A new method of solving traveling salesman problem is given by using simulated annealing algorithm. Based on the basic principle of simulated annealing algorithm, in view of the faults easily falling into local optimal solution when transformation only exchange two of the cities, puts forward an improved method by using reverse operation in the process of creating new solution. This makes the iterative process breakthrough circle of local optimum and jumps to another search space, which makes it more diversified and improves the local search ability of simulated annealing algorithm. And it is applied to solve TSP, which improves its ability of local search. The experiment results on several public test data show that the proposed approach is stable and feasible. Moreover, it has good performance in solving combinatorial optimization problems.