can be understood at least partly from their use in context. For example, depending at least partly on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood not necessarily to convey an exclusive set of factors, but rather, alternatively, depending at least partly on the context, to allow for the presence of other factors that are not necessarily explicitly described.

[0061] like Figures 1-2 As shown, the water pollution monitoring and treatment system based on multidimensional data analysis includes a dynamic data cube construction module, a pollution event slice extraction module, a multidimensional feature fingerprint coupling module, a pollution source reverse tracing module, and an adaptive early warning triggering module, wherein;

[0062] The dynamic data cube construction module receives water quality parameters, meteorological data from meteorological and hydrological stations, and satellite remote sensing spectral data uploaded in real time by the IoT sensor array deployed in the monitored water area, and generates a dynamic data cube by aligning it with spatiotemporal coordinates.

[0063] The pollution event slice extraction module performs a spatiotemporal sliding window scan on the dynamic data cube. If the entropy change value of the water quality parameter in the window exceeds the adaptive threshold, the data in that window is extracted as a pollution event slice.

[0064] The multidimensional feature fingerprint coupling module inputs pollution event slices into the feature coupling engine, simultaneously extracts spectral reflectance fluctuation patterns, physicochemical parameter gradient distributions, and biotoxicity response curves, and fuses them to generate pollution feature fingerprints.

[0065] The pollution source reverse tracing module performs dynamic time-warped matching between pollution feature fingerprints and pollution source feature database, and outputs pollution source location coordinates and pollution contribution weights.

[0066] The adaptive early warning triggering module generates graded early warning commands based on the pollution source location coordinates and pollution contribution weight, driving the aeration equipment or adsorption device in the target water area to start.

[0067] The dynamic data cube building blocks include:

[0068] Data standardization processing: We accept data from the following three sources:

[0069] The IoT sensor array deployed in the monitored water area uploads water quality parameters in real time, including dissolved oxygen, pH value, conductivity, ammonia nitrogen concentration and heavy metal ion concentration (such as copper, lead and mercury).

[0070] Meteorological data from meteorological and hydrological stations, including air temperature, wind speed, air pressure, and precipitation;

[0071] Satellite remote sensing spectral data acquired by the satellite remote sensing system includes red band reflectance, green band reflectance, near-infrared band reflectance, and reflectance time series curves.

[0072] Normalization processing is performed on the characteristics of data from different sources. Let the original observation data be... Where i represents the data type (water quality, meteorology, or remote sensing), j represents the j-th feature, and t represents the time index. Normalization is performed using the Min-Max standardization formula:

[0073]

[0074] The system outputs three types of standardized datasets: standardized water quality parameters, standardized meteorological data, and standardized satellite remote sensing spectral data. This eliminates dimensional differences and facilitates subsequent spatiotemporal fusion.

[0075] Time dimension alignment: To ensure that different data sources have a unified time resolution, the resampled data is resampled based on the unified timestamp set {T k}, and the linear interpolation method is used to complete the missing time points. Let the normalized data be X i,j,norm (t), then the value after time alignment is interpolated by the following formula:

[0076]

[0077] where t1 < T k < t2, and t1, t2 ∈ the original time point set.

[0078] The output results are: time-aligned water quality parameters, time-aligned meteorological data, and time-aligned satellite remote sensing spectral data.

[0079] Spatial dimension alignment: Through the spatial interpolation algorithm of the geographic information system, such as the inverse distance weighted interpolation (IDW) method, various time-aligned data are projected onto a unified grid coordinate system. Let the center point of the target grid cell be (x, y), and the N surrounding observation points be {(x k , y k}}, and their corresponding values be {v k}, then the interpolation result is:

[0080]

[0081] where p is the weight decay exponent, and the value is 2.

[0082] After interpolation, we get: spatially aligned water quality parameters, spatially aligned meteorological data, and spatially aligned satellite remote sensing spectral data, which have consistent geographical resolution and spatial structure.

