Data-Based Modelling of Chemical Oxygen Demand for Industrial Wastewater Treatment

The aim of wastewater treatment plants (WWTPs) is to clean before it discharged into the environment. Real-time monitoring and control will become more essential as regulations for effluent discharges are likely stricter in future. Model-based soft sensors provide a promising solution estimating important process variables such chemical oxygen demand (COD) help predicting performance WWTPs. This paper explores possibility using interpretable model structures influent mill WWTPs by systematically finding best parameters an exhaustive algorithm. Experimentation was conducted with regression models multiple linear (MLR) partial least squares (PLSR), well LASSO nonlinear scaling function account nonlinearities. Some autoregressive time series were also built. results showed decent modelling accuracy when tested test data acquired from process. most notable included exogenous inputs COD (correlation 0.89, mean absolute percentage error 8.1%) PLSR prediction 0.77, 7.6%) 20 h horizon. show that these accurate enough real-time industrial WWTP.