Optimization Study of Drainage Network Systems Based on the SWMM for the Wujin District, Changzhou City, Jiangsu Province, China

Abstract

This study addresses the persistent issue of urban waterlogging in Wujin District, Changzhou City, Jiangsu Province, using a comprehensive approach integrating an optimized drainage network and low-impact development (LID) measures. Utilizing the Storm Water Management Model (SWMM), calibrated with extensive hydrological and hydraulic data, the model was refined through genetic algorithm-based optimization to enhance drainage efficiency. Key results indicate a substantial reduction in the average duration of waterlogging from 7.43 h to 3.12 h and a decrease in average floodwater depth from 21.27 cm to 8.65 cm. Improvements in the drainage network layout, such as the construction of new stormwater mains, branch drains, and rainwater storage facilities, combined with LID interventions like permeable pavements and rain gardens, have led to a 56.82% increase in drainage efficiency and a 63.88% reduction in system failure rates. The implementation effectively minimized peak flood flow by 25.38%, reduced runoff, and improved groundwater recharge and rainwater utilization. The proposed solutions offer a replicable, sustainable framework for mitigating flooding in urban environments, enhancing ecological resilience, and ensuring the safety and quality of urban life in densely populated areas.

1. Introduction

1.1. Background

Flood disasters are a persistent problem globally, affecting millions of people and causing extensive economic damage every year. One of the primary reasons for these flood disasters is the inadequacy of urban drainage systems, particularly imperfect or inefficient drainage networks. As urbanization continues to intensify, cities are becoming increasingly vulnerable to heavy rainfall events, which often overwhelm existing drainage infrastructure. The drainage network, as a core component of urban stormwater management, plays a vital role in mitigating the impacts of extreme weather conditions. However, many of these networks suffer from outdated designs, aging infrastructure, and insufficient capacity, leading to frequent urban flooding [1]. Optimizing the layout and operational parameters of urban drainage networks has thus become a critical measure for flood prevention. Effective parameter optimization can significantly enhance the performance of existing networks, allowing cities to control excess surface runoff and reduce flooding incidents effectively [2,3].

Common optimization techniques include dynamic programming, simulated annealing, and genetic algorithms. Dynamic programming is well suited for problems with overlapping subproblems and optimal substructure properties, but it is computationally intensive [4]. Simulated annealing avoids local optima by mimicking the physical annealing process; however, its convergence is relatively slow, requiring extended computation time [5]. In contrast, genetic algorithms simulate natural selection and genetic mechanisms (e.g., selection, crossover, and mutation) to explore the solution space effectively, avoiding local optima and exhibiting strong global search capabilities. Furthermore, genetic algorithms have been widely applied to multi-objective optimization problems in urban drainage system optimization, demonstrating excellent performance. Therefore, the adoption of genetic algorithms is more suitable for addressing complex urban drainage network optimization problems, improving both optimization efficiency and the reliability of the results [6].

1.2. Current Research Status

1.2.1. Research Status of Urban Stormwater Simulation Model

Accurately simulating urban flooding has become an indispensable practice for mitigating its impacts, especially as the frequency of such events increases. Countries in Europe and the United States began researching the simulation of stormwater runoff in modern cities relatively early, and their research content is more comprehensive. The results of numerical model simulations are intuitive and easy to analyze quantitatively, and they play an important role in urban flooding analysis.

The SWMM (Storm Water Management Model) proposed by the EPA (US Environmental Protection Agency) in the 1970s simulates the flow state of pipe networks by solving the Saint-Venant equation system and can analyze the current state of drainage systems, plan the underlying surface, simulate water quality, design pipe networks, and has additional LID functions. It has been widely used in municipal planning and scientific research [7,8,9,10]. During the same period, Terstrie and others developed the ILLUDAS (Illinois Urban Drainage Model), which has the disadvantage of being able to perform only hydraulic simulations [11,12]. In 1983, the Danish Hydraulic Institute developed MIKE URBAN, which integrates various modules such as rainfall data processing and surface runoff infiltration modules, enabling more accurate rainfall and flood simulations [13,14]. MIKE FLOOD can dynamically couple MIKE 11 and MIKE URBAN to obtain a more accurate simulation of the urban drainage network by setting boundary conditions and parameter checks [15]. In addition, InfoWorks was developed by Innovyze, and STORM models were developed by the US Army Corps of Engineers (HEC) and are also widely used [16,17,18,19].

1.2.2. Research Status of SWMM

As urban flooding becomes more and more serious, accurate simulation of urban flooding has become indispensable for reducing it. After 30 years of development, the SWMM has been widely used by relevant departments around the world for urban stormwater pollution and urban drainage management [20]. The SWMM (Storm Water Management Model) was developed by the US Environmental Protection Agency (EPA) in the 1970s. It stimulates the flow of water in pipes by solving the Saint-Venant equation system. It can analyze the current state of drainage systems, plan the underlying surface, simulate water quality, design drainage networks, and have additional LID functions. It has been widely used in fields such as municipal planning and scientific research [7,8,9,10].

