A monitoring point arrangement method and system for sudden water pollution events of a canal

Abstract

The application discloses a kind of canal sudden water pollution event monitoring distribution point method and system, it is related to pollution monitoring technical field.For the lack of dynamic prediction of pollution plume in prior art, the deficiency of monitoring point information amount evaluation and scheme dynamic adjustment, the application predicts concentration field by water power-water quality coupling model, solves peak migration trajectory and front edge front;Based on information entropy, the information gain of candidate point is calculated, and the information amount map is generated by combining the trajectory and the front weight;Under the constraint of resources, the optimal monitoring point combination is solved by genetic algorithm;And through ensemble Kalman filtering assimilation real-time data, form prediction-evaluation-distribution point-sampling-assimilation rolling closed loop.The application improves the capture efficiency of pollution peak and boundary and emergency dynamic adaptability under limited resources.

Description

Technical Field

[0001] This invention relates to the field of canal pollution monitoring technology, and in particular to a monitoring point deployment method and system for sudden water pollution events in canals. Background Technology

[0002] As inland shipping routes and inter-basin water transfer carriers, canals play a crucial role in ensuring the water environment safety of industrial and agricultural production, residents' lives, and the stable operation of ecosystems along their routes. With the increasingly dense industrial layout along canals and the continuous growth in the volume of hazardous chemical transportation, the risk of sudden water pollution incidents has significantly increased, placing higher demands on emergency monitoring and rapid response capabilities. After a sudden pollution incident, quickly and accurately obtaining the peak concentration, migration path, and impact boundary of the pollution plume is a prerequisite for scientifically formulating emergency response plans and protecting sensitive downstream targets. Therefore, developing methods that can dynamically track the migration of pollution plumes and optimize the deployment of monitoring forces accordingly has broad application prospects.

[0003] The existing monitoring point deployment method has the following shortcomings: First, it lacks quantitative prediction of the spatiotemporal evolution of the pollution plume, making it difficult to accurately predict the peak position and boundary advance front before the pollution plume arrives, which can easily lead to monitoring points being spatially deviated from the core pollution area; Second, it does not quantitatively assess the information acquisition capabilities of candidate monitoring points, making it impossible to determine which locations can minimize the uncertainty of pollution distribution under limited monitoring resources, resulting in the omission of key pollution information; Third, once the monitoring plan is determined, it is difficult to dynamically adjust it based on on-site sampling data, making it impossible to form a rolling closed loop of prediction-sampling-updating, resulting in insufficient timeliness and targeting of emergency response.

[0004] Therefore, this invention is proposed. Summary of the Invention

[0005] To address these issues, the present invention provides a monitoring point deployment method and system for sudden water pollution events in canals, which overcomes the problems of existing methods that lack quantitative prediction of the peak value and spatiotemporal evolution of pollution plumes, cannot quantitatively assess the information acquisition capabilities of candidate monitoring points, and cannot dynamically update the monitoring scheme based on on-site sampling data.

[0006] To achieve the above objectives, the present invention provides a method for monitoring and deploying monitoring points for sudden water pollution incidents in canals, comprising: S1. Obtain pollution source information and real-time hydrological data of the canal. Based on the real-time hydrological data of the canal, construct and use the pollution source information as input to drive the hydrodynamic-water quality coupling model, make positive predictions on the spatiotemporal distribution of pollutants, and generate the pollutant concentration field for the target period. S2, from the pollutant concentration field, search the spatial location of the maximum pollutant concentration along the water flow direction time step by time to form the peak migration trajectory of the pollutant plume; extract the set of spatial locations where the concentration value first reaches the preset concentration threshold at each time moment to form the leading edge of the pollutant plume; S3. Generate a set of candidate monitoring points based on the leading edge front and peak migration trajectory, calculate the information gain of each candidate monitoring point based on the pollutant concentration field, and determine the comprehensive monitoring value of each candidate monitoring point by combining the peak migration trajectory and the area covered by the leading edge front, and generate an information map. S4, based on the information map, aims to maximize the total information gain of candidate monitoring points under given monitoring resource constraints, and uses a genetic algorithm to determine the final monitoring points in order to capture the peak concentration and boundary range of the pollution plume; S5, conduct on-site sampling according to the final monitoring point to obtain real-time monitoring data; use data assimilation technology to integrate the real-time monitoring data into the hydrodynamic-water quality coupling model, update the state variables and parameters of the hydrodynamic-water quality coupling model, and return to S1 to perform the next round of prediction and monitoring point adjustment until the leading edge front disappears.

[0007] Furthermore, S3 also includes: S3a discretizes the pollutant concentration field into a spatiotemporal grid and calculates the information entropy of the pollutant concentration value in each spatiotemporal grid. S3b, the total information entropy reduction after including the candidate monitoring points in the candidate monitoring point set is used as the information gain of the candidate points; S3c, increase the weight of the spatiotemporal grid covered by the peak migration trajectory and the leading edge front, multiply the information gain of the spatiotemporal grid where each candidate monitoring point is located by the corresponding weight to obtain the comprehensive monitoring value of each candidate monitoring point, and generate the information map based on the comprehensive monitoring value.

[0008] Furthermore, the acquisition of pollution source information and real-time hydrological data of the canal, the construction of a hydrodynamic-water quality coupling model based on the real-time hydrological data of the canal and using the pollution source information as input to positively predict the spatiotemporal distribution of pollutants, and the generation of a pollutant concentration field for the target time period, includes: S11, Obtain pollution source information, including pollutant type, total leakage amount, leakage location, and leakage method; Obtain real-time hydrological data of the canal, including flow rate, water level, and flow velocity profile. S12, invoke the pre-built hydrodynamic-water quality coupling model of the canal, and set the initial conditions and boundary conditions of the model based on the real-time hydrological data of the canal; S13, if the pollution source leakage is instantaneous, the total leakage amount is released instantaneously at the leakage location at the initial moment; if the leakage is continuous, the total leakage amount is set as the river inflow flux boundary for the corresponding time series. S14 outputs a pollutant concentration field covering the target period in the future.

[0009] Furthermore, the candidate monitoring point set is the set of coordinates of the center points of the spatiotemporal grids through which the peak migration trajectory and the leading edge front pass, and whose information entropy is higher than a preset information entropy threshold.

