Reinforcement Learning Based Strategies for Decision Support on Water Treatment Plants

Treatments to be applied for water purification must be dynamically adaptable to any raw water conditions. Currently, treatments are applied based on standards that are generally correct but not optimized for the circumstances of each drinking water treatment plant (DWTP), neither for critical events. This work presents a methodology for the creation of an Artificial Intelligence (AI) water treatment model, based on reinforcement learning techniques, that provides suggestions about the most efficient treatments for various raw water conditions, increasing their resilience to climate and water related risks. The model has been developed, optimised and validated in a DWTP replica. The results and evaluation of the model are promising as a first approach of a decision support system for drinking water treatments suggestion to be applied to 4.0 DWTPs, although next versions may include more water quality parameters to characterize raw water.