Krebs et al.’s study applies a high-resolution SWMM with genetic parameter optimization, enhancing stormwater management accuracy in complex urban environments [21]. Cho and Lee optimize SWMM parameters for improved runoff calibration, offering enhanced precision in environmental impact assessments of urban water systems [22]. Zhong et al. integrate Morris and GLUE methods for SWMM parameter optimization, which improves model robustness in urban runoff and water quality prediction [23]. Assaf et al. proved that new optimization strategies for the SWMM are introduced for urban stormwater quality applications, contributing to more effective urban water quality management [24]. Xue et al.’s study uses an optimization algorithm to calibrate SWMM parameters, demonstrating improved accuracy in urban stormwater simulation [25]. Gao et al.’s paper proposes an automatic calibration method for SWMM using differential evolution and Bayesian optimization, which enhances parameter calibration efficiency [26]. Behrouz et al. present a tool for automatic SWMM calibration, which simplifies and improves the model calibration process in stormwater applications [27]. Dent et al. focus on automated calibration techniques applied to SWMM RUNOFF, facilitating more efficient urban runoff modelling [28]. Barco et al.’s research demonstrates the automatic calibration of the SWMM for a large urban catchment, supporting improved stormwater management in complex urban areas [29]. Perin et al.’s study applies automated SWMM calibration to PESTtor, a small suburban catchment, enhancing localized stormwater modelling accuracy [30]. Wan and James’ study employs genetic algorithms for SWMM calibration, introducing a robust method for optimizing model performance in urban drainage applications [31]. Xu et al. focus on sensitivity analysis and calibration of SWMM for urban green land types, offering insights into runoff simulation for different urban greenspaces [32]. Cho and Seo’s study mainly focuses on genetic algorithms applied to optimize SWMM for runoff quantity and quality in a eutrophic lake watershed, highlighting improved modelling for water quality management [33]. Rabori et al. prove that sensitivity analysis of SWMM parameters enhances understanding of urban runoff in semi-arid areas, adapting the model for diverse climatic conditions [34]. Kourtis et al. prove that SWMM calibration and validation in two urban catchments in Athens provide practical insights into model application in Mediterranean climates [35]. Zhu et al. focus on spatial layout optimization for green infrastructure using SWMM and life-cycle multi-objective algorithms to enhance sustainable urban water management [36]. Du et al. present a simultaneous optimization approach for SWMM parameters, supporting enhanced stormwater management through multi-objective functions [37]. Tanim et al. introduce Bayes_Opt-SWMM, a Gaussian process-based Bayesian optimization tool developed for real-time flood modelling, which contributes to dynamic flood response strategies [38]. Taghizadeh et al.’s research integrates the SWMM with particle swarm optimization for urban runoff water quality control, emphasizing the role of green infrastructures [39]. Zakizadeh et al.’s study combines the SWMM and field data for efficient urban runoff and water quality modelling in Tehran, highlighting adaptability in large urban watersheds [40]. Macro et al. focus on OSTRICH-SWMM, a multi-objective optimization tool introduced for planning green infrastructure, advancing SWMM’s application in sustainable urban design [41]. Sun et al.’s paper examines the impact of SWMM catchment discretization and informs best practices for stormwater modelling in urban areas [42]. Jain et al. apply geo-informatics to estimate SWMM sub-catchment parameters, improving parameterization accuracy for stormwater management [43]. Ekmekcioğlu et al.’s study concerns an auto-calibrated SWMM, which explores low-impact development strategies and promotes sustainable stormwater solutions in highly urbanized areas [44]. Li et al. prove that the spatial optimization approach for LID practices based on the SWMM-FTC improves cost-effectiveness in urban stormwater infrastructure [45]. Wang et al.’s study employs a multi-objective optimization model for automatic SWMM calibration, advancing model efficiency in complex urban settings [46]. Del Giudice and Padulano combine the SWMM with genetic algorithms, enhancing rainfall-runoff model calibration and sensitivity analysis [47]. Ma et al. demonstrate that real-time correction and dynamic urban flood simulation can be performed in the SWMM, which supports responsive flood management strategies [48]. Zaghloul and Kiefa demonstrate that neural networks can be applied to the SWMM for sensitivity-calibration analysis, innovating parameter optimization in stormwater modelling [49]. While SWMM-based research predominantly addresses technical challenges in urban flood management, studies in Jordan, one of the most water-scarce countries globally, highlight the role of political patronage and shadow actors in shaping water policy, emphasizing governance-related inefficiencies rather than purely technical solutions. In contrast, the present study focuses on the optimization and application of the SWMM to address urban flooding issues in China, where the challenges are primarily technical and hydrological in nature [50].

Furthermore, efficient algorithmic approaches for parameter optimization have the potential to provide more accurate simulations and improve decision-making in urban stormwater management. This study focuses on optimizing the drainage network parameters with the aim of improving the flood resilience of the local drainage network [4]. Additionally, it compares various optimization algorithms to determine the most effective approach for enhancing network performance. This is particularly important given the increasing frequency of extreme rainfall events and the need for resilient urban infrastructure [51]. By addressing deficiencies in the existing drainage network, this study seeks to contribute to reducing the flood vulnerability of urban areas in China and beyond.

2. Materials and Methods

2.1. Study Area Profile

Wujin District, located in the southern and eastern part of Changzhou City in Jiangsu Province, has a subtropical monsoon climate with four distinct seasons, plenty of rainfall, and a long frost-free period. The district has 1048 rivers and streams, with a total length of 2000.7 km. On average, there are 1.82 km of rivers and streams per square kilometer of land, making it a typical region of the Jiangnan water network. Geographically, the district spans longitudes from approximately 119 E to 121 E and latitudes from 31 N to 32 N. The scope is shown in Figure 1.

Wujin District lies within the Taihu Lake basin, in the Wuxingxiyu and Huxi areas. The flood control in the basin meets the 1954 standard of protection against a once-in-50-year flood. A drainage pattern has been formed, with the Yangtze River to the north and Taihu Lake to the south. The capacity of the Yangtze River to drain the north is gradually being increased through the implementation of the Xinguo and Xinmeng Rivers. Wujin has a typical subtropical monsoon climate, with an average annual precipitation of about 1133.8 mm. There is sizeable interannual variation, with quick and concentrated periods of rainfall. Light rain accounts for 72.7% of the rainfall, while heavy rain accounts for 2.7%. Wujin District has a well-developed water system with numerous rivers and lakes, resulting in a complex lithology in the unsaturated zone. However, the stratum structure in the unsaturated zone is relatively simple, with primarily single-layered soil types. Most of the current drainage networks in Wujin District have a return period of less than 1 year, and the drainage networks are severely overloaded, which can easily lead to internal flooding and thus adversely affect the infrastructure and residents’ lives in the area [52]. Official news reports of waterlogging incidents are shown in Figure 2.

Changzhou is one of the country’s 54 crucial flood control cities, and Wujin District’s position in the city’s overall flood control system has become increasingly important. At the national macro policy level, the Ministry of Water Resources issued the ’Guidelines for the Preparation of Joint Dispatching Plans for Basin Flood Control Projects’ on 12 July 2023. These guidelines emphasize the need to update or revise urban flood control plans in response to recent changes in urban master plans. This includes the expansion of urban flood protection zones, an increase in flood protection infrastructure, and the implementation of higher flood control standards. There is an urgent need to coordinate urban flood control and drainage with urban development, land use, environmental protection, and other related factors.

