基于STFS-Urban的秀水河流域洪涝过程模拟

赵宇; 梁益银; 鹿鹏程; 黄欣; 张书亮

doi:10.13928/j.cnki.wrahe.2025.10.003

水利水电技术（中英文） ›› 2025, Vol. 56 ›› Issue (10) : 31 -45. DOI: 10.13928/j.cnki.wrahe.2025.10.003

复杂灾害链与水旱巨灾风险评估专栏

基于STFS-Urban的秀水河流域洪涝过程模拟

赵宇 ¹^,² ,
梁益银 ¹^,² ,
鹿鹏程 ¹^,² ,
黄欣 ³ ,
张书亮 ¹^,²

作者信息 +

The flood process simulation of the Xiushui River Basin based on STFS-Urban

Yu ZHAO ¹^,² ,
Yiyin LIANG ¹^,² ,
Pengcheng LU ¹^,² ,
Xin HUANG ³ ,
Shuliang ZHANG ¹^,²

Author information +

文章历史 +

PDF

摘要

【目的】自然流域通常具有复杂的干支流体系，短历时强降雨作用下，产汇流过程加剧，河道流量激增，水位上涨，易发生严重的洪涝灾害。为提高流域洪涝模拟的时效性与准确性，【方法】基于洪涝时空模拟模型STFS-Urban,进行集成模式下的时空交互机制分析与特征参数表达，进一步构建标准数据集，结合改进的Transformer深度学习算法，实现流域洪涝物理模型和深度学习模型的耦合集成。以沈阳市秀水河流域作为研究区域，基于“2022-07-06”“2022-07-28”两场典型暴雨监测数据，开展洪涝过程情景模拟，将模拟结果与公主屯水文站实测数据进行验证分析。【结果】结果显示，预测水位与实测水位差值均小于0.5 m,误差在1.5%以内；预测峰现时间与实测时间误差均小于1.85 h,所构建的流域洪涝集成模型预测淹没范围、洪水演进趋势与实际情况相符，模拟时效较物理模型提升约31~34倍。【结论】结果表明，所建立的STFS-Urban流域洪涝集成模型，能够较好地模拟洪水演进过程，在保证精度的同时有效提升了计算效率，能够为流域洪涝灾害的防治和对策制定提供科学依据。

Abstract

[Objective] Natural catchments typically have complex tributary systems. Under the influence of short-duration, intense rainfall, the process of runoff generation and concentration is intensified, leading to a surge in river discharge and rising water levels, which increases the likelihood of severe flooding. To improve the timeliness and accuracy of basin flood simulations, [Methods] the spatiotemporal flood simulation model STFS-Urban is used to analyze the spatiotemporal interaction mechanisms and characteristic parameter expressions under the integrated model. A standard dataset is constructed, and the coupling integration of the basin flood physical model with the improved Transformer deep learning algorithm is achieved. The Xiushui River Basin in Shenyang is selected as the study area, where flood process scenario simulations are conducted using monitoring data from two typical heavy rainfall events on “2022-07-06” and “2022-07-28.” The simulation result are validated through comparison with observed data from the Gongzhutun hydrological station. [Results] The results show that the difference between the predicted and observed water levels is less than 0.5 m, with an error within 1.5%. The error in the predicted peak time relative to the observed time is less than 1.85 hours. The constructed integrated basin flood model accurately predicts the inundation extent and flood evolution trends, which are consistent with the actual situation. The simulation efficiency is approximately 31 to 34 times higher than that of the physical model. [Conclusion] The results indicate that the established STFS-Urban basin flood integrated model is capable of effectively simulating the flood evolution process. While ensuring accuracy, the model significantly enhances computational efficiency and can provide a scientific basis for the prevention and control of river basin flood disasters and the formulation of countermeasures.