In biotechnological processes, the productivity and costs depend strongly on the control of the operating conditions. For this reason, sensors that allow the monitoring of variables of interest become quite important. 2D fluorescence spectroscopy is one promising option among those that are being applied for this purpose. In the present work, three methods were evaluated to select the best excitation/emission wavelength pairs of 2D fluorescence spectra to infer product, substrate and cellular concentrations throughout a fermentation using a multiple linear chemometric model: Exhaustive Search (ES), Stepwise Regression and Genetic Algorithm (GA). The Stepwise Regression presented unsatisfying results, while GA always led to good R² values in short computational times. However, for the proposed problem, the ES showed the best performance, finding the global optimum in a few minutes.