[0010] Furthermore, based on the information map, and under given monitoring resource constraints, the final monitoring points are determined using a genetic algorithm with the objective of maximizing the total information gain of candidate monitoring points, in order to capture the peak concentration and boundary range of the pollution plume. This includes: S41, Obtain the spatial coordinates of each candidate monitoring point and its corresponding comprehensive monitoring value from the information map; S42, Set monitoring resource constraints, including the upper limit of the number of final monitoring points and the minimum distance between adjacent final monitoring points; S43, Based on the comprehensive monitoring value of the candidate monitoring points, and with the objective of maximizing the total information gain of the selected monitoring points while satisfying the monitoring resource constraints, a fitness function is constructed. S44. A genetic algorithm is used to encode the selection scheme of candidate monitoring points. The population is iteratively optimized through selection, crossover and mutation operations to find the optimal combination of monitoring points that satisfies the fitness function, which is then used as the final monitoring point.

[0011] Furthermore, based on the comprehensive monitoring value of the candidate monitoring points, and with the objective of maximizing the total information gain of the selected monitoring points while satisfying the monitoring resource constraints, a fitness function is constructed, including: S431, obtain the information gain corresponding to each candidate monitoring point, wherein the information gain is the total information entropy reduced after including the candidate points in the monitoring; S432, For any selection scheme containing several candidate monitoring points, sum the information gain values ​​of all candidate monitoring points in the scheme to obtain the total information gain of the scheme; S433 defines a penalty function: when the number of selected monitoring points exceeds the upper limit, the penalty value increases with the number of points exceeding the limit; when there are adjacent points among the selected monitoring points whose distance is less than the minimum distance, the penalty value increases with the number of point pairs that violate the distance constraint. S434, Construct a fitness function. The fitness function is the total information gain minus the penalty value of the penalty function. The optimization objective of the genetic algorithm is to maximize the fitness function value.

[0012] Furthermore, the process involves on-site sampling at the final monitoring points to obtain real-time monitoring data; using data assimilation technology to integrate the real-time monitoring data into the hydrodynamic-water quality coupled model, updating the model's state variables and parameters, and returning to S1 for the next round of prediction and monitoring point adjustment, until the leading front disappears, including: S51, At the spatial coordinates determined at the final monitoring point, on-site sampling is carried out according to the preset sampling frequency to obtain real-time concentration data of pollutants at each final monitoring point; S52, using the ensemble Kalman filter method, the real-time concentration data is used as an observation and integrated into the hydrodynamic-water quality coupling model to update the state variables and parameters in the pollutant concentration field, including the longitudinal dispersion coefficient and degradation rate; S53, using the updated state variables and parameters as initial conditions, returns to S1 to perform the next round of positive prediction; S54. Repeat steps S1 to S53 until the concentration values ​​at all spatial locations in the pollutant concentration field are lower than the preset concentration threshold, i.e., the leading edge front disappears, then terminate the monitoring point adjustment process.

[0013] Furthermore, the information map is a monitoring value distribution map constructed in the spatial dimension using the spatial coordinates of each candidate monitoring point as an index and the comprehensive monitoring value corresponding to each candidate monitoring point as an attribute value, which is used to provide a quantitative basis for spatial decision-making.

[0014] This invention also provides a monitoring deployment system for sudden water pollution incidents in canals, the system being used to implement the monitoring deployment method for sudden water pollution incidents in canals as described in any one of the claims, the system comprising: The model prediction module is used to acquire pollution source information and real-time hydrological data of the canal. Based on the real-time hydrological data of the canal, a hydrodynamic-water quality coupling model is constructed and driven by the pollution source information as input. The model makes a positive prediction of the spatiotemporal distribution of pollutants and generates the pollutant concentration field for the target period. The pollution plume calculation module is used to search the spatial location of the maximum pollutant concentration in the pollutant concentration field along the water flow direction step by step, forming the peak migration trajectory of the pollution plume; and extracting the set of spatial locations where the concentration value first reaches the preset concentration threshold at each moment, forming the leading edge of the pollution plume. The monitoring value quantification module is used to generate a set of candidate monitoring points based on the leading front and peak migration trajectory, calculate the information gain of each candidate monitoring point based on the pollutant concentration field, and determine the comprehensive monitoring value of each candidate monitoring point by combining the peak migration trajectory and the area covered by the leading front, and generate an information map. The monitoring optimization deployment module is used to determine the final monitoring points based on the information map, under given monitoring resource constraints, with the goal of maximizing the total information gain of candidate monitoring points, in order to capture the peak concentration and boundary range of the pollution plume; The dynamic update module is used to perform on-site sampling according to the final monitoring point to obtain real-time monitoring data; it uses data assimilation technology to integrate the real-time monitoring data into the hydrodynamic-water quality coupling model, updates the state variables and parameters of the hydrodynamic-water quality coupling model, and sends the updated state variables and parameters back to the model prediction module for the next round of prediction and monitoring point adjustment until the leading edge front disappears.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: Firstly, this invention discretizes the pollutant concentration field into a spatiotemporal grid and calculates the information entropy of the concentration values ​​within each grid. The reduction in the total information entropy of the system after a candidate monitoring point is included in the monitoring is used as the information gain for that point. Furthermore, the spatiotemporal grids covered by the peak migration trajectory and the leading front are given increased weights and multiplied with the information gain to obtain the comprehensive monitoring value. This generates an information map with spatial coordinates as the index and comprehensive monitoring value as the attribute value. This solves the problems in existing emergency monitoring methods, such as the inability to quantitatively assess the contribution of each potential monitoring location to eliminating the uncertainty of the spatiotemporal distribution of pollution, and the failure to include the peak value and boundary of the pollution plume as priority capture targets in the site selection decision. This improves the capture efficiency of the site selection scheme for peak pollution concentration and impact boundary under limited monitoring resources, and reduces the risk of missing key pollution information due to the monitoring point being deviating from the core pollution area or the leading front. Secondly, this invention sets monitoring resource constraints, including an upper limit on the number of monitoring points and a minimum distance between adjacent points. Based on the information gain of candidate monitoring points, it constructs a fitness function that integrates quantity constraint penalties and distance constraint penalties. It then uses a genetic algorithm to iteratively optimize the selection scheme of candidate monitoring points through encoding, selection, crossover, and mutation, and solves the optimal combination of monitoring points that maximizes the total information gain under the constraints. This solves the problem in existing empirical point placement methods that cannot perform global optimization of candidate monitoring points under limited monitoring resource constraints, resulting in high information redundancy between monitoring points or missing coverage of key locations. It improves the comprehensive capture efficiency of a limited number of monitoring points for the peak concentration and boundary range of pollution plumes, and reduces the risk of missing emergency decision-making information due to unreasonable spatial layout of monitoring points. Third, this invention organically integrates the positive prediction of the hydrodynamic-water quality coupling model, the dynamic calculation of pollution plume characteristics, the quantification of monitoring value based on information entropy, the optimized placement of monitoring points using genetic algorithms, and the data assimilation and update using ensemble Kalman filtering into a rolling closed loop of prediction-evaluation-placement-sampling-assimilation-re-prediction. This solves the problems in existing methods where the monitoring scheme cannot be dynamically adjusted based on field sampling data once determined, and the timeliness and specificity of emergency response are insufficient. It improves the dynamic adaptability and accuracy of the entire process of emergency monitoring of sudden water pollution events and reduces the risk of monitoring schemes continuously failing due to deviations from the initial prediction in the spatiotemporal evolution of the pollution plume. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a method for monitoring and deploying monitoring points for a sudden water pollution incident in a canal, as provided in an embodiment of the present invention; Figure 2 This is a structural block diagram of a monitoring deployment system for a sudden water pollution incident in a canal, provided as an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0019] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0020] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0021] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0022] Example 1 like Figure 1 As shown, this invention provides a method for monitoring and deploying monitoring points for sudden water pollution incidents in canals, comprising: S1. Obtain pollution source information and real-time hydrological data of the canal. Based on the real-time hydrological data of the canal, construct a hydrodynamic-water quality coupling model using pollution source information as input to positively predict the spatiotemporal distribution of pollutants and generate a pollutant concentration field for the target time period, including: S11, Obtain pollution source information, including pollutant type, total leakage amount, leakage location, and leakage method; Obtain real-time hydrological data of the canal, including flow rate, water level, and flow velocity profile. S12, invoke the pre-built hydrodynamic-water quality coupling model of the canal, and set the initial conditions and boundary conditions of the model based on the real-time hydrological data of the canal; S13, if the pollution source leakage is instantaneous, the total leakage amount is released instantaneously at the leakage location at the initial moment; if the leakage is continuous, the total leakage amount is set as the river inflow flux boundary for the corresponding time series. S14 outputs a pollutant concentration field covering the target period in the future.

