Research on water environmental indicators prediction method based on EEMD decomposition with CNN-BiLSTM

Abstract

Abstract Water resources protection is related to the development of the social economy, and the monitoring and prediction of water environmental indicators have important practical significance. In view of the seasonality, periodicity, uncertainty, and nonlinear characteristics of water quality indicators data, traditional prediction models have poor performance. To address this issue, this paper introduces a hybrid water quality index prediction model based on Ensemble Empirical Mode Decomposition (EEMD), combined with Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory Network (BiLSTM). We have conducted out experiments to predict dissolved oxygen based on the water quality monitoring indicators of the Liaohe National Control Sanhongcun Village station in Yichun City. The results show that the model proposed in this paper improves the $$R^2$$ R 2 index by 5%, 7% and 5% compared to the suboptimal model in the 4-h, 1-day and 2-day index predictions, respectively.