| 摘要: |
| 【目的】揭示径流时间序列变化规律并进行预测，为水库调度提供指导。【方法】针对径流时间序列的非线性特点，利用重构相空间的嵌入维数确定神经网络的结构，建立了基于混沌相空间重构的径流量预测BP网络模型,并利用该模型对位于陕西省汉江上游的石泉水文站的径流时间序列进行了预测。【结果】实例计算结果表明，石泉水文站月平均流量的时间序列具有混沌性，最大嵌入维数为12，依此构建的BP神经网络收敛速度快、预测精度较好。【结论】利用重构相空间中的最佳嵌入维数，可合理确定BP神经网络的输入层节点数。 |
| 关键词: 混沌径流时间序列 径流预测 神经网络 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金项目(50709027,50779053);教育部重点研究项目(209125);陕西省教育厅科技项目(09JK664) |
|
| Run-off prediction model based on the chaotic and BP network |
|
|
| Abstract: |
| 【Objective】It is important for reservoir operation to obtain and predict run-off variation law.【Method】According to the nonlinear characteristic of runoff time series,the network structure is determined by using embedding dimensions of reconstructed embedding phase space,and runoff prediction BP network model is established based on the chaotic phase space reconstruction.This method has been applied to predict the runoff time series of Shiquan hydrologic station.【Result】The calculation results show that the monthly average discharge time series of Shiquan hydrologic station is chaos,and its maximum embedding dimension is 12.The BP network is structured based on the dimension,and convergence speed is improved,calculation error is reduced.【Conclusion】Embedding dimensions of reconstructed embedding phase space can provide evidence for determining input layer node number. |
| Key words: chaotic runoff time series runoff forecast neural network |
