A method for constructing a pipe network model using online monitoring network data
The automated integration of online monitoring data for SWMM pipe network models addresses registration and calibration inefficiencies, enhancing accuracy and reliability by aligning monitoring points with nodes and using advanced algorithms for parameter optimization.
Patent Information
- Application Number
- CN202010909734.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-09-02
- Publication Date
- 2024-12-31
- Estimated Expiration
- 2040-09-02
AI Technical Summary
Current SWMM pipe network models face challenges in manual map registration, inaccurate data input, and limited data availability, leading to inefficiencies and reduced model reliability, especially in large study areas with uneven rainfall and inflow distribution.
An automated method using online monitoring network data to align and calibrate SWMM pipe network models, integrating real-time monitoring points with pipe network nodes, employing thin plate spline interpolation for coordinate transformation, and a composite crossover evolutionary algorithm for parameter calibration, with weighted Nash efficiency coefficients for model verification.
Improves model registration accuracy, reduces manual input errors, enhances calibration efficiency, and increases model robustness and reliability through multi-cycle verification using diverse monitoring data.
Smart Images
Figure CN112016178B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of urban rain and flood process simulation, and in particular to an optimization method for a SWMM model. Background Art
[0002] With the advancement of urbanization and the frequent occurrence of extreme weather events, urban flooding has repeatedly occurred in China, severely impacting urban transportation and residents' lives. Therefore, drainage network models are becoming an important tool for improving urban stormwater management. The SWMM (Storm Water Management Model) network model is widely used to analyze the impact of rainfall on urban runoff, thereby helping to optimize urban drainage systems and reduce the economic losses caused by urban flooding.
[0003] SWMM pipe network models can simulate: hydrological processes, including surface evaporation, snow accumulation and melting, rainfall infiltration, groundwater recharge from rainfall, the interaction between groundwater and surface runoff, and water losses caused by various land-use facilities during runoff; hydraulic processes, including the simulation of various river channels and canals, water storage and drainage facilities, pumps, weirs, and other hydraulic facilities, and the corresponding management measures; and water quality processes, including the migration of pollutants related to hydrological processes, such as rainfall-borne pollutants and rainfall erosion. Therefore, SWMM pipe network construction is complex, involving numerous and distributed parameters.
[0004] Currently, there are a lot of studies on optimizing the construction of SWMM pipe network models, such as calibration of model parameters and sensitivity analysis, but there are the following problems:
[0005] (1) In practical applications, the model pipeline network is usually displayed in combination with an online map or base map, which requires map registration. The existing map registration method is to manually select control points, which is time-consuming and has accuracy issues.
[0006] (2) When performing model parameter calibration, monitoring data needs to be manually input. Each time the calibration is re-done, the monitoring data needs to be reselected. In addition, manual judgment of which network nodes the monitoring data is applied to will result in certain errors.
[0007] (3) In actual situations, due to the difficulty in obtaining monitoring data, only a small amount of monitoring data can usually be used to calibrate and verify the model. For pipe network models with large study areas and uneven rainfall and inflow distribution, the global effectiveness of the model will be greatly reduced. Summary of the Invention
[0008] The present invention provides a method for constructing a pipe network model using online monitoring network data. For the SWMM model, real monitoring points are associated with pipe network nodes, which can realize automatic alignment of the pipe network model and automatic calibration and verification of the pipe network model over multiple cycles, thereby improving the reliability of the model.
[0009] A method for constructing a pipe network model using online monitoring network data includes the following steps:
[0010] S1: Establishing a SWMM pipe network model, wherein the SWMM pipe network model includes pipe network nodes, catchment areas, pipe sections, and outlets;
[0011] S2: Generate an online monitoring point layer based on the monitoring point information provided by the online monitoring system;
[0012] S3: Associating the SWMM pipe network model established in step S1 with the online monitoring point layer generated in step S2 to obtain associated monitoring points;
[0013] S4: Automatically map-register the SWMM network model with the online monitoring point layer based on the associated monitoring points;
[0014] S5: Select monitoring parameters and calibration facilities, calibrate SWMM pipe network model parameters, and obtain the calibrated SWMM pipe network model;
[0015] S6: Use monitoring data from different sessions to verify the calibrated SWMM pipe network model.
