Optimal layout method of detention basin based on stormwater pipe network model and decision tree algorithm

Abstract

The application discloses a kind of based on rain flood pipe network model and decision tree algorithm's regulating and storing pool optimization layout method, first city topography pipe network basic data acquisition, drainage pipe network monitoring data integration, the construction of SWMM rain flood pipe network model, the automatic rating of SWMM model parameter based on decision tree algorithm, rainfall scenario database generation, regulating and storing pool layout scheme prefabrication, regulating and storing pool layout scenario simulation and generate database, establish regulating and storing pool benefit quantification evaluation system, simulate regulating and storing pool multiple combination layout under the condition of benefit index, establish evaluation index and regulating and storing pool layout scheme deep connection in combination with decision tree algorithm, by setting evaluation index inversion regulating and storing pool layout scheme.The application is aimed at city initial rain surface source pollution is serious, regulating and storing pool layout lacks, initial rain pollutant reduction effect is poor problem, by combining rain flood pipe network model and decision tree algorithm, the optimal layout of regulating and storing pool is realized, and the initial rain pollution reduction benefit of regulating and storing pool is maximized.

Description

Technical Field

[0001] This invention belongs to the field of pollutant reduction and control technology of urban drainage systems, and relates to a method for optimizing the layout of stormwater storage tanks, specifically a method for optimizing the layout of stormwater storage tanks based on stormwater pipe network models and decision tree algorithms. Background Technology

[0002] With the development of urbanization, my country's sewage and rainwater pipe networks and sewage treatment plants have been gradually improved, and urban point source pollution of rivers, lakes and other receiving water bodies has been effectively controlled. However, the problem of non-point source pollutants being carried into rivers and lakes by rainwater erosion and pipe networks, causing the deterioration of river and lake water quality, remains serious. Therefore, it is of great significance to study how to effectively control non-point source pollution.

[0003] Currently, constructing stormwater storage tanks at the end of drainage systems to collect and store initial rainwater with high pollutant content before discharging it to wastewater treatment plants for initial rainwater purification is an important means of reducing urban non-point source pollution emissions. However, due to varying pipe network catchment areas and different catchment times, the traditional practice of placing stormwater storage tanks at the end of drainage systems is difficult to effectively utilize their interception capacity, resulting in limited pollutant reduction capabilities. Therefore, there is an urgent need to develop more rational stormwater storage tank deployment methods to fully leverage their pollution reduction capabilities and maximize their initial rainwater pollution reduction benefits.

[0004] In recent years, with the significant improvement of computing power, machine learning algorithms and physical process-based numerical models of stormwater pipe networks have made great strides. Among them, decision tree algorithms have been widely used in fields such as strategy optimization and image recognition, while stormwater pipe network models also play an important role in urban stormwater process simulation. Summary of the Invention

[0005] The purpose of this invention is to provide an optimized layout method for stormwater storage tanks based on stormwater pipe network models and decision tree algorithms. This method addresses the problem of poor initial rainwater pollutant reduction efficiency in traditional stormwater storage tank layout methods by combining stormwater pipe network models and decision tree algorithms to optimize the layout of stormwater storage tanks and fully leverage their initial rainwater pollution reduction benefits.

[0006] The technical solution adopted in this invention is: a method for optimizing the layout of stormwater storage tanks based on a stormwater pipe network model and a decision tree algorithm, comprising the following steps:

[0007] Step 1: Collect basic data for the target area, including digital elevation data, land use data, soil infiltration capacity, pipeline vector data, node vector data, water storage facility data, water system distribution data, river water level and flow rate, and river cross-section data;

[0008] Step 2: Construct a three-dimensional array to integrate measured rainfall and monitoring data. The first dimension is the measured rainfall event, the second dimension is the time series, and the third dimension is the rainfall meteorological monitoring data, water quality and flow monitoring data, and drainage network monitoring data. The third dimension data specifically includes rainfall observation data at each station in the target area, water level data at drainage network monitoring points, flow velocity data at drainage network monitoring points, flow process data at drainage network monitoring points, pollutant concentration data at drainage network monitoring points, and total pollutant discharge data at drainage network monitoring points.

[0009] Step 3: Construct an SWMM model based on the data collected in Step 1, including a surface runoff generation module, an underground pipe network confluence module, and a pollutant growth and transport module;

[0010] Step 4: Automatic calibration of SWMM model parameters;

[0011] Using the rainfall observation data in step 2 as input, and the changes in flow and pollutant concentration monitored by the monitoring points as objectives, the automatic calibration of the SWMM model parameters constructed in step 3 is achieved by combining the grid search algorithm.

[0012] Step 5: By combining the Chicago rainfall pattern with the target area's rainfall intensity coefficient, or by combining the typical rainfall pattern in the target area's hydrological manual with the target area's rainfall intensity coefficient, adjust the return period P and rainfall duration t to simulate and construct several rainfall scenario data with different return periods and rainfall durations, and generate a rainfall scenario database.

[0013] The Chicago rain pattern formula is as follows:

[0014]

[0015] In the formula, i is the intensity of the rainstorm, in mm / h; P is the return period of rainfall, in years; t is the duration of rainfall, in minutes; A, b, c, and n are local rainfall pattern coefficients.

