地下水位预报的相空间重构神经网络模型研究

为有效揭示水资源系统复杂的非线性结构及变化规律,结合混沌理论、相空间重构理论与神经网 络,研究了地下水水位预测模型,即通过相空间重构,把一维地下水水位时间序列拓展为多维序列,从而挖掘更为 丰富的信息;运用混沌方法构造训练样本并确定神经网络的网络结构,用神经网络拟合相空间相点演化的非线性 关系。实例计算表明,该模型具有较高的预报精度。

The model of phase space reconstruction and neural network in the groundwater level

In order to discover the complicated nonlinear structure and movement law of the water-resource system, in this paper the forecasting models for the water level of groundwater were researched integrating chaos ,reconstruction of phase space and neural network. One dimension water level series was developed into multi-dimension water level series with reconstruction of phase space ,and that multi-dimension series included the ergodic information. Training data construction and networks structure were determined by chaotic phase space ,fitting nonlinear relationship of phase points with neural networks. The example indicates the model is of high forecasting precision.