Abstract: To achieve accurate control of alumina concentration in aluminum electrolysis process, an adaptively predictive control method of recursive subspace for alumina concentration was proposed based on subspace identification and model predictive control technology. First, the recursive subspace algorithm with forgetting factor was used to establish an online prediction model of alumina concentration, which could adaptively predict the alumina concentration according to the change of working conditions. Then, the established model was applied to predictive control to achieve adaptive control of alumina concentration. Finally, based on the actual production data of an aluminum plant, experimental research was carried out to verify the effectiveness and superiority of the proposed control method for precise control of alumina concentration.