混沌水文时间序列区间预测研究

本文提出了一种混沌水文时间序列区间预测算法。该算法首先利用关联积分法计算嵌入参数，重构水文时间序列的相空间，而后采用交叉迭代模糊聚类算法确定当前时刻相点的相似状态，并依据给定的不同区间风险度，动态得到未来某一时刻水文要素值的取值区间。作为分析研究，本文采用长江寸滩水文站的月径流时间序列作为研究对象，对其进行非线性水文中长期区间预测研究。结果表明该方法不但可以避免混沌点预测中局部邻域确定的任意性问题，而且还避免了混沌点预测中必须模拟确定性的混沌规则，无法控制其误差的问题。

Interval prediction of chaotic hydrological time series

An interval prediction algorithm for chaotic hydrological time series is proposed. First, two important parameters for reconstructing phase space, time delay and embedding dimension are estimated by correlation integral method. Then, the similar states of current point are determined by means of the cross iterative fuzzy clustering algorithm (CIFCA). The interval of the future value is forecasted under different interval risks. The monthly runoff series of Cuntan Hydrological Station in Yangtze River is analyzed as a case study. The result shows that the proposed method not only avoids the randomicity of neighborhood, but also avoids the forecasting error being getting out of control due to the necessity of simulating the deterministic chaotic rule.