With cloud computing, big data, artificial intelligence and other new information technologies going into space, the era of space intelligence is coming. The spacecraft is facing more frequent and multi-task tests in an unprecedented complex environment, which set higher requirements for self-management. The current challenge lies in how to further build an integrated system of the virtual and physical space for spacecraft. In this tendency, the integration of Digital twin(DT) with spacecraft is also in its evolutionary trend. DT provides a data and model-based systematic approach for operation and management in the entire service life of on-orbit spacecraft. In this context, as a basic space system entity, the method of constructing DT of spacecraft is explored in this paper. Firstly, the concept of Spacecraft Digital twin (SDT) is put forward and four stages of simulation development are discussed from the perspective of DT. Moreover, the conceptual structure of four-dimensional model is proposed to adapt spatial distribution. After that, the key problems of SDT are discussed, including data acquisition, system configuration, and data service model. Accordingly, the future intelligent applications of spacecraft based on DT in autonomous cognition, autonomous operation and maintenance, as well as cluster collaboration are presented respectively.