Machine learning strategy secures urban smart drinking water treatment plant through incremental advances

Abstract

The integration of machine learning into urban drinking water treatment plants (DWTPs) offers a transformative pathway to ensure drinking water safety while promoting the development of smart, low-carbon cities. However, the effectiveness of these systems is frequently hindered by challenges related to data security and reliability, including imprecise control logic, sensor inconsistencies, and data transmission errors. In this study, we introduce a novel progressive Step-by-Step (SBS) machine learning strategy, initially applied to precise disinfectant dosage control in drinking water treatment and subsequently extended to enhance the data security of the entire water supply system. Among eight evaluated methods, the deep neural network integrated with the SBS strategy demonstrated superior performance. In a real-world DWTP, the SBS model significantly outperformed manual fuzzy control, reducing disinfectant dosage by 22.0 % and effluent turbidity by 16.0 %. Furthermore, through simulations of extreme data-missing scenarios and the application of SBS-based corrections, the robustness and security of DWTPs were maintained. The integration of the SBS strategy has the potential to significantly improve emergency management in urban water systems and elevate the intelligence of water supply networks. This approach not only strengthens urban resilience but also supports the safe and sustainable evolution of smart urban water systems.