[0083] Three-dimensional data integration: After completing the spatio-temporal alignment, the three types of aligned data are fused based on a unified time axis, horizontal spatial grid axis, and feature data axis. The data structure of the dynamic data cube is defined as follows:

[0084] Time axis: Represented by the unified timestamp sequence {T k}

[0085] Spatial axis: Represented by the longitude and latitude pairs (x i , y j )

[0086] Feature axis: Includes subsets of water quality parameters (such as dissolved oxygen, pH, conductivity), subsets of meteorological data (such as temperature, precipitation, wind speed), and subsets of spectral data (such as reflectance in red, green, near-infrared bands, etc.);

[0087] Each dynamic data cube cell is defined as a five-tuple:

[0088]

[0089] Where W is a subset of water quality parameters, M is a subset of meteorological data, and R is a subset of spectral data.

[0090] Example Explanation: Taking a monitoring grid cell (x = 121.45°E, y = 31.23°N) at timestamp T = 2025-06-18 10:00 as an example, its dynamic data cube cell is stored as follows:

[0091] Water quality parameter subset: W = [6.5 (mg / L), 7.2 (pH), 320 (μS / cm)];

[0092] Meteorological data subset: M = [28.0°C, 3.5 (m / s), 1012 (hPa)];

[0093] Spectral data subset: R = [0.12, 0.14, 0.08] (reflectance in red, green, and near-infrared bands, respectively);

[0094] The pollution event slice extraction module includes:

[0095] Spatiotemporal sliding window initialization: On the dynamic data cube, set the size to W. t ×W x ×W y A three-dimensional spatiotemporal sliding window is used to represent the time dimension step (e.g., 6 hours) and the horizontal spatial dimension step (e.g., 5×5 grid cells). The sliding method is as follows:

[0096] Slide along the time axis with a sliding step size Δt

[0097] Slide along the spatial axis with step sizes Δx, Δy;

[0098] Each slide extracts data from the current spatiotemporal region within the window, forming a window dataset D. (w) .

[0099] The window dataset includes a subset of water quality parameters corresponding to all time points and grid cells within the window.

[0100] Water quality parameter entropy change calculation: Multivariate information entropy analysis is performed on a subset of water quality parameters in each window dataset to measure the complexity of water quality changes in the current window. Let the observed values ​​of m water quality parameters in the window form a joint distribution P = {p1, p2, ..., p...}. n}, and its joint information entropy is:

[0101]

[0102] Then calculate the entropy change ΔH between the current window and the previous time window. (w) :

[0103] ΔH (w) =|H (w) -H (w-1 )|;

[0104] Among them, H (w) H is the joint entropy of the current window dataset. (w-1) Let ΔH be the joint entropy of the previous window dataset. (w) The entropy change value represents the degree of change in water quality parameters.

[0105] Example: If the joint probability distribution of observed pH, conductivity, and dissolved oxygen parameters in window w is estimated to be {0.4, 0.3, 0.2, 0.1}, then:

[0106] H (w) =-(0.4log0.4+0.3log0.3+0.2log0.2+0.1log0.1)≈1.846;

[0107] Adaptive threshold generation: The adaptive threshold θ for the current window is calculated from the entropy value sequence of the same spatial region within the past 30 days using quantile statistics. (w) Let the historical entropy sequence be {ΔH}. (1) ,ΔH (2) ,...,ΔH (K)},but:

[0108]

[0109] Among them, Q p This represents the p-th quantile (e.g., the 90th percentile), used to dynamically reflect the background level of water quality disturbance in the region.

[0110] Example: If the 90th percentile of the entropy change sequence of a certain region over the past 30 days is 0.42, then the current window threshold is θ. (w) =0.42.

[0111] Pollution event slice determination and extraction: If the entropy change value ΔH of the water quality parameter in the current window is... (w) The following conditions must be met:

[0112] ΔH (w) >θ (w) ;

[0113] This means that a contamination disturbance event has occurred in the current window. At this point, the slice extraction mechanism is triggered to extract all data features of the current window from the dynamic data cube, including:

[0114] Time range: The time period covered by the current window;

[0115] Spatial extent: The grid area covered by the current window;

[0116] Data content: subsets of water quality parameters, meteorological data, and spectral data from all grids in this region.

