Method and system for tracing and positioning pollution of rainwater pipe network
By combining the SWMM model and Bayesian optimization algorithm, and using sparse Gaussian process to fit the dataset, the problem of sewage mixing in high-water-level rainwater pipe networks was accurately located, achieving efficient and low-cost pollution source tracing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-09
- Publication Date
- 2026-04-03
AI Technical Summary
Existing technologies for investigating sewage mixing problems in high-water-level stormwater pipe networks rely on manual labor and CCTV technology, which are inefficient and inaccurate, and cannot provide effective diagnosis under high-water-level conditions.
By combining the SWMM model with the Bayesian optimization algorithm, a pollution source tracing and localization method is constructed by monitoring water quality characteristic factors such as flow rate, oxygen isotope O18 and deuterium D. The dataset is fitted using a sparse Gaussian process to invert the optimal set of pollution inflow parameters. Combined with validation indicators such as MSE, NSE and R2, efficient and accurate localization is achieved.
It reduces monitoring data redundancy and costs, breaks through the limitations of high-water-level pipe network investigation, accurately identifies pollution sources, controls simulation errors within 5%, and shortens the optimization cycle.
Smart Images
Figure CN121787023A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of urban water environment management technology, specifically to a method and system for tracing and locating pollution sources in rainwater pipe networks. Background Technology
[0002] In recent years, China has proposed a campaign to tackle urban black and odorous water bodies, accelerating the improvement of urban water environment quality. Stormwater drainage networks are the lifeline of cities, serving as crucial municipal infrastructure for alleviating urban flooding and improving water quality. However, due to improper construction, aging pipe materials, corrosion, and other factors, sewage mixing is a common problem in my country's urban stormwater drainage networks, posing significant risks to water environment health and public safety. Scientific and efficient diagnosis of sewage mixing problems in stormwater drainage networks, pinpointing the locations of these problems, and identifying the types of issues can prevent blind diagnostic work and provide a reference for related work in other areas with high water levels, thus contributing to the high-quality development of the urban water environment management industry.
[0003] Currently, the investigation of mixed connections in rainwater pipe networks relies mainly on manual on-site operations and technologies such as CCTV and QV. This is greatly affected by subjective factors, and the efficiency and accuracy of the work cannot be guaranteed. In addition, some areas have pipelines operating at high water levels, making it impossible to conduct direct manual and equipment inspections.
[0004] Therefore, it is urgent to study a diagnostic and location method for pollution source tracing problems applicable to high-water-level rainwater pipe networks, in order to efficiently and accurately assess the mixed connection situation of high-water-level rainwater pipe networks. Summary of the Invention
[0005] This invention aims to at least solve one of the technical problems existing in the prior art. Therefore, one objective of this invention is to propose an efficient and accurate method for tracing and locating pollution sources in stormwater drainage networks, which solves the problem of tracing pollution sources in high-water-level drainage networks by integrating the SWMM model and Bayesian optimization algorithm.
[0006] To address the aforementioned problems, this invention provides a method for tracing and locating pollution sources in stormwater drainage networks. This method is based on the SWMM model and a Bayesian optimization algorithm, and includes the following steps: Step 1: Based on the topology, operating water level, and water quality characteristic factor data of the stormwater pipe network in the target area, construct the SWMM model of the pipe network water quality and hydrodynamics. The model configuration includes the division of the catchment area, sealing of the terminal outlet, setting the initial water level to full pipe state, setting the initial water quality to the average concentration of stormwater characteristic factors, and setting the operating step size and total duration. Step 2: Key monitoring points are strategically deployed in the stormwater drainage network through topology analysis to obtain measured water quality characteristic factors at these points. These water quality characteristic factors include flow rate and oxygen isotope O. 18 and deuterium D; Step 3: Based on the reasonable range of water quality characteristic factor concentrations, generate a node pollution inflow parameter file using a probability distribution random assignment as a priori probability distribution, where the flow rate follows a uniform distribution, O 18 The values of D and D follow a normal distribution, and the SWMM model is run to obtain simulated water quality data for each monitoring point; Step 4: Iterate through Step 3 n times to generate n pollution inflow parameter files and corresponding simulated water quality datasets; Step 5: Based on the measured water quality characteristic factors, the dataset is fitted using a Bayesian optimization algorithm and a sparse Gaussian process. The optimal pollutant inflow parameter set is obtained as the posterior probability distribution. The Bayesian optimization uses an objective function f(x) to measure the deviation between the simulation and the measured values. The objective function is defined as: ); ; Where: m is the amount of measured data, n is the number of monitoring points, and C H and C O These represent the concentrations of the characteristic factors, respectively. Step Six: Combining the SWMM model with measured data, calculate validation indices to verify the accuracy and rationality of the optimal pollutant inflow parameter set. Validation indices include mean square error (MSE), Nash efficiency coefficient (NSE), and determinism coefficient (R²). 2 ,in: MSE is defined as: MSE = ; NSE is defined as: NSE = 1 - ; R 2 Defined as: ; Where: N is the amount of test data; If the verification indicators meet the preset threshold, the pollution source is considered to have been successfully located; if the preset threshold is not met, the process returns to the parameter optimization iteration stage, where the algorithm parameters are adjusted or the number of iterations is increased for re-optimization until the target is met.
