Application of VMD-LSTM in Water Quality Prediction

Abstract

Water is an indispensable prerequisite and essential factor for human life and social development. Change in water quality is closely related to water pollution. Water quality prediction is the basic work to control water environmental pollution. Therefore, the establishment of effective models is of great value to social economy, environmental protection and human health. This paper aims at the problem that the prediction error of relative maximum and relative minimum values of Long Short-Term Memory (LSTM) in dealing with time series is large, combined with the advantages of Variational Mode Decomposition(VMD) to extract signal fluctuation characteristics and reduce noise interference, Taking the Dissolved oxygen(DO) and the Total phosphorus(TP) as the main research objects, A combined model based on VMD-LSTM is proposed. The three indexes of Mean Absolute Error(MAE), Root Mean Square Error(RMSE) and Mean Absolute Percentage Error(MAPE) were used to judge. Experiments show that the VMD-LSTM model has improved all indicators compared with the LSTM model. DO prediction results respectively reduced by 38.8%, 41.4% and 44.4%, TP prediction results respectively reduced by 30.4%, 34.4% and 32.2%, the validity of the model is verified.