This paper presents a hybrid genetic algorithm to solve a multi-depot homogenous locomotive assignment problem with time windows. The locomotive assignment problem is to assign a set of homogeneous locomotives locating in a set of dispersed depots to a set of pre-schedules trains that are supposed to be serviced in pre-specified hard/soft time windows. A mathematical model is presented, using vehicle routing problem with time windows (VRPTW) for formulation of the problem. A cluster-first, route-second approach is used to inform the multi-depot locomotive assignment to a set of single depot problems and after that we solve each problem independently. Each single depot problem is solved heuristically by a hybrid genetic algorithm that in which Push Forward Insertion Heuristic (PFIH) is used to determine the initial solution and @l-interchange mechanism is used for neighborhood search and improving method. A medium sized numerical example with different scenarios is presented and examined to more clarification of the approach as well as to check capabilities of the model and algorithm. Also some of the results are compared with the solutions produced by branch & bound technique to determine validity and quality of the model. The experiments with a set of 15 completely random generated instance problems indicate that this algorithm is efficient and solves the problem in a polynomial time.