An artificial ground freezing method is often applied to highly permeable gravel formations. Seepage flow increases energy consumption and engineering accidents under this condition. Based on the physical modeling tests, the numerical simulations were conducted. The physical modeling test examined the development of the temperature and frozen wall. Numerical modeling integrated with the ACO algorithm was established to optimize the layout parameters of the freezing pipes. The results indicate that the flowing water prolongs the closure time of the frozen wall. Meanwhile, the total thickness of the frozen wall is also reduced by the flowing water. There are significant differences in the development rates of the frozen wall in different zones. The thickness of the entire frozen wall is nonuniform owing to the seepage flow. Following optimization using the proposed algorithm, the closure time was shortened from 82.4 d to 56.9 d for the frozen wall. Moreover, the freezing efficiency increased by 30.95% after optimization, and consequently, the entire frozen wall was more uniform with a nonweak zone. A case study showed that this optimization system is an effective method for artificial ground freezing operations in geotechnical engineering.