Renmin Zhujiang (Jan 2022)
Research on Runoff Prediction Based on EMD-FBI-ELM Model
Abstract
To improve the accuracy of runoff prediction,this paper proposes a runoff prediction method based on empirical mode decomposition (EMD),forensic-based investigation (FBI) algorithm and extreme learning machine (ELM).Firstly,EMD is used to decompose the runoff series data into multiple more regular component series.The phase space of each component series is reconstructed by the autocorrelation function method (AFM) and the false nearest neighbor (FNN) method.Secondly,the FBI algorithm is used to optimize the ELM input layer weight and hidden layer bias,on the basis of which the EMD-FBI-ELM runoff prediction model is established.Three other models are also established for comparison,i.e.,EMD-FBI-SVM,FBI-ELM and FBI-SVM.Finally,the EMD-FBI-ELM,EMD-FBI-SVM,FBI-ELM and FBI-SVM models are verified and analyzed with the annual runoff at the Gulaohe River Hydrological Station in Yunnan Province as a prediction example.The results show that the average relative error of the EMD-FBI-ELM model is 3.97% for the annual runoff prediction,which is 53.9%,81.7% and 86.5% lower than those of the EMD-FBI-SVM,FBI-ELM and FBI-SVM models,respectively.The EMD-FBI-ELM model is feasible for runoff prediction,and the model and optimization method can provide reference for relevant prediction research.
