This paper presents a unifying methodology for optimization of biotechnological processes, namely optimal adaptive control which combines the advantages of both the optimal control and the adaptive control approaches. As an example, the design of a substrate feeding rate controller for a class of biotechnological processes in stirred tank reactors characterized by a decoupling between biomass growth and product formation is considered. More specifically, the most common case is considered of a process with monotonic specific growth rate and non-monotonic specific production rate as functions of substrate concentration. The main contribution is to illustrate how the insight, obtained by preliminary optimal control studies, leads to the design of easy-to-implement adaptive controllers. The controllers derived in this way combine a nearly optimal performance with good robustness properties against modeling uncertainties and process disturbances. Since they can be considered model-independent, they may be very helpful also in solving the model discrimination problem, which often occurs during biotechnological process modeling. To illustrate the method and the results obtained, simulation results are given for the penicillin G fed-batch fermentation process.