[0023] In one possible implementation, this embodiment takes a sudden pollution event of phenol leakage caused by a collision of transport ships in a canal as the application scenario. After the accident, the following pollution source information is obtained: the pollutant is phenol, the total amount of leakage is about 500 kg, the leakage location is at chainage K15 of the canal, and the leakage method is instantaneous leakage, that is, all the phenol in the cargo hold enters the water body in a short time after the ship collision.

[0024] Meanwhile, real-time hydrological data of the canal at the moment of the accident was obtained through an online hydrological monitoring system deployed along the canal: the river flow rate was 45 cubic meters per second, the water level at the leak location was 3.2 meters, and the average flow velocity at the cross-section was 0.8 meters per second. The online hydrological monitoring system includes shore-based radar level gauges and acoustic Doppler current profilers, and transmits data in real time via a network.

[0025] A pre-built hydrodynamic-water quality coupled model of the canal, constructed using a hydrodynamic simulation software platform, was invoked. This model included cross-sectional data of approximately 80 kilometers of the canal's channel topography, roughness parameters for each cross-section (calibrated using historical hydrological data, the roughness coefficient n for this section is approximately 0.025–0.032), and parameters for the degradation rate of phenol in the water. Based on the determined leakage mode as instantaneous, a total leakage of 500 kg of phenol was set as the instantaneous mass release at the leakage location at the initial moment of the model. Specifically, in the model input file, the corresponding calculation node of the channel cross-section where the leakage location was located was set as the pollutant concentration source term at the initial moment, and its initial concentration value was calculated by dividing the total leakage amount by the water volume controlled by that node. Real-time hydrological data of the canal was used as boundary conditions input into the model. Specifically, the upstream boundary was set as a flow boundary with a value of 45 cubic meters per second; the downstream boundary was set as a water level boundary with a value of 3.2 meters. The time of the accident is taken as the zero point of the model calculation, the model calculation time step is set to 5 minutes, and the total simulation time is the next 24 hours.

[0026] It should be noted that if the leakage is continuous, such as continuous leakage from a ruptured pipeline, the total leakage amount is allocated according to the leakage duration and set as the boundary condition for the inflow flux of the corresponding time series. That is, the mass flux of pollutants corresponding to the leakage rate is input to the node in each time step.

[0027] The driving model performs forward prediction calculations, outputting the phenol concentration values ​​at each calculation node along the entire river at each time step within the next 24 hours from the initial moment, forming a pollutant concentration field covering both spatial and temporal dimensions. The target time period, i.e., 24 hours, is determined based on the river flow velocity of 0.8 m / s and the nearest downstream water intake distance of 60 km, ensuring that the entire process of the pollution plume migrating from the leak location to downstream sensitive targets is fully covered.

[0028] The output format of the pollutant concentration field is a table of concentration distribution data for each cross section arranged in time series, including fields such as time label, river station number, and average concentration of the cross section.

[0029] This invention acquires pollution source information, including pollutant type, total leakage amount, leakage location, and leakage mode, as well as real-time hydrological data of the canal, including flow rate, water level, and velocity profiles. It generalizes the pollution source information into model input conditions for instantaneous mass release or time-series inflow flux boundaries according to different leakage modes, such as instantaneous or continuous leakage. Based on the real-time hydrological data of the canal, it sets the initial and boundary conditions of the model to drive the hydrodynamic-water quality coupled model for positive prediction. This solves the problems in existing methods, such as the mismatch between pollution source input conditions and actual leakage modes, and the dependence of model boundary conditions on historical data rather than real-time hydrological information, which leads to prediction results deviating from the actual diffusion process. It improves the consistency between the predicted pollutant concentration field and the actual spatiotemporal evolution of the pollution plume.

