Adaptation of activated sludge to varying hydraulic and nutrient load in a coke oven wastewater treatment plant

The effective biological treatment of concentrated and highly toxic coking wastewater (CWW) depends on several operational and technological parameters. This study investigates the adaptation of activated sludge to varying hydraulic and nutrient load, both on a technological, and a microbial scale, by evaluating six years’ operational data of a full-scale European coking wastewater treatment plant (CWWTP). Using machine learning ensemble methods (XGBoost, Random Forest, and Gradient Boosting), the volumetric Chemical Oxygen Demand (COD) load was identified as the most significant predictor of effluent COD compliance. The formerly unattainable Best Available Techniques (BAT) target of 220 mg COD/L in the effluent of the biological treatment step was kept stable under a volumetric load of 0.20 kg COD/m 3 d and a specific load of 0.04 kg COD/kg MLVSS ⋅ d. The identified loading threshold values are 0.30 kg COD/m 3 d, which equals 0.05 kg COD/kg MLVSS ⋅ d in the examined technology. Microbial analysis revealed significant shifts in community composition across loading periods, with functional genera adapting to phenol and SCN − loads. Further 39% reduction in the effluent residual COD is achievable with a well-chosen physico-chemical post-treatment, which basically affects the reuse potential of the treated effluent. • The volumetric COD load is the top predictor of the effluent COD in full-scale. • The load limit values in reaching 220 mg/L effluent COD were calculated. • It is 0.30 kg COD/m 3 d (0.05 kg COD/kg MLVSS ⋅ d) in AS CWWTPs. • Further 39% reduction in the effluent COD is achievable with post-treatment. • Microbial community adapted to phenolic and SCN − loads affected treatment performance.