[0117] The output is a slice of pollution events, with a structure consistent with the original dynamic data cube, possessing complete spatiotemporal and multidimensional feature attributes, which facilitates the use of the subsequent feature fingerprint coupling module.

[0118] Circular mechanism: If ΔH (w) ≤θ (w) Then the window slides along the time axis or space axis to the next position, and the spatiotemporal sliding window initialization, water quality parameter entropy change value calculation, and adaptive threshold generation are re-executed until the entire time and space region of the dynamic data cube is covered.

[0119] The multidimensional feature fingerprint coupling module includes:

[0120] Contamination Event Slice Parsing: Receives contamination event slices generated by the contamination event slice extraction module and inputs them into the feature coupling engine. During parsing, the data feature axes in the data structure are split as follows:

[0121] Satellite remote sensing spectral data corresponding to each spatiotemporal grid in the pollution event slice were extracted to form a spectral data matrix. Where n is the number of grid points and b is the number of spectral bands;

[0122] Extract all physicochemical indicators and toxic substance concentrations from the subset of water quality parameters to construct the water quality parameter tensor. Where m represents the number of water quality parameters.

[0123] Spectral reflectance fluctuation pattern extraction: satellite remote sensing spectral data vector s for each spatial grid point i =[s i,1 ,s i,2 ,...,s i,b Continuous wavelet transform is performed to extract the distribution characteristics of its spectral energy at different scales. The formula for continuous wavelet transform is expressed as:

[0124]

[0125] Where ψ is the mother wavelet function (such as the Morlet wavelet), a is the scaling parameter, τ is the translation parameter, and * is the complex conjugate.

[0126] Extracting the energy density E(a) at different scales = ∫|W(a,τ)| 2 dτ, and combined into a frequency band energy vector, constitute the spectral reflectance fluctuation mode: F spec=[E(a1),E(a2),...,E(a k )];

[0127] Example: After continuous wavelet transform, the energies of a spectral sample at scales a = 2, 4, and 8 are [0.12, 0.31, 0.48], respectively. Then, its spectral reflectance fluctuation mode is F. spec =[0.12,0.31,0.48].

[0128] Physicochemical parameter gradient distribution extraction: Spatial kriging interpolation is performed on the water quality parameters in the pollution event slice to generate a continuous distribution field of each physicochemical parameter on a spatial grid. Let the predicted value of the j-th water quality parameter at spatial coordinates (x, y) be:

[0129] Among them, Z j (x i ,y i Let be the observed value of the j-th parameter at the i-th sample point. The weights are obtained from the inversion of the variogram model and covariance matrix, and N is the number of sample points involved in the interpolation.

[0130] Then, the spatial gradient vector field for each parameter is calculated:

[0131] The gradient magnitudes of all parameters are integrated to form the gradient distribution vector of the physicochemical parameters. Each dimension represents the spatial variation intensity of a physicochemical index.

[0132] Biotoxicity response curve generation: Based on the concentration data of toxic substances in water quality parameters, such as ammonia nitrogen and heavy metal ions, biotoxicity response curves are constructed using standard dose-response relationships. The relationship between the organism's inhibition rate I and the toxicant concentration C is modeled using a logistic dose-response function.

[0133]

[0134] Where I(C) represents the biological inhibition rate at a toxicant concentration C, I max For the maximum inhibition rate (set to 100%), EC 50 The concentration at which a 50% inhibitory effect is produced, and n is the slope factor of the dose-response curve.

[0135] Using the above model, the toxicological effects of toxic substances at various points within a pollution event slice are estimated, resulting in a time-series biotoxicity response curve F. tox =[I(C1),I(C2),...,I(C T )).