[0016] Step 6: Based on the land use type and project construction budget of the study area, determine the possible locations of the water storage ponds and their total water storage volume;

[0017] Step 7: Simulate the deployment scenario of the water storage tank and generate a database;

[0018] Using the SWMM model calibrated in step 4, the changes in water level, flow rate, and pollutant concentration at the outlet under various rainfall scenarios and storage tank layout schemes are simulated, along with the number of overflows at pipeline nodes and the total amount of pollutant overflows. The data are then compiled and a database is generated.

[0019] Step 8: Establish a quantitative evaluation system for the benefits of water storage ponds, including evaluation indicators for pollution reduction capacity, flood drainage capacity, interception capacity, and comprehensive benefits.

[0020] Step 9: Calculate the benefit indicators under the condition of multiple combinations of storage tank layout;

[0021] Based on the simulation results in step 7, calculate the values ​​of the various evaluation indicators proposed in step 8 under the conditions of each storage tank layout.

[0022] Step 10: Establish a deep connection between evaluation indicators and storage tank layout schemes by combining decision tree algorithm. Analyze and integrate the evaluation indicator data obtained in step 9 with their corresponding storage tank layout schemes. Using the evaluation indicator data as input conditions and the storage tank layout scheme as the target result, establish a deep connection by combining decision tree algorithm to optimize the storage tank layout scheme through the optimization of evaluation indicators.

[0023] Compared with existing technologies, the beneficial effects of this invention are as follows: Addressing the problems of severe urban rainwater non-point source pollution, lack of basis for regulating reservoir layout, and poor rainwater pollutant reduction effects, this invention establishes an SWMM stormwater network model to simulate several regulating reservoir layout schemes, constructs a quantitative evaluation system for regulating reservoir benefits to analyze the effectiveness of the layout schemes, and then combines a decision tree algorithm to establish a deep connection between evaluation indicators and regulating reservoir layout schemes. By inputting target benefit indicators, the optimal layout scheme for corresponding regulating reservoirs can be quickly output, maximizing the rainwater pollution reduction benefits of regulating reservoirs. Attached Figure Description

[0024] Figure 1 This is a flowchart of the method according to an embodiment of the present invention;

[0025] Figure 2 This is a flowchart of the automatic calibration process for SWMM model parameters based on the grid search algorithm constructed in this invention;

[0026] Figure 3 This is a flowchart of the inversion process for optimizing the layout of water storage ponds based on the decision tree algorithm constructed in this invention. Detailed Implementation

[0027] To facilitate understanding and implementation of the present invention by those skilled in the art, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0028] Please see Figure 1 This invention provides a method for optimizing the layout of stormwater storage tanks based on a stormwater pipe network model and a decision tree algorithm, comprising the following steps:

[0029] Step 1: Collect basic data for the target area, including digital elevation data, land use data, soil infiltration capacity, pipeline vector data, node vector data, water storage facility data, water system distribution data, river water level and flow rate, and river cross-section data;

[0030] Step 2: Construct a three-dimensional array to integrate measured rainfall and monitoring data. The first dimension is the measured rainfall event, the second dimension is the time series, and the third dimension is the rainfall meteorological monitoring data, water quality and flow monitoring data, and drainage network monitoring data. The third dimension data specifically includes rainfall observation data at each station in the target area, water level data at drainage network monitoring points, flow velocity data at drainage network monitoring points, flow process data at drainage network monitoring points, pollutant concentration data at drainage network monitoring points, and total pollutant discharge data at drainage network monitoring points.

[0031] Step 3: Construct an SWMM model based on the data collected in Step 1, including a surface runoff generation module, an underground pipe network confluence module, and a pollutant growth and transport module;

[0032] The construction process of the surface runoff generation module in this embodiment is as follows:

[0033] Based on the topographic elevation and the distribution of pipeline inspection wells in the study area, the study area is divided into several sub-catchment areas. Each sub-catchment area consists of three parts: a permeable area, an impermeable area with depression storage capacity, and an impermeable area without depression storage capacity. The corresponding surface runoff is also divided into three parts: permeable area runoff R1, impermeable area runoff with depression storage capacity R2, and impermeable area without depression storage capacity R3.

[0034] 1) For permeable areas, once rainfall meets the surface infiltration conditions, water begins to accumulate on the surface until it exceeds the depression's storage capacity, thus forming surface runoff. The runoff calculation formula is:

[0035] R1 = (if)·Δt;

[0036] In the formula, R1 is the permeable flow rate in the permeable zone, in mm; i is the rainfall intensity, in mm / h; f is the surface infiltration rate, in mm / h; Δt is the calculation time interval, in h;

[0037] The variation of surface infiltration capacity over time is described using the Horton model, with the expression: f = (f0 - f ∞ )e -kt +f ∞

[0038] In the formula, f represents the surface infiltration capacity, in mm / h; f0, f ∞ , representing the initial infiltration rate and the steady-state infiltration rate, respectively, in mm / h; t, representing the rainfall time, in h; and k, representing the infiltration attenuation index, which is closely related to soil conditions.

[0039] 2) For impermeable depressions, runoff can form once the rainfall reaches the maximum surface water storage capacity. The runoff calculation formula is:

[0040] R2 = PD;

[0041] In the formula, R2 is the runoff of the impermeable zone with water storage capacity, in mm; P is the rainfall, in mm; and D is the water storage capacity, in mm.

[0042] 3) For impermeable areas without depressions, rainfall is essentially converted into runoff, except for surface evaporation. Runoff occurs when rainfall exceeds evaporation. The formula for calculating runoff generation is:

[0043] R3 = PE;

[0044] In the formula, R3 is the runoff of the permeable water storage area without depressions, in mm; P is the rainfall, in mm; and E is the evaporation, in mm.

