摘要
本文以倒传递神经网路模式,进行长期流量预测之应用研究,并以台湾中部德基水库上游松茂流量站历年观测资料进行探讨.初步得到预测结果与实测值比较尚有一倍以上的误差。但以神经网路之潜在应用潜力,寻找系统化网路参数值建立方法以得到更精确预测结果之经验,为寻人相当值得继续研究之课题。
This paper uses Back-Propagation Neural Network Model to proceed the research of long-term flow prediction. The observed flow data of Son-Mao station on the upstream of De-Ji Reservoir of central Taiwan are used. The primer result is the over prediction nearly 100 % with observed data. But for future potential of Neural Network application, the experience of searching the systematized method of network parameters construction, and to get more accurate prediction results, is worth to study.
出处
《水土保持研究》
CSCD
1995年第3期62-67,共6页
Research of Soil and Water Conservation
关键词
水土流失
神经网路模式
最陡坡降法
流量预测
back-propagation neural network model the gradient steepest descent method sigmoid function flow prediction