No abstract

No abstract

Disinfection is one of the most critical processes for municipal wastewater treatment. However, traditional chemical dosing approaches do not consider how changes in water quality and process operation can alter disinfection performance. This work aims to develop novel disinfection models for precise prediction of peracetic acid (PAA) performance that considers real-time changes in water quality. Artificial and recurrent neural networks (ANN and RNN, respectively) are trained to predict PAA at various locations throughout the disinfection basin, CT (a function of the active concentration and contact time), and preand postdisinfection Escherichia coli using online and laboratory data. An ANN is found to predict PAA concentrations at an error rate comparable to that of an online analyzer. Additionally, an ANN can predict CT more accurately than a conventional first-principles method both with and without an online analyzer. An ANN with a lagged response variable can predict E. coli in a fraction of the time of an RNN, but with a slightly increased error. The integration of the models developed in this work into existing monitoring and control systems could provide treatment facilities with more robust and dynamic disinfection control without the need for costly analyzers.

No abstract

Summary Decentralized waste water treatment facilities monitor many features that are complexly related. The ability to detect the onset of a fault and to identify variables accurately that have shifted because of the fault are vital to maintaining proper system operation and high quality produced water. Various multivariate methods have been proposed to perform fault detection and isolation, but the methods require data to be independent and identically distributed when the process is in control, and most require a distributional assumption. We propose a distribution-free retrospective change-point-detection method for auto-correlated and non-stationary multivariate processes. We detrend the data by using observations from an in-control time period to account for expected changes due to external or user-controlled factors. Next, we perform the fused lasso, which penalizes differences in consecutive observations, to detect faults and to identify shifted variables. To account for auto-correlation, the regularization parameter is chosen by using an estimated effective sample size in the extended Bayesian information criterion. We demonstrate the performance of our method compared with a competitor in simulation. Finally, we apply our method to waste water treatment facility data with a known fault, and the variables identified by our proposed method are consistent with the operators’ diagnosis of the fault's cause.