[0045] The construction process of the underground pipeline network manifold module in this embodiment is as follows:

[0046] The flow confluence process in the pipeline is solved using the dynamic wave method, which is a complete one-dimensional Saint-Venant equation. It is solved by simultaneously solving the continuity equations, combining the calculation of water level at the nodes and the flow rate in the pipeline. This method can be applied to complex flow calculations. Its governing equations are:

[0047]

[0048] In the formula, Q is the instantaneous flow rate, and the unit is m³ / s. 3 / s; A is the cross-sectional area of ​​the water passage, in square meters; x is the pipe length, in meters; t is time, in seconds; H is the water depth, in meters; g is the acceleration due to gravity, in meters per second. 2 S f denoted as ν, representing the energy gradient caused by frictional loss; ν is the overall roughness coefficient of the pipe; R is the hydraulic radius in meters; and v is the average flow velocity across the cross section in meters per second.

[0049] In the governing equations of dynamic waves, there is a pressure term passing through the unpressurized pipe. Characterization: Inflow and outflow are represented by the sign of the flow rate Q; energy loss mainly considers losses caused by friction, denoted by S. f After characterization, the equation can be solved using the finite difference method, and its finite difference form can be expressed as:

[0050]

[0051] In the formula, Q t+Δt Let Q be the flow rate at time t+Δt. t Let A2 be the flow rate at time t; A2 and A1 are the cross-sectional areas of the pipe segment at the upper and lower nodes, and H2 and H1 are the water depths at the upper and lower nodes of the pipe segment. Δt represents the average cross-sectional area and flow velocity during the time interval Δt; ΔA represents the change in cross-sectional area during the time interval Δt; L represents the length of the pipe segment; n represents the overall roughness of the pipe; R represents the hydraulic radius; and g represents the acceleration due to gravity.

[0052] The construction process of the pollutant growth and transport module in this embodiment is as follows:

[0053] Pollutant transport mainly includes source pollution and non-point source pollution. This invention focuses on non-point source pollution caused by street scouring and does not consider pollutant concentrations in rainfall. The pollutant growth and transport process is characterized by street pollutant growth and pollutant scouring; specifically, street pollutant growth is represented by a saturation function, meaning that pollutant growth begins at a linear rate, and the cumulative growth rate decreases continuously over time until a saturation value is reached, expressed as:

[0054]

[0055] In the formula, B represents the cumulative amount of pollutants, in kg / m²; C1 represents the maximum cumulative amount of pollutants per unit area, in kg / m². 2 C2 is the half-saturation constant, which is the number of days when half of the maximum cumulative pollutant amount per unit area is reached; t is the number of days.

[0056] The pollutant flushing is set as exponential flushing, and its flushing load capacity W is:

[0057]

[0058] In the formula, W1 is the scour coefficient; W2 is the scour index; q is the runoff rate per unit area, in mm / h; and B is the cumulative amount of pollutants, in kg / m². 2 .

[0059] Step 4: Automatic calibration of SWMM model parameters;

[0060] Using the rainfall observation data in step 2 as input, and the changes in flow and pollutant concentration monitored by the monitoring points as objectives, the automatic calibration of the SWMM model parameters constructed in step 3 is achieved by combining the grid search algorithm.

[0061] Please see Figure 2 Step 4 includes the following sub-steps:

[0062] Step 4.1: Determine the reasonable range of model parameters based on literature review;

[0063] Step 4.2: Automatically construct parameter combinations using a grid search algorithm, using the rainfall data from Step 2 as input conditions, and the flow rate and pollutant concentration changes monitored at the monitoring points as calibration targets for model training;

[0064] Step 4.3: Output the optimal combination of parameters for the model;

[0065] Step 4.4: Select ≥2 measured rainfall events to verify the flow process and pollutant concentration change process at the monitoring nodes. Use the Nash coefficient as the evaluation index. If the Nash coefficient is greater than the threshold (0.7 in this embodiment), output the model; otherwise, proceed to step 4.1, adjust the parameter range, and retrain the model.

[0066] This embodiment includes, but is not limited to, the changes in flow rate and pollutant concentration at the monitoring points, and can add calibration conditions for water level, flow velocity, and total pollutant amount;

[0067] Step 5: By combining the Chicago rainfall pattern with the target area's rainfall intensity coefficient, or by combining the typical rainfall pattern in the target area's hydrological manual with the target area's rainfall intensity coefficient, adjust the return period P and rainfall duration t to simulate and construct several rainfall scenario data with different return periods and rainfall durations, and generate a rainfall scenario database.

[0068] The Chicago rain pattern formula is as follows:

[0069]

[0070] In the formula, i is the intensity of the rainstorm, in mm / h; P is the return period of rainfall, in years; t is the duration of rainfall, in minutes; A, b, c, and n are local rainfall pattern coefficients.

[0071] Step 6: Based on the land use type and project construction budget of the study area, determine the possible locations of the water storage ponds and their total water storage volume;

[0072] Step 7: Simulate the deployment scenario of the water storage tank and generate a database;

[0073] Using the SWMM model calibrated in step 4, the changes in water level, flow rate, and pollutant concentration at the outlet under various rainfall scenarios and storage tank layout schemes are simulated, along with the number of overflows at pipeline nodes and the total amount of pollutant overflows. The data are then compiled and a database is generated.

