In view of that the operation characteristics of wind generation unit are nonlinear,coupled and with great inertia,a new system identification method based on sample modification and support vector machine (SVM) was put forward.By virtue of an application case,the proposed method of modified support vector machine was compared with the traditional SVM method and the back propagation (BP) neural network algorithm.The results indicate that,it is faster and more accurate than the initial one due to using modified samples.The study on the actual operation characteristics can provide reference to precise modeling and intelligent control of wind generation units.