[0136] Feature fusion and fingerprint generation: combining the above three contamination features:

[0137] Spectral reflectance fluctuation mode F spec ∈R k ;

[0138] Physicochemical parameter gradient distribution F chem ∈R m ;

[0139] Biotoxicity response curve F tox ∈R T ;

[0140] The input is fed into the fusion unit of the feature coupling engine, where principal component analysis (PCA) is performed for dimensionality reduction and fusion to generate a contamination feature fingerprint F of uniform length. finger ∈R d (e.g., d=20):

[0141] F finger =P·[F spec ;F chem ;F tox ];

[0142] Where P is the PCA mapping matrix trained by the feature coupling engine, and the concatenation symbol ; indicates vector concatenation.

[0143] Example: If the total dimension of the input features is 80, after PCA mapping, it is compressed to 20 dimensions, and the output fingerprint is in the form of: F finger = [0.12, 0.78, 0.43, ..., 0.05];

[0144] The output pollution feature fingerprint will serve as the input feature template for the subsequent pollution source reverse tracing module.

[0145] The pollution source reverse tracing module includes:

[0146] Contamination Feature Fingerprint Reception: First, acquire the contamination feature fingerprint F corresponding to the contamination event. finger ∈R d This vector contains dimension-reduced fusion features from spectral reflectance fluctuation patterns, physicochemical parameter gradient distributions, and biotoxicity response curves. To adapt to dynamic matching requirements, it is transformed into a time-series matrix T. pollute ∈R T×d Where T is the number of time steps within a slice, and d is the fusion dimension. If the original fingerprint is a single-time representation, a short-time sequence is generated by expanding the time window (e.g., by moving average or time series stacking).

[0147] Pollution Source Feature Database Query: Access the system's pre-built pollution source feature database, which contains historically labeled pollution sources and their corresponding feature fingerprint time series sets. Assume the database contains N pollution source samples, each represented as:

[0148] Where T i This represents the time length of the record for the i-th pollution source, where d is consistent with the dimension of the pollution feature fingerprint. A candidate set is formed:

[0149] Dynamic time warping matching execution: for contaminated feature fingerprint sequence T pollute With each candidate pollution source sequence Perform Dynamic Time Warping (DTW) matching to calculate the optimal bending path distance.

[0150] The DTW distance calculation formula is:

[0151]

[0152] in, Let p be the time step vector of the fingerprint sequence of the pollution event. Let be the time step vector of the i-th pollution source sequence, W be the set of matching paths, and ∥·∥2 be the Euclidean distance. The matching results of all candidate pollution sources constitute the matching distance matrix:

[0153]

[0154] Example: If the fingerprint length of a contamination event is 5 steps, and the fingerprint length of a contamination source is 7 steps, and the DTW matching distance on the path W = {(1,1),(2,2),...,(5,6)} is 1.82, then the matching result for the contamination source is:

[0155] Pollution source location coordinates determination: After matching is completed, the K pollution sources with the smallest distance values ​​are selected from the matching distance matrix as the most likely candidates, and their coordinates are set as (x... i ,y i The corresponding matching distance is D. i Then the pollution source location coordinates (x * ,y * The result is obtained by the weighted average method:

[0156]

[0157] Among them, w i To match the inverse weights, ∈ is a regularization term to prevent division by zero, and is set to 10. -6 .

[0158] Pollution contribution weight calculation: To quantify the impact of each candidate pollution source on the current pollution event, softmax normalization is performed on the matching distance, mapping it to a pollution contribution weight vector P = [p1, p2, ..., p N ];

[0159]

[0160] This weight reflects the degree to which the i-th pollution source is attributable to the pollution event, and is used as a reference for the decision-making system for scheduling.

[0161] Example: If the matching distance between three candidate sources is [1.0, 0.5, 2.0], then their pollution contribution weight is:

[0162]

[0163] The final output includes the location coordinates of the pollution source and the pollution contribution weight, which are then used by the adaptive early warning trigger module to coordinate and execute the governance response.

