Control systems in modern automatic engineering are nonlinear , time - changed and indefinite . lt is difficult to model by traditional method , even sometime impossible . under these circumstances we should apply model identification to gain the approximate model of object for effective control , there are many models to be chosen , fuzzy model is one of them , it is put forward with the development of fuzzy control . fuzzy model has characteristics of general approximation and strong nonlinear , it is fit for describing complex , nonlinear systems . to avoid rules expansion when the number of input values are very big . in this paper we apply hierarchical fuzzy model to resolve this problem , we also illustrate it has general approximation to any nonlinear systems . genetic algorithm is a algorithm to help find the best parameters of process . lt has abilities of global optimizing and implicit parallel , it can be generally used for all applications . in our paper we use fuzzy model as predictive model and apply ga to identify fuzzy model ( including hierarchical fuzzy model ) , we made experiments to nonlinear predictive systems and got very good results . the paper contains chapters as below : chapter 1 preface 現代控制工程中的系統多表現為非線性、時變和不確定性,采用傳統的建模方法比較困難,或者根本無法實現,在這種情況下,要實現有效的控制,必須采用模型辨識的方法來獲取對象的近似模型,并加以控制,目前用于系統辨識的模型種類很多,模糊模型是其中的一種,它隨著模糊控制的發展而被人提出,模糊模型具有萬能逼近和強非線性的特點,比較適合于描述復雜非線性系統,為了解決模糊模型在輸入變量較多時規則數膨脹的問題,文中引入遞階型模糊模型,并引證這種結構的通用逼近特性。遺傳算法是模擬自然界生物進化“優勝劣汰”原理的一種參數尋優算法,它具有隱含并行性和全局最優化的能力,并且對尋優對象的要求比較低,在工程應用和科學研究中,得到了廣泛的應用,本文將遺傳算法引入模糊模型的辨識,取得了很好的效果。