In this paper we study the differential evolution algorithm meta-heuristic to solve the travelling salesman problem with time windows. To create an initial population of the problem we use time windows. For every costumer we generate a random number in those time intervals. These time variables represent the visiting time of the related costumer. Our results for the benchmark problems show that with this initial population process and differential evolution algorithm that it could be a good alternative for the travelling salesman problem with time windows. In addition to describing the method, the experimental results of the benchmark test problems are presented.