In order to reduce the production cost of electrolytic aluminum, an optimization extreme method is proposed based on genetic algorithm-multiple extreme learning machine (GA-MELM) to find the optimal production cell voltage and the corresponding production conditions. First, kernel principal component analysis method is used to determine the key parameters that affect aluminum electrolysis production, and the MELM model of electrolytic aluminum is established. Then the genetic algorithm is used to optimize the electrolytic aluminum channel voltage model and obtain the optimized cell voltage and corresponding operation parameters. The simulation experiment struck by the actual production data, the results show that the GA-MELM can forecast and optimize the electrolysis cell voltage, and then it can provide effective theoretical guidance to reduce energy consumption for practical production process.