Hydrological temporal series of monthly runoff prediction by CEEMD-BP model

Monthly hydrological time series prediction plays an important role in the planning and management of water resources. Due to nonlinear and non-stationary nature of runoff sequences, it is difficult to predict accurately. Runoff sequence in the Huangshui River Basin of Qinghai Province from 1956 to 2013 was used to predict monthly runoff, combining complete ensemble empirical mode decomposition method (CEEMD) with BP neural network. The combined EEMD-BP and CEEMD-BP models were found to retain original data information better compared to single BP neural network, and prediction performance was better. CEEMD-BP was found to have better prediction accuracy in the combined model for hydrological monthly runoff prediction.