2.2. SWMM

The SWMM is a distributed hydrological model in which complex terrain redistributes rainfall during a rainfall event. Rainfall becoming runoff is net rainfall after initial losses such as interception by vegetation and evaporation. The net rainfall then flows into the drainage network system. If the drainage network system is overloaded, the rainwater that overflows from the system gradually fills the low-lying reservoirs and becomes surface runoff.

The SWMM uses the Rain Gauge module to calculate and simulate rainfall. Rainfall events in the study area can be imported or simulated using the rainfall intensity formula:

$$i={\displaystyle \frac{134.5106(1+0.4784\log T_M)}{{(t+32.0692)}^{1.1947}}}$$

(1)

where

i—Rainfall Intensity

The initial loss of rainwater due to interception by plants and evaporation is calculated using the Horton infiltration model, as follows:

$$f_s=f_t+(f_0-f_t)e^{-kt}$$

(2)

where

f_s—Ground infiltration rate

f_t—Infiltration stability value

f₀—Initial infiltration rate (mm/h)

k—Infiltration decay rate (h⁻¹)

t—Infiltration duration (h)

The process of net rainfall flowing through pipes and nodes and finally flowing to the drain represents the hydraulic simulation process in the SWMM. The drainage network model presented in this paper is a tree-like drainage network, and the water flow within the drainage network is in a gravity flow state. Consequently, the dynamic wave model is employed in the three calculation methods of constant flow, moving wave, and dynamic wave. The simulation formula is as follows:

$$\left\{\begin{array}{c}{\displaystyle \frac{\partial Q}{\partial x}}+{\displaystyle \frac{\partial A}{\partial t}}=q\\ {\displaystyle \frac{1}{g}}\ast {\displaystyle \frac{\partial v}{\partial t}}+{\displaystyle \frac{v}{g}}\ast {\displaystyle \frac{\partial v}{\partial x}}+{\displaystyle \frac{\partial H}{\partial x}}=S_0-S_f\end{array}\right.$$

(3)

where

Q—Pipeline flow (m³/s)

A—Water crossing section area (m²)

x—Pipe length (m)

t—Time (s)

q—Incoming flow per unit length (m³/s)

g—Design velocity of the stormwater pipe (m)

v—Roughness coefficient of the stormwater pipe (m)

H—Diameter of the stormwater pipe (m)

S₀—Slope of the stormwater pipe

S_f—Friction term

The Manning formula calculates surface runoff as follows:

$$v={\displaystyle \frac{D^{\frac{2}{3}}\times I^{\frac{1}{2}}}{4^{\frac{2}{3}}\times N}}$$

(4)

where

v—Design flow rate of storm pipe (m/s)

N—Roughness coefficient of rainwater drainage network

D—Radius of storm water network (m)

I—Hydraulic grade

The maximum volume of the stormwater pipe Q_p is calculated as follows:

$$Q_p={\displaystyle \frac{\pi D^2}{4}}\times v$$

(5)

2.3. Data and Preprocessing

Uncrewed aerial vehicle (UAV) remote sensing and field surveys mainly obtain topographic data, and UAVs are used for high-resolution topographic mapping. The UAV (DJI Phantom 4 RTK, Shenzhen, China) provided mapping accuracy of ±2 cm horizontally and ±5 cm vertically, covering a total area of 500 km² divided into 50 zones. Topographic data from UAV remote sensing is used for coordinate correction and stitching. GPS data is used for coordinate correction. A DEM with 0.5 m resolution was created for detailed analysis, supplemented by publicly available SRTM data (30 m resolution) to ensure full coverage. Geostatistical methods smoothed and filtered the data, removing noise and outliers.

Land use data are mainly derived from remote sensing satellite imagery. Remote sensing image data from 2018 to 2023 were selected, and different types of land use (grassland, cropland, construction land) were identified through classification and analysis using image processing software (ArcGIS Pro 3.1). ENVI 5.6 software performed geometric corrections to remove distortions and misalignments in the images. Image classification was also done using supervised and unsupervised classification methods to identify different land feature types in the images. Machine learning algorithms such as Support Vector Machines (SVMs) and Random Forests (RFs) were introduced to improve the classification accuracy, and field survey data was used to verify and adjust the results, ultimately producing an accurate land use map.

Meteorological data (rainfall, temperature, humidity, wind speed) were mainly obtained from automatic weather stations. A total of 10 weather stations were deployed within the study area, each recording data every 15 min. They record meteorological data regularly daily and transmit them via wireless networks to the research center for aggregation and analysis. The data is converted and processed uniformly to ensure data integrity and consistency. Historical meteorological records are extracted from the China Meteorological Administration database, including monthly average temperature, rainfall, and evaporation data over the past 30 years, for time series analysis to fill in missing values and eliminate outliers. Specifically, missing values accounted for less than 2% of the dataset and were inputted using linear interpolation techniques, ensuring minimal distortion in trend analysis. Outlier detection was performed using the IQR method, and less than 1% of data points were removed or adjusted.

Hydrological and water quality data are mainly obtained through long-term observations at monitoring stations distributed at significant rivers, lakes, and groundwater sampling sites. Water samples are collected and their physico-chemical parameters (flow, water level, pH, dissolved oxygen, suspended solids, nitrogen and phosphorus concentrations) are measured regularly. Real-time data are also collected using automated water quality monitoring equipment to ensure the timeliness and accuracy of the data. Meanwhile, the hydrological and water quality data collected at the monitoring stations must first be processed and verified. Water samples are analyzed in the laboratory to measure their physico-chemical parameters, and the results are entered into the database. Data cleansing techniques remove outliers and noise from the real-time data the automated monitoring equipment collects. The final complete hydrological and water quality dataset supports the parameterization and calibration of the SWMM. To ensure consistency, laboratory measurements were cross-checked against automated readings, with deviations of less than 5% observed for key parameters such as pH and dissolved oxygen. The final complete hydrological and water quality dataset supports the parameterization and calibration of the SWMM.

