Towards the digitalization of water treatment facilities: A case study on machine learning-enabled digital twins

The digital transformation of water treatment facilities through machine learning-enabled Digital Twins (ML-DTs) offers a paradigm shift in sustainable water management. This work introduces an embedded DT framework that seamlessly integrates water treatment facility equipment, middleware, cloud computing, and predictive analytics. A case study on centralized membrane bioreactor (MBR) wastewater treatment plants demonstrates ML-DTs' capability for proactive process control and maintenance (PC&M): specifically, a knowledge-based multi-objective particle swarm optimization (KBMOPSO) fuzzy controller reducing aeration energy consumption of the aerobic zone from 0.12 to 0.15 kWh/t to 0.06–0.12 kWh/t, while maintaining required effluent quality. In parallel, a long short-term memory (LSTM) encoder-decoder model achieved accurate forecasting of MBR membrane fouling (MAPE < 6.45 %, R² > 0.87), enabling operators to proactively determine the need for online chemical cleaning under dynamic operating conditions. Despite these promising outcomes, the broader adoption of ML-DTs faces several barriers, including limited data availability, technical integration challenges, and organizational and human resource constraints. This work also provides actionable insights to help facilitate the transition towards intelligent water treatment facilities through the implementation of ML-DTs.