A rapid forecasting method for mountain flood disaster based on machine learning algorithmsEI北大核心CSCD

基于高精度水动力模型与机器学习技术，运用极限随机树(ERT)及KNN算法，构建了高分辨率山洪灾害快速预报模型。利用确定系数、平均绝对误差和均方根误差3种指标评估模型的整体可靠性，同时，截取流域出口断面流量验证模型的预报性能。结果表明:所建模型预报结果与水动力模型模拟结果淹没范围基本一致，流域淹没范围平均相对误差低于5%，模型整体稳定可靠；流域出口断面流量平均相对误差低于10%，断面平均水深、流速平均相对误差低于5%，模型预报性能良好；模型可在10s内完成最大淹没情况计算并输出淹没范围图，能为紧急决策提供足够的前置时间，协助决策者更好地采取应对措施。

A rapid forecasting method for mountain flood disaster based on machine learning algorithms

Based on high-precision hydrodynamic model and machine learning technology, a high-resolution rapid forecasting model of mountain flood disaster was constructed by using extremely randomized trees and KNN algorithm. The overall reliability of the model was evaluated by three indexes: determination coefficient, mean absolute error and root mean square error. At the same time, the discharge of the outlet section of the basin was intercepted to verify the prediction performance of the model. The results show that the prediction results of the model are basically consistent with the inundation range of the hydrodynamic model simulation results, the average relative error of the basin inundation range is less than 5%, and the model is stable and reliable as a whole. The relative error of discharge at the outlet section of the basin is less than 10%, the average relative error of water depth and velocity at the section is less than 5%, and the prediction performance of the model is good. The model can calculate the maximum inundation situation and output the inundation range map within 10 s, which can provide sufficient lead time for emergency decision-making and assist decision-makers to take better countermeasures.