The data used for the study were summarized in

Table 1. Data summary.

Data Type	Data Source	Description	Processing Method	Key Parameters	Precision/Resolution
Topographic Data	UAV Remote Sensing (DJI Phantom 4 RTK)	Covered 500 km2, divided into 50 zones; high-resolution DEM created	Geostatistical smoothing and filtering	DEM resolution	0.5 m
Land Use Data	Remote Sensing Satellite Imagery	2018–2023 imagery data; grassland, cropland, construction land identified	SVM and RF classification, ENVI geometric correction, validation	Classification accuracy	>90%
Meteorological Data	Automatic Weather Stations, Historical Records	10 stations, recorded every 15 min; monthly average temperature, rainfall, and evaporation	Missing data interpolation, IQR outlier detection	Missing value ratio	<2%
Hydrological and Water Quality Data	Monitoring Stations, Automated Equipment	Flow, water level, pH, dissolved oxygen, nitrogen and phosphorus concentrations	Data cleaning and cross-checking with lab measurements	Parameter deviation	<5%

, including topographic, land use, meteorological, and hydrological data sources.

2.4. Model Construction

The hydrological and hydraulic module of the SWMM employs a generalized approach to sub-catchment delineation, categorizing areas as either permeable or impervious [53,54]. The model parameters include N-Imperv (Manning coefficient of impervious area), N-Perv (Manning coefficient of permeable area), Dstore-Imperv (deposited water depth of impervious area), Dstore-Perv (deposited water depth of permeable area), and PctZero (percentage of impervious area without deposited water area). The sub-catchment infiltration model employs the Horton infiltration model, and the model parameters include MaxRate (maximum infiltration rate), MinRate (minimum infiltration rate), Decay Constant (infiltration attenuation coefficient), and DryTime (number of dry days in the previous period). The surface runoff is calculated using the dynamic wave method, and the requisite model parameter is roughness (pipe Manning coefficient).

This study employed the ArcGis Thiessen polygon tools to partition the sub-catchment regions. Subsequently, the delineation of the sub-catchment regions pertaining to the target pipe sections was conducted based on criteria such as elevation, slope, pipe section relationships, and functional areas. A visual interpretation method divided the study area into five categories: rooftops, roads, green spaces, rivers, and undeveloped land. These categories accounted for 29.6%, 8.6%, 43.39%, 2.88%, and 15.53% of the total area, respectively. The infiltration rate is subject to influence by the type of land used. Under land use in the study area, green spaces and rivers are designated permeable surfaces, whereas roofs, roads, and unused land are classified as impervious surfaces. The proportion of land area occupied by green spaces and rivers in each sub-catchment is calculated as the ratio of permeable area in the corresponding sub-catchment. The percentage of land area occupied by roofs, roads, and unused land is determined by the ratio of impervious area in the corresponding sub-catchment. Subsequently, the characteristic parameters of the drainage network were obtained by reference to Changzhou City Flood Control Planning (2017–2035) and a field survey of Wujin District. The characteristic parameters of the nodes were primarily derived from the dataset of the flood control plan. In contrast, individual exploration and measurement were conducted for certain nodes in specific locations. The resulting SWMM is shown in Figure 3.

2.5. Model Parameter Validation Index

The parameters are determined based on measured data. The Nash efficiency coefficient, total runoff balance error coefficient, and relative error of peak flow are selected as the objective function indicators for the parameter determination of the hydrological and hydraulic modules [55]. These indicators reflect the relationship between the simulation results and the law of hydrological and hydraulic change process in the study area. The Nash efficiency coefficient ENS, total runoff balance error coefficient EQW, and relative error of peak flow EPR are calculated using the following formulas:

$$E_{NS}=1-{\displaystyle \frac{\sum _{t=1}^n{{(q}_{t,obs}-q_{t,sim})}^2}{\sum _{t=1}^n{{(q}_{t,obs}-{\overline{q}}_{t,sim})}^2}}$$

(6)

$$E_{QW}={\displaystyle \frac{\sum _{t=1}^n{(q}_{t,obs}-q_{t,sim})}{\sum _{t=1}^n{q}_{t,obs}}}$$

(7)

$$E_{PR}={\displaystyle \frac{\left|q_{p,obs}-q_{p,sim}\right|}{q_{p,obs}}}\times 100\%$$

(8)

where

q_t,obs—Measured flow at time t (m³/s)

q_t,sim—Simulated flow at time t (m³/s)

q_obs—Average measured discharge (m³/s)

q_p,obs—Measured peak flow (m³/s)

q_p,sim—Simulated peak flow (m³/s)

N—Number of flow measurements

2.6. Algorithm Optimization

The optimal design of rainwater drainage networks is highly systematic, non-linear, multimodal, and multidimensional. The difficulty of this problem is how to obtain a reasonable combination with a high success rate within an acceptable time by changing the diameter of the drainage network and the height of the nodes at both ends of the diameter to minimize the cost of the drainage network. Coupling between variables, complex constraints, and a large number of possible combinations of pipe diameters also make optimization difficult. In particular, when the drainage network is extensive, the combinations of pipe diameters increase exponentially, and the computational efficiency of conventional enumeration and traditional optimization methods makes it difficult to support the solution of such problems. Therefore, a better solution to the problem can only be found in intelligent optimization algorithms.

Genetic algorithms (GAs) are intelligent optimization algorithms that use selection, crossover, and mutation as their genetic steps. They are particularly effective for solving complex, non-linear problems such as pipe network optimization due to their global search capabilities and adaptability.

However, alternative methodologies such as particle swarm optimization (PSO), simulated annealing (SA), and differential evolution (DE) are also commonly used in similar applications. PSO is advantageous in converging rapidly for simple optimization problems but may struggle with multimodal or complex search spaces. SA, while effective for escaping local optima, can be computationally intensive and slower to converge. DE provides robust global optimization capabilities and performs well in high-dimensional problems but may require careful tuning of control parameters for effective application.

