Abstract: To deal with the fermentation process problem, the technique of parameter estimation is introduced to be combined with the model algorithm control (MAC), resulting in self-tuning MAC. It is then used for the control of penicillin fermentation, which is time-variant and non-linear usually. On the basis of impulse response model of CO_2 evolution rate(CER), the process is predicted and the optimal sugar feed rate is determined. Consequently, an optimal control is achieved with a good tracking of ideal CER change, thus leading to an optimal penicillin production. Simulation studies show that the control strategy developed is effective and feasible.