An automatic water quality monitoring method and system

Abstract

The present application relates to the technical field of water quality monitoring, and particularly relates to a water quality automatic monitoring method and system, comprising the following steps: combining the collected multiple water quality monitoring data two by two, calculating the dynamic coupling strength, generating a topological network atlas, realizing the automatic extraction and structured characterization of the dynamic coupling relationship between complex water quality parameters, overcoming the limitations of traditional manual feature recognition; then matching the topological network atlas with a preset pollution mode feature library, dynamically determining a new abnormal mode, outputting an abnormal feature code set, solving the key defect that a static model cannot identify unknown pollution modes; finally, by fusing the pollution diffusion prediction result and the abnormal feature code set, outputting a water quality monitoring early warning signal, realizing the bidirectional verification of data driving and mechanism model, and significantly improving the early warning accuracy of water quality abnormal phenomena.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water quality monitoring, in particular to a water quality automatic monitoring method and system. BACKGROUND

[0002] In the field of environmental monitoring and public health protection, water quality monitoring plays an irreplaceable role in early warning of pollution events and ensuring drinking water safety. The current mainstream method mainly relies on multi-parameter time series analysis system to collect key indicators such as pH value, dissolved oxygen, turbidity, and conductivity in real time through deployment of sensor networks, and to detect abnormalities based on threshold comparison or statistical models.

[0003] However, such methods have significant limitations: on the one hand, in the face of complex water environment with nonlinear and highly coupled water quality parameter interaction, expert experience is needed to manually design feature recognition; on the other hand, static models cannot adaptively learn new pollution patterns, resulting in delayed or even missed identification of unknown abnormal conditions.

[0004] Therefore, how to automatically extract key features of complex water quality data to accurately warn of water quality abnormalities is a problem that needs to be solved by those skilled in the art. SUMMARY

[0005] In order to automatically extract key features of complex water quality data to accurately warn of water quality abnormalities, the present application provides a water quality automatic monitoring method and system.

[0006] The water quality automatic monitoring method and system provided by the present application adopts the following technical solution:

[0007] A water quality automatic monitoring method, comprising:

[0008] calculating the dynamic coupling strength between each pair of the collected multiple water quality monitoring data, screening target water quality monitoring data with dynamic coupling strength exceeding a preset coupling strength, generating edges with direction on the network vertices mapped by the target water quality monitoring data, and outputting a topological network graph;

[0009] matching the topological network graph with a preset pollution pattern feature library, determining a new abnormal pattern, and outputting an abnormal feature code set;

[0010] obtaining a pollution diffusion prediction result, fusing the pollution diffusion prediction result and the abnormal feature code set, and outputting a water quality monitoring warning signal.

[0011] Further, the step of calculating the dynamic coupling strength between each pair of the collected multiple water quality monitoring data, screening target water quality monitoring data with dynamic coupling strength exceeding a preset coupling strength, generating edges with direction on the network vertices mapped by the target water quality monitoring data, and outputting a topological network graph comprises:

[0012] cross-dimensionally correlate the change trends of any two water quality monitoring data, and calculate the dynamic coupling strength between the two water quality monitoring data;

[0013] When the dynamic coupling strength exceeds the preset coupling strength, a directional edge is generated on the network vertices corresponding to the two target water quality monitoring data, respectively, until multiple water quality monitoring data are traversed.

[0014] According to the generated multiple directional edges and the network vertices corresponding to the edges, a topological network graph is output.

[0015] Further, the step of cross-dimensionally correlating the change trends of any two water quality monitoring data and calculating the dynamic coupling strength between the two water quality monitoring data includes:

[0016] Identify the change trends of any two water quality monitoring data to obtain first trend characteristics and second trend characteristics;

[0017] Obtain the consistency degree of the first trend characteristics and the second trend characteristics in the data change direction, the matching degree in the data change rate, and the coordination degree in the data fluctuation mode, and generate a comparison result;

[0018] According to the comparison result, the dynamic coupling strength is calculated and output.

[0019] Further, the step of matching the topological network graph with the preset pollution mode feature library to determine the newly added abnormal mode and output the abnormal feature code set includes:

[0020] Traverse the topological network graph and parse to obtain multiple key substructures;

[0021] Dynamically correlate each key substructure with the preset pollution mode feature library to obtain an association degree matching threshold;

[0022] Obtain a target key substructure whose association degree matching threshold is lower than a preset association degree matching threshold, and generate a newly added abnormal mode based on the target key substructure;

[0023] Aggregate the newly added abnormal mode and the fixed abnormal mode in the preset pollution mode feature library to output an abnormal feature code set.

[0024] Further, the step of traversing the topological network graph and parsing to obtain multiple key substructures includes:

[0025] Based on the gradient mutation of the edge weight, traverse the topological network graph, and extract a path with a gradient change greater than a preset gradient change as a candidate path;

[0026] aggregate adjacent nodes of each candidate path to generate a plurality of key substructures representing pollution transmission characteristics.

[0027] Further, the step of traversing the topology network graph and analyzing to obtain a plurality of key substructures includes:

[0028] Based on the time sequence transmission characteristics, analyze the inter-node associated path of the topology network graph, and identify the transmission path with time delay characteristics as a candidate path.

[0029] By backtracking the adjacent nodes of the candidate path, a pollution transmission cluster is generated and output as a plurality of key substructures.

[0030] Further, the step of fusing the pollution diffusion prediction result and the abnormal feature code set to output the water quality monitoring warning signal includes:

[0031] Correct the confidence weight of the abnormal feature code set, and spatially fuse the corrected abnormal feature code set and the pollution diffusion prediction result to generate a risk distribution map.

[0032] According to the risk gradient mutation boundary of the risk distribution map, output the water quality monitoring warning signal.

[0033] Further, the step of fusing the corrected abnormal feature code set and the pollution diffusion prediction result to generate a risk distribution map includes:

[0034] According to the spatial position correlation, dynamically weight fuse the corrected abnormal feature code set and the pollution diffusion prediction result to output a spatialized risk value distribution.

[0035] The spatialized risk value distribution is passed through a risk gradient generator to construct a continuous spatialized risk value distribution.

[0036] According to the spatial variation characteristics of the continuous spatialized risk value distribution, extract the risk level boundary to obtain the risk distribution map.

[0037] Further, the step of outputting the water quality monitoring warning signal according to the risk gradient mutation boundary of the risk distribution map includes:

[0038] Analyze the risk gradient mutation boundary of the risk distribution map to obtain boundary feature attributes and risk transmission attributes.

[0039] Through behavior semantic mapping, map the boundary feature attributes and risk transmission attributes into pollution behavior semantics.

[0040] After generating the warning instruction based on the pollution behavior semantic matching response strategy, output the water quality monitoring warning signal according to the warning instruction.

[0041] The application also proposes a water quality automatic monitoring system, comprising:

[0042] a data coupling module, configured to calculate dynamic coupling strengths between each pair of the collected multiple water quality monitoring data, filter target water quality monitoring data with dynamic coupling strengths exceeding a preset coupling strength, generate edges with directionality on network vertices mapped by the target water quality monitoring data, and output a topological network graph;

[0043] a data matching module, configured to match the topological network graph with a preset pollution mode feature library, determine a newly added abnormal mode, and output an abnormal feature code set;

[0044] a data decision module, configured to obtain a pollution diffusion prediction result, fuse the pollution diffusion prediction result and the abnormal feature code set, and output a water quality monitoring early warning signal.

