Focused on the problem that the common cyclostationary beamformer converges slowly, the robust cycliststionary beamformer based on the diagonal loading method with an adaptive shrinkage factor is proposed. The proposed method first utilizes the shrinkage factor to modify the sampling covariance matrix and naturally obtain the estimation of the covariance matrix. Then the shrinkage factor can be calculated by solving the optimal problem about the minimum mean square error between the real covariance matrix and the estimated covariance matrix. Finally using the cyclic adaptive beamforming (CAB) algorithm to achieve the weighting value of the array. Simulation results show that the proposed method converges faster compared with the traditional cyclostationary beamforming algorithm when it comes to high power of interferences or low power of interferences, and outputs higher SINR under the case of low snapshot.