摘要： The rapid, nondestructive, cost-effective NIR measurement method was used for final molasses quality monitoring to determine fermentable sugar content to optimize ethanol yield. Molasses is stored in temperature-controlled tanks during the cane crushing and remelt seasons to ensure molasses quality and availability. However, there is variation in molasses temperature during storage. The impacts of temperature variation on molasses NIR spectra and calibration performance were studied. About one hundred molasses samples were collected for spectral profiling (400–2500 nm) at three different temperatures (25, 35 and 45 °C) using a FOSS NIR DS2500 spectrometer. A partial least squares regression (PLSR) model was developed using full cross-validation. The predictive models were developed using molasses spectra at 25, 35 and 45 °C and used to determine sucrose, glucose, fructose (fermentable sugars) concentrations in the molasses. External validation was achieved using thirty percent of calibration samples for each validation set, 25, 35, and 45 °C. Variation of the sample spectra was observed for the visible region and NIR region (1450 and 1970 nm), due to O–H bonding. The root means squared standard error of cross-validation obtained varied depending on sample temperature. Root means squared standard error of prediction results for external validation samples tended to increase with increasing temperature. Predicted values were not statistically different (p > 0.05) to reference values using different temperatures of models and validation. Calibration models including three temperature spectra showed potential of fermentable sugar analysis in molasses without temperature compensation.