[0074] Step 8: Establish a quantitative evaluation system for the benefits of water storage ponds, including evaluation indicators for pollution reduction capacity, flood drainage capacity, interception capacity, and comprehensive benefits.

[0075] In this embodiment, the pollution reduction capacity evaluation indicators include, but are not limited to, the reduction of overflow pollution load, the reduction of pollution load into rivers, the reduction of influent to sewage treatment plants, and the reduction of influent pollution load to sewage treatment plants.

[0076] In this embodiment, the overflow pollution load reduction amount represents the difference between the total overflow pollution load before the operation of the stormwater and sewage storage tank and the total overflow pollution load after the operation of the stormwater and sewage storage tank. The total overflow pollution load is:

[0077]

[0078] Among them, Q 污 The total pollution load of the overflow outlet is given by n, where n is the number of overflow outlet inflow data points during the calculation period, m is the total number of simulation time intervals, and C is the total pollution load of the overflow outlet. ij Let Q be the pollutant concentration at the i-th confluence during the j-th time interval. ij Let t be the inflow rate of the i-th overflow outlet at the j-th time interval. j Let be the time interval for water to flow from the j-th overflow outlet;

[0079] In this embodiment, the reduction in pollution load entering the river represents the difference between the total pollution load entering the river before the operation of the stormwater and sewage storage tank and the total pollution load entering the river after the operation of the stormwater and sewage storage tank.

[0080] In this embodiment, the reduction in the influent volume of the wastewater treatment plant represents the difference in the total influent volume of the wastewater treatment plant before and after the operation of the stormwater and sewage storage tank.

[0081] In this embodiment, the reduction in influent pollution load of the wastewater treatment plant represents the difference in influent pollution load before and after the operation of the stormwater and sewage storage tank, expressed as:

[0082] ΔW 污水厂负荷 =Q 调蓄池运行前 C 调蓄池运行前 -Q 调蓄池运行后 C 调蓄池运行后 ;

[0083] Wherein, ΔW 污水厂负荷 Q represents the reduction in influent pollution load at the wastewater treatment plant. 调蓄池运行前 C represents the influent volume of the wastewater treatment plant before the regulating reservoir is put into operation. 调蓄池运行前 Q represents the influent concentration of the wastewater treatment plant before the regulating reservoir is put into operation. 调蓄池运行后 C represents the influent volume of the wastewater treatment plant after the regulating reservoir is put into operation. 调蓄池运行后 This indicates the concentration of the influent to the wastewater treatment plant after the regulating reservoir is in operation.

[0084] In this embodiment, the drainage capacity evaluation indicators include, but are not limited to, the reduction in the number of nodes overflowing, the reduction in the time of node overflowing, the reduction in the length of the full pipe, the reduction in the duration of the full pipe, and the reduction in runoff into the river.

[0085] In this embodiment, the reduction in the number of nodes overflowing is calculated as the difference between the sum of the overflow amounts of each node before the operation of the stormwater and sewage storage tank and the sum of the overflow amounts of each node after the operation of the stormwater and sewage storage tank.

[0086] In this embodiment, the reduction in node overflow time is calculated as the difference between the sum of the overflow times of each node before the operation of the stormwater and sewage storage tank and the sum of the overflow times of each node after the operation of the stormwater and sewage storage tank.

[0087] In this embodiment, the reduction in full pipe length is the difference between the total length of the full pipe section before the stormwater and sewage storage tank is put into operation and the total length of the full pipe section after the stormwater and sewage storage tank is put into operation.

[0088] In this embodiment, the reduction in the full pipe duration is the difference between the sum of the full pipe operating times of each pipe section before the operation of the stormwater and sewage storage tank and the sum of the full pipe operating times of each pipe section after the operation of the stormwater and sewage storage tank.

[0089] In this embodiment, the reduction in river runoff represents the difference in river runoff before and after the operation of the storage tank.

[0090] In this embodiment, the evaluation indicators for interception capacity include, but are not limited to, overflow reduction and interception volume.

[0091] In this embodiment, the overflow reduction amount is:

[0092]

[0093] In the formula, Q 减 Let Q be the total overflow rate of the overflow outlet, n be the number of overflow inflow data points during the calculation period, m be the total simulation time interval, and Q be the total overflow rate of the overflow outlet. ij调蓄池运行前 Let Q be the flow rate at the i-th overflow outlet at the j-th time before the storage tank starts operating. ij调蓄池运行后 Let t be the flow rate at the i-th overflow outlet at the j-th time after the storage tank has been put into operation. j For the j-th time interval;

[0094] In this embodiment, the intercepted water volume is the total inflow into the regulating reservoir:

[0095]

[0096] Among them, Q 截流 Let be the inflow rate of the rainwater storage tank, n be the number of overflow inflow data points during the calculation period, m be the total simulation time interval, and Q be the inflow rate of the rainwater storage tank. ij Let t be the inflow rate of the i-th regulating reservoir at time j. j Let j be the j-th time interval.

[0097] In this embodiment, the comprehensive benefit evaluation index includes, but is not limited to, the pollutant reduction ratio. The ratio of the total amount of pollutants intercepted by the storage tank to the total volume of the storage tank represents the benefit between the construction cost of the storage tank and the reduction of initial rainwater pollution by the storage tank.

