Error diagnostics and data reconciliation for activated sludge modelling using mass balances

• Published in: Water science and technology 2002-01, Vol.45 (6), p.145-156
• Main Authors: MEIJER, S. C. F, VAN DER SPOEL, H, SUSANTL, S, HEIJNEN, J. J, VAN LOOSDRECHT, M. C. M
• Format: Article
• Language: English
• Subjects: Accuracy; Activated sludge; Activated Sludge Process; Adaptation; Applied sciences; Balances (scales); Biological and medical sciences; Biological treatment of waters; Biotechnology; Calibration; Coefficients; Data; Datasets; Detection; Environment and pollution; Error correction & detection; Error detection; Errors; Exact sciences and technology; Fundamental and applied biological sciences. Psychology; General purification processes; Industrial applications and implications. Economical aspects; Mass; Methods; Modelling; Models, Theoretical; Pollution; Procedures; Reproducibility of Results; Sensitivity and Specificity; Sewage - microbiology; Sludge; Waste Disposal, Fluid; Wastewater treatment plants; Wastewaters; Water treatment and pollution
• ISSN: 0273-1223; 1996-9732
• DOI: 10.2166/wst.2002.0102

Research on activated sludge models is mainly directed towards the reliability and estimation of model coefficients. Model calculations however, rely heavily on accurate determination of operational conditions. Accurate measurement of operational conditions and mass flows is difficult, caused by large (full-scale) process flows and the absence of reliable measurements. Therefore operational data should be verified on (gross) errors before being implemented in model studies. Calibrating a model on erroneous mass flows leads to laborious calibration procedures and moreover, unjustified adaptation of the model (kinetic and stoichiometric) parameters. Gross error detection is possible when there are more measurements than strictly required to solve a system of linear conservation relations (mass balances). A simple method for error detection is evaluating the mass balance residuals. For over determined systems data accuracy can be improved using balancing methods (i.e. minimising balance residuals). This is referred to as data reconciliation. A reconciled data set contains fewer errors and is exactly in line with the mass balances of the system. In this paper we describe a method for gross error detection and data reconciliation. It is shown how data reconciliation improves the accuracy of the data set and how the use of a balanced data set simplifies the model calibration procedure. This is demonstrated on the basis of a modelling study of a full-scale WWTP.