The invention discloses a nuclear power plant large break accident uncertainty analysis method. The nuclear power plant large break accident uncertainty analysis method comprises selecting initial uncertainty input parameters and output parameters; performing primary sensitivity analysis on the initial uncertainty input parameters to obtain a primary analysis result; through a preset sensitivity analysis method, screening out important uncertainty input parameters from the initial uncertainty input parameters; performing secondary sensitivity analysis on the important uncertainty input parameters to obtain a secondary analysis result; determining whether the output parameters accord with normal distribution, if so, computing the permissible limit of the output parameters of 95% probability under a confidence coefficient of 95% according to the Owen factor statistical method, and if not, computing the permissible limit according to a Wilks formula statistical method. Compared with the prior art, the nuclear power plant large break accident uncertainty analysis method has the advantages of being concise and efficiency in analysis, avoiding deviation led to by response surface structuring and Monte Carlo sampling, having not limit to the number of initial uncertainty input parameters and enabling the computation-obtained permissible limit to be more conservative.