[0098] Step 9: Calculate the benefit indicators under the condition of multiple combinations of storage tank layout;

[0099] Based on the simulation results in step 7, calculate the values ​​of the various evaluation indicators proposed in step 8 under the conditions of each storage tank layout.

[0100] Step 10: Establish a deep connection between evaluation indicators and storage tank layout schemes by combining decision tree algorithm. Analyze and integrate the evaluation indicator data obtained in step 9 with their corresponding storage tank layout schemes. Using the evaluation indicator data as input conditions and the storage tank layout scheme as the target result, establish a deep connection by combining decision tree algorithm to optimize the storage tank layout scheme through the optimization of evaluation indicators.

[0101] Please see Figure 3 In this embodiment, step 10 is specifically implemented by including the following sub-steps:

[0102] Step 10.1: Analyze and integrate the evaluation index data obtained in Step 9 with their corresponding flood storage tank layout schemes. Using the evaluation index data as input conditions and the flood storage tank layout schemes as the target results, construct a model training dataset D = {(x1, y1), (x2, y2), ..., (x... n y n )}, where y n Let y represent the nth combination of water storage tank layouts. n =(v1,v2,…,v n ), where v n Let x be the storage volume of the nth storage tank; n For the nth combination of water storage tanks, step 9 simulates and calculates the corresponding benefit index, denoted as x. n =(para1,para2,...,para n ), where para n The evaluation indicators include, but are not limited to, those proposed in step 8, namely, the reduction of overflow pollution load, the reduction of pollution load into the river, the influent volume of the sewage treatment plant, the reduction of pollution load, the reduction of the number of overflows at nodes, the reduction of overflow time at nodes, the reduction of full pipe length, the reduction of full pipe duration, the reduction of runoff into the river, the reduction of overflow, the intercepted water volume, and the pollutant reduction ratio.

[0103] Step 10.2: A deep relationship between input conditions and output results will be established using the decision tree algorithm. The relationship between input conditions and output results will be represented as follows:

[0104]

[0105] In the formula, f(x) is the predictor variable, and M represents the input data feature space divided into M regions R1, R2, ..., R through spatial partitioning. M c m For R m The output value in the space, where I is the identity matrix and x is the input variable;

[0106] Step 10.3: Find the optimal split point and optimal split variable, using the squared error. To minimize the splitting error (Value), a heuristic algorithm is used to find the optimal splitting point, expressed as:

[0107]

[0108] In the formula, j represents the total number of variables, s represents the split point, and x represents the x-axis. i For the i-th index combination, y i For x i The corresponding combination of storage tanks under the indicator, R1 and R2 are the two variable spaces segmented, R1(j,s)={x|x (j) ≤s}, R2(j,s)={x|x (j) >s},x (j) c1 represents the j-th variable in the x-index combination; c1 and c2 are the output values ​​corresponding to the two variable spaces, c1 = ave(y i |x i ∈R1(j,s)), c2=ave(y i |x i ∈R2(j,s)), by traversing all input variables, determine the split point s for each variable j, and when the Value reaches the minimum value, the optimal variable j and the optimal split point s are determined;

[0109] Step 10.4: Calculate the output value of the divided region under the conditions of optimal segmentation feature j and optimal segmentation point s. Represented as:

[0110]

[0111] In the formula, x i ∈R m m = 1, 2, N m For space R m The number of combined storage tank schemes in the middle, y i For x i The corresponding combination of regulating reservoirs under the indicators;

[0112] Step 10.5: Repeat steps 10.3-10.4 until the maximum number of layers in the decision tree reaches the set threshold or all flood control pond schemes are individually partitioned. After pruning, the variable space is divided into M regions R1, R2, ..., R M Generate a decision tree model;

[0113] Step 10.6: Test the decision tree effect using a test set, i.e., input feature indicators, and invert the optimized layout scheme of the water storage tanks using the decision tree model generated in Step 10.5; further, use the optimized layout scheme of the water storage tanks obtained by inversion as a condition to drive the SWMM model and calculate the indicator evaluation system proposed in Step 8; finally, compare it with the input feature indicators of the initial decision tree model to verify the reliability of the constructed decision tree model; when the relative error of each indicator can be controlled within the threshold (15% in this embodiment), the constructed decision tree model is considered to meet the requirements, and the decision tree model is output and stored; otherwise, return to Step 10.1, readjust the input indicators, and train the model.

[0114] Step 10.7: Automatically generate an optimized layout plan for storage tanks by inputting the benefit evaluation indicators of the storage tanks.

[0115] This invention combines stormwater pipe network models with decision tree algorithms to optimize the layout of stormwater storage tanks, which has broad application prospects for fully utilizing the potential of stormwater storage tanks and further reducing non-point source pollution caused by initial urban rain.

[0116] It should be understood that the above description of the preferred embodiments is quite detailed, but it should not be considered as a limitation on the scope of protection of this invention. Those skilled in the art, under the guidance of this invention, can make substitutions or modifications without departing from the scope of protection of the claims of this invention, and all such substitutions or modifications fall within the scope of protection of this invention. The scope of protection of this invention should be determined by the appended claims.