[0016] Furthermore, in step S2, the monitoring point location information includes longitude, latitude, monitoring point equipment type and name, and monitoring data items;
[0017] The types of equipment at the monitoring points include: rain gauges, flow meters, liquid level meters, water quality meters and dissolved oxygen flow meters.
[0018] The monitoring data items include:
[0019] Monitoring data items unit Liquid level m rainfall mm flow L / s suspended matter mg / L dissolved oxygen ％
[0020] Furthermore, in step S3, the pipe network model is associated with the online monitoring point layer. The pipe network nodes in the pipe network model include inspection wells. The inspection well types include three types: rainwater wells, sewage wells and combined wells. Among them, the rainwater wells are associated with monitoring points whose monitoring point equipment types are rain gauges, liquid level meters and flow meters. The sewage wells are associated with monitoring points whose equipment types are water quality meters and dissolved oxygen flow meters. The combined wells are associated with monitoring points whose equipment types are rain gauges, liquid level meters, flow meters, water quality meters and dissolved oxygen flow meters.
[0021] Furthermore, in step S4, the automatic map registration of the pipe network model and the online monitoring point layer includes the following steps:
[0022] 1) Obtain a control point set based on the associated monitoring points in step S3;
[0023] 2) Check the control point residuals. If there is a control point residual greater than e -8 When , add a new control point;
[0024] 3) Use thin plate spline interpolation for coordinate transformation.
[0025] Furthermore, in step 1), the control points centrally select the network node coordinates in the SWMM network model as source coordinates, and the associated monitoring point coordinates as target coordinates.
[0026] Furthermore, the residual in step 2) represents the difference in distance between the source point and the target point in the control point; the point of the pipe network model is called the source point, and the point of the monitoring point layer is called the target point.
[0027] Furthermore, the thin plate spline interpolation in step 3) is, for the selected n pairs of marked points p i and q i , i = 1, ... n, according to the correspondence between the two control midpoint marks, find a suitable transformation function f in the two-dimensional space so that:
[0028]
[0029] where function U(r) = r 2 logr 2 It is the basis function for constructing thin plate splines. Through f(x), the coordinates (x, y) in all source images can be obtained by transforming the new corresponding points (x′, y′).
[0030] Furthermore, in step S5, the monitoring parameters include the water depth and total inflow of the pipe network node; the calibration facilities include the pipe network node associated with the monitoring point and the corresponding pipe section.
[0031] Furthermore, in step S5, a composite crossover evolutionary algorithm is used for parameter calibration, with the weighted average of the Nash efficiency coefficients of each associated monitoring point as the objective function value. A convergence threshold for the objective function value is also set. The calibration process is terminated if the parameter value cannot be improved for more than two times; if the objective function value cannot be improved for more than N times, the calibration process is terminated, with N typically set to 10; and if the objective function value exceeds 0.8, the calibration process is terminated.
[0032] Among them, whether the objective function value can be significantly improved can be expressed by the following mathematical expression:
[0033]
[0034] Among them, λ i-1 and λ iare the values of the objective function at step i-1 and step i, TOL λ is the convergence threshold, TOL λ The value is usually assigned to 0.0001. When the value calculated on the left side of the formula is less than or equal to the convergence threshold, it is considered that the objective function value cannot be significantly improved.
[0035] Furthermore, multiple monitoring points are selected, a weight is set for each monitoring point, and the Nash efficiency coefficients of all monitoring points are weighted averaged, which can be expressed mathematically as follows:
[0036]
[0037] Among them, w i is the weight of the monitoring point, E i is the Nash efficiency coefficient of the observation point, which can be expressed as follows:
[0038]
[0039] Among them, Q o Refers to the observed value, Q m Refers to the analog value, Q τ represents a value at time t, It represents the total average of the observations. When E is close to 1, it means that the model quality is good and the model credibility is high; when E is close to 0, it means that the simulation results are close to the average level of the observations, that is, the overall results are credible, but the simulation error is large; when E is much smaller than 0, it means that the model is unreliable.
[0040] After adopting the above technical solution, the present invention brings the following beneficial effects:
[0041] (1) This method is based on the QGIS 3.12.0 open source library and is implemented using C++ programming. A program is developed that has the functions of establishing a pipe network model, map registration, model parameter calibration, and displaying simulation results.