[0164] The adaptive early warning triggering module includes:

[0165] To determine the warning level, the pollution contribution weight vector output by the pollution source reverse tracing module is first analyzed. The pollution contribution weight vector is: P = [p 1， p2, ..., p N ], where p i Let represent the pollution contribution weight of the i-th candidate pollution source to the current pollution event, and satisfy the normalization condition. To facilitate subsequent tiered response decisions, the weight corresponding to the maximum pollution contribution is selected as the pollution intensity of the dominant pollution source, defined as: p * =max(P);

[0166] The pollution contribution weight p * As the basis for determining the pollution level, it is entered into the preset warning level mapping rule table for level classification and response equipment scheduling.

[0167] Pollution source data reception: Receives two types of output data from the pollution source reverse tracing module:

[0168] The coordinates of the pollution source location are denoted as (x * ,y * (), indicating the geographical location of the pollution source;

[0169] Pollution contribution weight: denoted as p * This represents the contribution of the main pollution sources to the current pollution incident.

[0170] This data serves as input for subsequent early warning level determination and spatial scheduling calculations.

[0171] Warning level determination: based on pollution contribution weight p * The system uses a lookup table to match the preset warning level mapping table and generates the corresponding graded warning instruction L. The specific rules are as follows:

[0172]

[0173] Example: If the matching contribution weight of a certain pollution source is p * =0.72, then the corresponding warning level is red-level warning instruction.

[0174] The output graded early warning command will be transmitted to the device matching unit;

[0175] Equipment matching: Based on the generated graded early warning instruction L and the pollution source location coordinates (x... * ,y * ), and execute the corresponding disposal equipment matching strategy.

[0176] Red-level early warning instruction: Response across the entire area, activating all aeration equipment and adsorption devices within the target water area;

[0177] Orange-level warning command: Local response, activated with (x * ,y * All adsorption devices within a radius of 500 meters centered on a circle;

[0178] Yellow-level warning instruction: Mild response, only activate all aeration equipment in the target water area, do not start the adsorption device.

[0179] Let the aeration equipment be... The adsorption device is assembled as follows Its coordinates are (x j ,y j Then, under an orange alert, the following conditions must be met:

[0180]

[0181] Device drive signal generation: Based on the determined set of processing devices, corresponding control signals are generated and converted into IoT control protocol signals. Control parameters are set as follows:

[0182] The aeration equipment receives the oxygenation intensity parameter α∈[0,1], whose default value is assigned according to the pollution level:

[0183]

[0184] The adsorption device receives the adsorbent dosage parameter β, which is linearly set according to the pollution contribution weight: β = β max ·p * , where β max This represents the maximum deployment capacity of the equipment.

[0185] The control commands are encapsulated into a communication message structure as follows:

[0186] Control device number;

[0187] Action type (start / stop);

[0188] Control parameters (α, β);

[0189] Timestamps and synchronization signals;

[0190] Target location coordinates;

[0191] Example: If aeration equipment A5 is activated under a red alert, its control signal is:

[0192] {Equipment No.: A5, Type: Start-up, Oxygenation Intensity: 1.0, Timestamp: 10:00, Coordinates: (121.45, 31.23)};

[0193] Emergency equipment activation: Control signals are transmitted to the target water area edge node via aquatic IoT protocols (such as LoRaWAN, NB-IoT, or 5G edge communication protocols) to drive the target equipment to start operation. Equipment activation actions include:

[0194] The aeration equipment adjusts the air volume and aeration intensity according to parameter α;

[0195] The adsorption device controls the adsorbent release motor and distribution valve according to parameter β;

[0196] The startup status and operating indicators of all equipment will be fed back to the central control system, forming a closed-loop response and scheduling mechanism.

