This paper describes a reduction in the peaks of traction energy for metro railways in timetabling. We develope a mixed integer programming (MIP) model, which minimizes the number of trains running simultaneously. We suggest two approaches. In the first approach, we use the commercial MIP solver, CPLEX. In the second approach, we propose a heuristic algorithm. We apply both methods to the current daily timetable of the Korea Metropolitan Subway. We determine an optimal solution, which results in an improvement of approximately 25% over the current timetable.