Abstract: In this paper, the objective function of radioactive plume diffusion inverse problem was constructed by using the theory of maximum frequency value. And with genetic algorithm, a solving model of radioactive plume diffusion inverse problem was formed, and the corresponding program was written with Fortran language. In order to improve the search efficiency of the genetic algorithm, the selection, crossover and mutation operators of the genetic algorithm were optimized. Finally, inverse problem solving experiments were carried out with simulation and wind tunnel data. The results show that the model has high accuracy when the adaptability of diffusion model and monitoring data is high, but when the adaptability decreases, the accuracy of solution also decreases.