This paper introduces and sets up some kinds of nonlinear growth models

the SI（Susceptible and infective） model and piecewise SI model

for forecasting clinical diagnose cumulative SARS（Severe acute respiratory syndrome） cases is Beijing. The 95% confidence interval of the time change point on piecewise SI model is made well which includes April 21

22 and 23. It means some control policies in Beijing at the end of this April 24 played an important role for anti-spreading of SARS

after change of increase rate for SARS cases is quite significant.