GA-BP组合预测方法在北洛河年径流量预测中的应用  

Application of GA-BP combination forecasting method into the annual runoff prediction of Beiluo River

作　　者：王广宇[1] 解建仓[1] 张建龙[2] 

机构地区：[1]西安理工大学教育部西北水资源与环境生态重点实验室,陕西西安710048 [2]山西省水利建设开发中心,山西太原030002

出　　处：《干旱地区农业研究》2014年第1期203-207,共5页Agricultural Research in the Arid Areas

基　　金：国家自然科学基金项目(51079120);国家自然科学基金项目(50979088)

摘　　要：为寻求较为合理的径流量预测方法,利用GA-BP组合模型进行径流量预测的研究,首先利用GA算法(Geneticalgorithm)全局搜索能力定位出一个较好的搜索空间,再利用BP算法(Backpropagation)神经网络的局部寻优能力,以陕西省北洛河状头站实测数据为基础进行径流量预测。结果表明:利用GA-BP组合模型预测2001—2010年结果的相对误差分别为-8.9%、-12.6%、-7.4%、9.4%、-7.6%、15.0%、-5.5%、10.7%、13.3%和-8.2%,预测精度较高,从相对误差和训练次数比较可以看出,GA-BP组合模型优于基本BP模型和GA模型。The paper has developed a GA-BP combination model to forecast the runoff, in order to seek more reasonable runoff forecasting methods. First, use the GA global search capability to locate a better search space, then use the local optimization ability of BP neural network. This forecast mod el was used the measured data of Zhuangtou Hydrometric Station in Beiluo River from 1960 to 2010. The results showed that： The relative error of the predicted results from 2001 to 2010 were -8.9%、-12.6%、-7.4%、9.4%、-7.6%、15.0%、-5.5%、10.7%、13.3% and -8.2%, respectively. The predicted results of GA-BP combination model were better than the basic BP model and the GA model from the relative err or and the training numbers.

关 键 词：径流量 预测 GA—BP组合模型 状头站 

分 类 号：TV121.2[水利工程—水文学及水资源]