[0030] S2, from the pollutant concentration field, search the spatial location of the maximum pollutant concentration along the water flow direction time step by time to form the peak migration trajectory of the pollutant plume; extract the set of spatial locations where the concentration value first reaches the preset concentration threshold at each time moment to form the leading edge of the pollutant plume; In one possible implementation, a preset concentration threshold is determined for extracting the leading edge front. According to the specific project standard limit for centralized drinking water surface water sources, the limit for phenol is 0.01 mg / L. Therefore, the preset concentration threshold is set to 0.01 mg / L, meaning that when the phenol concentration at a certain cross-section first reaches or exceeds 0.01 mg / L, the leading edge of the pollution plume is considered to have reached that cross-section.

[0031] The output concentration field data is scanned time-step by time. Specifically, the phenol concentration values ​​of each calculated section of the entire river channel are obtained at each output time step. Each section is traversed along the flow direction, comparing the concentration values ​​of each section within the current time step, and recording the maximum concentration and its corresponding river channel station number. For example, at a time step of 2 hours after the leak, the concentration values ​​along the river channel section are successively 0.12 mg / L, 0.38 mg / L, 0.41 mg / L, 0.35 mg / L, ..., then the peak position at that moment is station number K15, and the peak concentration is 0.41 mg / L. Connecting the peak spatial positions of each time step in chronological order constitutes the peak migration trajectory of the pollution plume. The resulting trajectory can be represented as a set of stations on a time series. The concentration value of each section is compared with a preset concentration threshold one by one along the flow direction from upstream to downstream. The section with the first concentration value reaching or exceeding 0.01 mg / L is marked as the leading edge point at that moment, and its river channel station number is recorded. For example, one hour after the leak, taking K15 as an example, the concentration at the cross-section from upstream is 0.000, 0.000, ..., 0.006, 0.009, 0.012 mg / L, then station K15+000 is the leading edge point at this time. If the concentration at all cross-sections is below the threshold at this moment, then there is no leading edge point at that moment. Connecting the station numbers of the leading edge points at all times in chronological order forms the leading front of the pollution plume, represented as a set of station numbers in a time series. Finally, the peak migration trajectory and the leading front are output, depicting the spatiotemporal path of the pollution plume core moving along the canal and defining the dynamic boundary of the pollution impact range extending downstream.

[0032] This invention searches for the spatial location of the maximum concentration along the water flow direction from the pollutant concentration field at time steps to form the peak migration trajectory of the pollution plume. At the same time, it extracts the set of spatial locations where the concentration value first reaches the preset concentration threshold at each moment to form the leading edge front of the pollution plume. This solves the problem in existing emergency monitoring methods that can only obtain discrete cross-sectional concentrations and cannot dynamically and synchronously quantitatively describe the migration path of the pollution plume core and the advancement process of the influence boundary. This improves the calculation accuracy of the peak position and the spatiotemporal evolution process of the pollution plume leading edge.

[0033] S3. Generate a set of candidate monitoring points based on the leading edge front and peak migration trajectory, calculate the information gain of each candidate monitoring point based on the pollutant concentration field, and determine the comprehensive monitoring value of each candidate monitoring point by combining the peak migration trajectory and the area covered by the leading edge front, and generate an information map. S3 also includes: S3a discretizes the pollutant concentration field into a spatiotemporal grid and calculates the information entropy of the pollutant concentration value in each spatiotemporal grid. S3b, the total information entropy reduction after including candidate monitoring points in the candidate monitoring point set is used as the information gain of the candidate points; S3c, increase the weight of the spatiotemporal grid covered by the peak migration trajectory and the leading edge front, multiply the information gain of the spatiotemporal grid where each candidate monitoring point is located by the corresponding weight to obtain the comprehensive monitoring value of each candidate monitoring point, and generate the information map based on the comprehensive monitoring value.

[0034] The candidate monitoring point set is the set of coordinates of the center points of the spatiotemporal grids through which the peak migration trajectory and the leading edge front pass, and whose information entropy is higher than a preset information entropy threshold.

[0035] The information map is a monitoring value distribution map constructed in a spatial dimension, using the spatial coordinates of each candidate monitoring point as an index and the comprehensive monitoring value corresponding to each candidate monitoring point as an attribute value. It is used to provide a quantitative basis for spatial decision-making.

[0036] In one possible implementation, the pollutant concentration field is gridded. Spatially, with the longitudinal direction of the canal as the x-axis, the entire 80-kilometer waterway is divided into 400 spatial grids at 200-meter intervals. Temporally, the next 24 hours are divided into 288 temporal grids at 5-minute intervals. This results in a total of 115,200 spatiotemporal grid units, each corresponding to a unique spatial location interval and a time interval.

[0037] Extract the phenol concentration values ​​of all computational nodes within each spatiotemporal grid to form the grid's concentration value set. Analyze the distribution characteristics of the concentration value set, divide the concentration value range into several intervals, calculate the probability p of the concentration value occurring in each interval, and then apply the formula: The information entropy value H of the spatiotemporal grid is calculated. A higher information entropy value indicates a more uneven concentration distribution and greater uncertainty within the spatiotemporal grid, and thus a greater potential for obtaining information through monitoring. For example, in the second hour after the leak, the spatial grid at chainages K15 to K16 covers the area where the pollution plume transitions from a low-concentration zone to a high-concentration core zone, with concentration values ​​spanning multiple orders of magnitude; the calculated information entropy value for this grid is 4.2 bits. In contrast, the upstream grid at chainages K5 to K6, far from the pollution plume, consistently shows concentration values ​​close to zero, with an information entropy value of only 0.3 bits. This indicates that the former has a higher potential for monitoring value.

[0038] The peak migration trajectory and all spatiotemporal grids traversed by the leading edge front at each time point are extracted. Grids with information entropy values ​​higher than a preset information entropy threshold are selected from these spatiotemporal grids; in this embodiment, the information entropy threshold is set to 1.0 bits. The center point coordinates of each selected spatiotemporal grid are used as candidate monitoring points to form a candidate monitoring point set. Following this rule, this embodiment selects approximately 860 candidate monitoring points from 115,200 spatiotemporal grids.