[0045] The beneficial effects achieved are as follows:

[0046] The application provides a water quality automatic monitoring method, which comprises the following steps: calculating dynamic coupling strengths between each pair of the collected multiple water quality monitoring data, filtering target water quality monitoring data with dynamic coupling strengths exceeding a preset coupling strength, generating edges with directionality on network vertices mapped by the target water quality monitoring data, and outputting a topological network graph; matching the topological network graph with a preset pollution mode feature library, determining a newly added abnormal mode, and outputting an abnormal feature code set; obtaining a pollution diffusion prediction result, fusing the pollution diffusion prediction result and the abnormal feature code set, and outputting a water quality monitoring early warning signal.

[0047] That is, in the application, the collected multiple water quality monitoring data are combined in pairs to calculate dynamic coupling strengths, generate a topological network graph, realize automatic extraction and structured representation of dynamic coupling relationships between complex water quality parameters, and overcome the limitations of traditional manual feature recognition; then the topological network graph is matched with a preset pollution mode feature library, a newly added abnormal mode is dynamically determined, and an abnormal feature code set is output, which solves the key defect that a static model cannot identify unknown pollution modes; finally, a pollution diffusion prediction result and an abnormal feature code set are fused to output a water quality monitoring early warning signal, realize bidirectional verification of data driving and mechanism model, and significantly improve the early warning accuracy of water quality abnormal phenomena. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 is a flowchart of a water quality automatic monitoring method provided in the application;

[0049] Figure 2 is a flowchart of generating a topological network graph based on multiple water quality monitoring data in the application;

[0050] Figure 3 is a schematic diagram of a topological network graph output in the application;

[0051] Figure 4 is a flowchart of the application based on the topological network atlas and the preset pollution mode feature library, outputting an abnormal feature code set;

[0052] Figure 5 is a flowchart of the application outputting a water quality monitoring early warning signal according to the abnormal feature code set;

[0053] Figure 6 is a structural schematic diagram of a water quality automatic monitoring system.

[0054] BRIEF DESCRIPTION OF DRAWINGS

[0055] 10, data coupling module; 20, data matching module; 30, data decision module. DETAILED DESCRIPTION

[0056] The following will be described in detail in combination with the accompanying Figures 1-6 The application will be further described in detail.

[0057] In the description of the application, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like indicate the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the application. In addition, the terms "first", "second", "third" are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0058] In the description of the application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the application can be understood according to the specific circumstances.

[0059] The embodiment of the application discloses a water quality automatic monitoring method.

[0060] Please refer to Figure 1 In an embodiment of the application, a water quality automatic monitoring method comprises steps S10-S30:

[0061] In step S10, the dynamic coupling strength between each pair of the collected water quality monitoring data is calculated, target water quality monitoring data with dynamic coupling strength exceeding a preset coupling strength is screened, a directional edge is generated on a network vertex mapped by the target water quality monitoring data, and a topological network graph is output.

[0062] The current water quality monitoring scheme mainly relies on single-parameter threshold alarm or multi-parameter statistical correlation analysis, and has the defect of being difficult to capture the nonlinear coupling between multiple parameters in a complex water quality system, resulting in insufficient recognition ability for a complex pollution event.

[0063] In view of the above problems, the embodiment combines each pair of the collected water quality monitoring data, such as dissolved oxygen, pH value, conductivity and the like, which can reflect the characteristics of the water environment, calculates the dynamic coupling strength between two water quality monitoring data, screens target water quality monitoring data with dynamic coupling strength exceeding a preset coupling strength, generates an edge with directionality on a network vertex mapped by the target water quality monitoring data, and constructs a topological network graph representing the dynamic correlation between parameters according to the edge and the corresponding network vertex. The core lies in converting discrete multiple water quality monitoring data into a network structure. Because pollution events in a real water environment often show coordinated changes in data clusters, such as a sharp drop in dissolved oxygen accompanied by a sharp rise in turbidity, reverse fluctuations in pH value and conductivity, and the like, traditional linear statistics cannot quantify such complex interactions.

[0064] The embodiment can automatically extract the implicit conduction path and strength relationship between each water quality monitoring data, wherein the network vertex maps the water quality monitoring data entity, and the edge weight dynamically encodes the spatiotemporal correlation of the water quality monitoring data changes, so as to effectively identify key features such as ammonia-nitrogen-dissolved oxygen correlation rupture caused by industrial wastewater, chlorophyll-a-temperature coupling oscillation in the early stage of algal bloom, and the like, provide high-dimensional feature input for pollution pattern matching, and significantly improve the early warning ability for hidden pollution and new pollutants.

[0065] In step S20, the topological network graph is matched with a preset pollution pattern feature library, a new abnormal pattern is determined, and an abnormal feature code set is output.

[0066] By matching the topological network graph with the preset pollution pattern feature library, the key defect that a conventional water quality pollution recognition model cannot recognize unknown pollution patterns is solved.

[0067] By dynamically comparing the abnormal substructure in the topology network graph with the known pollution mode in the preset pollution mode feature library, when the matching degree is lower than the set threshold, the abnormal mode generation mechanism is triggered, and the abnormal feature code set containing the coupling features and topology attributes of the water quality monitoring data is output, so that the adaptive expansion and incremental learning ability of the preset pollution mode feature library are realized, and when a sudden unknown pollution event occurs, the system can automatically generate a corresponding feature template and encode and store it, which significantly improves the real-time capture ability and early warning accuracy of new pollution modes, and avoids the false negative risk caused by model solidification in the traditional scheme.

[0068] In step S30, the pollution diffusion prediction result is obtained, the pollution diffusion prediction result and the abnormal feature code set are fused, and the water quality monitoring early warning signal is output.

[0069] By obtaining the pollution diffusion prediction result and fusing the pollution diffusion prediction result and the abnormal feature code set, the problem of high false positive rate and response lag caused by the disconnection between water quality monitoring data and physical model in traditional water quality early warning is solved. Because the abnormal feature code set in the actual pollution event may be affected by the pseudo signal generated by sensor noise or local interference, and the pollution diffusion result can provide physical rule constraints based on the water dynamic mechanism, the real-time abnormal mode represented by the abnormal feature code set and the pollution migration trend quantified by the pollution diffusion prediction result are dynamically weighted and fused, the credibility of the abnormal feature is verified by the physical rule, and the preset pollution mode feature library is corrected.

[0070] The water quality monitoring prediction signal is output to drive the equipment action under the mechanism constraint, which significantly improves the accuracy of identifying complex pollution events and avoids the low accuracy of early warning caused by the limitation of single water quality monitoring data.

[0071] It should be noted that the water dynamic mechanism can provide physical rule constraints, which are the pollution migration and diffusion rules established based on the principle of fluid mechanics, and are the forced boundary conditions for reasonable checking and correcting the water quality monitoring early warning signal.

[0072] In this embodiment, by combining the collected multiple water quality monitoring data in pairs, the dynamic coupling strength is calculated, the topology network graph is generated, the automatic extraction and structured representation of the dynamic coupling relationship between complex water quality parameters are realized, and the limitation of traditional manual feature recognition is overcome; then the topology network graph is matched with the preset pollution mode feature library, the new abnormal mode is dynamically determined, the abnormal feature code set is output, and the key defect that the static model cannot identify unknown pollution modes is solved; finally, the pollution diffusion prediction result and the abnormal feature code set are fused, the water quality monitoring early warning signal is output, the bidirectional verification of data driven and mechanism model is realized, and the early warning accuracy of water quality abnormal phenomenon is significantly improved.

[0073] In a feasible implementation, reference is made toFigure 2 As shown, step S10 includes steps A10-A30:

[0074] Step A10, cross-dimension correlation analysis is performed on the change trend of any two water quality monitoring data to calculate the dynamic coupling strength between the two water quality monitoring data.

