Research on lake water quality prediction method based on WT-VMD-LSTM

This paper presents a lake water quality prediction method that combines time series decomposition and machine learning to tackle lake eutrophication. Wavelet transform (WT) and variational mode decomposition (VMD) are used for data denoising and decomposition to extract multi-scale features, after which a long short-term memory (LSTM) network constructs the model. The study focuses on Lake data denoising west part, analyzing four parameters: DO, CODMn, TP, and TN. Results show the original LSTM model struggled with precise predictions due to the high volatility, noise, and limited samples of water quality data, only capturing the general trend. However, after WT and VMD preprocessing, the model's prediction accuracy significantly improved. For the four parameters, the R² of the prediction results rose by 6.31%, 29.28%, 25.58%, and 26.37% respectively.