Author(s): Wenhong Wu; Yunkai Kang; Yuexia Xu

Linked Author(s):

Keywords: Granger Causality Test; Graph Neural Networks; Transformer; Water Demand Prediction

Abstract: Accurately predicting water demand is pivotal in optimizing strategies for multiple water sources. In response to challenges, including the suboptimal accuracy observed in existing prediction models, the complexity of achieving precise predictions across diverse time scales and sensors. This study introduces the Ensemble Empirical Mode Decomposition Granger causality test Attention Transformer Network (EGATN) model, which combines the advantages of traditional statistical methods with deep learning models. Experimental results demonstrate that compared to baseline models, the proposed model improves MAPE metrics by 2.12%, 4.33%, and 6.32% at forecasting granularities of 15 minutes, 45 minutes, and 90 minutes, respectively. The model achieves an R2 score of 0.97, indicating outstanding predictive accuracy, generalization, and exceptional explanatory power for the target variable. This research exhibits significant potential applications in predictive tasks within smart water management systems.

DOI: https://doi.org/10.3850/iahr-hic2483430201-338

Year: 2024