关键词: 日径流, 小波分析, 粒子群优化算法, 支持向量回归, 金沙江

Abstract: In view of the high nonlinearity of runoff time-series process, we establish a Wavelet-PSO-Support vector regression (WPSO-SVR) coupled model to predict daily runoff by combining wavelet analysis, particle swarm optimization with support vector regression. The model can fully exploit the multi-resolution function of wavelet analysis and the nonlinear approximation ability of support vector regression, and decompose the daily runoff time series into different frequency sub-sequences with Wavelet analysis. Then we use particle swarm global search function to find the optimal model parameters and construct a model with these reconstructed sequences as inputs, and apply the model to daily runoff prediction at Shigu hydrologic station in the middle reaches of Jinsha River. The results show that the predictive effect of the WPSO-SVR coupled model is better than the single Support vector regression model and it has strong adaptability in daily runoff forecasting.

Key words: daily runoff, wavelet analysis, particle swarm optimization algorithm, support vector regression, Jinsha River