For the problem of estimation error and feedback delay in the process of acquiring channel information in Multiple-Input Multiple-Output (MIMO) systems, a robust interference alignment algorithm based on weight adjustment was proposed to improve the system performance at low Signal-to-Noise Ratio (SNR). Firstly, the system model was reconstructed by considering the influence of channel error on the basis of the ideal channel; secondly, the signal space at the receiving terminal was decomposed by matrix projection and divided into two parts, including desired signal subspace and interference signal subspace; thirdly, considering the interaction between the desired signal and the interference signal, the weighted sum of the power leaked into the corresponding subspaces was used as the objective function, and the iterative idea was used to calculate the precoding and interference suppression matrix. Finally, the calculated pre-coding and interference suppression matrix was used to derive the sum-rate expression of the channel error. Simulation results show that compared with the robust minimum interference leakage algorithm, when the SNR is 10 dB and the channel error variance value is 0.05, the spectrum efficiency of the system is increased by 25% and the energy efficiency is increased by 38%. Therefore, the proposed algorithm can effectively improve system performance at low SNR.