Research of the Multi-input EMD-Bi-LSTM for reservoir water level prediction

The water level of reservoirs plays a crucial role in the local ecological environment, influencing various aspects of surrounding ecosystems. Given the numerous factors that control reservoir water levels, this paper proposes an advanced reservoir water level prediction model, which utilizes a hybrid algorithm combining Multi-Input Empirical Mode Decomposition (EMD) and Bidirectional Long Short-Term Memory Networks (Bi-LSTM). By incorporating hydrometeorological data and reservoir operation conditions as input variables, the Multi-Input EMD-Bi-LSTM model effectively captures and addresses the inherent nonlinear and nonstationary characteristics of time series data, thereby improving the accuracy and stability of reservoir water level predictions. The proposed Multi-Input EMD-Bi-LSTM model is compared with other models, including EMD-RNN、EMD-LSTM and EMD-GRU. The results demonstrate that the Multi-Input EMD-Bi-LSTM model significantly outperforms these traditional models regarding prediction accuracy and reliability. This advantage is attributed to the model's ability to handle complex multi-scale temporal patterns in the data, which simpler models often overlook. The enhanced model supports water resource management and contributes to planning responses to extreme weather events, which are becoming increasingly frequent and severe due to climate change.