Least squares support vector regression-based modeling of ammonia oxidation using immobilized nanoFeCu

Abstract

The substantial release of ammonia (NH3) into water streams results in eutrophication, which is harmful to aquatic life. The effectiveness of immobilized iron‑copper bimetallic nanoparticles (nanoFeCu) in removing NH3 from sewage was proved to be effective. However, the study of immobilized nanoFeCu applications for NH3 removal using machine learning (ML) in wastewater treatment is limited. The objective of this study is to develop an intelligent soft sensor using real pilot-scale experimental data to predict NH3 concentration during NH3 removal over the catalytic process using immobilized nanoFeCu. In addition to implementing the new ML model alongside an empirical model in real-time conditions, the study also involves comparing the developed model with existing models for wastewater treatment prediction. The results showed that the developed model outperformed other models, including artificial neural networks and support vector regression. Additionally, the developed model provided an accurate prediction of NH3 concentration with a correlation coefficient of 0.9215, well above the accepted threshold of 0.7. For online NH3 concentration estimation, the developed model effectively addresses the impacts of real-time conditions and demonstrates its adaptability to process changes. Further study on other pollutants removals, such as heavy metals or dyes, using the developed model can be conducted for wastewater treatment applications.