[0039] For each candidate point i in the candidate monitoring point set, calculate its information gain ΔH. i Specifically, after simulating the inclusion of the candidate point into the monitoring network, the predicted concentration value of that point is used to constrain the uncertainty of the concentration distribution in the surrounding spatiotemporal grid. The total information entropy of the entire system before the update and the total information entropy after the update are calculated, and the difference ΔH between the two is calculated. i This is the information gain of the candidate point. ΔH i The larger the value, the more system uncertainties can be eliminated by deploying the monitoring point, and the stronger its information acquisition capability.

[0040] To prioritize the capture of peak concentrations and boundary ranges of pollution plumes, different weights are assigned to spatiotemporal grids located in different regions. Specifically, spatiotemporal grids located within the coverage of both the peak migration trajectory and the leading front are assigned a weight of 1.5; those located within only one of them are assigned a weight of 1.2; and all other spatiotemporal grids are assigned a weight of 1.0. The calculated information gain value of each candidate monitoring point is multiplied by the weight corresponding to its spatiotemporal grid to obtain the comprehensive monitoring value of that candidate monitoring point. For example, the candidate monitoring point at station K15 is located within both the peak migration trajectory and the leading front coverage, and its comprehensive monitoring value is 3.8 × 1.5 = 5.70; the candidate monitoring point at station K18 is located only within the leading front coverage, and its comprehensive monitoring value is 2.1 × 1.2 = 2.52.

[0041] Using the spatial coordinates of each candidate monitoring point as an index, the corresponding comprehensive monitoring value is assigned as an attribute value to that spatial coordinate, and a monitoring value distribution map is constructed along the longitudinal direction of the canal, thus obtaining the information map. This map intuitively displays the priority of monitoring deployment at various spatial locations along the canal; locations with higher monitoring value should be prioritized for monitoring point deployment.

[0042] This invention discretizes the pollutant concentration field into a spatiotemporal grid and calculates the information entropy of the concentration values ​​within each grid. The reduction in the total information entropy of the system after a candidate monitoring point is included in the monitoring is used as the information gain for that point. The spatiotemporal grids covered by the peak migration trajectory and the leading front are given increased weights and multiplied with the information gain to obtain the comprehensive monitoring value. This generates an information map with spatial coordinates as indexes and comprehensive monitoring value as attribute values. This solves the problems in existing emergency monitoring methods, such as the inability to quantitatively assess the contribution of each potential monitoring location to eliminating the uncertainty of the spatiotemporal distribution of pollution, and the failure to include the peak value and boundary of the pollution plume as priority capture targets in the site selection decision. It improves the capture efficiency of the site selection scheme for peak pollution concentration and influence boundary under limited monitoring resources, and reduces the risk of missing key pollution information due to the monitoring point being deviating from the core pollution area or the leading front.

[0043] S4, based on an information map, aims to maximize the total information gain of candidate monitoring points under given monitoring resource constraints. It uses a genetic algorithm to determine the final monitoring points to capture the peak concentration and boundary range of the pollution plume, including: S41, Obtain the spatial coordinates of each candidate monitoring point and its corresponding comprehensive monitoring value from the information map; S42, Set monitoring resource constraints, including the upper limit of the number of final monitoring points and the minimum distance between adjacent final monitoring points; S43, Based on the comprehensive monitoring value of the candidate monitoring points, and with the objective of maximizing the total information gain of the selected monitoring points while satisfying the monitoring resource constraints, a fitness function is constructed. S44. A genetic algorithm is used to encode the selection scheme of candidate monitoring points. The population is iteratively optimized through selection, crossover and mutation operations to find the optimal combination of monitoring points that satisfies the fitness function, which is then used as the final monitoring point.

[0044] Based on the comprehensive monitoring value of the candidate monitoring points, and with the objective of maximizing the total information gain of the selected monitoring points while satisfying the monitoring resource constraints, a fitness function is constructed, including: S431, obtain the information gain corresponding to each candidate monitoring point, wherein the information gain is the total information entropy reduced after including the candidate points in the monitoring; S432, For any selection scheme containing several candidate monitoring points, sum the information gain values ​​of all candidate monitoring points in the scheme to obtain the total information gain of the scheme; S433 defines a penalty function: when the number of selected monitoring points exceeds the upper limit, the penalty value increases with the number of points exceeding the limit; when there are adjacent points among the selected monitoring points whose distance is less than the minimum distance, the penalty value increases with the number of point pairs that violate the distance constraint. S434, Construct a fitness function. The fitness function is the total information gain minus the penalty value of the penalty function. The optimization objective of the genetic algorithm is to maximize the fitness function value.

[0045] In one possible implementation, data from all 860 candidate monitoring points are extracted from the generated information map. Each data point contains spatial coordinates and comprehensive monitoring value. Based on the actual monitoring resource allocation for this emergency response, two constraints are set: First, the emergency monitoring team can call upon a total of 8 sets of fixed monitoring equipment, so the maximum number of monitoring points is 8. Second, to avoid excessive information redundancy between adjacent monitoring points and to ensure the spatial coverage efficiency of the monitoring network, the minimum distance between adjacent final monitoring points is set to 1.0 kilometer.

[0046] Obtain the information gain corresponding to each candidate monitoring point. The information gain is the reduction in total information entropy after including the candidate point in the monitoring. Calculate the total information gain of the scheme. For any selection scheme containing m candidate monitoring points, sum the information gain values ​​of all candidate monitoring points in the scheme to obtain the total information gain of the scheme.

[0047] The penalty function is used to penalize schemes that violate constraints. The specific rules are as follows: when the number of monitoring points *m* selected in the scheme exceeds the upper limit (the excess is (m-8), and the penalty value increases by (m-8)×10; when there are adjacent monitoring points in the scheme whose river distance is less than the minimum spacing of 1.0 km, the penalty value increases by 5 for each pair of adjacent points violating the spacing constraint. If the scheme fails to satisfy both constraints simultaneously, the penalty value is 0.

