Abstract: The uncertainty evaluation is a key step of best estimate safety analysis for nuclear power plant. The origins of the uncertainty and the uncertainty evaluation methods were described. The methods could be categorized into statistical type and deterministic type, and the general procedure of statistical evaluation method was summarized. Various uncertainty evaluation methods were analyzed and compared from the aspect of computational cost and accuracy. The analysis result shows that the method coupling the nonparametric sampling with a sophisticated high-fidelity thermal-hydraulic model may be the best choice for the uncertainty evaluation in best estimate analysis at present. Under the condition that the “95/95 criteria” is satisfied, such a method is easy to implement and the cost is relatively low.