An intelligent cloud-enabled system for water quality index prediction of sewage treatment plant using deep learning models

 Riya, Anju Bala, Anamika Sharma, Neetu Singh,
An intelligent cloud-enabled system for water quality index prediction of sewage treatment plant using deep learning models,
Journal of Water Process Engineering,
Volume 78,
2025,
108787,
ISSN 2214-7144,
https://doi.org/10.1016/j.jwpe.2025.108787.
(https://www.sciencedirect.com/science/article/pii/S2214714425018604)
Abstract: With the rapid increase in population and urbanization, sewage water generation has increased, creating significant challenges for sewage treatment systems in terms of water quality. Monitoring water quality is essential to effective urban infrastructure, especially for sewage treatment systems. Many traditional water-quality models operate offline or rely on local computation. Existing water quality index prediction systems often lack real-time scalability and centralized management. To address this gap, the objective of the current work is to develop a cloud-enabled system integrating advanced deep learning models to improve the accuracy and accessibility of water quality index prediction in sewage treatment plants. Therefore, this paper proposes a three-layer intelligent, cloud-integrated architecture that utilizes deep learning models to predict the Water Quality Index effectively. It integrates IoT, cloud, and intelligent layers for effective water quality assessment for real-time and scalable decision-making for urban water resource management. Recurrent deep learning architectures, including LSTM, GRU, BiLSTM, BiGRU, along with a stacked ensemble framework, were implemented to exploit temporal dependencies inherent in time-series sensor data, thereby enabling robust classification of treated water quality as suitable or unsuitable for reuse. The models’ performance was evaluated using a wide range of metrics, including MAE, MSE, RMSE, R2, accuracy, precision, recall, and F1 Score. With an R2 value of 0.93, the stacked ensemble model outperformed all others, demonstrating exceptional predictive reliability. Along with this, the stacked ensemble model shows the lowest MAE (0.02), MSE (0.0029), and RMSE (0.0655) when compared with existing research work. The deep learning models were deployed and executed in both desktop and cloud computing environments, where cloud-based execution shows reduced computational latency. These results demonstrate that the stacked ensemble model exhibits enhanced predictive performance for water quality monitoring across both deployment environments, with the cloud environment representing improved execution efficiency.
Keywords: Sensors; Sewage treatment plant; Cloud infrastructure; Water quality index; Deep learning; Stacked ensemble