This paper presents a bottleneck identification based differential evolution algorithm for scheduling complex production lines. Operation priority sequences of bottleneck machine groups are determined by the differential evolution algorithm, while operation priority sequences of non-bottleneck machine groups are determined by predefined heuristic rules. The bottleneck identification method is presented based on the average flow time of all operations on each machine groups. The machine groups with longer average flow time are considered to be bottlenecks of manufacturing lines. In the differential evolution algorithm, mutation process is constructed by the operation priority differences and crossover process is constructed by operation priority swap. DE/best/1/bin and DE/rand/1/bin are used together to improve the efficiency while avoiding prematurity. Simulation results indicate the validity and efficiency of the algorithm presented in this paper.