多输入单输出BP网络预测的非线性预处理研究
NLP of Day-flow Prediction in MISO of BP ANN
作者：刘政(四川大学 水利水电学院, 四川 成都, 610065)；李存军(四川大学 建筑与环境学院, 四川 成都, 610065)；邓红霞(四川省紫坪铺开发有限责任公司, 四川 成都, 610091)
Author：(School of Water Resource and Hydropower, Sichuan Univ., Chengdu 610065, China)；(School of Architecture and Environment, Sichuan Univ., Chengdu 610065,China)；(Sichuan Province Zipingpu Development Co. Ltd., Chengdu 610091,China)
收稿日期：2007-09-17 年卷（期）页码：2008,40(4):53-57
期刊名称：工程科学与技术
Journal Name：Advanced Engineering Sciences
关键字：水文预测；非线性预处理；MISO；神经网络
Key words：hydrology prediction;non-linear pretreatment;MISO;ANN
基金项目：四川大学青年基金资助项目(06016);四川交通职业技术学院2007科研资助项目
中文摘要
非线性预处理在人工神经网络多输入单输出系统日径流预测中的性能主要取决于各输入序列和输出序列的变换组合方式，具有类似变化特征的输入和输出各序列采用变换特性相近或相同的变换方式相比于采用不同的变换组合具有更好的预测效果。以广西桂江流域阳朔、恭城和平乐3个水文站1973年~2001年的日径流量为例，研究了不同组合变换下的MISO非线性预处理预测效果。结果表明，非线性预处理预测均比线性预处理相对误差
英文摘要
The day-flow prediction result in MISO of BP ANN with NLP is mainly determined by transfer combination of series. The same transfers with similar characteristic are better than different transfers of series . This paper predicted in 7 cases of MISO with daily runoff of Yangshuo, Gongcheng and Pingle Station in Guangxi province during 1973~2001. The results show that eligibility ratio of relative error in 10%, in 20% and in 30% by NLP are averagely 17.87%, 15.85% and 8.29% more than LP respectively, that eligibility ratio of relative error in 10%, in 20% and in 30% by same NLP combination were averagely 3.52%, 2.19% and 1.24% more than by different NLP combination respectively.
【关闭】
