Abstract: When a nuclear accident occurs, the local weather conditions play an important role in the diffusion of airborne radionuclides in the atmosphere. In this paper, a 10×8 multi-fan wind tunnel was used to simulate the wind field during the nuclear accident. Based on the improved BP neural network algorithm, the wind speed profile in the multi-fan wind tunnel was quickly and accurately predicted. The setting of fan speeds to create the wind speed profile was then obtained by applying the genetic algorithm. The results indicate that the proposed method can dramatically improve the efficiency of the diffusion experiments in a multi-fan wind tunnel and can provide the environment for the diffusion simulations under various wind speed conditions.