Abstract-During the internal model control of multivariate multiple time delay system, process model is very important in the design of controller, but complicated mathematical model is often encountered in decoupled multivariable multiple time delay system and needs to be reduced order. So this paper proposes a low order identification structure and an optimization method for model approximation of the complex model containing the time delay and non-minimum phase parts. In this paper, with the adoption of Pade approximation, a suboptimal approximate algorithm is used for the model approximation and model identification. The integral square error index and the frequency-domain integral square error index, as well as the integral time absolute error index is used to evaluate the approximate model comprehensively. Simulation results show that using the proposed model identification structure and adopting the suboptimal approximate algorithm to deal with this kind of approximation, that can get an approximate model that well reveals the dynamic characteristics of system and has a high approximation precision.Index Terms-Model approximation, complicated model, suboptimal approximate algorithm, error performance index.