Claims

1. A method for optimizing the layout of stormwater storage tanks based on a stormwater pipe network model and a decision tree algorithm, characterized in that, Includes the following steps: Step 1: Collect basic data for the target area, including digital elevation data, land use data, soil infiltration capacity, pipeline vector data, node vector data, water storage facility data, water system distribution data, river water level and flow rate, and river cross-section data; Step 2: Construct a three-dimensional array to integrate measured rainfall and monitoring data. The first dimension is the measured rainfall event, the second dimension is the time series, and the third dimension is the rainfall meteorological monitoring data, water quality and flow monitoring data, and drainage network monitoring data. The third dimension data specifically includes rainfall observation data at each station in the target area, water level data at drainage network monitoring points, flow velocity data at drainage network monitoring points, flow process data at drainage network monitoring points, pollutant concentration data at drainage network monitoring points, and total pollutant discharge data at drainage network monitoring points. Step 3: Construct an SWMM model based on the data collected in Step 1, including a surface runoff generation module, an underground pipe network confluence module, and a pollutant growth and transport module; Step 4: Automatic calibration of SWMM model parameters; Using the rainfall observation data in step 2 as input, and the changes in flow and pollutant concentration monitored by the monitoring points as objectives, the automatic calibration of the SWMM model parameters constructed in step 3 is achieved by combining the grid search algorithm. Step 5: By combining the Chicago rainfall pattern with the target area's rainfall intensity coefficient, or by combining the typical rainfall pattern in the target area's hydrological manual with the target area's rainfall intensity coefficient, adjust the return period P and rainfall duration t to simulate and construct several rainfall scenario data with different return periods and rainfall durations, and generate a rainfall scenario database. The Chicago rain pattern formula is as follows: In the formula, i The intensity of the rainstorm is expressed in mm / h. P The return period for rainfall is expressed in years (a). t The duration of rainfall is expressed in minutes; A, b, c, and n are local rainfall pattern coefficients. Step 6: Based on the land use type and project construction budget of the study area, determine the possible locations of the water storage ponds and their total water storage volume; Step 7: Simulate the deployment scenario of the water storage tank and generate a database; Using the SWMM model calibrated in step 4, the changes in water level, flow rate, and pollutant concentration at the outlet under various rainfall scenarios and storage tank layout schemes are simulated, along with the number of overflows at pipeline nodes and the total amount of pollutant overflows. The data are then compiled and a database is generated. Step 8: Establish a quantitative evaluation system for the benefits of water storage ponds, including evaluation indicators for pollution reduction capacity, flood drainage capacity, interception capacity, and comprehensive benefits. Step 9: Calculate the benefit indicators under the condition of multiple combinations of storage tank layout; Based on the simulation results in step 7, calculate the values ​​of the various evaluation indicators proposed in step 8 under the conditions of each storage tank layout. Step 10: Establish a deep connection between evaluation indicators and storage tank layout schemes by combining decision tree algorithm. Analyze and integrate the evaluation indicator data obtained in step 9 with their corresponding storage tank layout schemes. Using the evaluation indicator data as input conditions and the storage tank layout scheme as the target result, establish a deep connection by combining decision tree algorithm to optimize the storage tank layout scheme through the optimization of evaluation indicators.

2. The method for optimizing the layout of stormwater storage tanks based on stormwater pipe network model and decision tree algorithm according to claim 1, characterized in that: The surface runoff generation module described in step 3 is constructed as follows: Based on the topographic elevation and the distribution of pipeline inspection wells in the study area, the study area is divided into several sub-catchment areas. Each sub-catchment area consists of three parts: a permeable area, an impermeable area with depression storage capacity, and an impermeable area without depression storage capacity. The corresponding surface runoff is also divided into three parts: permeable area runoff R1, impermeable area runoff with depression storage capacity R2, and impermeable area without depression storage capacity R3. 1) For permeable areas, once rainfall meets the surface infiltration conditions, water begins to accumulate on the surface until it exceeds the depression's storage capacity, at which point surface runoff is formed. The runoff calculation formula is: In the formula, The flow rate of the permeable zone, in mm; i Rainfall intensity, in mm / h; Surface infiltration capacity, unit: mm / h; The time interval is calculated in hours (h). The variation of surface infiltration capacity over time is described using the Horton model, and its expression is as follows: In the formula, Surface infiltration capacity, unit: mm / h; , These are the initial infiltration rate and the steady infiltration rate, respectively, in mm / h; t Rainfall duration, in hours (h). k It is the infiltration attenuation index, which is closely related to soil conditions; 2) For impermeable depressions, runoff can form once the rainfall reaches the maximum surface water storage capacity. The runoff calculation formula is as follows: In the formula, R 2 represents the flow rate of the impermeable zone with water storage capacity, in mm; P Rainfall amount, in mm; D The storage capacity of the depression is expressed in mm. 3) For impermeable areas without depressions, rainfall is converted into runoff after surface evaporation. Runoff occurs when rainfall exceeds evaporation. The runoff calculation formula is: In the formula, R 3 represents the flow rate of the permeable storage zone without depressions, in mm; P Rainfall amount, in mm; E Evaporation rate, in mm.

