Hall-Héroult process is a complex electrolysis procedure to produce aluminum at industrial scale. Due to the extreme operational conditions, it is difficult to continuously measure and precisely control this process. This results into limitations in the efficiency. This paper proposes a linear predictive control strategy for process states regulation. The solution relies on the use of an estimator and considers the process constraints to enhance productivity. Based on real industrial conditions, a tuning procedure is discussed and the proposed control approach is illustrated and tested in different simulation scenarios.