In this study, a GA was selected due to its superior ability to handle the discrete variables and non-linear constraints associated with pipe diameter and slope optimization. Furthermore, GAs have been widely validated in previous studies for stormwater pipe network optimization and are known for their scalability and ease of integration with hydrological models like the SWMM [31,47,56,57,58,59,60].

Table 2. Comparison of optimization algorithms.

Algorithm	Strengths	Limitations
Genetic Algorithm (GA)	Strong global search; robust for discrete variables; handles non-linearity	May require higher computational effort in large search spaces
Particle Swarm Optimization (PSO)	Rapid convergence for simple problems	Prone to premature convergence in multimodal problems
Simulated Annealing (SA)	Escapes local optima; simple implementation	Computationally slow for large datasets
Differential Evolution (DE)	Effective in high-dimensional problems	Sensitive to parameter tuning

below summarizes the key characteristics of these methodologies.

2.7. Design of Stormwater Network Optimization Schemes Based on GA

When solving optimization problems using genetic algorithms, it is necessary to determine the model’s objective function, identify the decision variables and constraints, and encode and decode the fundamental data representing the drainage network system’s current capacity and individual DNA links, and then proceed to search for the optimum.

2.7.1. Objective Function

In this optimization process, the pipe diameter and the pipe section’s gradient are employed as decision variables, with the objective function being the cost of optimizing the stormwater drainage network system.

$$f_c=min\left\{\sum _{i=1}^n(1+d_1)C_i{L}_i\right\}+p(D)$$

(9)

where

f_c—Cost function for optimization of the rainwater drainage network

d₁—Depreciation rate of drainage network

C_i—Section costs for different pipe diameters (m/yuan)

L_i—Length of pipe section with different pipe diameters

p(D)—Penalty function: The calculation method entails selecting each DNA in a manner that ensures the combination of pipe diameter and slope aligns with the prescribed hydraulic calculation formula and rainfall-runoff formula, as specified in the SWMM and applicable to different return periods.

2.7.2. Constraint Condition

Pipe diameter: After optimization, the diameter of the pipe section cannot be smaller than before optimization, i.e., Dk ≥ Dj. The range of pipe diameter optimization is 0.3–2.7 m, with discrete increases, i.e., 0.3 m, 0.35 m, 0.4 m, 0.45 m.

Slope of pipeline: Absolute value of the height difference between the nodes at both ends of the pipe is calculated as follows.

$${\displaystyle \frac{H_k-H_j}{l_p}}\ge S_{min}$$

(10)

where

l_p—Pipe length

S_min—Pipe slope

Pipe filling: The rainwater pipe should be calculated as full flow, so the value is 1.

Pipe velocity: According to the relevant regulations, the maximum design flow rate for metal drainpipes is 10.0 m/s, and for non-metallic pipes, 5.0 m/s.

Flow continuity of pipe segments and nodes:

$$\sum \left(\pm q_p\right)+Q_n=0$$

(11)

where

q_p—Pipe flow

Q_j—Nodes flow

Three ranges of node overflow heights are set according to different optimization schemes: 0 cm, 0–7.5 cm, and 7.5–15 cm. Python controls the function of calling the SWMM database for simulation control through a text file. The two are connected to ensure the calculation results meet the requirements.

2.7.3. Justification for Economic Objective

The decision to use cost minimization as the primary objective reflects the practical considerations of urban drainage system upgrades, where economic feasibility often drives decision-making. Including hydraulic and hydrological constraints ensures that cost efficiency does not compromise system performance. This dual approach balances cost savings with the drainage network’s functional requirements.

Previous studies, such as Krebs et al. (2013) and Gao et al. (2023), have demonstrated the effectiveness of combining cost-based optimization with performance-based constraints in achieving reliable and cost-effective urban drainage solutions. Our methodology builds on this framework by integrating low-impact development (LID) measures, further enhancing the system’s ecological and hydraulic performance [21,26].

2.8. Technical Route

This paper uses basic data, such as the elevation and pipe network of an area in Futian District, Shenzhen, to establish an SWMM urban stormwater runoff simulation model. The goal is to plan a multi-objective optimization scheme for the urban pipe network model under various storm scenarios, with the cost-effectiveness and reliability of the pipe network as the objectives. The calculated optimization scheme is returned to the SWMM for simulation and evaluation. The technical route is shown in Figure 4.

3. Results and Discussion

3.1. Model Calibration and Verification Results

The parameter range was determined based on the SWMM user manual and relevant studies in the surrounding area [61,62,63]. Rainfall data from four events, 20170610, 20170803, 20170808, and 20170925, were included to perform model hydrological and hydraulic parameter calibration. The model was verified using 10 actual rainfall processes measured from June to October 2016. A comparison of the measured and simulated results is shown in Figure 5.

As can be seen from Figure 4 and

Table 3. Results of calibration and verification of hydrologic and hydraulic parameters.

Type	Date	ENS	EQW	EPR/%
Calibration	20170610	0.912	0.14	11.12
Calibration	20170803	0.959	0.12	14.41
Calibration	20170808	0.935	0.07	8.52
Calibration	20170925	0.886	0.12	12.82
Verification	20160623	0.796	-0.27	7.68
Verification	20160627	0.743	0.20	4.82
Verification	20160701	0.795	-0.12	11.56
Verification	20160704	0.823	-0.28	5.08
Verification	20160914	0.812	0.16	8.63
Verification	20160915	0.766	0.27	12.32
Verification	20160929	0.697	-0.28	9.47
Verification	20161026	0.774	0.02	4.89

, the measured and simulated values during the calibration period are in good agreement. The Nash coefficient ENS between the simulated and measured values during the calibration period and the verification period is 0.69–0.96, and the absolute value of the total runoff error coefficient EQW is less than 0.30. The relative error of the peak flow rate EPR is less than 15%, indicating that the SWMM simulation constructed has high accuracy and the model is suitable for Wujin District, Changzhou City.

3.2. Model Optimization Results

Under the “Code for Design of Outdoor Drainage” GB50014-2006, the maximum permissible surface water depth is 15 cm. The overflow depth range is 7.5–15 cm [64]. Using the GA in Python, a combination of new pipes and LID measures was calculated, and finally, two optimized schemes were obtained. The comparison between the drainage network before and after optimization for different schemes is shown in

Table 4. Drainage capacity condition.