[0197] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0198] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A water pollution monitoring and treatment system based on multidimensional data analysis, characterized in that, It includes a dynamic data cube construction module, a pollution event slice extraction module, a multi-dimensional feature fingerprint coupling module, a pollution source reverse tracing module, and an adaptive early warning triggering module, among which; The dynamic data cube construction module receives water quality parameters, meteorological data from meteorological and hydrological stations, and satellite remote sensing spectral data uploaded in real time by the IoT sensor array deployed in the monitored water area, and generates a dynamic data cube by aligning them with spatiotemporal coordinates. The pollution event slice extraction module performs a spatiotemporal sliding window scan on the dynamic data cube. If the entropy change value of the water quality parameter within the window exceeds an adaptive threshold, the data of that window is extracted as a pollution event slice. The calculation of the entropy change value of the water quality parameter includes performing multivariate information entropy analysis on a subset of water quality parameters in each window dataset to measure the complexity change of the water quality in the current window. It is assumed that the observed values ​​of the water quality parameters in the window constitute a joint distribution. Its joint information entropy is: ； Then calculate the entropy change between the current window and the previous time window. : ； in, The joint entropy of the current window dataset. The joint entropy of the previous window dataset, The entropy change value represents the degree of change in water quality parameters; The multidimensional feature fingerprint coupling module inputs the pollution event slices into the feature coupling engine, and simultaneously extracts the spectral reflectance fluctuation pattern, physicochemical parameter gradient distribution and biotoxicity response curve, and fuses them to generate pollution feature fingerprints. The pollution source reverse tracing module performs dynamic time-normalization matching between the pollution feature fingerprint and the pollution source feature database, and outputs the pollution source location coordinates and pollution contribution weight. The adaptive early warning triggering module generates graded early warning commands based on the pollution source location coordinates and pollution contribution weight, driving the aeration equipment or adsorption device in the target water area to start.

2. The water pollution monitoring and treatment system based on multidimensional data analysis according to claim 1, characterized in that, The dynamic data cube construction module includes: Data standardization processing: Receive water quality parameters, meteorological data from meteorological and hydrological stations, and satellite remote sensing spectral data uploaded in real time by IoT sensor arrays deployed in the monitored water area. Normalize the three types of data respectively and output standardized water quality parameters, standardized meteorological data, and standardized satellite remote sensing spectral data. Time dimension alignment: Based on a unified timestamp, the standardized water quality parameters, standardized meteorological data and standardized satellite remote sensing spectral data are resampled to generate time-aligned water quality parameters, time-aligned meteorological data and time-aligned satellite remote sensing spectral data with consistent time resolution; Spatial dimension alignment: Using the spatial interpolation algorithm of the geographic information system, the time-aligned water quality parameters, time-aligned meteorological data and time-aligned satellite remote sensing spectral data are mapped to a unified grid coordinate system to generate spatially aligned water quality parameters, spatially aligned meteorological data and spatially aligned satellite remote sensing spectral data with spatial coordinate matching; Three-dimensional data integration: The spatially aligned water quality parameters, meteorological data, and satellite remote sensing spectral data are integrated along the time axis, horizontal spatial axis, and data feature axis to generate a dynamic data cube containing spatiotemporal attributes and multidimensional features. Its data structure satisfies the following: Each grid cell stores a triplet of longitude, latitude, and timestamp; The data feature axis integrates subsets of water quality parameters, meteorological data, and spectral data.

3. The water pollution monitoring and treatment system based on multidimensional data analysis according to claim 2, characterized in that, The pollution event slice extraction module includes: Spatiotemporal sliding window initialization: Set the initial size of the spatiotemporal sliding window, and perform a sliding scan along the time axis and horizontal spatial axis of the dynamic data cube. Each slide extracts the window dataset within the current window. Water quality parameter entropy change calculation: Perform multivariate information entropy analysis on a subset of water quality parameters in each window dataset, calculate the entropy difference between the current window and the previous time window, and output the water quality parameter entropy change value; Adaptive threshold generation: Based on the fluctuation range of water quality parameter entropy values ​​in the dynamic data cube of the same region over the past 30 days, the adaptive threshold is dynamically calculated using the quantile statistical method. Pollution event slice determination and extraction: If the entropy change value of the water quality parameter exceeds the adaptive threshold corresponding to the current window, the slice extraction mechanism will be triggered to completely capture the spatiotemporal coordinate range and all data features of the dataset in the window and generate a pollution event slice. If the threshold is not exceeded, the sliding window moves to the next position and repeats the spatiotemporal sliding window initialization, water quality parameter entropy change calculation, and adaptive threshold generation.

