To study the universality of applying TL-moments method to flood frequency distribution analysis, we took the annual maximum stream flow series at six stations in Shenmu, Suide, Xinghe, Liujiahe, Huangling and Zhidan in northern Shaanxi Province as examples, and selected generalized Pareto distribution, generalized extreme value distribution and generalized Logistic distribution to evaluate the fitting effect of the 3 parameter estimation methods, namely L-moments method, maximum likelihood method and TL-moments method. The results show that, while Xinghe stations annual maximum stream flow series are better described with generalized extreme value distribution, generalized Pareto distribution fits the flow series of the other five stations more accurately. Compared with L-moments method and maximum likelihood method, TL-moments method shows a better fitting effect at each station, which is a better way to estimate the distribution parameters of flood frequency.