The invention relates to the field of optimal position planning of movable guide stars, and provides a real-time planning method for an optimal position of a movable guide star of an adaptive opticalsystem. The method comprises the steps that: on the basis of sufficient actual measurement atmospheric turbulence profile data accumulation, profile statistical distribution characteristics are obtained through clustering analysis; on the basis of analyzing statistical distribution characteristics, the positions of the guide stars corresponding to different types of profiles are obtained by utilizing a Monte Carlo simulation and optimization algorithm, and the mapping relationship between different turbulence profiles and the optimized positions of the movable guide stars is further obtained through adoption of a machine learning method. The optimal position of the corresponding movable guide star can be quickly calculated according to the actually measured turbulence data in practical application.