Based on the artificial bee colony(ABC) algorithm, a double population co-learning(DPCL) algorithm is proposed. A population is divided into two populations according to their fitness. The individuals of each population are updated according to the given learning rules. With a test on ten benchmark functions, the proposed DPCL algorithm is proved to have significant improvement over canonical ABC and several other comparison algorithms. The DPCL algorithm is then employed for permutation flow-shop scheduling problem(PFSP). Twenty-one Reeves instances and forty Taillard instances are used. The results show that the DPCL algorithm can obtain better results than other algorithms, and is a competitive approach for PFSP.