[0048] The fitness function is calculated by subtracting the penalty value from the total information gain of the scheme. A larger fitness function value indicates a greater amount of total information that the scheme can acquire while satisfying the constraints, making the scheme more optimal. The genetic algorithm uses maximizing this fitness function value as its optimization objective. Specifically, it encodes the selection schemes for candidate monitoring points. A binary encoding method is used, with a chromosome length equal to the total number of candidate monitoring points (860). Each gene position on the chromosome corresponds to a candidate monitoring point; a gene value of 1 indicates that the candidate point is selected, and a gene value of 0 indicates that it is not selected. An initial population of 50 individuals is randomly generated, with each individual represented by a binary string of length 860, representing a monitoring point selection scheme. For each individual in the initial population, it is checked whether the quantity constraint is met. Individuals that do not meet the constraint are randomly repaired, i.e., monitoring points exceeding the allowed number are randomly removed to ensure the rationality of the initial population. The iterative optimization process then begins. Calculate the fitness function value of each individual in the current population; employ a roulette wheel selection strategy, where individuals with higher fitness values ​​have a greater probability of being selected as parents; perform a single-point crossover operation on the selected parent individuals, setting the crossover probability to 0.8, and a basic mutation operation, i.e., setting the mutation probability to 0.05, to generate the offspring population; impose penalties or perform repairs on individuals in the offspring population that violate constraints, and calculate the offspring population fitness; retain the elite individuals with the highest fitness to directly enter the next generation, and replace the remaining individuals with offspring.

[0049] Repeat the above iterative process, setting the maximum number of iterations to 200 generations. When the algorithm converges or reaches the maximum number of iterations, select the individual with the highest fitness value from the final population and decode to obtain the optimal combination of monitoring points. For example, after optimization by the genetic algorithm, the optimal combination of monitoring points includes 8 fixed monitoring points with spatial coordinates of K15+200, K15+800, K17+000, K18+600, K20+400, K22+800, K25+600, and K29+200. In this scheme, K15+200 and K15+800 are located in the initial mixing zone downstream of the leak location and the peak core zone on the peak migration trajectory, respectively. The monitoring points from K17+000 to K25+600 uniformly cover the main channel of the pollution plume migration along the peak migration direction at intervals of approximately 1.5 to 3.2 kilometers. K29+200 is located slightly ahead of the expected arrival location of the leading front and is used to capture the pollution boundary. The scheme has a total information gain of 24.3 bits, fully satisfies the quantity and spacing constraints, and the corresponding penalty value is 0.

[0050] This invention obtains the spatial coordinates and comprehensive monitoring value of each candidate monitoring point from an information map, sets monitoring resource constraints including an upper limit on the number of points and a minimum distance between adjacent points, constructs a fitness function that integrates quantity constraint penalties and distance constraint penalties based on the information gain of the candidate monitoring points, and uses a genetic algorithm to iteratively optimize the selection scheme of candidate monitoring points through encoding, selection, crossover, and mutation to find the optimal combination of monitoring points that maximizes the total information gain under the constraints. This solves the problem in existing empirical point placement methods that cannot perform global optimization of candidate monitoring points under limited monitoring resource constraints, resulting in high information redundancy between monitoring points or missing coverage of key locations. It improves the comprehensive capture efficiency of a limited number of monitoring points for peak concentration and boundary range of pollution plumes, and reduces the risk of missing emergency decision-making information due to unreasonable spatial layout of monitoring points.

[0051] S5, conduct on-site sampling according to the final monitoring points to obtain real-time monitoring data; use data assimilation technology to integrate the real-time monitoring data into the hydrodynamic-water quality coupling model, update the state variables and parameters of the hydrodynamic-water quality coupling model, and return to S1 to perform the next round of prediction and monitoring point adjustment until the leading front disappears, including: S51, At the spatial coordinates determined at the final monitoring point, on-site sampling is carried out according to the preset sampling frequency to obtain real-time concentration data of pollutants at each final monitoring point; S52, using the ensemble Kalman filter method, the real-time concentration data is used as an observation and integrated into the hydrodynamic-water quality coupling model to update the state variables and parameters in the pollutant concentration field, including the longitudinal dispersion coefficient and degradation rate; S53, using the updated state variables and parameters as initial conditions, returns to S1 to perform the next round of positive prediction; S54. Repeat steps S1 to S53 until the concentration values ​​at all spatial locations in the pollutant concentration field are lower than the preset concentration threshold, i.e., the leading edge front disappears, then terminate the monitoring point adjustment process.

[0052] In one possible implementation, the spatial coordinates of the eight monitoring points were transmitted to the site. Monitoring personnel deployed sampling points along the centerline of the river cross-section. The preset sampling frequency was once every 30 minutes. A portable spectrophotometer was used to measure the phenol concentration in the water samples on-site, and the data was transmitted back in real time via satellite communication. Monitoring data showed that the highest phenol concentration, 1.23 mg / L, was found at monitoring point 2 (K15+800), representing the peak of the pollution plume at the current moment. The concentration at monitoring point 5 (K20+400) was 0.03 mg / L, slightly higher than the preset concentration threshold of 0.01 mg / L, indicating that the leading edge of the pollution plume had advanced to the vicinity of this cross-section. The concentrations at monitoring points 6 to 8 were below the threshold or undetectable, indicating that the pollution impact had not yet reached the area approximately 22.8 kilometers downstream. The sampling results verified the predicted trends of the peak migration trajectory and the leading edge front, but there was a certain deviation between the absolute concentration values ​​and the model predictions.

[0053] To construct the background set of model states, specifically in the hydrodynamic-water quality coupled model, random perturbations are applied to key parameters and state variables to generate a model state set containing 50 members. The perturbed parameters include the longitudinal dispersion coefficient Ex and the phenol degradation rate k; the state variables are the phenol concentration values ​​at each calculation cross section. The longitudinal dispersion coefficient is a dimensionless parameter characterizing the intensity of diffusion and mixing of pollutants or solutes along the flow direction in open water flows such as rivers due to uneven flow velocity distribution at cross sections; it is essentially a shear-induced dispersion effect. The phenol degradation rate is the amount of phenol converted into other substances in water per unit time through natural processes such as biodegradation, photolysis, and hydrolysis, usually expressed as the decrease in phenol concentration per unit time, with dimensions of (concentration / time). In environmental hydraulics and water quality models, first-order reaction kinetics equations are often used to describe the phenol degradation process.