4. The water pollution monitoring and treatment system based on multidimensional data analysis according to claim 3, characterized in that, The multidimensional feature fingerprint coupling module includes: Pollution event slice analysis: Receive the pollution event slice generated by the pollution event slice extraction module, input it into the feature coupling engine, and separate the satellite remote sensing spectral data and water quality parameters from it; Spectral reflectance fluctuation pattern extraction: In the feature coupling engine, continuous wavelet transform is performed on satellite remote sensing spectral data to extract frequency band energy distribution features and generate spectral reflectance fluctuation patterns; Physicochemical parameter gradient distribution extraction: In the feature coupling engine, spatial kriging interpolation is performed on the water quality parameters, and the rate of change vector field of each parameter in three-dimensional space is calculated to generate the physicochemical parameter gradient distribution.

5. The water pollution monitoring and treatment system based on multidimensional data analysis according to claim 4, characterized in that, The multidimensional feature fingerprint coupling module also includes: Biotoxicity response curve generation: In the feature coupling engine, based on the concentration data of toxic substances in the water quality parameters, the biological inhibition rate is calculated through the dose-response function to generate a biotoxicity response curve; Feature fusion and fingerprint generation: The spectral reflectance fluctuation mode, physicochemical parameter gradient distribution and biotoxicity response curve are input into the fusion unit of the feature coupling engine. The feature dimensionality reduction and fusion are performed by principal component analysis to output the pollution feature fingerprint.

6. The water pollution monitoring and treatment system based on multidimensional data analysis according to claim 5, characterized in that, The pollution source reverse tracing module includes: Contamination feature fingerprint reception: Obtain the contamination feature fingerprint output by the multidimensional feature fingerprint coupling module and convert its format into a time series matrix; Pollution source feature database query: Access the pre-built pollution source feature database, retrieve the feature fingerprint time series of all potential pollution sources, and generate a candidate pollution source feature set; Dynamic time warping matching execution: Perform dynamic time warping matching on the pollution feature fingerprint and each feature sequence in the candidate pollution source feature set, calculate the optimal curved path distance between sequences, and output the matching distance matrix.

7. The water pollution monitoring and treatment system based on multidimensional data analysis according to claim 6, characterized in that, The pollution source reverse tracing module also includes: Determining the location coordinates of pollution sources: Select the K candidate pollution sources with the smallest distance values ​​in the matching distance matrix, and calculate the location coordinates of the pollution sources based on their weighted average geographical coordinates, with the weights being inversely proportional to the matching distances; Pollution contribution weight calculation: Based on the distance values ​​of the matching distance matrix, the pollution contribution weight of each pollution source to the current pollution event is generated by normalization using the softmax function.

8. The water pollution monitoring and treatment system based on multidimensional data analysis according to claim 7, characterized in that, The adaptive early warning triggering module includes: Pollution source data reception: Obtain the pollution source location coordinates and pollution contribution weights output by the pollution source reverse tracing module; Warning level determination: Query the preset warning level mapping table based on the pollution contribution weight: If the pollution contribution weight is greater than 0.6, a red-level warning instruction will be generated; If 0.3 ≤ pollution contribution weight ≤ 0.6, an orange-level warning instruction will be generated; If the pollution contribution weight is less than 0.3, a yellow-level warning instruction will be generated.

9. The water pollution monitoring and treatment system based on multidimensional data analysis according to claim 8, characterized in that, The adaptive early warning triggering module also includes: Equipment matching: Based on the pollution source location coordinates and the graded early warning command level, equipment matching is performed. Red Order: Activate all aeration equipment and adsorption devices in the target water area; Orange command: Activate adsorption devices within a 500-meter radius of the pollution source's location coordinates; Yellow instruction: Activate only the aeration equipment; Equipment drive signal generation: Convert equipment activation commands into IoT control protocol signals, wherein the aeration equipment receives oxygenation intensity parameters and the adsorption device receives adsorbent dosage parameters; Emergency equipment activation: The drive signal is transmitted to the aeration equipment and / or adsorption device in the target area through the water IoT, triggering them to start operation according to the command parameters.

Images