相空间重构神经网络在洪水灾害损失预报中的应用

Application of Phase Space Reconstruction and Neural Network in Flood Disaster Losing Forcasting

在灾害领域中引入混沌理论,将相空间重构理论与神经网络相结合,提出了洪灾成灾面积预测模型。通过相空间重构,把一维成灾面积时间序列拓展为多维序列,而多维序列包含着各态历经的信息,从而可挖掘更为丰富的信息,有利于神经网络的训练。利用神经网络模型可以较好地求解非线性问题,因而使预测结果更符合实际。实例表明,该模型预报精度较高。

Introducing chaos theory in the disaster resources field, the forecasting models for the inundated area of flood disaster were brought forward integrating reconstruction of phase space and neural network. One-dimension inundated area series is developed to multi-dimension inundated area series with reconstruction of phase space, and the multi dimension series include ergodic information, so that more abundant information can be found in favor of ANN training. With neural network, non-linear problem can be solved better, as a result, forecasting can accord well with practice even more. The example indicates that the model has highly forecasting precision.