[0042] (2) Automatically pairing monitoring points with model nodes reduces the subjective influence of manual input of monitoring data, reduces the time of manual selection of control points, and improves the registration accuracy;
[0043] (3) According to the corresponding relationship between monitoring points and model nodes, when calibration is required, only the time range needs to be adjusted to realize the automatic input of monitoring data. By integrating the online monitoring system, the generation of monitoring layers and the automatic import of monitoring data can be realized. The monitoring data can be used as the input and monitoring data of the model, reducing the error of manual selection, realizing convenient, efficient, and multi-cycle model calibration, and improving the calibration efficiency.
[0044] (4) Use monitoring data from different sessions to verify the calibrated model effect, improve the model robustness, and increase the model reliability;
[0045] (5) When verifying the model, for all model pipe network nodes associated with the online monitoring points, multiple observation points are selected and set with different weights. The Nash efficiency coefficients of these monitoring points are weighted averaged to evaluate the model and improve the reliability of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] Figure 1 is a flow chart of the method of the present invention;
[0047] Figure 2 Schematic diagram of the SWMM pipe network model of the present invention;
[0048] Figure 3 This is a pipeline network association flow chart of the present invention;
[0049] Figure 4 This is a schematic diagram of the SWMM pipe network model and the online monitoring point layer before registration of the present invention;
[0050] Figure 5 This is a schematic diagram of the alignment of the SWMM pipe network model and the online monitoring point layer of the present invention.
[0051] Figure 6 This is the SWMM pipe network model parameter calibration flow chart of the present invention. DETAILED DESCRIPTION
[0052] Please refer to the attached Figures 1 to 6 The present invention is described in further detail.
[0053] A method for using online monitoring network data to assist in building a pipe network model. For a specific flow chart, refer to Figure 1 , including the following steps:
[0054] S1: Establish a SWMM pipe network model. The specific reference of the completed SWMM pipe network model is as follows: Figure 2 The SWMM pipe network model includes pipe network nodes, catchment areas, pipe sections, and outfalls, and includes the following steps:
[0055] (1) Collect background information and basic information of drainage network:
[0056] The research area is a sponge pilot area located in central my country, with a total area of 27.6 km 2 , of which the drainage network covers an area of 20.6km 2 There are a total of 27 rainwater outlets at the end of the river.
[0057] (2) Establishing SWMM pipe network model
[0058] Based on the pipeline network survey data, combined with the topographic map and land use type map of the study area, a SWMM pipeline network model was constructed. The study area was divided into 8013 pipeline nodes, 8012 catchment areas, 7807 pipeline sections, and 27 outfalls.
[0059] S2: Generate an online monitoring point layer based on the monitoring point information provided by the online monitoring system;
[0060] The monitoring point information stored in the online monitoring system includes longitude, latitude, monitoring point equipment type and name, and monitoring data items. The monitoring point equipment types include: rain gauge, flow meter, liquid level meter, water quality meter, and dissolved oxygen flow meter. The monitoring data items include:
[0061] Monitoring data items unit Liquid level m rainfall mm flow L / s suspended matter mg / L dissolved oxygen ％
[0062] In the online monitoring system, monitoring points refer to real-time monitoring devices with GPS positioning. By accessing the online monitoring system interface, monitoring point information can be obtained in batches. Based on the monitoring point information (mainly longitude and latitude), a vector layer is generated through a platform independently developed based on the QGIS open source library.
[0063] QGIS is an open source library. The platform is independently developed based on the QGIS 3.12.0 version of the open source library. It is programmed in C++ and the generation of vector layers is achieved by calling the methods in the QGIS open source library.
[0064] S3: Associate the SWMM pipe network model established in step S1 with the online monitoring point layer generated in step S2 to obtain associated monitoring points.
[0065] According to the monitoring point equipment type and name stored in the online monitoring system, the corresponding network node can be found in the SWMM network model. The monitoring point equipment type is associated with the network node. The network node coordinates are the source coordinates, and the monitoring point equipment type coordinates are the target coordinates. The associated monitoring point is obtained, and the monitoring data of the associated monitoring point is used as the input data of the network node and the reference value in the calibration of the SWMM network model.