[0075] First, the change trend of the first water quality monitoring data and the second water quality monitoring data is extracted respectively by sliding window processing combined with linear regression to generate the first trend feature and the second trend feature. Each trend feature includes a data change direction component, a data change rate scalar value, and a data fluctuation pattern sequence.

[0076] Among them, cross-dimension correlation analysis is to compare different dimension parameters, such as the characteristics of ammonia nitrogen and dissolved oxygen. Specifically, the consistency degree of the first trend feature and the second trend feature in the data change direction is calculated by cosine similarity, the matching degree of the data change rate is obtained by difference calculation and subtraction operation after normalization, and the fluctuation pattern coordination degree is quantified by dynamic time warping algorithm. The comparison results of the three dimensions are input into the integration function, and the dynamic coupling strength which comprehensively represents the mutual influence degree between the two water quality monitoring data is output.

[0077] This flow directly supports automatic extraction of the implicit conduction path and strength relationship between each water quality monitoring data, and all calculation results are updated in real time through the stream processing framework.

[0078] Optionally, step A10 can include steps A11-A13:

[0079] Step A11, identify the change trend of any two water quality monitoring data to obtain the first trend feature and the second trend feature.

[0080] First, through real-time time series analysis, such as sliding window processing combined with linear regression, the change trend of any two water quality monitoring data is identified. The purpose is to extract the dynamic behavior features of each water quality monitoring data, such as calculating the short-term change direction (upward or downward tendency), change rate (numerical change amount per unit time), and fluctuation pattern (sequence periodicity or amplitude characteristics) of each water quality monitoring data, thereby obtaining the first trend feature and the second trend feature.

[0081] Step A12, obtain the consistency degree of the first trend feature and the second trend feature in the data change direction, the matching degree in the data change rate, and the coordination degree in the data fluctuation pattern to generate comparison results.

[0082] The consistency degree of the first trend feature and the second trend feature in the data change direction, the matching degree of the first trend feature and the second trend feature in the data change rate, and the coordination degree of the first trend feature and the second trend feature in the data fluctuation mode are obtained, so as to generate a comprehensive comparison result, and the similarity and correlation of the two trend features are compared.

[0083] It should be noted that the consistency degree of the first trend feature and the second trend feature in the data change direction can be calculated by the cosine similarity; the matching degree of the first trend feature and the second trend feature in the data change rate can be calculated by the difference calculation rate; and the coordination degree of the first trend feature and the second trend feature in the data fluctuation mode can be obtained by applying the dynamic time warping quantification fluctuation coordination.

[0084] Specifically, the consistency degree of the first trend feature and the second trend feature in the data change direction can be calculated by the cosine similarity, and the process is as follows:

[0085] The data change direction extracted from the first trend feature and the second trend feature is converted into a direction vector, respectively, wherein the dimension of each direction vector corresponds to the sequence position of the real-time sliding window, and the numerical change slope of each real-time sliding window is calculated as a direction component; the dot product of the direction vector of the first trend feature and the direction vector of the second trend feature is calculated, and then divided by the product of the module lengths of the two vectors, and the cosine similarity is output.

[0086] The calculated cosine similarity represents the consistency degree of the first trend feature and the second trend feature in the data change direction, which specifically shows that when the directions of the two vectors are completely the same, 1 is output, when the directions are orthogonal, 0 is output, and when the directions are opposite, -1 is output. When new water quality monitoring data arrives, the direction vector in the sliding window is dynamically updated, and the cosine similarity is recalculated, so as to automatically quantify the consistency degree of the trend direction of any two water quality monitoring data.

[0087] Specifically, the matching degree of the first trend feature and the second trend feature in the data change rate can be calculated by the difference calculation rate, and the process is as follows:

[0088] The data change rate scalar value contained in the first trend feature and the data change rate scalar value contained in the second trend feature are extracted, respectively, the absolute difference of the two data change rate scalar values is calculated, the absolute difference is compared with the maximum possible rate change range in the current monitoring period, and a relative difference value in the range of 0 to 1 is obtained by using the normalization method. A minus operation is performed on the relative difference value, so that the output value is converted into a matching degree measurement value.

[0089] The matching degree value is defined as the matching degree of the first trend feature and the second trend feature in the data rate of change, which is specifically represented as: output 1 when the two scalar values are completely equal, and output zero when the rate difference reaches the maximum range.

[0090] By updating the preset rate range in real time, the calculation results in each monitoring period are dynamically adapted to the actual water quality monitoring data change amplitude, ensuring the timeliness and feasibility of the rate matching degree quantization results, and supporting the subsequent dynamic coupling strength calculation.

[0091] Specifically, the coordination degree of the first trend feature and the second trend feature in the data fluctuation mode can be quantified by the dynamic time warping process as follows:

[0092] The data fluctuation mode sequence of the first trend feature and the second trend feature is extracted by a sliding time window, wherein each data fluctuation mode sequence is composed of amplitude features at multiple time points.

[0093] The dynamic time warping algorithm is applied to calculate the minimum warping path cumulative distance between the two data fluctuation mode sequences, which reflects the difference in fluctuation patterns between the sequences. The minimum warping path cumulative distance is divided by a preset maximum possible fluctuation distance range for normalization, and then a subtraction operation is performed on the normalized result to convert it into a coordination degree value.

[0094] The coordination degree value is defined as the coordination degree of the first trend feature and the second trend feature in the data fluctuation mode, which is represented as: when the two data fluctuation mode sequences are completely the same, the dynamic time warping cumulative distance is zero, and the normalized subtraction operation outputs a coordination degree value of 1, representing the highest coordination degree; when the fluctuation mode difference reaches the preset range boundary, the coordination degree value is output as 0, representing the lowest coordination degree.

[0095] The distance matrix of dynamic time warping is used to calculate the matching of non-equal length sequences, and the maximum fluctuation distance range reference value is updated in real time, ensuring that the quantization process of fluctuation coordination degree can effectively capture the asynchronous fluctuation correlation between water quality monitoring data.

[0096] Step A13, according to the comparison result, the dynamic coupling strength is calculated and output.

[0097] According to the consistency degree of the generated first trend feature and the second trend feature in the data change direction, the matching degree of the data change rate, and the coordination degree of the data fluctuation mode, the consistency degree in the data change direction, the matching degree of the data change rate, and the coordination degree of the data fluctuation mode are combined into a single value by a weighted average method. The single value is the output dynamic coupling strength, which aims to quantitatively integrate the mutual influence degree between the two water quality monitoring data, serve the subsequent automatic extraction of the implicit conduction path and strength relationship between the water quality monitoring data, and provide input for network vertex mapping and edge weight generation.

[0098] It should be noted that the above is realized by the real-time numerical calculation engine, which ensures that the dynamic coupling strength is automatically recalculated after each new data update comparison result, meets the feasibility of stream data processing, and supports pollution pattern matching and early warning capability.

[0099] Step A20, when the dynamic coupling strength exceeds the preset coupling strength, a directional edge is generated between the network vertices corresponding to the two target water quality monitoring data, respectively, until multiple water quality monitoring data are traversed.

[0100] It should be noted that the preset coupling strength is a preset coupling strength set in advance, which is used to filter the water quality monitoring data correlation pairs with actual conduction significance.

[0101] Through the real-time calculation of the dynamic coupling strength of the trend of any two water quality monitoring data, it is determined that there is a significant implicit conduction relationship between the two target water quality monitoring data only when the dynamic coupling strength exceeds the preset coupling strength. At this time, according to the dominant influence direction when the first trend feature and the second trend feature are generated, for example, through the direction vector slope sign or the rate change sequence, a directional edge is generated between the network vertices mapped by the target water quality monitoring data entities, which is pointed to by the dominant change data vertex to the affected data vertex, so as to automatically construct a directed topological network graph that can express the pollution conduction path. In the specific implementation, the preset coupling strength can be configured through historical data analysis or expert experience, and the logic of generating directional edges is embedded in the real-time stream processing process, which ensures the technical feasibility, and finally serves the pollution conduction mode recognition and early warning.

