Multi-project multi-site location problems are multi-objective combinational optimization ones with discrete variables which are hard to solve. To do so, the case of particle swarm optimization is considered due to its useful characteristics such as easy implantation, simple parameter settings and fast convergence. First these problems are transformed into ones with continuous variables by defining an equivalent probability matrix in this paper, then multi-objective particle swarm optimization based on the minimal particle angle is used to solve them. Methods such as continuation of discrete variables, update of particles for matrix variables, normalization of particle position and evaluation of particle fitness are presented. Finally the efficiency of the proposed method is validated by comparing it with other methods on an eight-project-ten-site location problem.