3. The method for optimizing the layout of stormwater storage tanks based on stormwater pipe network model and decision tree algorithm according to claim 1, characterized in that: The underground pipeline manifold module described in step 3 is constructed as follows: The flow confluence process in the pipeline is solved using the dynamic wave method, which combines the determination of water level at the nodes and flow rate in the pipeline. The governing equations are as follows: In the formula, Q This is the instantaneous flow rate, expressed in m³ / s. A The cross-sectional area of ​​the water passage is expressed in square meters (m²). x This refers to the pipe length, in meters (m). t Time, in seconds; H Water depth, in meters (m). g This is the acceleration due to gravity, measured in m / s². 2 ; This refers to the energy gradient caused by frictional losses. n The overall roughness of the pipeline; R The hydraulic radius is expressed in meters (m). v The cross-sectional average velocity is expressed in m / s. In the governing equations of dynamic waves, there is a pressure term passing through the unpressurized pipe. Characterization: Inflow and outflow are represented by the sign of the flow rate Q; energy loss considers losses caused by friction, through... The equation is characterized and solved using the finite difference method, and its finite difference form is expressed as follows: In the formula, for Flow rate at any moment Let be the flow rate at time t; The cross-sectional area of ​​the water passage at the upper and lower nodes of the pipe section. The water depth at the upper and lower nodes of the pipe section; They are respectively The average cross-sectional area and flow velocity of the water flow over a given period; for The change in the cross-sectional area of ​​the water passage during the time period; L is the length of the pipe segment; n The overall roughness of the pipeline; R The hydraulic radius; g This is the acceleration due to gravity.

4. The method for optimizing the layout of stormwater storage tanks based on stormwater pipe network model and decision tree algorithm according to claim 1, characterized in that: The construction process of the pollutant growth and transport module described in step 3 is as follows: The processes of pollutant growth and transport are characterized by street pollutant growth and pollutant scouring. Street pollutant growth is represented by a saturation function, whereby pollutant growth begins at a linear rate, and the cumulative growth rate decreases over time until a saturation value is reached, expressed as: In the formula, B represents the cumulative amount of pollutants, expressed in kg / m³. 2 ; C 1 represents the maximum cumulative amount of pollutants per unit area, expressed in kg / m². C 2 is the half-saturation constant, which is the number of days when half of the maximum cumulative amount of pollutants per unit area is reached; t For the number of days; The pollutant flushing is set as exponential flushing, and its flushing load capacity W is: In the formula, W 1 represents the scouring coefficient; W 2 represents the scouring index; q B is the runoff rate per unit area, in mm / h; B is the cumulative amount of pollutants, in kg / m². 2 .

5. The method for optimizing the layout of stormwater storage tanks based on stormwater pipe network model and decision tree algorithm according to claim 1, characterized in that: Step 4 includes the following sub-steps: Step 4.1: Determine the reasonable range of model parameters based on literature review; Step 4.2: Automatically construct parameter combinations using a grid search algorithm, using the rainfall data from Step 2 as input conditions, and the flow rate and pollutant concentration changes monitored at the monitoring points as calibration targets for model training; Step 4.3: Output the optimal combination of parameters for the model; Step 4.4: Select ≥2 measured rainfall events to verify the flow process and pollutant concentration change process at the monitoring nodes. Use the Nash coefficient as the evaluation index. If the Nash coefficient is greater than the threshold, output the model; otherwise, proceed to step 4.1, adjust the parameter range, and retrain the model.

6. The method for optimizing the layout of stormwater storage tanks based on stormwater pipe network model and decision tree algorithm according to claim 1, characterized in that: In step 8, the pollution reduction capacity evaluation indicators include the reduction of overflow pollution load, the reduction of pollution load entering the river, the reduction of influent to the sewage treatment plant, and the reduction of influent pollution load to the sewage treatment plant. The overflow pollution load reduction amount represents the difference between the total overflow pollution load before the operation of the stormwater and sewage storage tank and the total overflow pollution load after the operation of the stormwater and sewage storage tank. The total overflow pollution load is: in, Let n be the total pollution load of the overflow outlet, n be the number of overflow outlet inflow data points during the calculation period, and m be the total number of simulation time intervals. Let be the pollutant concentration at the i-th confluence during the j-th time interval. Let i be the inflow rate of the i-th overflow port during the j-th time interval. Let be the time interval for water to flow from the j-th overflow outlet; The reduction in pollution load entering the river is the difference between the total pollution load entering the river before the operation of the stormwater and sewage storage tank and the total pollution load entering the river after the operation of the stormwater and sewage storage tank. The reduction in influent volume of the wastewater treatment plant represents the difference in the total influent volume of the wastewater treatment plant before and after the operation of the stormwater and sewage storage tank. The reduction in influent pollution load at the wastewater treatment plant represents the difference in influent pollution load before and after the operation of the stormwater and sewage storage tank, expressed as: in, This indicates the reduction in the influent pollution load of the wastewater treatment plant. This indicates the influent volume of the wastewater treatment plant before the regulating reservoir is put into operation. This indicates the concentration of the influent from the wastewater treatment plant before the regulating reservoir is put into operation. This indicates the influent volume of the wastewater treatment plant after the regulating reservoir is put into operation. This indicates the concentration of the influent to the wastewater treatment plant after the regulating reservoir is in operation.

7. The method for optimizing the layout of stormwater storage tanks based on stormwater pipe network model and decision tree algorithm according to claim 1, characterized in that: In step 8, the drainage capacity evaluation indicators include the reduction in the number of nodes overflowing, the reduction in the time of node overflowing, the reduction in the length of the full pipe, the reduction in the duration of the full pipe, and the reduction in the runoff into the river. The reduction in the number of overflows at each node is the difference between the sum of the overflow amounts at each node before the operation of the stormwater and sewage storage tank and the sum of the overflow amounts at each node after the operation of the stormwater and sewage storage tank. The reduction in overflow time at each node is calculated as the difference between the sum of the overflow times of each node before the operation of the stormwater and sewage storage tank and the sum of the overflow times of each node after the operation of the stormwater and sewage storage tank. The reduction in full pipe length is the difference between the total length of the full pipe section before the stormwater and sewage storage tank is put into operation and the total length of the full pipe section after the stormwater and sewage storage tank is put into operation. The reduction in the full pipe duration is the difference between the sum of the full pipe operating times of each pipe section before the operation of the stormwater and sewage storage tank and the sum of the full pipe operating times of each pipe section after the operation of the stormwater and sewage storage tank. The reduction in river runoff refers to the difference in river runoff before and after the operation of the regulating reservoir.