[0066] Specific association rules: In the SWMM pipe network model, pipe network nodes primarily refer to manholes. Manhole types include rainwater wells, sewage wells, and combined wells. Rainwater wells are associated with monitoring point equipment types such as rain gauges, liquid level gauges, and flow meters. Sewage wells are associated with monitoring point equipment types such as water quality meters and dissolved oxygen flow meters. Combined wells are associated with monitoring point equipment types such as rain gauges, liquid level gauges, flow meters, water quality meters, and dissolved oxygen. In the SWMM pipe network model, the inflow data of the manhole is set as the flow data for the associated monitoring point. Simultaneously, the rainfall data for the catchment area whose outlet is the manhole is set as the rainfall data for the associated monitoring point.
[0067] Association process: Determine whether the network node type and the monitoring point device type meet the association rules. If they meet, further determine whether the network node name and the monitoring point device name are the same. If both meet, establish the corresponding relationship between the network node and the monitoring point for subsequent search. For the specific process, see Figure 3 .
[0068] S4: Automatically map the SWMM network model and the online monitoring point layer based on the associated monitoring points. After registration, the monitoring point layer and the network model layer are superimposed and displayed, and the network and the physical parts of the map such as roads, rivers, and lakes can overlap (see Figure 4 、 Figure 5 );
[0069] In the current SWMM
[0070] When performing map registration on a pipe network model, the following steps are involved:
[0071] (1) Based on the associated monitoring points in step S3, at least four control points are obtained to form a control point set. After the network nodes are associated with the monitoring point equipment types, the coordinates of the monitoring point equipment types and the network nodes are already known. When implementing map registration, the control point set selects the network node coordinates as the source coordinates and the monitoring point equipment type coordinates as the target coordinates. The corresponding relationship previously stored in the online monitoring system is read, thereby realizing automatic import of control points without manual selection.
[0072] The associated pipe network nodes and monitoring point equipment types are used as the source coordinates and target coordinates of the control points respectively, replacing the existing manual selection of control points. This avoids the problem of inaccurate map registration caused by manual selection of control points, solves the problems of time-consuming and poor accuracy, improves the registration accuracy, and improves the calibration efficiency.
[0073] (2) Check the residuals. If the residual of a control point is greater than e -8 When a new control point is added, the residual represents the difference between the source and target coordinates of the control point; when a new control point is added, the residual of the existing control points is readjusted.
[0074] (3) Use thin plate spline interpolation to perform coordinate transformation. The thin plate spline interpolation is to select n pairs of marker points p i and q i , i = 1, ... n, according to the specific correspondence of the two control point marks, find a suitable transformation function f in the two-dimensional space so that:
[0075]
[0076] where function U(r) = r 2 logr 2is the basis function for constructing thin plate splines, w=(w1,w2,...,w n ) T ,
[0077] a=(a1,a x , a y ) T ;
[0078] Use the following linear equations to find vectors w and a:
[0079] Among them, K is the n*n order vector matrix of the function U(r), P is the n*3 order matrix of the source coordinates x and y, V is the matrix of the target coordinates x and y, and n is the number of selected control points. Substituting the obtained a and w vectors into the function f(x), the new corresponding points (x′, y′) obtained after the coordinates (x, y) in all source images are transformed can be obtained.
[0080] The new corresponding points (x′, y′) obtained after the transformation of the coordinates (x, y) in all source images can be obtained through f(x).
[0081] S5: Select monitoring parameters and calibration facilities, calibrate SWMM pipe network model parameters, and obtain the calibrated SWMM pipe network model, which specifically includes the following steps:
[0082] (1) Set 8 parameters that are more sensitive to the model
[0083] Including parameters such as permeability roughness coefficient (N-Perv, NP), impermeability roughness coefficient (N-Imperv, NImp), permeable depression storage (Dstore-Perv, SP), impermeable depression storage (Dstore-Imperv, S-Imp), impermeability without depression storage (%Zero-Imperv, Z-Imp), maximum infiltration rate (Max.Infil.Rate, Max-In), minimum infiltration rate (Min.Infil.Rate, Min-In), and decay coefficient (Decay Constant, De-Con);
[0084] (2) Select monitoring parameters and calibration facilities
[0085] Select the network nodes and corresponding pipelines with associated monitoring points as calibration facilities; the monitoring parameters can be selected from the water depth, total inflow, etc. of the network nodes; the level meter monitoring data is used as the measured value of the monitoring parameter water depth, and the flow meter monitoring data is used as the measured value of the monitoring parameter total inflow. The time range can be determined and adjusted, and the system can import the data automatically.