[0102] Step A30, according to the generated multiple directional edges and the network vertices corresponding to the edges, output the topological network graph.

[0103] By traversing all water quality monitoring data pairs, a trend correlation analysis is performed on each pair of water quality monitoring data, and the dynamic coupling strength is calculated. Whenever the dynamic coupling strength of any pair of target water quality monitoring data exceeds the preset coupling strength, a directional edge is generated between the two network vertices it maps to. After traversal, all directed edges of vertex pairs that meet the correlation strength condition are integrated into an edge set, and all monitored water quality monitoring data entities are automatically mapped into a vertex set.

[0104] At this point, the vertex set and edge set are combined to construct a complete topological network graph structure, where each vertex represents a specific water quality monitoring data entity, such as the ammonia nitrogen concentration at a certain monitoring point, and each directional edge represents a conduction relationship exceeding the preset coupling strength, such as the unidirectional impact of industrial wastewater discharge leading to a decrease in dissolved oxygen downstream.

[0105] For reference Figure 3 As shown, by traversing all pairs of water quality monitoring data, such as "industrial wastewater discharge" and "dissolved oxygen at downstream point B", and "ammonia nitrogen concentration at monitoring point A" and "directional change", a trend correlation analysis is performed on each pair of water quality monitoring data and the dynamic coupling strength is calculated. Whenever the dynamic coupling strength of any pair of target water quality monitoring data exceeds the preset coupling strength, a directional edge is generated between the two network vertices it maps to. For example, if the correlation strength between industrial wastewater discharge and dissolved oxygen at downstream point B exceeds the standard, a pointing arrow (i.e., a directional edge) is generated between industrial wastewater discharge and dissolved oxygen at downstream point B.

[0106] After the traversal is complete, all directed edges between vertex pairs that meet the association strength condition are integrated into an edge set, containing Figure 3 All solid lines with arrows, such as “Directional Change” → “Ammonia Nitrogen Monitoring”, “Downstream Point B Dissolved Oxygen” → “pH Value”, “Downstream Point B Dissolved Oxygen” → “Monitoring Point B”, “Downstream Point B Dissolved Oxygen” → “Dissolved Oxygen”, and “Monitoring Point C” → “pH Value”, are automatically mapped to a set of vertices, namely “Industrial Wastewater Discharge”, “Downstream Point B Dissolved Oxygen”, “Ammonia Nitrogen Monitoring”, “pH Value”, “Directional Change”, “Dissolved Oxygen”, “Monitoring Point B”, “Monitoring Point C”, and “Monitoring Point A Ammonia Nitrogen Concentration” water quality monitoring data entities.

[0107] At this point, the vertex set and edge set are combined to construct a complete topological network graph structure, where each vertex represents a specific water quality monitoring data entity, such as the vertex "ammonia nitrogen concentration at monitoring point A" representing the ammonia nitrogen concentration index at monitoring point A, and each directional edge represents a transmission relationship exceeding the preset coupling strength, such as "dissolved oxygen at downstream point B" → "industrial wastewater discharge" clearly indicating that industrial wastewater discharge leads to a decrease in the dissolved oxygen index at downstream point B.

[0108] It should be noted that,Figure 3 The arrowless connecting lines in the middle, such as the straight line between "ammonia nitrogen concentration of monitoring point A" and "monitoring point B", essentially represent the static binding relationship between the water quality monitoring data entity and its attribute label, and only represent the dependent relationship between the water quality monitoring data entity and its exclusive index.

[0109] The topology network graph is stored and rendered in real time through a graph database or a graph computing engine, forming a visual high-dimensional network structure, the edge weight of which is automatically updated by the dynamic coupling strength, ultimately realizing the automatic extraction of the implicit conduction path between water quality parameters, and providing dynamic topological feature support for pollution diffusion pattern matching and early warning.

[0110] In a feasible implementation, referring to FIG. 10, step S20 includes steps B10-B40. Figure 4

[0111] Step B10, traverse the topology network graph and parse to obtain a plurality of key substructures.

[0112] By systematically scanning and traversing the constructed topology network graph, key substructures with specific pollution conduction significance are identified and separated.

[0113] The key substructures mainly include three categories: the first category is a one-way conduction cluster, such as the vertex sequence from the ammonia nitrogen concentration of the industrial wastewater monitoring point to the downstream dissolved oxygen, which reveals the directional diffusion path of the pollutant; the second category is a high-coupling-strength subnetwork, such as the bidirectional strong coupling vertex group formed by the chlorophyll a and temperature monitoring points, which represents the collaborative oscillation mechanism between parameters; and the third category is a topological abnormal feature, such as an isolated vertex or a missing edge formed by the broken edge between ammonia nitrogen and dissolved oxygen, which indicates a pollution conduction blocking event.

[0114] By driving the extraction of key substructures through dynamic coupling strength values, key patterns such as the chlorophyll a-temperature closed-loop oscillation network formed in the early stage of algal blooms and the downstream dissolved oxygen collapse chain caused by industrial pollution are automatically output, providing high-dimensional spatiotemporal features for pollution pattern matching and significantly improving the early warning capability for hidden pollution.

[0115] In the first feasible implementation, step B10 is implemented through steps B11-B12.

[0116] Step B11, traverse the topology network graph based on edge weight gradient mutation, and extract the path with a gradient change greater than a preset gradient change as a candidate path.

[0117] ​Firstly, record the time series data of the edge weight of each edge in the topology network graph. When the network is updated with new water quality monitoring data, traverse the topology network graph to calculate the instantaneous change gradient of all edge weights. The specific algorithm is as follows: subtract the dynamic coupling strength values calculated continuously twice, and then divide by the previous dynamic coupling strength value to obtain the normalized gradient value.

[0118] When a sudden change in the gradient value of the edge weight on any path is detected, that is, the normalized gradient value of a certain edge and the normalized gradient value of its adjacent edge simultaneously exceed the conventional fluctuation range within a short time, then immediately calculate the gradient change amount of all adjacent edge weights on this path, defined as the absolute difference between the current normalized gradient value and the historical average normalized gradient value. If the gradient change amount of more than three consecutive edges in the path exceeds the preset gradient change amount, mark the path as a candidate path, so as to automatically capture the chain weight surge on the ammonia-nitrogen-dissolved oxygen conduction path downstream of the industrial pollution source.

[0119] Meanwhile, perform dynamic coupling strength continuous threshold value polling: monitor the dynamic coupling strength values of the edges in each subnet of the topology network graph. When it is detected that more than 70% of the edges in a certain subnet have a dynamic coupling strength greater than the preset coupling strength for three consecutive calculation periods, directly mark the subnet as a high coupling strength candidate substructure.

[0120] Synchronously implement topology anomaly feature special scanning: perform cliff drop detection on the dynamic coupling strength of each edge in the topology network graph. When the current dynamic coupling strength value is less than or equal to 0.2 and the gradient change amount is less than -0.7, mark the associated edge breakage event, and simultaneously count the vertex correlation degree. When the vertex correlation degree is zero for two consecutive periods, capture the isolated vertex.

[0121] Step B12, aggregate the adjacent nodes of each candidate path to generate a plurality of key substructures representing pollution conduction features.

[0122] Firstly, based on the candidate path extracted by the edge weight gradient mutation detection, traverse all nodes and perform adjacent node clustering operation: if any two nodes in the candidate path meet the preset topology continuity condition, such as spatial coordinate distance below threshold, hydrological flow connectivity or time window synchronization, etc., then aggregate them into the same substructure.

