In order to improve the accuracy of monthly runoff prediction and solve the problem that the traditional de-composition and integration of runoff prediction method introduces"future information"in advance,which cannot be real-ized in practical engineering,a monthly runoff prediction model(ASEEMD-BES-ELM)based on the coupling of adaptive sliding ensemble empirical mode decomposition(ASEEMD),bald eagle search(BES)algorithm,and extreme learning machine(ELM)was proposed.Taking the monthly runoff sequence of Manas River from 1957 to 2014 as an example,firstly,the original monthly runoff sequence was adaptively decomposed using ASEEMD to obtain several sub-sequences.Secondly,each sub-sequences were inputted into the ELM model optimized by combining the BES algorithm and the grid search for prediction,respectively.Finally,the prediction results of each sub-sequence were accumulated to obtain the fi-nal monthly runoff prediction value.Comparison with the ELM*,BES-ELM*,BES-ELM and EEMD-BES-ELM(tradi-tional"bundle decomposition")models show that the Nash-Sutcliffe efficiency coefficient of the ASEEMD-BES-ELM model was 0.971,the mean absolute error was 5.173 m3/s,and the root-mean-square error was 8.282 m3/s,and the mean absolute percentage error was 16.033%,which has the highest prediction accuracy in line with the practical applica-tion.The results can provide a reference for the monthly runoff prediction in arid areas.
万方数据
维普
