利用小波变换原理将具有非平稳特征的径流序列进行分解 ,使其平稳项和随机项分离 .对平稳项采用传统的 AR模型加以预测 ,而通过对随机项的混沌特征研究 ,发现其具有明显的混沌特征 ,进而提出了基于非线性混沌动力学的预测模型方法 .最后通过小波对所提出的 AR+NCDF预测模型预测结果予以重构 ,实现对原始径流序列的预测 .该方法通过实例验证具有较高的精度 ,是一种实际可行的方法 Using the principle of wavelet transform, the runoff series with non-stationary features are decomposed to separate the stationary term from the stochastic term. The stationary AR term is predicted by the traditional AR model. By studying the chaotic characteristics of stochastic terms, Chaotic features, and then put forward the prediction model method based on nonlinear chaotic dynamics.At last, the prediction results of the proposed AR + NCDF prediction model are reconstructed by wavelet to realize the prediction of the original runoff sequence.This method is verified by an example of High precision, is a practical method