Total reflection X-ray fluorescence spectrometry (TXRF) is a widely used, economical and rapid method for analysing trace elements. With the rapid development of machine automation in modern science and technology, sample preparation has become a key issue in TXRF quantitative analysis. In this paper, tea powder was used as the analysis object, the influence of dispersant, sample quantity and particle size on the sample preparation, repeatability and measurement accuracy of powder suspension samples in the process of TXRF analysis was discussed. The results showed that: (1) Global precision of the TXRF method was tested by analyzing five independent replicates of tea powder samples with particle size larger than 180 mesh. The stability of the instrument and the uncertainty in the sample preparation process were analyzed by error propagation. The results show that the uncertainty associated with the sample preparation step has a significant contribution (>60%) to the global precision of the obtained results regardless of the element and concentration range. Therefore sample preparation is the main source of analysis error; (2) By dispersing tea powder samples with a particle size range larger than 180 mesh into 1% Triton X-100 and deionized water, the effect on the dispersant was studied. Compared with the 1% Triton X-100 nonionic surfactant, deionized water has better repeatability. Its RSD was between 2.45%~11.64%, which is more suitable for dispersing tea powder samples with a particle size larger than 180 mesh, and could make the quantitative determination of medium and high atomic number elements more accurate; (3) The influence of the sample quantity was evaluated by adding powder samples of different masses in 5 mL deionized water. If the sample quantity is too low, the sample preparation repeatability will be poor, and if the sample quantity is too high, the sample film thickness will exceed the measured thickness of the X-ray, which may no longer be in the total reflection condition. 20 mg/5 mL is a suitable sample quantity for plant powder samples; (4) Through the measurement and analysis of powder samples in 7 types of particle size ranges, the influence of particle size on the measurement results was studied. The counts increased with the decrease of particle size when the particle size is less than 180 mesh; the precision increase with the decrease of particle size when the particle size is less than 200 mesh; the particle size has no significant effect on the accuracy except Mn; the uncertainty decreased rapidly in the range of 80~200 mesh, and less than 10% when the particle size more than 200 mesh. It is suggested that the particle size range should be between 200 and 300 mesh in the sample preparation process. The research results can provide an effective reference for the sample preparation method of plant powder samples.