In traditional Space-Frequency Adaptive algorithms, calculating the optimal weight vector involves directly inverting the covariance matrix. For covariance matrices with larger conditions, the results obtained by directly inverting are unstable and inaccurate, leading to parasitic nulls on the synthesized beam. To solve this problem, this article proposes using QR decomposition to indirectly inverse, thereby improving the ill condition of the covariance matrix and obtaining more accurate and stable weight vectors. 100 Monte Carlo simulation results show that the proposed method can not only generate large main nulls in interference, but also effectively eliminate parasitic nulls which can reduce the impact on satellite signals while suppressing interference, therefore, the overall anti-interference ability is effectively improved.