To solve the problems of low production efficiency and high production energy consumption of cell manufacturing systems under multiple flexible routes, a multi-objective model was established that considered the completion time, total production energy consumption, and the number of intercell transportations from the perspective of optimizing the number of intercell transportations. According to the characteristics of the model, a variable neighborhood non-dominated sorting genetic algorithm was proposed for solving this model. Based on the combination of global and random solution generation methods, individuals with fewer intercell transportations were selected as the initial solutions. A decoding method based on intercell transportation was designed, and three variable neighborhood operators were integrated to reduce the impact of transportation and idle time on processing. Then, the crossover and mutation non-dominated sorting methods were improved to achieve a high-quality evolutionary process. Finally, the rationality of the improved algorithm was demonstrated through simulation examples. In the case study section, it was shown that balancing the number of intercell transportations helps coordinate and optimize the two objectives of completion time and total energy consumption, further demonstrating the superiority of the established model.