Fault detection in a wastewater treatment plant

Abstract

In this paper, Unscented Kalman filter (UKF) based Exponentially Weighted Moving Average (EWMA) is proposed for fault detection in a Wastewater Treatment Plant (WWTP). In the developed UKF-based EWMA, the UKF technique is used to compute the residual between the true and the estimated variable and the EWMA control chart is applied to detect the faults. The fault detection technique will be tested using simulated COST wastewater treatment ASM1 model. The detection results of the UKF-based EWMA technique are evaluated using three fault detection criteria: the false alarm rate (FAR), Average Run Length (ARL1) and the missed detection rate (MDR).