Computational time in indoor Ray Tracing(RT) would rise sharply if more reflections are considered. The upper limit value of reflection number is the key to the efficiency of indoor RT. However, the convergence of reflection in indoor RT has not been clearly discussed yet. A convergence analysis of RT based on statistically studying digital data of indoor scenario is presented aiming to obtain the appropriate upper limit value of reflection number to improve the efficiency of RT without losing accuracy. Convergence study focuses on the distribution of indoor scene, derives the probabilities and the power of different reflection paths, and finally finds out the convergence result of the power of receiver and the upper limit value of reflection number. The comparisons with RT simulation show that the convergence of this model is consistent with that of RT simulation under the same computational accuracy and this study is feasible to improve the efficiency of RT.