Antenna array beamforming algorithm is simulated using Matlab2019b to optimize antenna array excitation for obtaining wide nulls and low sidelobe patterns. Comparative pattern analysis shows that adding quadratic constraints based on the original constraints and improving the covariance have the drawback of widening the main lobe. This study proposes a linear constrained minimum variance (LCMV) pattern correction algorithm based on parallel optimization of amplitude decile particle swarms to maintain the width of the main lobe while maintaining wide nulls and low sidelobes. The algorithm compares the characteristics of the antenna array excitation obtained using several algorithms, introduces preliminary information, optimizes only the magnitude of the weight vector to narrow the feasible solution space, improves the optimization mechanism of the particle swarm algorithm, and adopts the decile optimization method. The algorithm convergence is made more stable, and the particle swarm algorithm code is vectorized. A graphics processing unit is used to simultaneously update each particle, realize the particle swarm parallel algorithm, and speed up the computational time of the algorithm. The simulation results show that the algorithm can achieve a widening of nulls and low sidelobes while maintaining the width of the main lobe. In addition, it has the best effect among the compared algorithms. Simultaneously, the convergence of the particle swarm parallel algorithm based on the amplitude decile requires fewer iterations and it has faster computational speed, and the improvement is more obvious when the antenna scale is larger.