8. The method for optimizing the layout of stormwater storage tanks based on stormwater pipe network model and decision tree algorithm according to claim 1, characterized in that: In step 8, the interception capacity evaluation indicators include overflow reduction and interception volume; The overflow reduction amount is: In the formula, Let n be the total overflow rate of the overflow outlet, n be the number of overflow outlet inflow data points during the calculation period, and m be the total number of simulation time intervals. Let i be the flow rate of the i-th overflow outlet at time j before the storage tank starts operating. Let i be the flow rate of the i-th overflow outlet at time j after the storage tank has been put into operation. For the j-th time interval; The intercepted water volume is the total inflow into the regulating reservoir: in, denoted as , where n is the inflow rate of the rainwater storage tank, n is the number of overflow inflow data points during the calculation period, and m is the total number of simulation time intervals. Let be the inflow rate of the i-th regulating reservoir at time j. Let j be the j-th time interval.

9. The method for optimizing the layout of stormwater storage tanks based on stormwater pipe network model and decision tree algorithm according to claim 1, characterized in that: In step 8, the comprehensive benefit evaluation index includes the pollutant reduction ratio, which is characterized by the ratio of the total amount of pollutants intercepted by the storage tank to the total volume of the storage tank, representing the relationship between the construction cost of the storage tank and the benefit of the storage tank in reducing initial rainwater pollution.

10. The method for optimizing the layout of stormwater storage tanks based on a stormwater pipe network model and decision tree algorithm according to any one of claims 1-9, characterized in that, Step 10 includes the following sub-steps: Step 10.1: Analyze and integrate the evaluation index data obtained in Step 9 with their corresponding storage tank layout schemes. Using the evaluation index data as input conditions and the storage tank layout schemes as the target results, construct the model training dataset D = {( x 1, y 1),( x 2, y 2),…,( x n , y n )},in y n Let y represent the nth combination of water storage tank layouts. n = ( v 1, v 2,…, v n ), where v n Let n be the storage volume of the nth storage tank; x n For the nth combination of regulating reservoirs, step 9 simulates and calculates the corresponding benefit index, which is expressed as follows: , where para n The evaluation indicators proposed in step 8 are as follows: overflow pollution load reduction, river pollution load reduction, sewage treatment plant influent volume, pollution load reduction, number of nodes overflowing, node overflow time reduction, full pipe length reduction, full pipe duration reduction, river runoff reduction, overflow reduction, intercepted water volume, and pollutant reduction ratio. Step 10.2: A deep relationship between input conditions and output results will be established using the decision tree algorithm. The relationship between input conditions and output results will be represented as follows: In the formula, f ( x Let M be the predictor variable, and M be the region into which the input data feature space is divided by spatial partitioning. R 1. R 2…, R M , c m for R m The output value in space, where I is the identity matrix. x For input variables; Step 10.3: Find the optimal split point and optimal split variable, using the squared error. To minimize the value, a heuristic algorithm is used to find the optimal splitting point by minimizing the splitting error Value, which is represented as: In the formula, j represents the total number of variables, s represents the split point, and x represents the x-axis. i For the first i A combination of indicators, y i for x i The corresponding combination of water storage tanks under the indicators. R 1. R 2 represents the two variable spaces resulting from the partitioning. , , Indicates the first in the x-index combination j One variable; c 1. c 2 represents the output value corresponding to the two variable spaces. , By traversing all input variables, the splitting point s of each variable j is determined. When the value reaches the minimum value, the optimal variable j and the optimal splitting point s are determined. Step 10.4: Calculate the output value of the divided region under the conditions of optimal segmentation feature j and optimal segmentation point s. , is represented as: In the formula, x i ∈ R m m=1,2 N m For space R m The number of combined storage tank schemes in the project. y i for x i The corresponding combination of regulating reservoirs under the indicators; Step 10.5: Repeat steps 10.3-10.4 until the maximum number of layers in the decision tree reaches the set threshold or all flood control pond schemes are individually partitioned. After pruning, the variable space is divided into M regions. R 1. R 2…, R M Generate a decision tree model; Step 10.6: Test the decision tree performance using a test set, i.e., input feature indicators, and use the decision tree model generated in step 10.5 to invert the optimal layout scheme of the water storage tanks; further, use the optimized layout scheme of the water storage tanks obtained from the inversion as a condition to drive the SWMM model and calculate the indicator evaluation system proposed in step 8; finally, compare it with the input feature indicators of the initial decision tree model to verify the reliability of the constructed decision tree model; when the relative error of each indicator is controlled within the threshold, the constructed decision tree model is considered to meet the requirements, and the decision tree model is output and stored; otherwise, return to step 10.1, readjust the input indicators, and train the model. Step 10.7: Automatically generate an optimized layout plan for storage tanks by inputting the benefit evaluation indicators of the storage tanks.

Images