To solve the problem that the output signal-to-interference-plus-noise ratio (SINR) of the traditional beamforming algorithm decreases seriously under the condition of the mismatch of the steering vector and the error of the covariance matrix, a kind of robust adaptive beamforming algorithm based on the combination of steering vector optimization and covariance matrix reconstruction was proposed. By estimating the power and direction of signal and interference, interference plus noise covariance matrix was reconstructed. At the same time, combined with projection and spatial integral, the assumed steering vector was optimized to make it approximate to the actual steering vector. Then the complex weight was obtained by related calculation and the beamforming could be realized. The proposed algorithm can effectively suppress interference and improve the output SINR. For comparison, the performances of the proposed algorithm were simulated in six experiments. The simulation results show that the proposed algorithm has better performances. Compared with the worst-case performance optimization algorithm, the proposed algorithm has about 5 dB improvement in output SINR under the condition that the number of snapshots is fixed and the steering vector is mismatched. Compared with algorithms for comparison, in the case where the signal-to-noise ratio (SNR) is fixed and the steering vector is mismatched, the performance is improved by more than 4 dB. The simulation results verify the effectiveness of the proposed algorithm.