Hydrological time series forecast by ARIMA+PSO-RBF combined model based on wavelet transform

Aiming at the nonlinear and time-varying complexity of hydrological time series, a hydrological time series pretreatment algorithm based on wavelet transform is designed. By analyzing laws of flow variation, non-stationary characteristics and the mechanism of ARIMA model and RBF model, we know that the ARIMA model is suitable for linear time series forecast, neural network is suitable for dealing with nonlinear problems, so we combine these two models to build the ARIMA-RBF forecast model and propose a particle swarm optimization algorithm to optimize the RBF neural network to improve the forecast accuracy and convergence rate. Finally, the forecast of hydrological time series is realized. Experiments show that the combined model with proper wavelet decomposition function and combined model parameters can significantly improve the forecast accuracy of water level compared with the traditional RBF neural network. The combined model provides a useful reference for the practical hydrological forecast.