基于FFA-GRNN模型的土石坝溃坝洪峰流量预测  

Predicting Peak Discharge at Earth Rock Dam Break Based on FFA-GRNN Model

作　　者：严新军[1,2] 王雪虎 赵蕊婷 庄培源 王红徐 马俊玲 YAN Xin-jun;WANG Xue-hu;ZHAO Rui-ting;ZHUANG Pei-yuan;WANG Hong-xu;MA Jun-ling(College of Water Conservancy and Civil Engineering,Xinjiang Agricultural University,Urumqi 830052,China;Xinjiang Key Laboratory of Hydraulic Engineering Security and Water Disasters Prevention,Urumqi 830052,China)

机构地区：[1]新疆农业大学水利与土木工程学院,乌鲁木齐830052 [2]新疆水利工程安全与水灾害防治重点实验室,乌鲁木齐830052

出　　处：《长江科学院院报》2025年第3期99-106,共8页Journal of Changjiang River Scientific Research Institute

基　　金：新疆维吾尔自治区重点研发任务专项(2022B03024-3);新疆水利工程安全与水灾害防治重点实验室研究项目(ZDSYS-YJS-2022-09)。

摘　　要：为提高溃坝洪峰流量预测精度,提出了一种基于GRNN的预测模型,结合耳廓狐优化算法FFA进行超参数优化,实现对溃坝洪峰流量的预测。以国内外堤坝溃决数据库为基础,用溃口底部以上库容、溃口底部以上水深和溃口深度3种因子作为输入变量,构建FFA-GRNN溃坝洪峰流量预测模型。为验证模型在溃坝洪峰流量预测精确度和拟合度,与其他4种智能算法进行对比。结果表明:提出的FFA-GRNN模型相较于其他模型具有更低的RMSE、MAE和更高的拟合度R^(2),证明所建模型在整体上具有更好的计算精度与拟合效果。通过分析模型在溃坝洪峰流量预测中的适用性,可为溃坝分析提供技术支撑。The accuracy of predicting the peak flood flow at the breach of earth-rock dam is crucial for dam break analysis.To improve the prediction accuracy of the post-breach peak flood flow,this paper presents a prediction model based on the General Regression Neural Network(GRNN),optimized by the Fennec Fox Optimization(FFA)algorithm for hyperparameters,to forecast the peak flood flow caused by dam breaches.Using a database of domestic and international dam failure cases,the model selects three factors as input variables:the reservoir capacity above the breach bottom,the water depth above the breach bottom,and the breach depth,to build the FFA-GRNN prediction model.To evaluate the model’s precision and fitting accuracy in predicting peak flood discharge at dam break,we compared it with four other intelligent algorithms.Results show that the proposed FFA-GRNN model has a lower Root Mean Squared Error(RMSE),Mean Absolute Error(MAE),and a higher coefficient of determination(R^(2))than other models,indicating superior computational precision and fitting performance.

关 键 词：溃坝 洪峰流量 土石坝 耳廓狐算法 广义回归神经网络 

分 类 号：TV122[水利工程—水文学及水资源]