[0086] "Determine the time range" means specifying the start and end time range of the monitoring data. The monitoring data within this time range will be automatically read from the online monitoring system and imported into the pipe network model as the measured value of the monitoring parameter.
[0087] "Automatically import by adjusting the time range" is because when selecting the measured value of the monitoring parameter, the monitoring data corresponding to the pipeline node is selected (as mentioned before, the monitoring parameters selected are the water depth, total inflow, etc. of the pipeline node). When the time range is changed, you only need to re-request the monitoring data within that time range from the online monitoring system.
[0088] (3) Parameter calibration
[0089] The Shuffled Complex Evolution (SCE-UA) algorithm is used to calibrate the parameters of the pipe network model, which includes the following steps (see Figure 6 ):
[0090] a. Use Latin hypercube sampling to obtain 1000 samples;
[0091] b. Import 1000 sets of samples into the SWMM pipe network model through the C++ program to generate 1000 model .INP files. Then, call the SWMM calculation engine through the C++ program to obtain the simulation result .OUT file.
[0092] c. Read the simulated value in the .OUT file and calculate the Nash efficiency coefficient with the measured value of the monitoring parameter input previously. Take the weighted average of the Nash efficiency coefficient of each monitoring point to obtain the average Nash efficiency coefficient;
[0093] d. Start iterative evolution. Sort the samples in ascending order based on the average Nash efficiency coefficient obtained in the previous step. Select 2*n+1 samples from the initial sample to form a composite sample, where n=8, which is the number of sensitivity parameters.
[0094] e. Generate a model .INP file for each composite sample and calculate the simulation result .OUT file. Then calculate the average Nash efficiency coefficient of all composite samples. The average Nash efficiency coefficient of the composite samples is called the objective function value.
[0095] Each time the objective function value is calculated, the platform described in step S2 will record the number of iterations and parameter values;
[0096] f. Determine whether the number of iterations exceeds the upper limit of 100. If it exceeds 100, stop calibration. If it does not exceed 100, continue to determine whether the parameter value cannot be significantly improved. If the parameter value cannot be significantly improved for more than 2 times, stop calibration. If the parameter value can be significantly improved for more than 2 times, continue to determine whether the objective function value cannot be significantly improved for more than 10 times. If the objective function value cannot be significantly improved for more than 10 times, it is considered that the objective function has converged and stop calibration. If the objective function value can be significantly improved for more than 10 times, determine whether the objective function value is greater than 0.8. If so, stop calibration, otherwise return to step e.
[0097] Among them, whether the objective function value can be significantly improved can be expressed by the following mathematical expression:
[0098]
[0099] Among them, λ i-1 and λ i are the objective function values at step i-1 and step i, TOL λ is the convergence threshold, TOL λ The value is usually assigned to 0.0001. When the value calculated on the left side of the formula is less than or equal to the convergence threshold, it is considered that the objective function value cannot be significantly improved.
[0100] In the SCE-UA algorithm, each vertex of a composite is a potential parent and may participate in the evolutionary computation of the next generation of the population. Each sub-composite acts like a parent. At the beginning of the algorithm, a randomly distributed point group is introduced into the parameter feasible domain and these point groups are divided into several composite samples. The specific steps are as follows:
[0101] (1) Select the number p of complexes participating in iterative evolution (p>=1) and the number m of vertices contained in each complex (m>=n+1), and calculate the number of sample points s=pm, where n is the dimension, and n=8 in this invention;
[0102] (2) Randomly generate s sample points x1, ..., x1 in the parameter feasible region. s , calculate each point x separately i The simulation result value of
[0103] (3) Calculate the average value of the Nash efficiency coefficient based on the obtained simulation result value and the measured value of the monitoring parameter input previously;
[0104] (4) Sort the s sample points in ascending order according to the average value of the Nash efficiency coefficient, and record D = {(x i , f i ), i=1,...,s};
[0105] (5) Divide D into p composite types A1,…,Ap , each complex type contains m points, among which
[0106] (6) Evolving each composite sample separately using a competitive composite evolutionary algorithm;
[0107] (7) All vertices of each evolved composite sample are combined into a new point set, and are again arranged in ascending order of the average value of the Nash efficiency coefficient.