[0123] The specific clustering method uses the breadth-first search algorithm, taking the starting node of the candidate path as the root node, expanding the search adjacent nodes according to the gradient change direction of the dynamic coupling strength between nodes, and stopping until the boundary node whose gradient change amount is below the preset gradient change amount. Each aggregated local network unit constitutes a key substructure representing a specific pollution conduction feature, such as the one-way cluster of ammonia-nitrogen conduction cluster pointing to the dissolved oxygen collapse area representing industrial pollution event, and the closed loop of chlorophyll a-temperature synchronous oscillation representing the early stage of algal bloom.

[0124] Meanwhile, the high coupling strength candidate substructure is continuously polled by the dynamic coupling strength exceeding the threshold value, the chlorophyll a-temperature synchronous oscillation closed loop represents the early stage characteristics of algal blooms, and the key substructure generated by the topological abnormality feature is formed by the isolated vertex or missing edge captured by the special scanning of the topological abnormality feature, which indicates that the pollution transmission is blocked.

[0125] The finally generated key substructure provides space-time transmission information for pollution pattern matching, and significantly enhances the early identification ability of hidden pollution paths.

[0126] In the second feasible implementation, step B10 can also be implemented by steps B13-B14:

[0127] Step B13: Based on the time sequence transmission characteristics, the correlation paths between the nodes of the topological network graph are analyzed, and the transmission paths with time delay characteristics are identified as candidate paths.

[0128] First, the historical monitoring data time series of the starting node and the target node connected by each edge in the topological network graph are extracted, the fluctuation patterns of the two sequences are aligned by the dynamic time warping algorithm, and the optimal warping path is calculated. The time difference between the peak value of the starting node data fluctuation and the response peak value of the target node is identified as the dynamic delay value, and the covariance of the two sequences in the fluctuation mode is analyzed to eliminate random fluctuation interference.

[0129] When the dynamic delay values of the consecutive edges of a certain path are all greater than the preset minimum dynamic delay threshold value, and the covariance values of all adjacent node pairs in the path exceed the preset covariance threshold value, it is determined that the path is a transmission path with time delay characteristics, i.e. a candidate path.

[0130] For example, after the ammonia nitrogen concentration of the industrial wastewater monitoring point increases sharply, the downstream dissolved oxygen monitoring point shows a downward trend after a fixed delay time. The path from the industrial wastewater monitoring point to the downstream dissolved oxygen monitoring point is identified by the above algorithm to meet the delay condition. When traversing the topological network graph, the node pairs with continuous delay transmission are automatically aggregated by breadth-first search combined with delay constraints.

[0131] The finally extracted candidate path needs to meet two core characteristics: all edge dynamic delay values are positive, and the overall covariance value of the path reflects stable transmission directionality, such as the covariance value of the one-way transmission cluster continuously exceeding the threshold value, which realizes the automatic capture of abnormal delay transmission or unknown delay rules of new pollutants migration caused by illegal sewage caused by illegal sewage, and provides hidden transmission mode characteristics for early warning.

[0132] The preset minimum dynamic delay threshold value is a time window boundary value set based on the physical law of pollutant migration; and the preset covariance threshold value is a statistical critical value quantifying the stable correlation of fluctuation patterns.

[0133] Step B14, generate pollution conduction clusters by backtracking the adjacent nodes of the candidate path and output as multiple key substructures.

[0134] Based on the candidate path identified by the preset minimum dynamic delay threshold and the preset covariance threshold, a backtracking algorithm is performed from the path endpoint of the candidate path, as described below:

[0135] First, detect the direct upstream node B of node C (i.e., the path endpoint of the candidate path) based on the directed edge in the topological network graph pointing to node C, and verify whether the conduction relationship between node B and node C satisfies the pollution conduction characteristics, such as the dynamic delay value exceeding the threshold and the covariance being greater than 0.65.

[0136] If the conditions are met, continue backtracking to the upstream node A of node B and verify the effectiveness of the conduction from node A to node B. During the backtracking process, based on the preset spatial adjacency constraints such as hydrological flow direction topological path or geographical coordinate proximity, adjacent nodes are aggregated, and if all nodes on the path satisfy the conduction characteristic continuity, the backtracking path (node A→B→C) and spatially adjacent homogeneous nodes are merged into a pollution conduction cluster.

[0137] For example, when generating key substructures for industrial pollution conduction clusters, node A (ammonia nitrogen from the discharge outlet) is the cluster core, and nodes B, C, and D form the cluster members of the collapse area. The pollution impact intensity value is calculated by the average dynamic coupling strength between nodes within the cluster. Each pollution conduction cluster is a key substructure representing a unified pollution event. This process performs dynamic clustering in real time through the path backtracking API of the graph database, such as apoc.path.expandConfig of Neo4j, and finally outputs multiple key substructures carrying spatial range, pollution intensity, and conduction direction to provide fine pollution pattern fingerprints.

[0138] Step B20, dynamically associate each key substructure with the preset pollution pattern feature library to obtain an association degree matching threshold.

[0139] First, extract the predefined standardized pollution pattern feature templates from the preset pollution pattern feature library, such as the algal bloom outbreak pattern which is a bidirectional strong coupling closed loop structure composed of chlorophyll a and temperature vertices, and the industrial pollution pattern which is a chain structure with ammonia nitrogen vertex pointing to the dissolved oxygen collapse cluster through a one-way edge. For each key substructure to be matched, GraphSAGE (Graph Sample and Aggregated, a graph neural network model for graph node embedding learning) is used to convert its topological structure into a vectorized representation, including vertex type, edge weight average, conduction direction distribution, and other topological structure feature encoding. At the same time, the dynamic coupling strength gradient change quantity and other time series conduction features of the key substructure are injected.

[0140] The vectorized representation is calculated with the cosine similarity of all templates in the preset pollution pattern feature library, and the dynamic correlation value of the current key substructure to each pollution pattern is output, with a value range of -1 to 1.

[0141] Assuming that the dynamic correlation matching threshold is 0.8, which is generated by statistical calibration according to the matching rate of historical pollution events and pattern library templates, when the dynamic correlation value of the key substructure to a certain pollution template exceeds the threshold, such as a chlorophyll a-temperature closed loop similarity of 0.9 matching an algae bloom outbreak pattern, it is determined that it is successfully associated with the corresponding pollution pattern and triggers an early warning.

[0142] Step B30, obtain target key substructures with a correlation matching threshold lower than the preset correlation matching threshold, and generate new abnormal patterns based on the target key substructures.

[0143] After completing the dynamic correlation matching of each key substructure with the preset pollution pattern feature library, filter all target key substructures with a correlation matching threshold lower than the preset correlation matching threshold, for example, the historical calibration value is 0.8.

[0144] For each target key substructure, its core conduction features are automatically extracted by the topological structure analysis engine, including vertex combination type (such as coupling of abnormal appearing biological toxicity monitoring points and heavy metal vertices), edge weight distribution features (such as intermittent pulse oscillation of dynamic coupling strength), and spatiotemporal conduction patterns (such as reverse conduction timing of non-hydrological flow direction). Based on the extracted conduction features, new abnormal patterns are dynamically generated outside the preset pollution pattern feature library, and the specific method is:

[0145] The topological skeleton of the target key substructure (retaining its directional edge and vertex association) is abstracted into a general pattern template, and its conduction dynamic features (such as pulse oscillation parameters and non-reverse conduction timing features) are injected, while recording the spatiotemporal conditions (such as during heavy rain in the catchment area of the chemical plant) when the pattern is first triggered. The generated new abnormal pattern is finally stored in the preset pollution pattern feature library in the form of an encoded vector after spatial neighborhood verification and temporal stability analysis, realizing the automatic expansion of pollution identification capability.

