关键词: 混沌理论, 分解系数, 相空间神经网络模型, BP神经网络, 水文尺度分析

Abstract: In this paper, the problem of hydrological scale was discussed based on the chaos theory and artificial neural networks, and a method on the chaos theory and neural networks was used to analyze runoff downscaling. After comparing the chaotic characteristics of decomposition coefficient of annual runoff in large scale with decomposed monthly runoff in small scale, it is concluded that the decomposition coefficient series belong to chaotic series. A BP neural networks model based on chaotic phase space was used in forecasting the decomposition coefficient, and the predicted result used in calculating monthly runoff downscaling. Test results demonstrated that it is reasonable to use BP artificial neural networks for the description of nonlinear relation in the phase change of decomposition coefficients and to use a BP neural networks model based on chaotic phase space for the analysis of runoff downscaling.

Key words: chaos theory, decomposition coefficient, neural networks model based on chaotic phase space, backpropagation artificial neural network, hydrological scale analysis