The three identical lab-scale upflow anaerobic sludge blanket (UASB) reactors were operated continuously for treating cannery seafood wastewater at seven hydraulic retention times (HRTs) of 5, 4, 3, 2, 1, 0.5 and 0.25 days. The different of granule sizes from three sources: a cassava factory (CS), a seafood factory (SS), and a palm oil mill (PS), average sizes in the range 1.5-1.7, 0.7-1.0 and 0.1-0.2 mm respectively were used as inocula for anaerobic digestion. The UASB-R1 used only granules from seafood factory (R1-SS), the UASB-R2 used mixed granules from seafood with cassava factory (R2-SS+CS) and the UASBR3 used mixed granules from seafood factory with palm oil mill (R3-SS+PS). In this study selected mathematical models including Monod, Contois, Grau second-order and modified Stover-Kicannon kinetic models were applied to determine the substrate removal kinetics of UASB reactor. Kinetic parameters were determined through linear regression using experimental data obtained from the steady-state experiment and subsequently used to predict effluent COD. The results showed that Grau second-order and modified Stover-Kicannon kinetic models were more suitable than that of others for predicting the effluent COD, with high the correlation coefficient (R2). In addition, the UASB-R2 from mixed granules with cassava factory (SS+CS) gave the best performance and highest coefficient value.