[0007] Preferably, in step one, the construction of the SWMM model also includes determining key partitions through topological analysis and deploying monitoring points at the catchment nodes.
[0008] Preferably, in step three, the generation of the node contamination inflow parameter file is dynamically written through a self-iterative program, and the SWMM model is called to extract simulated feature factor values.
[0009] Preferably, in step five, the Bayesian optimization algorithm uses variational inference of a sparse Gaussian process to fit the posterior distribution and implements a batch sampling evaluation mechanism.
[0010] Preferably, in step five, the objective function is adjusted according to the actual calibration requirements.
[0011] Preferably, in step six, the preset thresholds include NSE greater than 0.75 and R 2 Greater than 0.8.
[0012] A source tracing and location system used in a method for tracing and locating pollution sources in a stormwater pipe network includes: The model building module is used to construct an SWMM model of the water quality and hydrodynamics of the stormwater pipe network based on the topology, operating water level and water quality characteristic factor data of the target area's stormwater pipe network. The monitoring module is used to deploy key monitoring points in different zones within the stormwater pipe network and obtain measured water quality characteristic factor indicators at the monitoring points. The parameter generation module is used to generate node pollution inflow parameter files as prior probability distributions by randomly assigning probability distributions based on a reasonable range of water quality characteristic factor concentrations, and to run the SWMM model to obtain simulated water quality data for each monitoring point. The iteration module is used to iteratively execute the operations of the parameter generation module n times, generating n pollution inflow parameter files and corresponding simulated water quality datasets; The optimization module is used to fit the dataset based on measured water quality characteristic factors using Bayesian optimization algorithm and sparse Gaussian process, and inversely obtain the optimal set of pollutant inflow parameters as the posterior probability distribution. The validation module is used to combine the SWMM model with measured data to calculate validation indicators to verify the accuracy and rationality of the optimal set of pollution inflow parameters.
[0013] Preferably, the model building module is also configured to determine key partitions through topology analysis and deploy monitoring points at the catchment nodes.
[0014] Preferably, the optimization module uses a variational inference sparse Gaussian process to achieve batch sampling evaluation.
[0015] Preferably, the preset thresholds for the verification module include NSE greater than 0.75 and R 2 Greater than 0.8.
[0016] The advantages of this invention compared to the prior art are: This invention limits the detection index water quality characteristic factors to two isotopes, which can effectively reduce the burden of cumbersome monitoring data indicators and data redundancy. At the same time, the method breaks through the limitations of the traditional mixed connection investigation mode, and uses the physical model SWMM combined with the Bayesian optimization algorithm of sparse elevation process to reduce the dependence on high-density monitoring points and save costs significantly. Moreover, it can accurately identify the parameter combination with the smallest deviation between simulation results and measured results with fewer iterations, and efficiently and accurately locate the mixed connection of high water level rainwater pipe network. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a schematic diagram of the method flow provided in an embodiment of the present invention; Figure 2 This is an SWMM network model diagram of a certain operating network in an embodiment of the present invention. Detailed Implementation
[0019] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.
[0020] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "joining" 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; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.
[0021] The present invention will now be described in further detail with reference to the accompanying drawings.