[0146] Step B40, aggregate the new abnormal patterns and the fixed abnormal patterns in the preset pollution pattern feature library, and output the abnormal feature code set.

[0147] First, all fixed abnormal patterns in the preset pollution pattern feature library, such as defined industrial pollution unidirectional conduction cluster templates and algae bloom outbreak bidirectional closed loop templates, are merged into a unified set with new abnormal patterns dynamically generated through the new abnormal pattern process, such as detected dissolved oxygen-pH atypical conduction ring pulse oscillation templates. Perform multi-layer feature abstraction coding on each pattern in this unified set:

[0148] For example, the first layer extracts its topological skeleton features, including vertex type combinations capable of representing vertex category distribution such as ammonia nitrogen, dissolved oxygen, or heavy metals, directional edge connection relationships representing the proportion of one-way / two-way conduction, and substructure scale values representing the number of vertices and edge density.

[0149] The second layer injects dynamic behavior features, including dynamic coupling strength gradient change amounts, such as the weight mutation slope of the collapsed cluster, time series conduction delay statistical quantities representing the mean and variance of the path delay value, and wave mode covariance values.

[0150] The third layer based on GraphSAGE converts the aforementioned feature combination into a fixed-dimensional vectorized abnormal feature encoding.

[0151] In this way, redundant dimensions are eliminated, and a lightweight abnormal feature encoding set is finally generated, each encoding corresponding to a unique pollution conduction mode fingerprint. The encoding set is updated to the real-time detection engine, providing an expandable pattern matching benchmark library for streaming water quality monitoring data, and ensuring the dynamic capture ability of the early warning system for hidden pollution and new pollution characteristics.

[0152] In a feasible implementation, referring to Figure 5 As shown in FIG. 10, step S30 includes steps C10-C20:

[0153] Step C10, correct the confidence weight of the abnormal feature encoding set, and spatially fuse the corrected abnormal feature encoding set and the pollution diffusion prediction result to generate a risk distribution map.

[0154] First, based on the verification results of the spatial coverage and time recurrence degree of the abnormal mode, the confidence weight of each encoding in the abnormal feature encoding set is linearly weighted and corrected, for example, the weight is increased by 0.1 for each increase in the number of three-day recurrence of the new abnormal mode, and the weight is increased by 0.05 for each expansion of the coverage range of the fixed abnormal mode.

[0155] The corrected abnormal feature encoding set carries the confidence weight vector, the time series conduction dynamic feature, and the topological structure fingerprint. The spatial fusion engine is input into the pollution diffusion prediction result calculated by the hydrological diffusion model. The spatial fusion engine aligns the position of the physical migration grid of the pollution diffusion prediction result and the spatial projection point of the abnormal feature encoding set by the Kriging spatial interpolation algorithm, and then superimposes the abnormal feature strength value with the confidence weight as the adjustment coefficient. The fusion calculation formula is the weighted product of the pollution prediction concentration value and the abnormal feature strength value of each geographic coordinate point. Finally, a risk distribution map is generated by integrating the comprehensive physical diffusion law and the statistical abnormal feature.

[0156] It should be noted that the pollution diffusion prediction result can include pollution concentration spatial distribution, migration direction and hot spot probability; the physical migration grid is the ammonia nitrogen diffusion concentration contour surface; and the spatial projection point of the abnormal feature code set can be the geographical coordinates of the dissolved oxygen collapse cluster.

[0157] The risk distribution map is rendered in real time by the vector tile technology, dynamically marks the enhanced area of the hidden pollution feature, and realizes the fine visual early warning of the pollution risk in the time and space dimensions.

[0158] Optionally, step C10 can include steps C11-C13:

[0159] Step C11, according to the spatial position correlation, dynamically weight fusion is performed on the modified abnormal feature code set and the pollution diffusion prediction result, and a spatialized risk value distribution is output.

[0160] First, a geographical spatial index network of water quality monitoring sites is established, and the spatial coordinates of the modified abnormal feature code set are mapped with the preset coupling strength of the spatial grid of the pollution diffusion prediction result in terms of position correlation.

[0161] The correlation calculation adopts the inverse distance weighting algorithm, and each spatial grid node within the preset spatial influence radius is weighted.

[0162] In the dynamic weight fusion stage, according to the modified confidence weight of the abnormal feature code set and the basic weight of the pollution diffusion prediction result, weighted superposition operation is performed according to the spatial correlation strength, and a vector-type spatialized risk value distribution covering the whole domain is output.

[0163] Among them, the spatial coordinates can be the center of the dissolved oxygen collapse cluster; the spatial grid node of the pollution diffusion prediction result can be the concentration point of every hundred meters interval output by the hydrological diffusion model; and the spatial grid node weighting can be that the closer the grid node within the 500-meter range around the center point of the collapse cluster, the higher the weight.

[0164] In addition, the specific formula for performing weighted superposition operation according to the spatial correlation strength is:

[0165]

[0166] It should be noted that R is the spatialized risk value, C is the pollution diffusion prediction concentration, is the basic weight, B is the abnormal feature strength, W is the modified confidence weight, and a is the position correlation coefficient, wherein the position correlation coefficient is obtained by linearly attenuating the distance from the grid node to the abnormal feature core point, the fusion process is continuously dynamically updated, and the abnormal feature strength and the pollution diffusion prediction value are adjusted synchronously when the water quality monitoring data is refreshed every round.

[0167] Step C12, the spatialized risk value distribution is input into the risk gradient generator to construct a continuous spatialized risk value distribution.

[0168] The discrete spatialized risk value distribution is input into the risk gradient generator, which first converts the discrete risk points into a continuous field based on the Kriging spatial interpolation algorithm, specifically by calculating the risk value of each unsampled point as a weighted average of the risk values of the adjacent monitoring points, with the weights determined dynamically by the spatial covariance function and the distance from the monitoring points, and then applies a gradient smoothing algorithm to eliminate gradient transition noise caused by uneven monitoring point density and enhance the diffusion continuity along the hydrological flow path.

[0169] The risk gradient generator synchronously injects spatial constraints to ensure that the gradient field forms a continuous risk transmission gradient along the river flow direction, and finally outputs a continuous spatialized risk value distribution that provides a centimeter-precision dynamic spatialized risk feature map for hidden pollution migration path identification and emergency decision-making.

[0170] It should be noted that the gradient smoothing algorithm in this embodiment is anisotropic diffusion filtering; the spatial constraint condition is the geographical boundary rule of river network topological structure, terrain elevation obstacles, etc.; the continuous risk transmission gradient is the gradient decrease downstream of industrial pollution sources, the ring-shaped gradient in reservoir detention areas, etc.

[0171] Step C13, according to the spatial variation characteristics of the continuous spatialized risk value distribution, extract the risk level boundary to obtain a risk distribution map.

[0172] First, calculate the risk gradient field of the continuous spatialized risk value distribution through the Sobel convolution kernel, identify the spatial position where the risk value changes abruptly, and based on the gradient extreme points, draw the candidate boundary line of the preset risk level threshold, while integrating the geographical constraints to correct the boundary topological form.

[0173] According to the preset risk level threshold, perform isosurface segmentation on the continuous field to generate risk level boundaries that conform to the natural diffusion form. Finally, superimpose the risk level boundaries and the risk gradient mutation lines onto the basic geographic base map, and output the risk distribution map with centimeter-precision through vector-raster fusion technology, providing dynamic spatialized decision-making basis for hidden pollution interception and emergency resource allocation.

