The power inversion (PI) algorithm based on the linear constrained minimum variance(LCMV) criterion is used for anti-jamming, and the deep null opening in the direction of interference is wider, and when the strong and weak interference coexists, the nulling of the algorithm is weak in the weak interference. By studying and deducing the mathematical principle of the power inversion algorithm, the covariance matrix of the observed signal is decomposed to obtain the orthogonal subspace and the noise subspace. The reason of the above problem is that the algorithm contains the weighted vector sum of the interference space. Based on this, this paper introduces the subspace projection method to avoid the influence of the interference subspace on the algorithm. The simulation results show that the algorithm can form a deeper null in the direction of weak interference and improve the anti-jamming performance of the algorithm effectively.