Complex electromechanical product assembly lines often adopt modular design to achieve parallel production of multiple types of products. During the assembly process of each module， there are complex situations such as batch processing of some process groups and parallel processing of multiple machines， which brings huge challenges to the formulation of workshop scheduling plans. This paper takes the flexible assembly workshop of complex electromechanical products as the research object， considers the parallel batch processing of multiple machines under the multi-variety small batch production mode， and establishes a multi-objective integer programming model with the goals of minimizing the makespan and the delay time， and maximizing the average equipment utilization. According to the characteristics of the problem， a fast non-dominated sorting genetic algorithm solution model embedded with heuristic rules is designed. Experiments show that the scheduling strategy and model proposed in this paper can provide a reasonable batch plan for batch production units. Compared with the traditional non-dominated sorting genetic algorithm， the improved algorithm shortens the maximum completion time by 4.25% on average， reduces the total delay time by 32.30% on average， and increases the average equipment utilization by 0.69%.