摘要
针对水文模型参数的不确定性,对洪水进行分类预报,不同类型洪水采用不同预报参数,旨在提高洪水预报精度。基于BP神经网络模型,依据分类因子选取原则,选取6项具有代表性的影响因子作为模型输入,可将洪水划分成高、中、低3类。基于遗传算法,对3类洪水进行参数率定,获得3组不同的参数组,最终利用训练好的分类预报模型实现不同类型洪水的变参数预报。以大伙房水库25场历史典型洪水进行实例验证与分析,结果表明:分类预报结果的洪峰误差、峰现误差、确定性系数及典型洪水过程的拟合效果明显优于分类前。经训练后的基于BP神经网络与遗传算法的洪水分类预报模型可较好适用于大伙房水库,结果更贴合实测值,效果整体上优于分类前,方法可行、有效。
The purpose of this paper is to classify and forecast floods according to the uncertainty of hydrological model parameters.Different types of floods adopt different forecast parameters to improve the accuracy of flood forecast.Based on BP neural network model and the principle of selecting classification factors,six representative influence factors are selected as model inputs,which can be divided into high,medium and low 3 class.Based on the genetic algorithm,the parameters of three kinds of floods are calibrated,and three groups of different parameters are obtained.Finally,the trained classification prediction model is used to realize the variable parameter prediction of different types of floods.25 historical typical floods of Dahuofang reservoir are used for example verification and analysis,and the results show that:flood peak error,peak error,certainty coefficient of classification prediction results and typical flood process.The fitting effect is better than that before classification.It can be concluded that the trained flood classification and prediction model based on BP neural network and genetic algorithm can be better applied to Dahuofang reservoir,the results are more suitable for the measured values,and the overall effect is better than that before classification.The method is feasible and effective.
作者
刘恒
LIU Heng(Research Institute Limited Liability Company of Water Resources and Hydropower, Shenyang 110003, Liaoning, China)
出处
《水利水电技术》
北大核心
2020年第8期31-38,共8页
Water Resources and Hydropower Engineering
基金
水利部公益性行业科研专项经费项目(200801040)
水利部科技推广计划项目(TG1142)
辽宁省农业攻关计划项目(2011216001)。
关键词
BP神经网络
遗传算法
分类预报
大伙房水库
BP neural network
genetic algorithms
classified forecast
Dahuofang reservoir
