Fourier transform mid-infrared (FT-MIR) spectroscopy is a fast, accurate, and practical technique for the quantitative analysis of compounds during ethanol fermentation. However, it is very temperature sensitive. This study aimed to develop a methodology for the real-time monitoring of fed-batch ethanol fermentations, taking into account the thermal fluctuations in the industrial process (in the range 30-35 degrees C). The use of FT-MIR combined with partial least squares (PLS) regression was evaluated as a tool to quantify sucrose, glucose, fructose, ethanol, and glycerol. The PLS models, based on fermentations carried out at constant temperatures in the range from 26 to 36 degrees C, presented high correlation coefficients (R-2) and low root-mean-square errors of cross-validation and prediction that were below 7% of the calibration ranges for all compounds. The good performance of the models demonstrated their suitability for accurate process monitoring, covering a wide range of thermal variations observed in industrial operations.