Engineering science Analysis of Chaotic Runoff Data Based on Chebyshev Polynomials Local Model

In this paper, we analyze dynamics of runoff data from actual hydrologic station on the basis of the chaos theory. Through analyzing Characteristics of power spectrum, geometric structure of attractors and calculating Lyapunov index, it confirmed that actual runoff has chaos dynamics behavior. Local prediction model based on chaotic runoff data is inferior for heavy computation load and poor accuracy under high order of polynomial and big sample size. To address these problems, combined the phase-space reconstruction technology, we propose a chaos local prediction model for runoff data based on Chebyshev Polynomials. Integrating the advantage of approximation effect of Chebyshev Polynomials with the nonlinear prediction ability of local prediction model, the proposed model could express dynamic characteristics of hydrologic runoff system well, increases prediction accuracy and enhances anti-noise performance of the model. Result of simulation shows that the model still gets relatively satisfying prediction effect under data pollution by white noise.