Download PDF (external access)

Mathematical and Computer Modelling of Dynamical Systems

Publication date: 2006-10-01
Volume: 12 Pages: 489 - 503
Publisher: Taylor & Francis Group

Author:

Smets, Ilse
Verdickt, Liesbeth ; Van Impe, Jan

Keywords:

asm1, model complexity reduction, linearization, cost benchmark, time-series analysis, water treatment plants, model, identification, Science & Technology, Technology, Physical Sciences, Computer Science, Interdisciplinary Applications, Mathematics, Applied, Computer Science, Mathematics, ASM1, COST benchmark, TIME-SERIES ANALYSIS, WATER TREATMENT PLANTS, MODEL, IDENTIFICATION, 0802 Computation Theory and Mathematics, 1702 Cognitive Sciences, Industrial Engineering & Automation, 46 Information and computing sciences, 49 Mathematical sciences

Abstract:

In the search for a reliable but simple model for the biodegradation processes of an activated sludge wastewater treatment plant, this paper presents a multi-model which is valid for the global operating region of a standard carbon and nitrogen removing facility. In a first step, locally valid linear models are derived. Two linearization procedures are compared. The first procedure is the classical Taylor series expansion, while the second is a newly developed linearization procedure based on weighted linear combinations. In a second step, the locally valid models are combined to obtain one globally valid multi-model. Previous work has focused on the most basic configuration of one anoxic and one aerated tank followed by a point settler [Smets, I.Y., Haegebaert, J.V. and Carrette, R. and Van Impe, J.F., 2003, Water Research , 37 , 1831 - 1851]. Refinements to the methodology are however needed (and presented here) once the influent flow rate range is increased and the benchmark configuration, proposed by the COST 682 working group no. 2, is taken as the simulation protocol. The main advantage of the obtained linear model (structure) remains the alliance of high predictive power with low complexity, rendering the multi-model fit for on-line optimization and control schemes.