A large number of subsystems and lots of parameters make it difficult to build a Statistical Energy Analysis (SEA) model for a complex system. A reduced statistical energy analysis with substitutive subsystems is proposed in this paper. Some subsystems in complex structures can be combined into one subsystem or are implicitly substituted by additional terms in reduced SEA model. In such a SEA model only the key subsystems are retained, the total number of subsystems is reduced and the procedure of SEA modelling would be simple. It is found that new terms are added to original parameters such as the coupling loss factors, damping loss factors. The additional terms related with the substituted subsystems representing theirs coupling properties, which could be measured by experiments. Numerical simulations are performed to verify the proposed method.