[0054] To calculate the Kalman gain, specifically, the simulated concentration values ​​of each member in the background set of the model state at eight monitoring points are compared with the actual observed values. The ensemble covariance matrix and observation error covariance matrix within the observation space are calculated, with the standard deviation of the observation error set to 10% of the measured value. The Kalman gain matrix is ​​then solved. Using the Kalman gain matrix, the concentration field state variables, longitudinal dispersion coefficients, and degradation rates of each member in the ensemble are updated synchronously. The average longitudinal dispersion coefficient is adjusted from 50 m² / s to 62 m² / s, and the average degradation rate is adjusted from 0.15 d... -1 Adjusted to 0.18 d -1 Furthermore, the ensemble mean of the concentration fields at each cross-section is closer to the measured values. The updated ensemble mean concentration field is the assimilated pollutant concentration field.

[0055] Using the updated pollutant concentration field, along with the adjusted longitudinal dispersion coefficient and degradation rate, as new initial conditions, the process returns to step S1 to re-drive the hydrodynamic-water quality coupled model for positive predictions over the next 24 hours, generating an updated pollutant concentration field. Steps S2 through S4 are then executed sequentially. Based on the new concentration field, the peak migration trajectory and leading front are recalculated, the candidate monitoring point set and information map are updated, and the optimal monitoring point scheme for the next round is re-solved using a genetic algorithm. After completing the first round of S1 through S53, the updated pollutant concentration field is checked: if any spatial location still has a phenol concentration higher than the preset concentration threshold of 0.01 mg / L, it indicates that the leading front has not yet disappeared, and the next round of sampling and updating continues; if the concentration values ​​at all spatial locations are below 0.01 mg / L, it indicates that the impact of the pollution plume has decayed to below a safe level, the leading front has disappeared, and the monitoring point adjustment process is terminated.

[0056] In this embodiment, after the first round of data assimilation and update, the model prediction showed that the peak concentration of the phenol pollution plume was gradually decreasing, and the leading front continued to advance downstream, but the concentration gradient tended to flatten. Based on the updated information map, monitoring points 7 and 8, where no concentration was detected downstream, were adjusted to the new predicted location of the leading front, while the core monitoring points on the peak migration trajectory were retained, and the second round of monitoring cycle was initiated. After four consecutive cycles, the phenol concentration throughout the river channel had dropped below 0.01 mg / L, meeting the termination condition, and the emergency monitoring point adjustment process was completed.

[0057] Example 2 This invention provides a monitoring deployment system for sudden water pollution incidents in canals. The system is used to implement the monitoring deployment method for sudden water pollution incidents in canals as described in any of Embodiment 1. The system includes: The model prediction module is used to acquire pollution source information and real-time hydrological data of the canal. Based on the real-time hydrological data of the canal, a hydrodynamic-water quality coupling model is constructed and driven by the pollution source information as input. The model makes a positive prediction of the spatiotemporal distribution of pollutants and generates the pollutant concentration field for the target period. The pollution plume calculation module is used to search the spatial location of the maximum pollutant concentration in the pollutant concentration field along the water flow direction step by step, forming the peak migration trajectory of the pollution plume; and extracting the set of spatial locations where the concentration value first reaches the preset concentration threshold at each moment, forming the leading edge of the pollution plume. The monitoring value quantification module is used to generate a set of candidate monitoring points based on the leading front and peak migration trajectory, calculate the information gain of each candidate monitoring point based on the pollutant concentration field, and determine the comprehensive monitoring value of each candidate monitoring point by combining the peak migration trajectory and the area covered by the leading front, and generate an information map. The monitoring optimization deployment module is used to determine the final monitoring points based on the information map, under given monitoring resource constraints, with the goal of maximizing the total information gain of candidate monitoring points, in order to capture the peak concentration and boundary range of the pollution plume; The dynamic update module is used to perform on-site sampling according to the final monitoring point to obtain real-time monitoring data; it uses data assimilation technology to integrate the real-time monitoring data into the hydrodynamic-water quality coupling model, updates the state variables and parameters of the hydrodynamic-water quality coupling model, and sends the updated state variables and parameters back to the model prediction module for the next round of prediction and monitoring point adjustment until the leading edge front disappears.

[0058] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for monitoring and deploying monitoring points for sudden water pollution incidents in canals, characterized in that, include: S1. Obtain pollution source information and real-time hydrological data of the canal. Based on the real-time hydrological data of the canal, construct and use the pollution source information as input to drive the hydrodynamic-water quality coupling model, make positive predictions on the spatiotemporal distribution of pollutants, and generate the pollutant concentration field for the target period. S2, from the pollutant concentration field, search the spatial location of the maximum pollutant concentration along the water flow direction time step by time to form the peak migration trajectory of the pollutant plume; extract the set of spatial locations where the concentration value first reaches the preset concentration threshold at each time moment to form the leading edge of the pollutant plume; S3. Generate a set of candidate monitoring points based on the leading edge front and peak migration trajectory, calculate the information gain of each candidate monitoring point based on the pollutant concentration field, and determine the comprehensive monitoring value of each candidate monitoring point by combining the peak migration trajectory and the area covered by the leading edge front, and generate an information map. S4, based on the information map, aims to maximize the total information gain of candidate monitoring points under given monitoring resource constraints, and uses a genetic algorithm to determine the final monitoring points in order to capture the peak concentration and boundary range of the pollution plume; S5, conduct on-site sampling according to the final monitoring point to obtain real-time monitoring data; use ensemble Kalman filtering to integrate the real-time monitoring data into the hydrodynamic-water quality coupling model, update the state variables and parameters of the hydrodynamic-water quality coupling model, and return to S1 to perform the next round of prediction and monitoring point adjustment until the leading edge front disappears; S3 further includes: S3a discretizes the pollutant concentration field into a spatiotemporal grid and calculates the information entropy of the pollutant concentration value in each spatiotemporal grid. S3b, the total information entropy reduction after including the candidate monitoring points in the candidate monitoring point set is used as the information gain of the candidate points; S3c, increase the weight of the spatiotemporal grid covered by the peak migration trajectory and the leading edge front, multiply the information gain of the spatiotemporal grid where each candidate monitoring point is located by the corresponding weight to obtain the comprehensive monitoring value of each candidate monitoring point, and generate the information map based on the comprehensive monitoring value.