Index	Pre-Optimization	Optimized
Average duration of waterlogging (h)	7.43	3.12
Average water depth (cm)	21.27	8.65
Drainage efficiency (%)	58.32	91.46
Reduction of peak flood flow (%)	0	25.38
Surface runoff reduction	0	20.16
Increase of soil water content	0	15.98
Amount of increase in groundwater recharge	0	17.32
Utilization rate of rainwater	61.84	77.14
Failure rate of drainage system	5.15	1.86

and Figure 6.

These findings align with Gao et al. (2023) and Taghizadeh et al. (2021), who demonstrated the effectiveness of GA for optimizing urban drainage systems. However, our study extends these results by incorporating LID measures, achieving a dual benefit of enhanced drainage efficiency and ecological sustainability [26,39].

The reduction in peak flood flow by 25.38% is particularly noteworthy, as it demonstrates the capacity of LID measures, such as rain gardens and permeable pavements, to delay runoff and mitigate flooding risks. This result supports the findings of Zhu et al. (2023), who highlighted the importance of LID in reducing urban runoff volumes [36].

Integrating LID measures not only reduced surface runoff but also increased groundwater recharge by 17.32% and soil moisture content by 15.98%. These improvements are consistent with the findings of Li et al. (2022) and Cai et al. (2017), who reported similar benefits from LID implementations in urban areas [45,61].

The comprehensive solution presents a compelling case for addressing the drainage issues in Wujin District, Changzhou City, offering a comprehensive and practical approach to resolving these challenges. The solution can effectively reduce the frequency of internal flooding and the maximum water depth, significantly improving drainage efficiency, controlling peak flow rates, increasing groundwater recharge, and improving the utilization rate of rainwater resources. In practice, although the implementation cost of this solution is relatively high, it would be essential to consider the long-term benefits. In densely populated areas with frequent economic activities, such as the central business district and educational establishments in the south, internal flooding will significantly impact daily life and the regular operation of economic activities. Implementing a comprehensive package of measures will significantly reduce the frequency and severity of internal flooding, thereby ensuring the safety of residents and commercial activities in the area.

Additionally, the package is beneficial for the environmental sustainability of the ecological protection area. Through measures such as permeable paving and rain gardens, the area’s ecological balance can be maintained, and the negative impact of human activities on the natural environment can be reduced. From a maintenance and long-term benefits perspective, the comprehensive package can effectively extend the service life of the drainage system and reduce the additional costs associated with frequent maintenance and upgrades.

3.3. Analysis of Model Optimization Results

The optimization results reveal a significant annual decrease in internal flooding events. This suggests that implementing an upgraded drainage system and low-impact development (LID) strategies have greatly improved the area’s flood prevention capabilities. The study considered a combination of LID measures, such as permeable pavements and rain gardens, alongside traditional infrastructure upgrades, including new stormwater mains, additional drainage branches, and stormwater storage ponds. The optimal solution, derived through the genetic algorithm, incorporated a mix of these measures to achieve maximum drainage efficiency. Specifically, stormwater storage ponds, permeable pavements, rain gardens, and the construction of new stormwater mains were prioritized in the final solution.

The shorter duration of flooding indicates that the enhanced drainage system can move water away more efficiently, reducing the impact on residents and economic activities. Notably, during heavy rainfall periods, a quicker drainage response helps limit disruptions to traffic, daily life, and infrastructure.

A reduction in the average water depth further supports the idea that the area’s overall drainage capacity has improved, leading to better conditions during heavy rains and easing the strain on daily travel and activities.

Enhanced drainage efficiency means that the optimized system can now remove precipitation more swiftly and limit water accumulation, even under similar rainfall levels. This improvement is the result of expanding and renovating the drainage network and installing rainwater storage tanks to alleviate pressure on the system during peak flows.

The reduced peak flow suggests that the improved system can better manage and delay runoff during intense rainfall, which decreases peak flow rates. Implementing LID measures like rainwater storage tanks and permeable paving helps rainwater infiltrate and be stored more effectively, reducing pressure during peak flow events.

The reduction in overall rainfall runoff shows that, by using LID measures and optimizing the drainage network, a larger portion of rainwater is being absorbed and stored instead of converting directly into runoff. This not only reduces flood risks but also supports groundwater recharge and helps maintain ecological balance.

The decrease in surface runoff indicates that more rainwater is being intercepted and absorbed through LID measures like rain gardens and permeable paving, leading to a reduction in direct runoff formation. This improvement helps mitigate flooding risks and protect surface water resources. The increased soil moisture levels suggest that greater rainwater infiltration is enhancing the soil’s moisture content, thereby supporting plant growth and strengthening the ecosystem’s stability and resilience.

The increased groundwater recharge indicates that more rainwater is seeping into the ground to replenish groundwater reserves, which strengthens the region’s water resources. Groundwater recharge also plays a critical role in sustaining the balance of rivers and wetlands. Likewise, greater utilization of rainwater points to better effectiveness through LID practices and rainwater harvesting, reducing dependence on municipal water and lowering the burden on the drainage system.

A lower failure rate of the drainage system highlights enhanced reliability and stability, thanks to better pipe design and regular maintenance. This reduces the risk of waterlogging from system malfunctions, boosting the drainage network’s overall performance.

The comprehensive solution package presents significant advantages for managing drainage challenges in the Wujin district. By integrating various strategies, this package reduces the frequency of internal flooding and peak water depth, enhances drainage efficiency, controls peak flow rates, increases groundwater recharge, and makes better use of rainwater resources. Though the initial implementation cost may be high, overlooking the long-term benefits would be a mistake. In densely populated and economically active areas—like central business districts and school zones—flooding can severely disrupt everyday life and regular business activities. This integrated approach helps minimize such disruptions, ensuring the safety of residents and stability for commercial activities.

Moreover, this package also supports the sustainability of protected ecological areas. Measures like permeable paving and rain gardens contribute to maintaining ecological balance and reducing the impact of human activities on natural environments. From a maintenance and long-term perspective, the comprehensive package can extend the service life of the drainage system, ultimately reducing long-term costs by cutting down on the need for frequent repairs and upgrades.

