The batch machine scheduling of continuous caster with flexible jobs is first proposed based on its character. The difference with classical problem is first summarized. The preprocess of the problem based on rule is proposed, and then the mathematical model with minimizing setup costs is proposed. To improve the efficiency of local search, the particle swarm optimization (PSO) is introduced, and then the PSO combined with heuristic is embedded into the iterated local search (ILS) to solve the problem. The solution space is divided into many subspaces based on the charge information, and then the subspaces are integrated after local search. The maximum number of iteration is introduced as stopping condition. Then, the proposed hybrid algorithm and ILS are compared, and the changing of number of generation, learning factors, and the maximum number of iteration are also tested, respectively. Finally, the simulation results show that the proposed algorithm can solve the problem efficiently.