基于SPOTPY包的流域水文模型动态率定研究

胡友兵; 马亚楠; 王凯; 徐建; 陈绪伟; 杜宏杰

doi:10.13928/j.cnki.wrahe.2025.S1.035

水利水电技术（中英文） ›› 2025, Vol. 56 ›› Issue (S1) : 230 -233. DOI: 10.13928/j.cnki.wrahe.2025.S1.035

水文水资源

基于SPOTPY包的流域水文模型动态率定研究

胡友兵 ¹^,² ,
马亚楠 ¹^,² ,
王凯 ¹^,² ,
徐建 ³ ,
陈绪伟 ³ ,
杜宏杰 ³

作者信息 +

Research on dynamic calibration technology of watershed hydrological model based on SPOTPY package

Youbing HU ¹^,² ,
Yanan MA ¹^,² ,
Kai WANG ¹^,² ,
Jian XU ³ ,
Xuwei CHEN ³ ,
Hongjie DU ³

Author information +

文章历史 +

PDF (4884K)

摘要

数字孪生流域建设是智慧水利最鲜明的特征，模型平台为数字孪生流域建设提供算法支撑，模型高效实例化是模型平台实践应用中亟需解决的一项重要技术瓶颈。聚焦模型实例化过程中动态在线率定需求，引入SPOTPY工具包，选取我国广泛应用的新安江模型为实例对象，按照水利部、淮委通用化模型建设颗粒规范要求，对该工具包提供的基于SCEUA优化算法的算例模块进行改造，将模型参数、优化方法参数、目标函数、雨水情数据等转变为传入变量。同时采用FastAPI、Docker等微服务构建、容器化打包技术杖，构建流域水文模型动态率定微服务实例。通过在数字孪生淮河模型平台中实践应用表明，该技术方法高效成熟可靠，为模型实例化在线动态率定实施提供了重要技术参考。

Abstract

The construction of digital twin watersheds is the most distinctive feature of smart water conservancy. The model platform provides algorithmic support for the construction of digital twin watersheds, and the efficient model instantiation is a critical technical bottleneck that urgently needs to be addressed in the practical application. Focusing on the demand for dynamic online calibration during model instantiation, it introduce the SPOTPY toolkit and select the widely used Xin'anjiang model in China as the case study. In accordance with the generalized model construction specifications set by the Ministry of Water Resources and the Huai River Commission, the example module provided by the toolkit was modified based on the SCEUA optimization algorithm. Model parameters, optimization method parameters, objective functions, rainfall datum, etc., were transformed into method input variables. Additionally, technologies were utilized such as FastAPI and Docker for microservice construction and containerization to build a dynamic calibration microservice instance for watershed hydrological models. The practical application of this technology in the construction of the digital twin Huai River model platform demonstrates that this method is efficient, mature, and reliable, providing important technical references for the implementation of online dynamic calibration of model instantiation.