Water Quality Prediction Based on the CEEMDAN-ABC-VMD-LSTM Model

Water pollution has a wide range of impacts, and accurate water quality prediction can improve the effectiveness of pollution control efforts. In this research, we developed a two-stage decomposition model for water quality prediction, optimized using the Artificial Bee Colony (ABC) algorithm. The model calculates the component complexity of Complete EEMD with Adaptive Noise (CEEMDAN) decomposition by sample entropy and aggregates the components according to the component complexity. Subsequently, the Variational Mode Decomposition (VMD) algorithm conducts a quadratic decomposition on the high complexity combinatorial components. Lastly, the outcomes are integrated into a Long Short-Term Memory (LSTM) model for forecasting. Water quality data from the Fu River within China's Yangtze River basin, including permanganate (COD), ammonia nitrogen (NH<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</inf>-N), and total nitrogen (TN) levels, were utilized to evaluate the model. As the results show, the ABC algorithm effectively optimized the decomposition parameters of VMD, and the two-stage decomposition successfully minimized the model's prediction error. Overall, the model exhibited robust predictive capabilities.