Based on the multivariable dynamic matrix control algorithm, we optimize the feedback correction part. When the prediction vector is corrected, a new function is introduced to make a dynamic feedback correction module. Furthermore, we optimize the parameters in the displacement matrix, compensate the system and improve the control effect. Taking a distillation column as the research subject, we analyze the process, select reasonable variables, and conduct simulations of the optimized multivariable dynamic matrix control algorithm. Simulations show that this algorithm can effectively improve system response speed, reduce system overshoot, improve the overall performance with strong robustness, good control effect and good application prospect.