Process Monitoring of Operational Cost for Wastewater Treatment Processes Using Variants of ARMA Models Based Soft-sensors

Process monitoring of operation cost index (OCI) is of great importance for wastewater treatment plants (WWTPs), which is not only able to support financial budget, but also to optimize local operation. This paper proposed four variants of auto-regressive and moving average (ARMA), based on recursive least squares algorithm (RLS), ARMA based on recursive extended least squares algorithm (RELS), nonlinear auto-regressive neural network (NARNN), and nonlinear auto-regressive neural network with external input (NARXNN) respectively, to predict the operating cost in WWTPs. The proposed methods were validated in the simulation platform, Benchmark Simulation Model No.2-P (BSM2-P). On account of the strong nonlinearity of the wastewater treatment process, the nonlinear model, like NARXNN, achieved better performance in terms of mean square error (MSE) and correlation coefficient (R).