【作者】 韩敏； 席剑辉； 许士国；

【Author】 HAN Min, XI Jian-hui, XU Shi-guo ( School of Electronic and Information Engineering ,Dalian University of Technology,Dalian , 116023 ; School of Civil Engineering and Architecture, Dalian University of Technology, Dalian, 116023)

【机构】 大连理工大学电子与信息工程学院； 大连理工大学土木建筑工程学院；

【摘要】 本文旨在讨论如何用人工神经网络(ANN)的方法预测河川年径流量,采纳一种新型的神经网络——通用学习网络(Universal Learning Network,ULN)。该网络所有节点互连,而且每两个节点之间是带有任意时间延迟的多重连接。以黄河年径流量预测为例,仿真表明,通过学习,ULN可成功地记忆年径流量的样本信息,建立非线性的年径流模型。通过与传统的卡尔曼滤波算法预测比较,ULN对年径流量的预测达到了较高的精度。更多还原

【Abstract】 This paper is concerned with the use of artificial neural networks (ANN) aimed at prediction of river annual runoff. A new type of neural network -- Universal Learning Network (ULN) is proposed. All nodes of this network are inter-connected and each pair of nodes can be connected with multiple branches that may have arbitrary time delays. Simulation to prediction of Yellow River annual runoff shows that ULN can memory the sample information successfully and can also build the nonlinear model of annual runoff. Compared with the traditional prediction using Kalman filter algorithm, LILN can get a prediction with higher precision to annual runoff.更多还原