To extend the coverage of traditional terrestrial communication networks and serve more diverse application scenarios, employing unmanned aerial vehicles (UAV) as aerial base stations has emerged as a viable solution. However, due to the mobility of users and the dynamic nature of UAV base stations (UAV-BSs), the optimal UAV-BS deployment problem is dynamic and complex. In this paper, the dynamic location adjustment problem of UAV-BS in an urban scenario is studied. At each time step, the locations adjustment problem is modeled as a particle swarm optimization (PSO) problem, in which the position of a particle represents a specific value of change in locations of all the UAV-BSs. The improved particle swarm optimization (IPSO) algorithm incorporating simulated annealing principle is adopted to search for the better positions of particles, and then, the locations of UAV-BSs are dynamically adjusted to improve the communication quality. The movement of users and the complex wireless channel in urban scenario are simulated. Results show that the proposed PSO model and the IPSO algorithm can effectively achieve dynamic deployment of UAV-BS. The proposed algorithm exhibits superior performance in practice.