RESEARCH ON PREDICTION AND ANALYSIS OF STORM SURGE LEVEL BASED ON CEEMDAN-LSTM

In this paper，an innovative approach is proposed，leveraging the complete ensemble empirical mode decomposition with adaptive noise and long short-term memory networks（CEEMDAN-LSTM）model，to forecast short-term time series of storm surge levels. The method is compared to commonly used machine learning models to assess its efficacy. The findings demonstrate that the CEEMDAN-LSTM neural network excels in accurately forecasting the short-term characteristics of storm surge levels within project areas. Notably，this model exhibits superior stability and precision compared to conventional machine learning models. © 2024 Science Press. All rights reserved.