Author(s): Cesar Arenas Prado; Sofia Jaray-Valdehierro; Benat Elduayen-Echave; Tamara Fernandez-Arevalo; Aitor Domec; Itxaro Errandonea; Saioa Arrizabalaga; Eduardo Ayesa; Luis Valcarcel Garcia

Linked Author(s):

Keywords: Digital transformation; Digital twin; Energy-efficient control; Multi-objective optimization; Decision support system; Real-time prediction

Abstract: Water Resource Recovery Facilities (WRRFs) have traditionally been modelled with mechanistic frameworks that use systems of differential equations to represent biochemical, chemical or physico-chemical transformations. As these frameworks have matured to reflect evolving operations, their growing complexity has increased. In practice, models are often deployed offline for scenario analysis rather than continuous decision-making. This has motivated the rise of data-driven approaches and hybrid strategies that couple mechanistic structures with machine-learning components to balance interpretability and adaptability. Within DARROW European Project (https://www.wastewater.ai/), we present a lightweight, flexible framework that delivers real-time predictions of key effluent variables and provides operator-friendly optimization under regulatory and operational constraints running fully online with minimal infrastructure.

DOI: https://doi.org/10.64697/iahr.proc.hic2026.291

Year: 2026