With the breakthrough of AlphaFold2, nearly all single-domain protein structures can be built at experimental resolution. However, accurate modelling of full-chain structures of multidomain proteins, particularly all relevant conformations for those with multiple states remain a challenging problem. In this study, we develop M-SADA, a method for assembling multiple conformational states of multidomain proteins through a population-based evolutionary algorithm, with multiple energy functions constructed by combining homologous and analogous templates with deep learning predicted inter-domain distance. On a development set containing 72 multidomain proteins with multiple conformational states, the performance of M-SADA is significantly better than that of AlphaFold2 on multiple conformational states modelling, where M-SADA models achieve TM-score >0.90 for two highly distinct conformational states on 29 proteins. Furthermore, M-SADA is tested on a development set containing 296 multidomain proteins with single conformational state, and results show that the average TM-score of M-SADA on the best models is 0.913, which is 5.2% higher than that of AlphaFold2 models (0.868).