Rapid determination of thermal conductivity is of great significance for realizing high-efficient and value-added utilization of crop straw. The feasibility of infrared photoacoustic spectroscopy coupled with chemometrics for developing the quantitative models of main crop straws’ thermal conductivity in China was investigated. The representative samples of wheat, corn and rice straws were initially acquired from North China, and the full-band models of single and mixed kinds of straws were then developed by using the partial least squares regression (PLSR) and Gaussian kernel support vector regression (RBF-SVR). By comparing the model effects of PLSR and RBF-SVR, it was found that the fullband RBF-SVR models of wheat stalk and rice straw had better performances, while the fullband PLSR models were more appropriate for corn and mixed straws. Moreover, based on the combination of abovementioned optimal modeling method and the ant colony algorithm, the new feature models of wheat, corn, rice and mixed straws showed better performances, which yielded determination coefficient of prediction set (R2p) of 0.77, 0.83, 0.96 and 0.79, root mean square error of prediction set (RMSEP) of 0.0078, 0.015, 0.0059 and 0.014W/(m·K), relative percent deviation of prediction set (RPD) of 2.81, 2.41, 7.39 and 2.15, respectively. Results showed that FTIR-photoacoustic spectroscopy coupled with applicable chemometrics had good potential for rapid quantitative analysis of main crop straws’ thermal conductivity in China.