[0108] (8) If the convergence condition is met, stop the calculation, otherwise continue with step (5).
[0109] Select multiple monitoring points, set different weights for each monitoring point, and take the weighted average of the NASH efficiency coefficients of each monitoring point as the objective function, and take the weighted average of the NASH-SUTCLIFFE NASH efficiency coefficients of all monitoring points as the objective function. The objective function can be expressed as follows:
[0110]
[0111] Among them, w i is the weight of the monitoring point, E i is the Nash efficiency coefficient of the observation point, which can be expressed as follows:
[0112]
[0113] Among them, Q o Refers to the observed value, Q m Refers to the analog value, Q t represents a value at time t, It represents the total average of the observations. When E is close to 1, it means that the model quality is good and the model credibility is high; when E is close to 0, it means that the simulation results are close to the average level of the observations, that is, the overall results are credible, but the simulation error is large; when E is much smaller than 0, it means that the model is unreliable.
[0114] After the pipe network model is established, some parameters are difficult to set through measurement, so calibration is needed to optimize these parameters. Before parameter calibration, we first need to determine: (1) the settings of the parameters to be optimized (such as the sensitive parameters in step (1)); (2) determine the objective function (this step requires selecting monitoring points, which can be multiple monitoring points), and then coordinate the proportion of each monitoring point in the objective function by setting weights. For example, there are two monitoring points: monitoring point 1 and monitoring point 2. Assuming that monitoring point 1 is more important, the weight of monitoring point 1 can be set to 0.7, and the weight of monitoring point 2 can be set to 0.3 to construct the final objective function. The Nash efficiency coefficient of each point in the formula is calculated based on the measured value (i.e., online monitoring data) and the simulated data of the monitoring point after the pipe network model simulation; (3) the same model input (the above 1 and 2 points need to be carried out under the same input conditions, such as the same rainfall input, etc.). The execution steps of the composite cross evolutionary algorithm continuously adjust the parameters to be optimized to perform SWMM simulation and then calculate the objective function until the convergence condition is reached.
[0115] S6: Compare the simulation results with monitoring data from different sessions. The simulation results are used to verify the accuracy of the calibrated SWMM network model. Models with poor calibration results will be recalibrated as the initial model. Using monitoring data from different sessions to verify the calibrated model improves model robustness and reliability.
[0116] First, the parameters must be calibrated. After calibration, the calibrated parameters are subjected to model simulation calculations, and the calculated values are compared with the measured values to verify the accuracy of the parameter calibration.
[0117] In the calibrated model, rainfall and inflow data from different episodes were used as model inputs. A four-hour simulation period was set, with the start and end times encompassing the time range of the rainfall and inflow data. The simulation was performed using the Horton infiltration method. The simulated water depth and total inflow were compared with the monitoring data, and a weighted average of the Nash efficiency coefficients at the monitoring points was calculated. The average value was used to verify the validity of the results. When the coefficient value was greater than 0.8, the model was considered reliable. When the value was less than 0.8, the model parameters were further calibrated using the monitoring data from that episode.
[0118] The Horton infiltration method is used to simulate the water flow. The SWMM engine is responsible for this. We only need to tell the SWMM engine to select the Horton infiltration method. The details are as follows:
[0119] The Horton infiltration method is expressed by the curve equation:
[0120] f(t)=f c +(f0-f c )e-kt
[0121] f is the infiltration rate, f c To stabilize the infiltration rate, f0 is the initial infiltration rate, k is the infiltration constant, and t is the time in seconds. This is calculated by the SWMM engine based on the model and rainfall and inflow data.
[0122] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.
