On the background that the bonded port area implemens the tax rebate policy for departure ports to the neighbouring inland dry ports and domestic trade ports, a port (port of departure) selection model was constructed to implement the policy of tax rebate at the port of departure based on the theory of user equilibrium cargo flow distribution. The model can determine the number and location of the port of departure under the constraints of the fixed total investment and to maximize the regional socio-economic benefits. A genetic algorithm model embedded with the Frank-Wolfe algorithm was designed and solved, and the export data of Dalian port hinterland were used for case analysis. Results show that after the implementation of the tax refund policy for the port of departure relying on the Dalian Bonded Port Area, seven places such as Harbin were selected as the port of departure, and the enterprises with export goods valued at USD 38.86 billion can receive a tax refund of USD 4.67 billion 10 ~ 25 days in advance each year. The above models and algorithms can provide necessary decision support for the government to formulate and implement the tax rebate policy for the port of departure.