[0022] The present invention provides a method and system for tracing and locating pollution sources in rainwater pipe networks, comprising the following core steps: 1. Model Construction: Based on the stormwater pipe network topology, operating water level, and water quality characteristic factors (such as flow rate, oxygen isotope O). 18 (and deuterium D), build an SWMM hydraulic model, and configure parameters such as catchment area division, end outlet sealing, and initial full pipe water level.
[0023] 2. Monitoring point layout: Key zones are delineated through topological analysis, and monitoring points are deployed at the nodes to obtain measured water quality characteristic factor indicators.
[0024] 3. Parameter Generation and Simulation: Based on a reasonable range of characteristic factor concentrations, a probability distribution is used (flow rate follows a uniform distribution, O...). 18The node pollution inflow parameter file (prior distribution) is randomly generated (D follows a normal distribution), and the SWMM model is run to output simulated water quality data.
[0025] 4. Iterative dataset construction: Repeat the above process n times to form a prior dataset containing n sets of pollution inflow parameters and simulation data.
[0026] 5. Bayesian Optimization Inversion: Based on measured data, a Bayesian optimization algorithm combined with a sparse Gaussian process is used to fit the dataset and invert the optimal set of pollutant inflow parameters (posterior distribution). The optimization objective function is defined as: ; The sparse Gaussian process achieves batch sampling through variational inference, thereby improving computational efficiency.
[0027] 6. Verification and Positioning: Combining the SWMM model with measured data, calculate verification indicators (such as mean squared error (MSE), Nash efficiency coefficient (NSE), and coefficient of determination (R²). 2 If the indicator meets the threshold (e.g., NSE > 0.75, R...), 2 If the value is greater than 0.8, the pollution source is considered to have been successfully located.
[0028] The core technical effect of this invention is as follows: 1. By limiting water quality characteristic factors to isotopic indicators (such as O2) 18 (and D), reduce data redundancy and lower monitoring costs.
[0029] 2. By integrating physical models with intelligent algorithms, the limitations of high-water-level pipe network investigation can be overcome, and pollution sources can be accurately identified with fewer iterations.
[0030] 3. The examples show that the optimized node parameter error can be controlled within 5%, with strong anti-noise interference capability and fluctuation amplitude of less than 3%.
[0031] To more clearly illustrate the specific embodiments of the present invention, an example is provided below: An embodiment of the present invention provides a method and system for tracing and locating pollution sources in rainwater pipe networks, the specific process of which is as follows: Figure 1As shown: A basic database including the topological relationships of stormwater pipelines, operating water levels, and water quality characteristic factor parameters is established in the study area; key monitoring points are deployed in different zones of the stormwater network in the study area to obtain the water quality characteristic factor indicators of the monitoring points, and a hydrodynamic model of the network water quality is established; based on a reasonable range of water quality characteristic factor concentrations, random values are assigned using a normal distribution, and n sets of pollution inflow parameter files (parameter prior probability distribution) are generated iteratively for each node, and the SWMM model is batch-processed to obtain simulated water quality data for each monitoring point; a prior dataset is constructed by comparing with actual measured indicators, and the parameter set is optimized by fitting and optimizing the records (posterior probability distribution); the accuracy and rationality of the optimal pollution inflow parameter set are verified to achieve pollution mixing location calculation. This invention addresses the problem of difficulty in tracing pollution sources in stormwater networks under high water levels by proposing a method for tracing and locating pollution sources in stormwater networks based on the SWMM model and Bayesian optimization algorithm. The core of this method is the Bayesian optimization method of sparse Gaussian processes, which can efficiently fit and optimize the large amount of stormwater and sewage mixing and the location of mixing points in the drainage network and present it in the form of a posterior probability distribution. Compared to traditional methods that limit water quality characteristic factors to two isotopes, this method effectively reduces the burden of