An Ensemble Hybrid Forecasting Model for Annual Runoff Based on Sample Entropy, Secondary Decomposition, and Long Short-Term Memory Neural Network

Accurate and consistent annual runoff prediction in a region is hot topic management, optimization, monitoring of water resources. A novel model (ESMD-SE-WPD-LSTM) presented this study. Firstly, extreme-point symmetric mode decomposition (ESMD) used to produce several intrinsic functions (IMF) residual (Res) by decomposing the original series. Secondly, sample entropy (SE) method employed measure complexity each IMF. Thirdly, wavelet packet (WPD) adopted further decompose IMF with maximum SE into appropriate components. Then long short-term memory (LSTM) model, deep learning algorithm based recurrent approach, predict all Finally, forecasting results components are aggregated generate final prediction. The proposed which applied seven series from different areas China, evaluated on four evaluation indexes (R, MAE, MAPE RMSE). Results indicate that ESMD-SE-WPD-LSTM outperforms other benchmark models terms indexes. Hence can provide higher accuracy consistency for prediction, rendering it an efficient instrument scientific management planning