The invention provides a spatial search method for multi-modal protein conformations. According to the spatial search method, the thoughts of a spatial locality principle and an assembly process are integrated based on a crowding differential evolution algorithm, and the protein conformations obtained through experiments are processed by adopting an energy minimization process. Due to the spatial locality principle, the convergence rate of the algorithm is increased, and the problems of local convergence and modal diversity in multi-modal optimization are effectively solved; in the assembly process, different crossover strategies are randomly selected, so that better segments in the conformations are prevented from being damaged by the algorithm, and the diversity of the multi-modal protein conformations is improved; and due to the energy minimization process, the complexity of space solution of the protein conformations is reduced, and the search space of feasible zones of the protein conformations is effectively shortened. According to the spatial search method, enkephalin is taken as an example, not only is the well-known most global stable structure of the enkephalin obtained, but also a series of high-quality local stable structure are obtained, so that the problem of multi-gene and multi-target paths of diseases which cannot be solved by the traditional single-target-directed single-modal research method is solved, and the requirements of computer-aided drug design on multi-modal protein structures at the present stage are met.