Artificial colony algorithm is a kind of simulation in the group of honey bees group of intelligent optimization algorithms. the traditional artificial colony algorithm has slow convergence speed, and Limits to local optimum .An improved algorithm is proposed, which introduces disturbance control frequency to guide employed bees searching for source, enhance algorithm local searching ability; introducing the adaptive dynamic mutation operator factor to improve the algorithm convergence speed; the strategy of choice for Boltzmann is adopted with dynamic adjustment of the search area of the algorithm, and enhancement the diversity of population, the algorithms of this paper are successfully applied to practical animal feed in proportion problems, the experimental results show that above algorithm is better than other compared algorithms in efficiency, the optimal solution quality and stability.