A Stochastic Expectation Maximization Algorithm for the Estimation of Wastewater Treatment Plant Ammonium Concentration

In this study, we address the intricate challenge of reconciling environmental sustainability with economic viability within wastewater treatment plants (WWTPs). Our primary objective is to minimize fossil energy consumption and reduce nitrogen concentrations. Current controllers struggle to adapt to fluctuating electricity prices and the variable conditions within WWTPs. While Model Predictive Control and Dynamic Programming offer promising control strategies, their effective deployment hinges on the availability of a robust system dynamics model. To address the stochastic and nonlinear nature of WWTP processes, we introduce a stochastic model and estimation method combining a Monte Carlo Sequential smoothing algorithm with a Stochastic Expectation Maximization method. The proposed methodology results in accurate 24-hour confidence interval predictions, outperforming the conventional estimation method, Prediction Error Minimization (PEM), offering a reliable model for control of WWTPs.