The invention provides a simulated annealing algorithm-based electronic compass calibration method, and belongs to the technical fields of intelligent algorithms and nonlinear regression. Compared with traditional electronic compass calibration methods and some common optimization algorithms, a simulated annealing algorithm which is used as an algorithm for extensively searching the optimal solution in a solution space has a stronger global convergence superiority, and meets the global convergence characteristic of the solution at an enough high initial temperature, an enough low termination temperature, an enough slow cooling rate and an enough long thermal balance time, that is the simulated annealing algorithm can always find the global minimum point in the solution space under suitableinitial parameter conditions. In actual experiments, it is found that the nonlinear fitting solution of an electronic compass can be generated at a relative good time and space complexity under the proper initial parameters and a new solution generation function, and the accuracy reaches 10E-3. The method adopting the simulated annealing algorithm has a higher precision than the traditional calibration methods and a least square method.