The solution of resource allocation is a significant issue in cognitive wireless network. For minimizing total power of second users, an subcarrier allocation algorithm of multi-colony adaptive immune optimization in NC-OFDM system is proposed. The basic idea and processes of multi-colony adaptive immune algorithm are given. And suitable adaptive operators are designed to speed up the convergence rate and simultaneously avoid the remaining local optimal. The simulation results show that the use of the algorithm can advance 20 percent in the reach of convergence state and make the total power of whole system decline 10 percent, and so it is an efficient subcarrier allocation algorithm of fast convergence speed.