A Study on Water Quality Index Prediction Based on Frequency Adaptive Normalization and TCN-Informer Model

Water resource protection is essential for sustainable social and economic development, making the monitoring and prediction of water quality indicators extremely important in practice. However, traditional prediction models often perform poorly due to the seasonal, cyclical, uncertain, and non-linear characteristics of water quality data. To overcome these challenges, this paper proposes a hybrid prediction model that incorporates the Frequency Adaptive Normalization (FAN), the Temporal Convolutional Network (TCN), and the Informer. Experiments on dissolved oxygen prediction were conducted using water quality monitoring data from the Wohu Mountain Reservoir in Jinan, Shandong Province. The results show that, compared with existing benchmark models, the proposed method improves the coefficient of determination (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{R}^{2}$</tex>) by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$7 \%, 10 \%$</tex>, and 11% for 4hour, 1-day, and 2-day predictions, respectively.