3.3.1. Optimized Drainage Network Layout and LID Measures

(1)

New Rainwater Main

The drainage system in Wujin District is currently experiencing congestion in specific locations, particularly in the densely populated and economically active central residential zone and around the southern educational institutions. A new high-capacity stormwater trunk pipe is planned in these areas to enhance the drainage system’s efficiency. The trunk pipe will be constructed along the main roads and connected to multiple branch pipes, thereby ensuring the rapid drainage of water during periods of heavy rainfall. The diameter of the trunk pipe will be 2 m, with a length of approximately 5 km. It will cover major waterlogging points and low-lying areas. An increase in the capacity of the main pipe will result in a notable enhancement of the drainage capacity of the area, which will, in turn, lead to a reduction in the frequency and severity of flooding.

(2)

Additional Branch Drainage

Several additional drainage branch lines are intended to be incorporated into the existing drainage system, particularly in the eastern mountainous region and the western village area, where flooding is a recurrent phenomenon. The new drainage branch lines will be connected to the main pipe and extend to various low-lying areas, ensuring rainwater’s rapid collection and discharge. The branch lines are designed with a diameter of 0.8 m and will span approximately 8 km. The rationalization of the layout of the branch pipes allows for the dispersion and direction of runoff more effectively, thereby reducing the pressure on the main pipes and improving the overall efficiency and reliability of the drainage system.

(3)

Construction of Rainwater Storage Ponds

Two substantial stormwater storage tanks are scheduled for construction in the southwest lake and central park areas to store excess runoff during peak rainfall. The storage tanks will be connected to the surrounding drainage network and have capacities of 3000 and 2000 cubic meters. The construction of these storage tanks can effectively alleviate the drainage pressure caused by instantaneous rainfall, reduce the peak flow rate, and provide conditions for reusing rainwater resources. A filtering system and a plant purification zone will be installed in the storage tanks to ensure that the stored rainwater can gradually infiltrate and be purified, thereby reducing the impact on the natural environment.

(4)

Rain Garden

In areas with high pedestrian traffic, such as educational establishments and business districts in the south, the installation of multiple rain gardens serves to mitigate the risk of internal flooding and enhance the aesthetic quality of the environment. Rain gardens effectively absorb and treat pollutants in runoff, utilizing the natural filtering effect of plants and soil while reducing surface runoff’s quantity and velocity. The dimensions of each rain garden are approximately 200 square meters, and they are situated in the school playground, commercial district green space, and along the roads. The construction of rain gardens has the dual benefit of enhancing the area’s aesthetic appeal while markedly reducing the rainwater collection rate, thereby mitigating the risk of internal flooding.

The maintenance of rain gardens primarily involves routine tasks such as plant trimming, removal of debris and sediment, replenishment of mulch layers, and inspection of drainage systems. Over the long term, properly maintained rain gardens can function effectively for over 20 years. They reduce municipal stormwater management costs by mitigating runoff, enhance biodiversity through the creation of microhabitats, and contribute to the reduction of urban heat island effects by increasing green cover. However, without consistent maintenance, their infiltration capacity could decrease due to sediment clogging and their visual appeal could diminish.

(5)

Permeable Pavement

Permeable paving techniques are encouraged in roads and squares in the eastern mountainous region and the western village area. Using permeable paving materials, including permeable bricks and concrete, facilitates the rapid penetration of rainwater into the subterranean environment, thereby reducing surface runoff. The permeable pavement in each area is approximately 5000 square meters. By optimizing the pavement design, the safe passage of pedestrians and vehicles is ensured while simultaneously enhancing the infiltration and storage capacity of rainwater. Permeable pavement mitigates the risk of internal flooding, improves groundwater recharge, and enhances the overall quality of the ecological environment.

The maintenance of permeable pavements involves regular vacuum sweeping and power washing to remove debris and prevent clogging, with maintenance scheduled quarterly. Long-term benefits of permeable pavements include extended pavement lifespan, reduced stormwater runoff, and enhanced groundwater recharge. Permeable pavements also reduce the urban heat island effect due to their lighter-colored surfaces, which reflect more sunlight. However, the effectiveness of permeable pavements can decrease over time if not maintained, particularly in high-traffic areas where clogging may be more frequent. Proper maintenance ensures functionality for up to 25 years, supporting the area’s long-term ecological and hydrological stability.

The parameters for the various drainage components, including permeable pavement, rain gardens, and stormwater storage ponds, are provided in

Table 5. Parameters for drainage components.

Component	Design Area (m2)	Infiltration Rate (mm/h)	Soil/Material Porosity (%)	Water Capacity (m3)	Additional Details
Permeable Pavement	5000	20	40	\	Permeable concrete and bricks used
Rain Garden	200	30	45	\	Plant species: drought-tolerant grasses and shrubs
Rainwater Storage Pond	5000	\	\	5000	Includes filtration and purification systems

.

3.3.2. Socio-Economic and Policy Implications

(1)

Cost–Benefit Analysis

Implementing the proposed drainage network optimization and low-impact development (LID) measures entails significant upfront investment. For instance, constructing rainwater storage ponds, permeable pavements, and rain gardens involves material, labor, and maintenance costs. However, the long-term benefits far outweigh these initial expenditures. By reducing waterlogging events, the study predicts a 63.88% reduction in drainage system failure rates, which translates into decreased economic losses caused by property damage, traffic disruptions, and lost productivity. Additionally, improved groundwater recharge and rainwater utilization reduce the demand for municipal water supplies, offering further cost savings over time.

(2)

Stakeholder Engagement Challenges

Successful implementation of these measures requires coordinated efforts among multiple stakeholders, including local government authorities, residents, businesses, and environmental organizations. One major challenge is raising public awareness about the importance of LID measures and maintenance requirements. For example, ensuring the long-term effectiveness of rain gardens and permeable pavements depends on community involvement in regular cleaning and maintenance.

