Application of Recursive Principal Component Analysis to a Continuous Biological Phosphorus Removal Process

Abstract

Abstract Rapid detection, identification and diagnosis of faults can contribute to minimising operating costs on biological wastewater treatment plants. Principal Component Analysis (PCA) is not optimal for monitoring purposes because biological wastewater treatment processes are non-stationary. In this paper recursive PCA (RPCA) is evaluated in a samplewise updating mode on a case-study of a biological phosphorus removal plant, aiming at the detection of load changes, i.e. rain events and low loaded periods. The RPCA model adapts efficiently to slow changes in the process behaviour, and is, therefore, shown to perform considerably better than conventional PCA