cumbersome and redundant monitoring data. Furthermore, it overcomes the limitations of traditional mixed-connection investigation models by utilizing a physical model (SWMM) combined with a Bayesian optimization algorithm based on sparse elevation processes, reducing reliance on high-density monitoring points and significantly saving costs. Moreover, it can accurately identify the parameter combination with the smallest deviation between simulated and measured results with fewer iterations, efficiently and accurately locating mixed connections in high-water-level stormwater pipe networks. Taking a certain operating stormwater pipe network as an example, the specific steps for tracing and locating pollution sources are as follows: Collect stormwater drainage network data (including dimensions, elevation, etc.) and water quality data (obtained through field measurements) for the study area. 18 (D data); Based on survey data such as pipe length, diameter, elevation, well depth, and pipe connection relationships, pipe network parameters are set, initial boundary conditions for the model are established, and model characteristic parameters such as flow rate q, O18, and D are determined. Model feature vectors and matrices are constructed, and finally, an SWMM model is established. Figure 2 As shown; Based on the reasonable range of water quality characteristic factor concentrations, normal distribution was used to randomly assign values, and node pollution inflow parameter files (prior probability distribution) as shown in Table 1 were generated for each node. The SWMM model was run to obtain simulated water quality data for each monitoring point. Table 1 Prior Dataset
[0032] Where, iteration: a sample in a certain iteration; node: node name; flow: flow rate; O18, D: concentration of water quality characteristic factors; Objective Function: objective function value; Iterate the S3 process n times to obtain n theoretical pollutant inflow parameter files and the water quality data simulated by the monitoring point model; Construct a prior data set by calculating the objective function based on the measured data at the monitoring points; Define the objective function to measure the deviation between the simulation and the measurement, and adopt the form of mean square error. The smaller the function value, the higher the model accuracy. The formula is as follows: )(m is the amount of measured data); ; Select the Bayesian optimization algorithm in the scikit-optimize library, and fit the data set in combination with the sparse Gaussian process to inversely obtain the optimal pollutant inflow parameter set (posterior probability distribution), as shown in Table 2 below; Table 2 Optimal pollutant inflow parameter set
[0033] where, iteration: the iteration sample where the optimal parameter set is located; objective_value: the optimal objective function of the fit; node: node name; flow: optimized flow; O18, D: concentrations of optimized water quality characteristic factors Combine the SWMM model with the measured data to verify the accuracy and rationality of the optimal pollutant inflow parameter set, and finally realize the positioning and calculation of pollution mixing connection. The calculation test indexes include mean square error (MSE), Nash efficiency coefficient (NSE) and determination coefficient (R2), and the formulas are as follows: MSE = (N is the amount of test data) NSE = 1 -
[0034]
[0035] If the indexes meet the thresholds (such as NSE > 0.75, > 0.8), determine that the model is qualified; otherwise, return to the parameter optimization iteration stage, adjust the algorithm parameters (such as increasing the number of iterations) and re-optimize until it meets the standard. Finally, it can accurately identify the parameter combination with the smallest deviation between the simulation result and the measured result, and efficiently and accurately locate the mixing position of the high-water-level rainwater pipe network.
[0036] The results of this embodiment demonstrate that Bayesian optimization combined with sparse Gaussian processes can achieve efficient optimization of SWMM model node parameters in stormwater pollution source tracing projects. For nodes such as stormwater inlets and discharge outlets, balancing computational cost and accuracy, the simulation error of key node water quality characteristic parameters is controlled within 5% after optimization, and the deviation between node parameters and measured values is less than 10%. Under 10% data noise interference, the fluctuation range of the optimization results is <3%, ensuring the stability and reliability of pollution source tracing parameters, providing accurate parameter support for pollution source tracing, shortening the optimization cycle, and facilitating rapid location of pollution sources.
[0037] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.