[0174] Among them, the spatial position can be, for example, the risk value steep drop edge downstream of an industrial pollution source; the candidate boundary line can be the boundary of the area where the risk value corresponding to the high risk level threshold is greater than 0.8; the preset risk level threshold can be divided according to the loss degree of historical pollution events, for example, divided into low risk 0~0.3, medium risk 0.3~0.6, and high risk 0.6~1.

[0175] Step C20, output water quality monitoring warning signals according to the risk gradient mutation boundary of the risk distribution map.

[0176] In the embodiment, by analyzing the continuous risk gradient field obtained by gradient smoothing algorithm in the risk distribution map in real time, the spatial gradient operator is applied to detect the risk gradient mutation boundary position of sharp change in risk value.

[0177] When the detected risk gradient mutation boundary strength exceeds the preset mutation threshold, the geographical coordinate set covered by the boundary is automatically extracted, and the topological network vertices mapped by the water quality monitoring entities in the spatial range thereof are associated. The risk difference inside and outside the risk gradient mutation boundary and the pollutant type identified by the associated abnormal feature code set generate a water quality monitoring warning signal containing a spatial positioning signal, a pollution feature signal and a timeliness signal.

[0178] Among them, the spatial positioning signal is formed by extending the circular warning area with the risk gradient mutation boundary geographical coordinates as the center; the pollution feature signal marks the dominant pollution parameter and risk level; the timeliness signal predicts the pollution diffusion arrival time according to the boundary gradient change rate.

[0179] Optionally, step C20 can include steps C21-C23:

[0180] Step C21, analyze the risk gradient mutation boundary of the risk distribution map to obtain boundary feature attributes and risk transmission attributes.

[0181] The spatial gradient operator is used to calculate the continuous risk gradient field to detect the risk gradient mutation boundary position, and the spatial geometric features of the boundary are extracted as the boundary feature attributes, including the boundary shape type, the gradient intensity distribution extreme point, the boundary line total length and the curvature average. At the same time, based on the angle between the boundary trend and the hydrological flow direction, the risk transmission attributes are calculated, and the transmission direction, the transmission stability and the dominant pollution type determined by the associated abnormal feature code set are extracted.

[0182] The boundary feature attributes quantify the spatial abnormal pattern features, and the risk transmission attributes represent the dynamic behavior mode of pollution diffusion, which provide dual decision inputs of spatial structure and transmission dynamics for subsequent water quality monitoring warning signal generation.

[0183] Among them, the boundary shape type can be tongue-shaped protrusion or ring-shaped closure; the transmission direction can be unidirectional downstream or reverse diffusion; the transmission stability is the standard deviation of the boundary gradient intensity; and the dominant pollution type can be heavy metal or dissolved oxygen collapse type.

[0184] Step C22, map the boundary feature attributes and risk transmission attributes to pollution behavior semantics through behavior semantics mapping.

[0185] Specifically, by calling the preset pollution behavior semantic rule library, the boundary shape type in the boundary feature attribute is matched to the preset rule item with the conduction direction in the risk conduction attribute, for example, tongue-shaped protrusion + downstream unidirectional mapping of industrial pollution is "industrial pollution downstream protruding behavior", and the composite condition judgment is performed by combining the conduction stability and the dominant pollution type, for example, high stability + dissolved oxygen collapse type is mapped as "biochemical oxygen consumption type pollution retention behavior".

[0186] The mapping process converts the geometric features and dynamic conduction modes into understandable natural language descriptions through a real-time semantic matching engine, and outputs pollution behavior semantics such as "heavy metal reverse osmosis behavior", "algal bloom ring diffusion behavior", and "industrial pollution downstream protruding behavior". The pollution behavior semantics are directly associated with the early warning instruction generation module to realize the explainable expression of the pollution situation.

[0187] Step C23, based on the pollution behavior semantic matching response strategy, after generating the early warning instruction, a water quality monitoring early warning signal is output according to the early warning instruction.

[0188] The pollution behavior semantics output by the behavior semantic mapping are matched with the associated response rules in the preset response strategy library, for example, the industrial pollution protruding behavior triggers a three-level interception response, and the early warning instruction set including the target space positioning instruction, the pollution feature regulation instruction and the timeliness instruction is generated.

[0189] The spatial positioning instruction of the early warning instruction is used to determine the coordinate range of the circular early warning area, the pollution feature regulation instruction is used to load the dominant pollution parameter list and the risk level label, and the diffusion time prediction model is driven by the timeliness instruction to output the pollution arrival time.

[0190] Finally, the above elements are integrated to generate a standardized water quality monitoring early warning signal. The water quality monitoring early warning signal is pushed to the supervision terminal in real time through the Internet of Things protocol and simultaneously triggers the automatic sampling equipment to perform targeted monitoring, realizing the closed-loop conversion from pollution behavior semantics to executable early warning signals.

[0191] The application also provides a water quality automatic monitoring system, as shown in Figure 6 The water quality automatic monitoring system comprises:

[0192] The data coupling module 10 is used for coupling the dynamic coupling strength between each two of the plurality of collected water quality monitoring data, screening target water quality monitoring data with a dynamic coupling strength exceeding a preset coupling strength, generating edges with directionality on the network vertex mapped by the target water quality monitoring data, and outputting a topological network graph;

[0193] The data matching module 20 is used for matching the topological network graph with a preset pollution mode feature library, determining a new abnormal mode, and outputting an abnormal feature code set;

[0194] The data decision module 30 is configured to obtain the pollution diffusion prediction result, fuse the pollution diffusion prediction result and the abnormal feature code set, and output the water quality monitoring early warning signal.

[0195] Optionally, the data coupling module 10 is further configured to:

[0196] perform cross-dimension correlation analysis on the change trend of any two water quality monitoring data, and calculate the dynamic coupling strength between the two water quality monitoring data;

[0197] when the dynamic coupling strength exceeds a preset coupling strength, generate a directional edge on the network vertex corresponding to each of the two target water quality monitoring data, and traverse multiple water quality monitoring data;

[0198] output the topological network graph based on the generated directional edges and the network vertices corresponding to the edges.

[0199] Optionally, the data coupling module 10 is further configured to:

[0200] identify the change trend of any two water quality monitoring data to obtain a first trend feature and a second trend feature;

[0201] obtain the consistency degree of the first trend feature and the second trend feature in the data change direction, the matching degree of the first trend feature and the second trend feature in the data change rate, and the coordination degree of the first trend feature and the second trend feature in the data fluctuation mode, and generate a comparison result;

[0202] calculate and output the dynamic coupling strength based on the comparison result.

[0203] Optionally, the data matching module 20 is further configured to:

[0204] traverse and analyze the topological network graph to obtain multiple key substructures;

[0205] perform dynamic correlation degree matching between each key substructure and a preset pollution mode feature library to obtain a correlation degree matching threshold;

[0206] obtain a target key substructure whose correlation degree matching threshold is lower than a preset correlation degree matching threshold, and generate a new abnormal mode based on the target key substructure;

[0207] aggregate the new abnormal mode and a fixed abnormal mode in the preset pollution mode feature library to output the abnormal feature code set.

[0208] Optionally, the data matching module 20 is further configured to:

[0209] traverse the topological network graph based on the gradient mutation of the edge weight, and extract a path with a gradient change amount exceeding a preset gradient change amount as a candidate path;

[0210] The adjacent nodes of each candidate path are aggregated to generate a plurality of key substructures representing the pollution transmission characteristics.

[0211] Optionally, the data matching module 20 is further configured to:

[0212] Based on the timing transmission characteristics, the associated paths between the nodes of the topology network graph are traversed and analyzed, and a transmission path with time delay characteristics is identified as a candidate path.

[0213] By backtracking the adjacent nodes of the candidate path, a pollution transmission cluster is generated and output as a plurality of key substructures.

