Flight approach sequencing is a typical NP-hard optimization problem, and its main difficulty lies in how to efficiently solve the optimal scheduling scheme under multi-constraint conditions. Aiming at the shortcomings of the traditional simulated annealing algorithm with insufficient global search ability, a two-stage neighborhood solution strategy was proposed, the first stage determined the flight landing sequence, and the second stage carried out the runway use time scheduling to meet the constraints of flight scheduling. By adjusting the maximum position offset limit that adapted to the temperature, the switching, inverting and inverting operators were introduced to further enhance the neighborhood search ability of the algorithm. Through the test of multiple sets of peak hour operation data(the time to solve the exact solution is greater than 300 s), the proposed improved algorithm can effectively improve the global search ability, and the simulation results have significant advantages in solution efficiency and solution quality: compared with the first-come, first-served method, the total flight delay cost is reduced by more than 65%. Compared with genetic algorithm, the delay cost is reduced by more than 34%. Compared with the integer programming model solver, which limits the calculation time of 300 s, the objective function solution error is less than 1%, and the solution time can be shortened to less than 1 s. When the delay cost per unit time of different models changes with the delay time, the proposed improved algorithm gives priority to the delay cost to increase the order of the larger models to reduce the total time deviation cost. Therefore, the algorithm can significantly reduce flight delays and provide fast and efficient decision support for runway scheduling during peak hours.