To predict the performance and evaluate the optimized process operation, a number of kinetic models were conducted in batch experiments for UASB-anammox biomass and EGSB-anammox biomass. Following a time series with substrate variations, the reaction of the mixed culture was separated into three phases: the anammox reaction, denitrification and cell lysis. Among the six selected kinetic models, the Hanlev and Luong models were found to be the most appropriate, with a prospected rmax of 0.28, 0.30 gN gVSS-1 d-1, a Ks of 53.38, 52.52 mg NH4+-N L-1 and inhibition coefficient of 900 and 928 mg N L-1, respectively. Significant differences were found in the simulated specific anammox activity (SAA) in the two reactors following longitudinal distribution. The EGSB-anammox biomass had the highest rmax of 0.30 gN gVSS-1 d-1, and a Ks of 53.38 mg NH4+-N L-1 (123.84 mg TN L-1), both validated in the models and experimentally. In contrast, a large variation was found in the UASB-anammox biomass, from 0.1 to 0.6 gN gVSS-1 d-1 from the top to the bottom of the reactor, and the removal efficiency of the whole system was lower. It was also found that a second feeding tended to increase the SAA for higher purity anammox biomass.