[1]康得军,林益贤,林男津,等.基于BP-SA优化算法的SWMM高精度参数率定[J].中国给水排水,2026,42(13):130-136.
　KangDejun,LinYixian,LinNanjin,et al.High-precision Parameter Calibration of SWMM Based on BP-SA Optimization Algorithm[J].China Water & Wastewater,2026,42(13):130-136.
点击复制

基于BP-SA优化算法的SWMM高精度参数率定

中国给水排水[ISSN:1000-4062/CN:12-1073/TU] 卷: 第42卷 期数: 2026年第13期 页码: 130-136 栏目: 出版日期: 2026-07-01
Title:
High-precision Parameter Calibration of SWMM Based on BP-SA Optimization Algorithm
作者:
康得军1， 林益贤1， 林男津2， 赖李保壹3， 赵颖4
（1.福州大学 土木工程学院，福建 福州 350108；2.福州市闽侯县水利局，福建 福州 350199；3.龙岩市长汀县濯田镇政府，福建 龙岩 366300；4.中国环境科学研究院，北京 100012）
Author(s):
Kang　Dejun1， Lin　Yixian1， Lin　Nanjin2， Lai　Libaoyi3， Zhao　Ying4
（1. College of Civil Engineering， Fuzhou University， Fuzhou 350108， China； 2. Minhou County Water Conservancy Bureau， Fuzhou 350199， China； 3. Zhuotian Town Government of Changting County， Longyan 366300， China； 4. Chinese Research Academy of Environmental Sciences， Beijing 100012， China）
关键词:
雨洪管理模型(SWMM);  参数率定;  反向传播（BP）神经网络;  模拟退火算法
Keywords:
storm water management model (SWMM);  parameter calibration;  back propagation (BP) neural network;  simulated annealing algorithm
摘要:
在雨洪管理模型（SWMM）构建中，参数率定是确保模型模拟精度与可靠性的关键环节。但传统参数率定方法因参数复杂、不确定性高和计算效率低，难以获取全局最优解。为此，提出了融合反向传播（BP）神经网络与模拟退火算法的混合优化算法。首先，用修正的Morris筛选法筛选出对径流峰值及总量敏感的水文水力参数，降低率定时的参数空间维度；其次，利用BP神经网络快速预测不同参数组合下的模拟结果，显著减少计算成本；最后，引入模拟退火算法，借助Metropolis准则与温度衰减机制，有效平衡全局探索与局部开发能力。实验结果表明，该方法在不同降雨重现期下均表现出高精度且具有优异的适应性，综合径流系数率定误差在0.1%左右，纳什效率系数率定至0.99以上，可为复杂城市雨洪模型的自动化率定提供可靠的技术路径。
Abstract:
In the construction of the storm water management model (SWMM), parameter calibration is a critical step to ensure the accuracy and reliability of model simulations. However, traditional calibration methods struggle to obtain the global optimal solution due to parameter complexity, high uncertainty, and low computational efficiency. To address this issue, a hybrid optimization algorithm integrating a back propagation (BP) neural network with a simulated annealing algorithm was proposed. First, a modified Morris screening method was applied to identify the hydrological and hydraulic parameters that were sensitive to peak runoff and total runoff, thereby reducing the dimensionality of the parameter space during calibration. Second, a BP neural network was used to rapidly predict simulation outcomes under different parameter combinations, significantly reducing computational costs. Finally, a simulated annealing algorithm was introduced, which leveraged the Metropolis criterion and a temperature decay mechanism to effectively balance global exploration and local exploitation capabilities. Experimental results demonstrated that this method achieved high accuracy and excellent adaptability under various rainfall return periods. Specifically, the relative error in calibrating the comprehensive runoff coefficient was maintained at approximately 0.1%, and the Nash-Sutcliffe efficiency coefficient was calibrated above 0.99, providing a reliable technical pathway for the automated calibration of complex urban stormwater models.

相似文献/References:

[1]陈刚,王琳,张寒.利用SWMM进行低影响开发模拟的模型误差分析[J].中国给水排水,2021,37(13):98.
　CHEN Gang,WANG Lin,ZHANG Han.Model Error Analysis of Low Impact Development Simulated by SWMM[J].China Water & Wastewater,2021,37(13):98.

更新日期/Last Update: 2026-07-01