A recursively identified model for short-term predictions of NH4/NO3 - concentrations in alternating activated sludge processes

One of the stumbling blocks in the operation of alternatingly aerated activ ated sludge processes (ASPs) for nitrogen removal is the limited knowledge of both the varying influent composition and the complex dynamics of the bi ological process. This paper presents a simple physical N-removal model for alternatingly aerated, continuously mixed ASPs. The simplicity is achieved by capturing the slower process dynamics in recursively estimated time-var ying model parameters. Both seasonal and diurnal parameter variations are t racked. Also the influent ammonium concentration is treated as a recursivel y estimated model parameter. The method performs excellently on real data c ollected from an alternatingly aerated pilot scale ASP fed with municipal w astewater. Simulation of the resulting time-varying model yields accurate a nd computationally cheap predictions of ammonium and nitrate concentrations in the specific plant under operation over the next hours. Simulation for different control input scenarios can be used to optimize process performan ce, either manually by operators or automatically by model based optimizing controllers. Another possible application is optimization of the sludge (b iomass) concentration, as the estimated parameters contain information rega rding process load and concentrations and activities of the N-removing biom ass. From this information it can be computed whether there is an excess/sh ortage of sludge in the reactor. (C) 1998 Elsevier Science Ltd. All rights reserved.