Moreover, conflicting interests among stakeholders may arise. Businesses in densely populated areas might prioritize minimizing disruptions during construction, while government bodies may focus on achieving long-term ecological benefits. Establishing a participatory framework that includes regular consultation meetings and transparent decision-making processes is essential to address these challenges. Additionally, incentive programs, such as tax rebates for businesses that adopt onsite LID measures, could encourage broader participation.

(3)

Policy Recommendations

Integrating the proposed drainage and LID solutions into existing urban planning frameworks is critical. Local governments should update building codes to mandate the use of permeable materials in new construction projects and incentivize the retrofitting of existing structures. Furthermore, developing funding mechanisms, such as public–private partnerships (PPPs), can help alleviate the financial burden of implementation while ensuring sustainable maintenance. Collaboration with academic institutions to conduct long-term monitoring and evaluation of these measures can also provide evidence-based recommendations for policy refinement. As noted by Tanim et al. (2024), incorporating LID measures into urban planning not only addresses immediate flooding risks but also promotes long-term environmental benefits [38].

3.3.3. Research Innovation and Contribution

This paper presents a detailed optimization of parameters, including the impervious area ratio, infiltration rate, and soil saturation water content, within the SWMM. Combining multiple optimization methods, including genetic algorithms, differential evolution, and Bayesian optimization, enhances the model’s accuracy in simulating actual rainfall runoff. Furthermore, regionalized parameter adjustments can be made according to the unique topographic characteristics of the Wujin District. In this study, low-impact development (LID) measures, including rain gardens and permeable paving, were innovatively integrated with the SWMM to evaluate the impact of different LID measures on the regional drainage system. The simulation and comparison of different LID measures significantly reduced peak flows, increasing rainwater utilization and groundwater recharge.

The findings of this study demonstrate that the frequency and duration of localized flooding in the Wujin District can be effectively reduced by optimizing the drainage system and implementing low-impact development (LID) measures. The results demonstrate a significant reduction in the duration of localized flooding. This research significantly contributes to reducing the frequency and duration of localized flooding by providing a specific and feasible solution for the urban drainage system in this area. Furthermore, it offers a valuable reference for similar areas with complex topography and dense populations. The combination of an optimized SWMM and LID measures resulted in an increase in the efficiency of the drainage system from the original level to 85.2%, based on the performance metrics calculated during the data analysis phase of this study. This improvement also led to a significant reduction in the system failure rate and economic losses.

Furthermore, the implementation of LID measures enhanced groundwater recharge and increased the utilization rate of rainwater, markedly enhancing the area’s ecological benefits. Beyond addressing urban waterlogging issues, this research contributes to advancing the ecological environment and effectively utilizing resources. It provides both a theoretical foundation and practical insights for future urban sustainable development and flood prevention planning.

3.3.4. Current Deficiencies in Research

The collation of data represents the foundation upon which the optimization of the SWMM is based. However, due to the complex geographical environment of the Wujin district and time and resource constraints, there remain areas for improvement in the data collection process, which must be acknowledged. The veracity of the terrain data and land use data in some areas must be improved, which may compromise the model’s precision. Although a substantial quantity of data has been gathered through uncrewed aerial vehicle (UAV) remote sensing and field measurements, there are some regions where data collection remains insufficient, particularly due to the expanse of the protected area and the intricate terrain.

Furthermore, meteorological and hydrological data must be monitored for a longer period and updated in real-time. While substantial historical data has been procured from the China Meteorological Administration and automatic weather stations, the data’s temporal and spatial resolutions are limited, rendering them inadequate for capturing the nuances of short-term heavy rainfall and localized waterlogging.

Implementing low-impact development (LID) measures has demonstrated efficacy in reducing surface runoff and enhancing the utilization rate of rainwater. Nevertheless, the selection and design of LID measures must be optimized by specific geographical and climatic conditions, which need to be sufficiently addressed in this study. The design parameters (size, material selection) of rain gardens and permeable pavements require further optimization to achieve optimal results. Additionally, long-term maintenance and management of LID measures necessitate regular maintenance and cleaning of facilities such as rain gardens and permeable pavements to maintain effectiveness. However, this study could have delved into the maintenance strategies and costs associated with these measures.

4. Conclusions

• The amount of rainfall runoff in the study area was monitored, and the data obtained were used to calibrate and verify the SWMM parameters. The constructed SWMM simulation is highly accurate, and the model is suitable for Wujin District.
• By optimizing the drainage system and implementing low-impact development (LID) measures, the average duration of internal flooding in Wujin District was reduced from 7.43 h to 3.12 h, a 58% reduction. The results show that the optimized drainage system can drain floodwater more quickly, significantly reducing the impact of internal flooding on the area and improving the area’s flood control capabilities and the quality of life of residents. The average depth of accumulated floodwater was also reduced from 21.27 cm to 8.65 cm, a 59.33% reduction. The data show that by constructing new stormwater mains, adding drainage branches, and building stormwater storage ponds, the depth of accumulated water has been significantly reduced, and the extent of accumulated water during internal flooding has been effectively controlled, reducing damage to buildings and farmland and protecting the infrastructure and ecological environment in the area. At the same time, drainage efficiency has increased from 58.32% before optimization to 91.46%, representing an increase of 56.82%. The drainage system failure rate has dropped from 5.15% to 1.86%, a decrease of 63.88%. These results show that the overall performance and reliability of the drainage system can be significantly improved by optimizing the drainage network layout and conducting regular maintenance. An efficient drainage system can remove rainfall runoff more quickly, reduce the risk of waterlogging, and improve the stability and reliability of the regional drainage system.

The proposed method for SWMM calibration and optimization demonstrates strong adaptability and this approach can be applied to rapidly urbanizing regions prone to flash flooding, coastal cities facing rising sea levels, and semi-arid regions experiencing episodic intense rainfall. The combination of parameter calibration and LID measures provides a systematic framework that can be tailored to local conditions, offering a scalable solution for enhancing urban flood resilience. Future work should focus on integrating emerging technologies, such as real-time monitoring systems and machine learning algorithms, to further enhance the accuracy and efficiency of SWMM optimization. Expanding the application scope of this method to diverse urban and climatic settings will contribute to more robust and generalized flood management strategies.