Abstract: To improve the overall efficiency and reduce cost of food enterprises, this paper addreess the integrated production and distribution scheduling (IPDS) problem in a multi-factory and multi-customer scenario. A multi-agent simulation model is developed with the objectives of minimizing makespan and total cost, and an improved NSGA-II algorithm is proposed for solving the problem. According to the two-stage integration characteristics of the problem, a two-dimensional encoding structure is designed based on order allocation and processing sequence, while the simulation model interprets the behavior of each agent as a decoder. In the improved NGSA-II algorithm, a combination of random and heuristic rules is used to generate the initial population. Four kinds of crossover and mutation operators are introduced according to the characteristics of the problem to enhance the genetic diversity and expand the search space. Additionally, a neighborhood search operator based on the key factories is designed to improve the local search capability of the algorithm. Finally, a real-word case from a food enterprise is used to generate test instances for simulation experiments. Results show that the improved NSGA-II algorithm is superior to existing multi-objective optimization algorithms. The integration of simulation model and the algorithm enables better diversity in optimization objectives, demonstrating strong application potential in integrated scheduling of production and distribution.