The optimal glucose concentration for the high-density culture of recombinant yeasts was obtained using dynamic simulation. An adaptive and predictive algoritilm complimented by the rule base was proposed for the control of the fed-batch fermentation process. The measurement of process variables has relatively long sampling period and relatively long time delay characteristics. As one of the solution on these problems, prediction techniques and rule bases were added to a classical recursive identification and control algorithm. Rule bases were used in the determination of control input considering the difference between the predicted value and the measured value. A mathelnatical model was used in the estimation and interpretation of the changes of state variables and parameters. Better performances were obtained by employing the control algorithm proposed in the present study compared to the conventional adaptive control method.