2. The monitoring point deployment method for sudden water pollution incidents in canals according to claim 1, characterized in that, The process of acquiring pollution source information and real-time hydrological data of the canal, constructing a hydrodynamic-water quality coupling model based on the real-time hydrological data of the canal and using the pollution source information as input to positively predict the spatiotemporal distribution of pollutants, and generating a pollutant concentration field for the target time period includes: S11, Obtain pollution source information, including pollutant type, total leakage amount, leakage location, and leakage method; Obtain real-time hydrological data of the canal, including flow rate, water level, and flow velocity profile. S12, invoke the pre-built hydrodynamic-water quality coupling model of the canal, and set the initial conditions and boundary conditions of the model based on the real-time hydrological data of the canal; S13, if the pollution source leakage is instantaneous, the total leakage amount is released instantaneously at the leakage location at the initial moment; if the leakage is continuous, the total leakage amount is set as the river inflow flux boundary for the corresponding time series. S14 outputs a pollutant concentration field covering the target period in the future.

3. The monitoring point deployment method for sudden water pollution incidents in canals according to claim 1, characterized in that, The candidate monitoring point set is the set of coordinates of the center points of the spatiotemporal grids through which the peak migration trajectory and the leading edge front pass, and whose information entropy is higher than a preset information entropy threshold.

4. The monitoring point deployment method for sudden water pollution incidents in canals according to claim 1, characterized in that, Based on an information map, and under given monitoring resource constraints, the method aims to maximize the total information gain of candidate monitoring points. A genetic algorithm is used to determine the final monitoring points to capture the peak concentration and boundary range of the pollution plume. This includes: S41, Obtain the spatial coordinates of each candidate monitoring point and its corresponding comprehensive monitoring value from the information map; S42, Set monitoring resource constraints, including the upper limit of the number of final monitoring points and the minimum distance between adjacent final monitoring points; S43, Based on the comprehensive monitoring value of the candidate monitoring points, and with the objective of maximizing the total information gain of the selected monitoring points while satisfying the monitoring resource constraints, a fitness function is constructed. S44. A genetic algorithm is used to encode the selection scheme of candidate monitoring points. The population is iteratively optimized through selection, crossover and mutation operations to find the optimal combination of monitoring points that satisfies the fitness function, which is then used as the final monitoring point.

5. The monitoring point deployment method for sudden water pollution incidents in canals according to claim 4, characterized in that, Based on the comprehensive monitoring value of the candidate monitoring points, and with the objective of maximizing the total information gain of the selected monitoring points while satisfying the monitoring resource constraints, a fitness function is constructed, including: S431, obtain the information gain corresponding to each candidate monitoring point, wherein the information gain is the total information entropy reduced after including the candidate points in the monitoring; S432, For any selection scheme containing several candidate monitoring points, sum the information gain values ​​of all candidate monitoring points in the scheme to obtain the total information gain of the scheme; S433 defines a penalty function: when the number of selected monitoring points exceeds the upper limit, the penalty value increases with the number of points exceeding the limit; when there are adjacent points among the selected monitoring points whose distance is less than the minimum distance, the penalty value increases with the number of point pairs that violate the distance constraint. S434, Construct a fitness function. The fitness function is the total information gain minus the penalty value of the penalty function. The optimization objective of the genetic algorithm is to maximize the fitness function value.

6. The monitoring point deployment method for sudden water pollution incidents in canals according to claim 1, characterized in that, The process involves on-site sampling at the final monitoring points to obtain real-time monitoring data; then, using ensemble Kalman filtering, the real-time monitoring data is integrated into the hydrodynamic-water quality coupled model to update the model's state variables and parameters, and the process returns to S1 for the next round of prediction and monitoring point adjustment, until the leading front disappears. This includes: S51, At the spatial coordinates determined at the final monitoring point, on-site sampling is carried out according to the preset sampling frequency to obtain real-time concentration data of pollutants at each final monitoring point; S52, using the ensemble Kalman filter method, the real-time concentration data is used as an observation and integrated into the hydrodynamic-water quality coupling model to update the state variables and parameters in the pollutant concentration field, including the longitudinal dispersion coefficient and degradation rate; S53, using the updated state variables and parameters as initial conditions, returns to S1 to perform the next round of positive prediction; S54. Repeat steps S1 to S53 until the concentration values ​​at all spatial locations in the pollutant concentration field are lower than the preset concentration threshold, i.e., the leading edge front disappears, then terminate the monitoring point adjustment process.

7. The monitoring point deployment method for sudden water pollution incidents in canals according to claim 1, characterized in that, The information map is a monitoring value distribution map constructed in a spatial dimension, using the spatial coordinates of each candidate monitoring point as an index and the comprehensive monitoring value corresponding to each candidate monitoring point as an attribute value. It is used to provide a quantitative basis for spatial decision-making.

8. A monitoring deployment system for sudden water pollution incidents in canals, characterized in that, The system is used to implement the monitoring point deployment method for sudden water pollution events in canals as described in any one of claims 1 to 7, and the system comprises: The model prediction module is used to acquire pollution source information and real-time hydrological data of the canal. Based on the real-time hydrological data of the canal, a hydrodynamic-water quality coupling model is constructed and driven by the pollution source information as input. The model makes a positive prediction of the spatiotemporal distribution of pollutants and generates the pollutant concentration field for the target period. The pollution plume calculation module is used to search the spatial location of the maximum pollutant concentration in the pollutant concentration field along the water flow direction step by step, forming the peak migration trajectory of the pollution plume; and extracting the set of spatial locations where the concentration value first reaches the preset concentration threshold at each moment, forming the leading edge of the pollution plume. The monitoring value quantification module is used to generate a set of candidate monitoring points based on the leading front and peak migration trajectory, calculate the information gain of each candidate monitoring point based on the pollutant concentration field, and determine the comprehensive monitoring value of each candidate monitoring point by combining the peak migration trajectory and the area covered by the leading front, and generate an information map. The monitoring optimization deployment module is used to determine the final monitoring points based on the information map, under given monitoring resource constraints, with the goal of maximizing the total information gain of candidate monitoring points, in order to capture the peak concentration and boundary range of the pollution plume; The dynamic update module is used to perform on-site sampling according to the final monitoring point to obtain real-time monitoring data; the real-time monitoring data is integrated into the hydrodynamic-water quality coupling model using ensemble Kalman filtering, the state variables and parameters of the hydrodynamic-water quality coupling model are updated, and the updated state variables and parameters are sent back to the model prediction module for the next round of prediction and monitoring point adjustment until the leading edge front disappears.

Images