摘要
洪水的非线性变化特征使其预测结果在防汛指挥中难以让决策者实时有效利用,基于此,提出两种小波-ANFIS模型,将其应用于张家庄水库的年最大洪峰流量预测中,从而选出预测效果较好者。首先拟合1973—2002年的原始数据S,利用不同小波分解函数db5和sym8,分别对S信号进行尺度为4的分解,得到两组系列的低、高频信号;然后利用各低、高频信号对ANFIS模型进行训练,调试得到最佳模型并用来预测各低、高频信号在2003—2007年的信息;最后将各系列低、高频预测信号进行重构,即生成由sym8-ANFIS和db5-ANFIS模型预测的两个序列。经比较分析,sym8-ANFIS模型具有较快的收敛速度,且精度更高。
Aiming at the difficulty for decision-makers to make effective and real-time decision-making in flood control,this paper put forward two models based on the Wavelet-Anfis method and these were applied in prediction of Zhangjiazhuang Reservoir about its annual maximum peak discharge. Firstly,the original data S from 1973 to 2002 was fit,then respectively decomposed with the db5 and sym8 at scale 4 for the low-frequency and high-frequency signals. Secondly,the signals were trained and debugged for the respective best model to forecast the low-frequency and high-frequency signals from 2003 to 2007. Finally,two prediction sequences were generated by the predicted signals rebuilt by sym8-Anfis and db5-Anfis model. By comparison analysis,sym8-Anfis model has a faster convergence rate and higher accuracy,and it has important reference value for the annual maximum flood peak flow prediction and the corresponding decision.
出处
《人民黄河》
CAS
北大核心
2015年第10期42-45,共4页
Yellow River
基金
山西省水利厅基金资助项目(087)
