Economic and social perspectives of implementing artificial intelligence in drinking water treatment systems for predicting coagulant dosage: A transition toward sustainability

Abstract

The traditional determination of optimum coagulant doses in drinking water treatment plants (WTPs) using the jar test technique is time-consuming, expensive, significantly influenced by variations in raw water quality, and susceptible to human error. Therefore, this study employed an artificial intelligence (AI) technique, namely, artificial neural network (ANN), to predict the coagulant dosage and explore the economic and social perspectives of its implementation in waterworks. Data on raw water quality parameters obtained from a drinking WTP were utilized to develop prediction models. The results show that the proposed ANN model performed slightly better than the multiple linear regression (MLR) model regarding coagulant dosage prediction. With test data, the ANN model provided the mean absolute percentage error (MAPE), root mean square error (RMSE), and coefficient of determination (R2) values of 5.59, 2.49 mg/L, and 0.94, which were improved by 52.1%, 44.2%, and 17.1%, in comparison to the MLR model. The economic benefits from AI implementation in coagulation-flocculation operations could produce total savings of US$ 8036.44 per year, net present value (NPV) of US$ 18,154.81, and a discounted payback period of 3.4 years. Moreover, social benefits included safeguarding human health, decent job creation, and reliability and resilience of water production infrastructure. The adoption of AI in WTPs showed a strong linkage with multiple technological- and water-related sustainable development goals (SDGs) and associated targets. Identifying SDG correlations would aid in the sustainable planning and management of operational processes in WTPs worldwide.