This paper presents a solution of the Optimal Tolerance Design Problems for the production of power transformers. Based on statistical and probabilistic methods, it establishes a multi-objective and multi-constraint mathematical model. We then present a solution that maximizes the effective utilization rate of sheet material, for producing core columns of power transformers, and the passing rate of quality controls of matching coils and cores. It also minimizes the Tolerance cost of the product. The solution uses Particle Swarm Optimization (PSO), which is very effective as shown by simulation results.