Claims
1. A method for tracing and locating the source of pollution in a rainwater pipe network, characterized in that: This source tracing and localization method is based on the SWMM model and Bayesian optimization algorithm, and includes the following steps: Step 1: Based on the topology, operating water level and water quality characteristic factor data of the stormwater pipe network in the target area, construct the SWMM model of the pipe network water quality and hydrodynamics. The model configuration includes the division of the catchment area, sealing of the terminal outlet, setting the initial water level to full pipe state, setting the initial water quality to the average concentration of stormwater characteristic factors, and setting the operating step size and total duration. Step 2: Key monitoring points are strategically deployed in the stormwater drainage network through topology analysis, and measured water quality characteristic factors at these points are obtained. These water quality characteristic factors include flow rate and oxygen isotope O. 18 and deuterium D; Step 3: Based on the reasonable range of water quality characteristic factor concentrations, generate a node pollution inflow parameter file using a probability distribution random assignment as a priori probability distribution, where the flow rate follows a uniform distribution, O 18 The values of D and D follow a normal distribution, and the SWMM model is run to obtain simulated water quality data for each monitoring point; Step 4: Iterate through Step 3 n times to generate n pollution inflow parameter files and corresponding simulated water quality datasets; Step 5: Based on the measured water quality characteristic factors, the dataset is fitted using a Bayesian optimization algorithm and a sparse Gaussian process to obtain the optimal pollutant inflow parameter set as the posterior probability distribution. The Bayesian optimization uses an objective function f(x) to measure the deviation between the simulation and the measured values. The objective function is defined as: )； ； Where: m is the amount of measured data, n is the number of monitoring points, and C H and C O These represent the concentrations of the characteristic factors, respectively. Step Six: Combining the SWMM model with measured data, calculate validation indices to verify the accuracy and rationality of the optimal pollution inflow parameter set. These validation indices include the mean square error (MSE), Nash efficiency coefficient (NSE), and determinism coefficient (R²). 2 ,in: MSE is defined as: MSE = ; NSE is defined as: NSE = 1 - ; R 2 Defined as: ; Where: N is the amount of test data; If the verification indicators meet the preset threshold, the pollution source is considered to have been successfully located; if the preset threshold is not met, the process returns to the parameter optimization iteration stage, where the algorithm parameters are adjusted or the number of iterations is increased for re-optimization until the target is met.
2. The method for tracing and locating pollution sources in a rainwater pipe network according to claim 1, characterized in that: In step one, the construction of the SWMM model also includes determining key partitions through topology analysis and deploying monitoring points at the catchment nodes.
3. The method for tracing and locating pollution sources in a rainwater pipe network according to claim 1, characterized in that: In step three, the generation of the node pollution inflow parameter file is achieved through dynamic writing via a self-iterative program, and the SWMM model is called to extract simulated feature factor values.
4. The method for tracing and locating pollution sources in a rainwater pipe network according to claim 1, characterized in that: In step five, the Bayesian optimization algorithm uses variational inference of a sparse Gaussian process to fit the posterior distribution and implements a batch sampling evaluation mechanism.
5. The method for tracing and locating pollution sources in a rainwater pipe network according to claim 1, characterized in that: In step five, the objective function is adjusted according to the actual calibration requirements.
6. The method for tracing and locating pollution sources in a rainwater pipe network according to claim 1, characterized in that: In step six, the preset threshold includes NSE greater than 0.75 and R 2 Greater than 0.
8.
7. A source tracing and location system used in the rainwater pipe network pollution source tracing and location method as described in claim 1, characterized in that, include: The model building module is used to construct an SWMM model of the water quality and hydrodynamics of the stormwater pipe network based on the topology, operating water level and water quality characteristic factor data of the target area's stormwater pipe network. The monitoring module is used to deploy key monitoring points in the rainwater pipe network in different zones and obtain the measured water quality characteristic factor indicators of the monitoring points. The parameter generation module is used to generate node pollution inflow parameter files as prior probability distributions by randomly assigning probability distributions based on a reasonable range of water quality characteristic factor concentrations, and to run the SWMM model to obtain simulated water quality data for each monitoring point. The iteration module is used to iteratively execute the operations of the parameter generation module n times, generating n pollution inflow parameter files and corresponding simulated water quality datasets; The optimization module is used to fit the dataset based on the measured water quality characteristic factor index using a Bayesian optimization algorithm and a sparse Gaussian process, and inversely obtain the optimal pollution inflow parameter set as the posterior probability distribution. The verification module is used to combine the SWMM model with measured data to calculate verification indicators to verify the accuracy and rationality of the optimal pollution inflow parameter set.
8. The source tracing and location system used in the method for tracing and locating pollution sources in a rainwater pipe network according to claim 7, characterized in that: The model building module is also configured to determine key partitions through topology analysis and deploy monitoring points at the catchment nodes.
9. The source tracing and location system used in the method for tracing and locating pollution sources in a rainwater pipe network according to claim 7, characterized in that: The optimization module uses a variational inference sparse Gaussian process to achieve batch sampling evaluation.
10. The source tracing and location system used in the method for tracing and locating pollution sources in a rainwater pipe network according to claim 7, characterized in that: The verification module has preset thresholds including NSE greater than 0.75 and R 2 Greater than 0.8.