In view of the fact that the traditional simulated annealing algorithm has a long running time, easily been trapped in local optimum in solving the traveling salesman problem, especially the defects become more obvious with the increase of the problem scale, the inner loop and annealing mechanism of the traditional algorithm are improved, so that the search strength of the inner loop can be adjusted adaptively according to the change of temperature. At the same time, a fluctuation temperature control mechanism is proposed, so that the algorithm can realize multiple heating processes while maintaining the general trend of decreasing in temperature amplitude, which enhances the solution effect and shortens the solution time. It is verified by a large number of examples from the TSPLIB database.