The objective of this study was to investigate the effects of different scanning temperatures and accessories on the near-infrared reflectance spectroscopy (NIRS) prediction of straw and to assess the potential of calibration transfer of the NIRS model for calorific value prediction. Three scanning temperatures (15, 25, and 35 degrees C) and two accessories were used, and the typical temperature of 25 degrees C was denoted as the master condition. Three correction and transfer methods were tested and evaluated: slope/bias, difference spectra, and local centering. Our results showed that the scanning temperature exhibited a more significant effect on the NIRS prediction of the calorific value for the straw samples than the accessory, and the prediction errors for 15 degrees C were higher than those for 35 degrees C. Both the slope/bias and local centering methods could improve the predictions, but not all of the results were acceptable. The best result was provided by the local centering method for transferring the calibrations between different scanning temperatures and accessories, and it was determined to be efficient and useful. These results are relevant for the broader application of using near-infrared spectroscopy for energy quality control of straw, not only for the raw material but also for the conversion processing products.