Comprehensively understanding the characteristics of chemical reactions that occur during the combustion process is necessary to improve the design of modem internal combustion engine. However, huge computational cost limits the application of detailed chemical mechanism for practical combustion technologies. Therefore, the objective of this study is to improve the computational efficiency of the chemical reaction in practical engine simulations based on cell agglomeration (CA). In this study, in situ adaptive tabulation (ISAT), dynamic adaptive chemistry (DAC), and CA algorithms are implemented into the KIVA code. CA coupled with DAC and ISAT methods are carried out to validate the optimal acceleration method for practical engine simulation. The results show that CA can significantly reduce the cost of chemistry calculations by effectively reducing the number of CFD cells in which the chemistry calculation need to be performed during HCQ engine and diesel engine simulations. A maximum speed-up factor of 261-fold can be achieved using a 171-species mechanism. The error tolerance can affect the computational accuracy, particularly while using larger chemical mechanism. For non-premixed combustion, the selected coordinates in the reduced-space species composition can further affect the computational accuracy. Overall, the CA-DAC method can achieve the best computational efficiency without lowering the computational accuracy.