基于BP神经网络对七里街测站洪峰的预报与分析  

The analysis and forecast of flood crest in Qilijie Station based on the BP neural network

作　　者：肖恭伟 刘国林[1] 曹淑敏[1] 孙志阳[1] 

机构地区：[1]山东科技大学测绘科学与工程学院,山东青岛266590

出　　处：《山东理工大学学报（自然科学版）》2016年第4期49-52,共4页Journal of Shandong University of Technology:Natural Science Edition

基　　金：山东省自然科学基金项目(ZK2012DM001)

摘　　要：在分析预报误差的时间分段递推修正方法的基础上,以建溪流域东游、水吉、建阳三个水文站点的水位监测数据为基础,计算得到BP神经网络隐含层最优节点数目为10,建立了BP神经网络对七里街测站水位预报的数学模型.在此基础上,利用时间分段递推修正方法对预报的结果进行修正,计算结果表明,时间分段递推修正方法使得预报精度提高很多,其结果与实际更加符合.According to the recursion in different section of time,and the monitoring data of water level in three hydrological sites including the east reach of Jianxi,Shuiji,and Jianyang,we find that the optimal node number of hidden layer of BP neural network is 10 through complex calculation,and establish a mathematical model to forecast the water level in Qilijie Station using the BP neural network.On this basis,we can amend the forecast results with recursion in different section of time.The caculation results show that this method improves the forecast accuracy.It can be pointed out through calculation that the correction method of recursion in different section of time is a better choice due to the coincidence between results and facts.

关 键 词：BP神经网络 洪峰预报 隐含层节点 误差修正 

分 类 号：P426.616[天文地球—大气科学及气象学] TP183[自动化与计算机技术—控制理论与控制工程]