Advancing the Comprehension of Iron Ions with Different Valences in Anammox-Based Processes: Insight into the Response and Mechanism Based on Prediction and Classification Machine Learning Models
Zhicheng Jiang, Xinxin Xu, Yuhang He, Ming Zeng, Meng Zhang, Wei Liu, Nan Wu
Abstract
The responses of anammox to Fe 2+ and Fe 3+ have been widely discussed; however, the critical concentration of Fe 2+ and Fe 3+ is not certain due to the different culture environments. A reliable and widely applicable predictive classification system based on the anammox-based system with iron-containing wastewater needs to be established, which can surmount the independence between different experiments and be combined with statistical analysis to elucidate the mechanism of the effect of Fe 2+ and Fe 3+ based on critical concentrations. The results confirmed that 5 mg/L iron promoted the nitrogen removal process, while higher concentrations inhibited nitrogen removal with a ratio of above 80%. Moreover, the results of Spearman correlation analysis proved that Fe 2+ showed a more obvious effect on the nitrogen removal rate than Fe 3+ . To precisely predict the nitrogen removal performance of anammox, Support Vector Machine, XGBoost, and Random Forest (RF) were compared, and the RF model was confirmed as the preferable model ( R 2 = 0.99). According to the classification model, the influence of iron ions on the anammox performance could be successfully traced back to the valent states of iron with an accuracy of 97.7%. Furthermore, a mechanism of the nitrogen removal process in the anammox-based system under iron ion stress was proposed.