In this paper, co-evolutionary augmented lagrangian method (CEALM) is applied to the constrained optimization problem of electromagnetic devices. To verify the performance of the CEALM, it is applied to the superconducting magnetic energy storage (SMES) optimization benchmark problem. As a result, accurate global solution can be found with less computational efforts. However, the conventional CEALM has a problem that the convergence speed decreases as the solution approaches near a global optimum. In this paper, to solve the problem, an improved scheme for CEALM is proposed. The effectiveness of the improved CEALM will be proven by the comparison with the conventional one.