The Activated Sludge Model No. 2 (ASM2) is applied to modeling process dyna
mics (nitrogen and phosphorus) in a sequencing batch reactor (SBR). Additio
nally, simplified process model (SPM) and neural net (NN) approaches are pr
oposed and introduced into the SBR modeling. Simulation results show that t
he ASM2 can offer better predictions and explanations to certain batches of
SBR data than those of the SPM; frequent calibrations, however, are needed
. The hybrid SPM and NN model show model robustness and improve model predi
ctions. The results of this study demonstrate that the ASM2 is good for det
ailed process simulations while the proposed hybrid approach is more suitab
le for on-line prediction and process control. (C) 1999 IAWQ Published by E
lsevier Science Ltd. All rights reserved.