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

Abstract

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

Robust Runoff Prediction With Explainable Artificial Intelligence and Meteorological Variables From Deep Learning Ensemble Model

“…In summary, while scholars have conducted considerable research in runoff prediction, there are still some shortcomings that need to be addressed, including: The commonly used approach among scholars to obtain the final prediction results after acquiring each predicted IMF component is through linear summation. However, this approach may exclude crucial information, thereby limiting the model's accuracy (W. C. Wang et al., 2021). Most researchers only consider past runoff and precipitation records as inputs for the model, overlooking the fact that runoff is closely related to other meteorological variables like temperature and evaporation (Gao et al., 2022; Samantaray et al., 2022; J. Zhang & Yan, 2023).…”

“…The commonly used approach among scholars to obtain the final prediction results after acquiring each predicted IMF component is through linear summation. However, this approach may exclude crucial information, thereby limiting the model's accuracy (W. C. Wang et al., 2021).…”

Monthly Runoff Forecasting Using Particle Swarm Optimization Coupled with Flower Pollination Algorithm-Based Deep Belief Networks: A Case Study in the Yalong River Basin

“…For example, Ren et al [34] adopted the RNN, LSTM, and GRU models for mid-to long-term runoff prediction and obtained good effects. Wang et al [35] proposed the SMD-SE-WPD-LSTM hybrid forecasting model for annual runoff and achieved higher accuracy and consistency.…”

Runoff Forecasting Based on Combining Multiple Neural Networks and Modal Decomposition

“…In addition, the non-stationary volatility characteristics of hydrological series bring di culties to prediction, and it is di cult for a single model to obtain accurate and stable prediction results (Li et al, 2019;Ramos-Scharrón, 2021;Turunen et al, 2020;Zahmatkesh et al, 2015). Time-frequency analysis methods such as wavelet packet decomposition (WPD) and singular spectrum analysis (SSA) can decompose a hydrological sequence with strong nonlinearity into multiple components with different periods, extract the detailed features, and achieve the purpose of simplifying the original sequence (Hou et al, 2020;Moosavi et al, 2017;Wang et al, 2019Wang et al, , 2021. The "decomposition-prediction-reconstruction" prediction model proposed thereby has stronger predictive capabilities (Chen et al, 2021;Liu et al, 2021; Q. J. .…”