The collection of fugitive fumes from aluminum electrolysis is a major technical issue that has hampered the green production of aluminum electrolysis for a long time. The collection process of fugitive fumes from aluminum electrolysis also is a complex flow process containing multifield coupling. There are various structural parameters that affect the collection efficiency in this procedure. In this paper, a multifield coupling simulation method based on a computational fluid dynamics discrete phase model was developed. To achieve the best collection efficiency, a multifactor interaction effect analysis was executed with the length <i>L<sub>W</sub></i> of the exhaust hood edge, the vertical distance <i>D<sub>H</sub></i> between the exhaust hood and the surface of the residual carbon block, and the horizontal distance <i>D<sub>L</sub></i> between the exhaust hood and the surface of the residual carbon block as the optimization parameters. The response surface methodology and multi-island genetic algorithm were used to carry out the optimization study of maximizing the collection efficiency. The results showed that for the flue gas and dust collection efficiency, the <i>D<sub>H</sub></i> contributes the most, followed by the <i>L<sub>W</sub></i>, and the <i>D<sub>L</sub></i> is the smallest; when the <i>L<sub>W</sub></i> is 748.99 mm, <i>D<sub>H</sub></i> is 224.06 mm, and <i>D<sub>L</sub></i> is 227.66 mm, it can effectively improve the flue gas and dust collection efficiency, in which the flue gas collection efficiency is increased by 6.01% and the dust collection efficiency is increased by 5.97%.