Claims
1. A method for using online monitoring network data to assist in constructing a pipe network model, characterized in that: The steps include: S1: Establishing a SWMM pipe network model, wherein the SWMM pipe network model includes pipe network nodes, catchment areas, pipe sections, and drainage outlets; S2: Generate an online monitoring point layer according to the monitoring point information provided by the online monitoring system; S3: Associating the SWMM pipe network model established in step S1 with the online monitoring point layer generated in step S2 to obtain associated monitoring points; S4: Automatically map the SWMM network model and the online monitoring point layer based on the associated monitoring points; S5: Select monitoring parameters and calibration facilities, calibrate SWMM pipe network model parameters, and obtain the calibrated SWMM pipe network model; S6: Use monitoring data from different sessions to verify the calibrated SWMM network model; Among them, in step S4, automatically mapping the pipe network model and the online monitoring point layer includes the following steps: 1) Obtain a control point set according to the associated monitoring points obtained in step S3; 2) Check the control point residuals. If there are control point residuals greater than e -8 When , add a new control point; 3) Use thin plate spline interpolation for coordinate transformation; In step S5, the monitoring parameters include the water depth and total inflow of the pipe network node; the calibration facilities include the pipe network node associated with the monitoring point and the corresponding pipe section; A composite crossover evolutionary algorithm is used for parameter calibration, and the weighted average value of the Nash efficiency coefficient of each associated monitoring point is used as the objective function value. At the same time, a convergence threshold of the objective function value is set. When the parameter value cannot be improved for more than 2 times, the calibration procedure is terminated; when the objective function value cannot be improved for more than N times, the calibration procedure is terminated, and N is assigned to 10; when the objective function value is greater than 0.8, the calibration is stopped; Among them, whether the objective function value can be improved is expressed by the following mathematical expression: Among them, λ i-1 and λ i are the objective function values at step i-1 and step i, respectively, TOL λ is the convergence threshold, TOL λ Assign a value of 0.0001, max is the maximum objective function value in the previous i iterations, λ min is the minimum objective function value in the previous i iterations; Select at least one monitoring point, set a weight for each monitoring point, and take the weighted average of the Nash efficiency coefficients of all monitoring points. The mathematical expression can be expressed as: Among them, w i is the weight of the monitoring point, E i is the Nash efficiency coefficient of the observation point, which can be expressed mathematically as follows: Among them, Q o refers to the observed value, Q m Refers to the analog value, Q t represents a value at time t, It represents the total average of the observations. When E is close to 1, it means that the model quality is good and the model credibility is high.
2. The method according to claim 1, characterized in that In step S2, the monitoring point location information includes longitude, latitude, monitoring point equipment type and name, and monitoring data items; The types of equipment at the monitoring points include: rain gauges, flow meters, liquid level meters, water quality meters and dissolved oxygen flow meters; The monitoring data items include: 。 3. The method according to claim 1, characterized in that In step S3, the pipe network nodes in the pipe network model include inspection wells, and the types of inspection wells include three types: rainwater wells, sewage wells and combined wells. Among them, the rainwater wells are associated with monitoring points whose equipment types are rain gauges, liquid level meters and flow meters. The sewage wells are associated with monitoring points whose equipment types are water quality meters and dissolved oxygen flow meters. The combined wells are associated with monitoring points whose equipment types are rain gauges, liquid level meters, flow meters, water quality meters and dissolved oxygen flow meters.
4. The method according to claim 1, characterized in that: The control points in step 1) are concentratedly selected from the pipe network node coordinates in the SWMM pipe network model as the source coordinates, and the associated monitoring point coordinates are the target coordinates.
5. The method according to claim 1, characterized in that The residual in step 2) represents the difference in distance between the source point and the target point in the control point; the point of the pipe network model is called the source point, and the point of the monitoring point layer is called the target point.
6. The method according to claim 1, characterized in that The thin plate spline interpolation in step 3) is, for the selected n pairs of marked points p i and q i , i = 1, ... n, according to the correspondence of the marks in the two control points, find the transformation function f in the two-dimensional space so that: The function U(r) = r 2 logr 2 is the basis function for constructing thin plate splines, w=(w1,w2,...,w n ) T , a=(a1,a x , a y ) T , find the new corresponding points (x′, y′) obtained after the coordinates (x, y) in all source images are transformed.