关键词: 有色噪声, 子空间辨识, 状态空间模型, 连续搅拌反应釜（CSTR）

Abstract: In order to deal with non-stationary noise in the industrial production processes, an improved subspace identification method is proposed. The identification result is not efficient in presence of non-stationary noise through traditional subspace identification method. So the transformational system model form is used to eliminate the effect of non-stationary noise on system through improved subspace identification method. The accurate state-space model is obtained through identification using the transformational system model data. Practice has proved, the state-space model is more suitable for industrial processes. CSTR is a typical industrial production system and the subspace identification method is applied to the process simulation on a CSTR. Through comparisons of the predication error before and after improved, the effectiveness of the proposed method is showed.

Key words: non-stationary noise, subspace identification, state-space model, Continuous Stirred Tank Reactor（CSTR）