流域产沙量预测的神经网络模型

Forecasting Model of Drainage Basin Sediment Yielding Based on Neural Networks

流域产沙量演变规律的研究对于工程设计、水土流失规划与治理具有重要作用。由于其影响因素多、演变过程复杂,目前,对流域产沙量的研究主要侧重于定性分析,这就造成了缺乏理论基础及精度不高的缺陷。人工神经网络能以非显式表示产沙量与其影响因素之间的非线性复杂关系,将其应用到流域产沙量的拟合与预测中,在改进BP网络不足及优化确定网络结构的基础上,建立了云南楚雄州龙川江流域产沙量预测模型,通过对预测样本的检验,表明其具有比较高的精度,基本能够反映龙川江流域产沙量的演变规律。

The research on evolvement law of drainage basin sediment yielding plays an important role in design of a project, planning and control of soil erosion. Because of multi - factor＇ s impact and complex evolvement process the research on sediment yielding is at present focused on qualitative analysis. It is not based on the sound theory and its resuits are not of high precision. The artificial neutral networks can show the non - linear relation between sediment yielding and its impact factors in a blurry way and can be applied to simulating and forecasting of sediment yielding. Having improved the BP networks and optimized the networks framework the forecasting model of sediment yielding in the Long- chuan river basin in Yunnan province has been established. The test of the samples indicates that the model can basically reflect the evolvement law of sediment yielding with high accuracy in the Longchuan river basin.