Predicting energy performance of the drinking water treatment process and its determinants

Abstract

Within the global climate change framework, enhancing energy efficiency presents a significant challenge for water utilities. Drinking water treatment is energy-intensive, involving several physicochemical processes to remove multiple pollutants from raw water. This study employs artificial neural networks (ANNs) and decision tree methods to gain a deeper understanding of the water-energy nexus in drinking water treatment processes. The energy efficiency of a sample of Chilean drinking water treatment plants (DWTPs) was estimated, resulting in an average score of 0.343. This indicates that on average, DWTPs could potentially save 65.7% of their current energy consumption if they were operating at an efficient level while producing the same quantity and quality of drinking water. The main source of raw water and the technology for treating water have been identified as critical factors influencing energy efficiency. Specifically, using surface water for producing drinking water, energy efficiency can increase to 0.514, whereas using groundwater would regress energy efficiency to 0.240. The use of predictive tools such as ANNs provides relevant information to support decision-making processes for a transition toward a sustainable urban water cycle