[0214] Optionally, the data decision module 30 is configured to:

[0215] The confidence weight of the abnormal feature code set is corrected, and the corrected abnormal feature code set is spatially fused with the pollution diffusion prediction result to generate a risk distribution map.

[0216] According to the risk gradient mutation boundary of the risk distribution map, a water quality monitoring warning signal is output.

[0217] Optionally, the data decision module 30 is further configured to:

[0218] According to the spatial position correlation, the corrected abnormal feature code set is dynamically weighted and fused with the pollution diffusion prediction result to output a spatialized risk value distribution.

[0219] The spatialized risk value distribution is input into the risk gradient generator to construct a continuous spatialized risk value distribution.

[0220] According to the spatial variation characteristics of the continuous spatialized risk value distribution, a risk level boundary is extracted to obtain a risk distribution map.

[0221] Optionally, the data decision module 30 is further configured to:

[0222] The risk gradient mutation boundary of the risk distribution map is analyzed to obtain boundary feature attributes and risk transmission attributes.

[0223] The boundary feature attributes and risk transmission attributes are mapped to pollution behavior semantics through behavior semantic mapping.

[0224] After generating a warning instruction based on the pollution behavior semantic matching response strategy, a water quality monitoring warning signal is output according to the warning instruction.

[0225] The above are preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, therefore: any equivalent changes made in the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method of automatically monitoring water quality, characterized by, The method comprises the following steps: By calculating the dynamic coupling strength between each pair of collected water quality monitoring data, target water quality monitoring data with dynamic coupling strength exceeding a preset coupling strength is screened out, edges with directionality are generated on the network vertices corresponding to the two target water quality monitoring data, and a topological network graph is output; The topological network graph is matched with a preset pollution mode feature library to determine a new abnormal mode, and an abnormal feature code set is output; A pollution diffusion prediction result is obtained, and the pollution diffusion prediction result and the abnormal feature code set are fused to output a water quality monitoring early warning signal; The step of calculating the dynamic coupling strength between each pair of collected water quality monitoring data, target water quality monitoring data with dynamic coupling strength exceeding a preset coupling strength is screened out, edges with directionality are generated on the network vertices corresponding to the two target water quality monitoring data, and a topological network graph is output comprises the following steps: Cross-dimension correlation analysis is performed on the change trend of any two water quality monitoring data, and the dynamic coupling strength between the two water quality monitoring data is calculated; When the dynamic coupling strength exceeds the preset coupling strength, edges with directionality are generated on the network vertices corresponding to the two target water quality monitoring data, and the multiple water quality monitoring data are traversed; The topological network graph is output according to the generated multiple edges with directionality and the network vertices corresponding to the edges; The step of performing cross-dimension correlation analysis on the change trend of any two water quality monitoring data and calculating the dynamic coupling strength between the two water quality monitoring data comprises the following steps: The change trend of the two water quality monitoring data is identified to obtain first trend features and second trend features; The consistency degree in the data change direction, the matching degree in the data change rate, and the coordination degree in the data fluctuation mode of the first trend features and the second trend features are obtained, and a comparison result is generated; The dynamic coupling strength is calculated and output according to the comparison result.

2. The water quality automatic monitoring method according to claim 1, characterized in that, The step of matching the topological network graph with a preset pollution mode feature library to determine a new abnormal mode and output an abnormal feature code set comprises the following steps: The topological network graph is traversed and parsed to obtain multiple key substructures; Each key substructure is dynamically associated with the preset pollution mode feature library to obtain an association degree matching threshold; A target key substructure with an association degree matching threshold lower than a preset association degree matching threshold is obtained, and the new abnormal mode is generated based on the target key substructure; The new abnormal mode and a fixed abnormal mode in the preset pollution mode feature library are aggregated to output the abnormal feature code set.

3. The water quality automatic monitoring method according to claim 2, characterized in that, The step of traversing the topological network graph and parsing to obtain multiple key substructures comprises the following steps: The topological network graph is traversed based on edge weight gradient mutation, and a path with a gradient change greater than a preset gradient change is extracted as a candidate path; Adjacent nodes of each candidate path are aggregated to generate multiple key substructures representing pollution transmission characteristics.

4. The water quality automatic monitoring method according to claim 2, characterized by, The step of traversing the topological network graph and parsing to obtain multiple key substructures comprises the following steps: Based on the time sequence conduction characteristics, the correlation paths between nodes of the topology network graph are analyzed, and a conduction path with time delay characteristics is identified as a candidate path; By backtracking the adjacent nodes of the candidate path, a pollution conduction cluster is generated and output as a plurality of key substructures.

5. The water quality automatic monitoring method according to claim 1, characterized in that, The step of fusing the pollution diffusion prediction result and the abnormal feature code set to output the water quality monitoring warning signal comprises: The confidence weight of the abnormal feature code set is corrected, and the corrected abnormal feature code set is spatially fused with the pollution diffusion prediction result to generate a risk distribution map; According to the risk gradient mutation boundary of the risk distribution map, the water quality monitoring warning signal is output.

6. The water quality automatic monitoring method according to claim 5, characterized in that, The step of fusing the pollution diffusion prediction result and the abnormal feature code set to output the water quality monitoring warning signal comprises: According to the spatial position correlation, the corrected abnormal feature code set and the pollution diffusion prediction result are dynamically weighted and fused to output a spatial risk value distribution; The spatial risk value distribution is input into a risk gradient generator to construct a continuous spatial risk value distribution; According to the spatial variation characteristics of the continuous spatial risk value distribution, a risk level boundary is extracted to obtain the risk distribution map.

7. The water quality automatic monitoring method according to claim 5, characterized by, The step of fusing the pollution diffusion prediction result and the abnormal feature code set to output the water quality monitoring warning signal comprises: The risk gradient mutation boundary of the risk distribution map is analyzed to obtain boundary feature attributes and risk conduction attributes; Through behavior semantic mapping, the boundary feature attributes and the risk conduction attributes are mapped into pollution behavior semantics; Based on the pollution behavior semantics matching response strategy, a warning instruction is generated, and the water quality monitoring warning signal is output according to the warning instruction.

8. An automatic water quality monitoring system, characterized by, It comprises: A data coupling module is configured to calculate the dynamic coupling strength between each pair of collected water quality monitoring data, filter target water quality monitoring data with a dynamic coupling strength exceeding a preset coupling strength, generate edges with directionality on network vertices mapped by the target water quality monitoring data, and output a topology network graph; A data matching module is configured to match the topology network graph with a preset pollution mode feature library, determine a new abnormal mode, and output an abnormal feature code set; A data decision module is configured to obtain a pollution diffusion prediction result, fuse the pollution diffusion prediction result and the abnormal feature code set, and output a water quality monitoring warning signal; The data coupling module is further configured to perform cross-dimension correlation analysis on the change trend of any two water quality monitoring data, and calculate the dynamic coupling strength between the two water quality monitoring data; When the dynamic coupling strength exceeds the preset coupling strength, edges with directionality are generated on the network vertices mapped by the corresponding two target water quality monitoring data, and the process is repeated until all the water quality monitoring data are traversed; According to the generated edges with directionality and the network vertices corresponding to the edges, the topology network graph is output; The data coupling module is further configured to identify the change trend of the any two water quality monitoring data to obtain first trend characteristics and second trend characteristics; obtaining a consistency degree of the first trend feature and the second trend feature in a data change direction, a matching degree of the first trend feature and the second trend feature in a data change rate, and a coordination degree of the first trend feature and the second trend feature in a data fluctuation mode, and generating a comparison result; calculating and outputting the dynamic coupling strength according to the comparison result.

Images