Deciphering organic substrate impacts in Anammox systems: A machine learning driven framework for predictive classification and process mechanism analysis

The response mechanisms of anaerobic ammonium oxidation (Anammox) systems to organic compounds have been extensively characterized, yet the synergistic impacts of organic classification and concentration gradients remain debated due to inconsistencies in operation conditions and microbial cultivation environments. This study investigates the critical factors governing nitrogen removal efficiency in Anammox processes, with particular emphasis on the combined effects of influent organic concentration and organic matter characteristics. Three datasets were constructed based on organic types: biodegradable organic compounds, biorefractory organic compounds and combined two types organic compounds. Two machine learning models were employed to predict Anammox performance, with Random Forest (RF) identified as the optimal model, subsequently validated using real coking industry wastewater treatment data. SHapley Additive exPlanations (SHAP) analysis revealed differential regulatory dominance across systems: organic concentration, influent ammonium nitrogen ( NH 4 + -N ) manifaested as the primary governing factors in biodegradable system, whereas organic type emerged as the most critical factors in biorefractory and combined systems. The results demonstrate that evaluating organic impacts solely through concentration levels is oversimplified, necessitating concurrent consideration of both organic type and concentration. Mechanistic interpretation through molecular-level inhibition analysis further validates the scientific rationale for employing Biochemical Oxygen Demand/Chemical oxygen demand (BOD/COD) ratio as a comprehensive indicator of carbon source effects on Anammox systems. In summary, machine learning framework effectively integrates and optimizes the regulation of material stoichiometry, environmental parameters, and microbial functionality, thereby advancing the development of energy-efficient nitrogen removal technologies and enhancing the evaluation system for wastewater treatment processes.