Abstract: In the laser induced breakdown spectroscopy (LIBS) detection process of coal particle flow, the unstable fluctuation of particle flow affects the stability of laser-material interaction, which has a negative effect on the quantitative analysis results. In this work, a quantitative analysis method for coal particle flow by LIBS based on the TrAdaBoost algorithm is proposed. The stable spectral data from the pressed pellet coal samples are migrated during calibration modeling to correct the fluctuation bias of the particle flow data. It can improve the accuracy of the measurement of the coal particle flow. By investigating the similarity of the spectral properties in the coal pellet and particle flow, a regression prediction model incorporating both data is developed. The results show that the coefficients of determination of the volatile matter content, calorific value, and ash content of coal quantification analysis models using TrAdaBoost reach 0.996, 0.934, and 0.937, respectively, which are significantly improved compared with the conventional prediction models for particle flow using the PLSR algorithm. The measurements demonstrate that the proposed transfer learning method has a promising prospect in improving particle flow quantitative analysis of LIBS.