Predictive monitoring of wastewater treatment performance: Seasonal microbial activity and data-informed water quality model

초록

Biological treatment processes in wastewater treatment plants (WWTPs) require effective monitoring systems to rapidly detect process anomalies and identify causes. Traditional statistical approaches struggle with capturing dynamic microorganism characteristics influenced by seasonal variations and fluctuating conditions. To address this, a predictive and adaptive water quality monitoring system was developed, integrating an adaptive quality monitoring chart (AQUA) with a water quality auto-regressive variational mode enhanced model (WAVE). The WAVE model combines partial least-squares regression, autoregression, and variational mode decomposition to capture temporal and seasonal dynamics in microbial activity. Data from a full-scale M-city WWTP were analyzed to identify significant features and seasonal patterns. The WAVE model showed high prediction performances for chemical oxygen demand (COD) and total nitrogen (TN) removal efficiencies with root mean square error (RMSE) values of 0.45 and 1.12, respectively, while the AQUA chart detected abnormal changes in microbial activity. This system effectively accounts for seasonal fluctuations resulting from microbial activity's variations, reduces false alarm rates, and enhances process monitoring, contributing to stable effluent water quality and optimized biological treatment processes in full-scale WWTPs.

키워드