Abstract: In sequential simulation for distribution system reliability assessment, randomicity of sampling result in uncertainty of solution, thus the solving process must be expressed by probability knowledge, the convergence criterion for reliability indices is also different from which of common numerical iteration algorithm. Directed against solving uncertainty problems for sequential simulation of distribution system, choosing independent events was studied, independent generalized events were constituted. By analyzing confidence interval of general average value on the condition of a given confidence probability, a conclusion was reached that confidence interval length is related to absolute error of sample mean, and then a practical convergence criterion that expressed with absolute error was proposed, and the unequal relational expression about confidence interval length and a given calculation precision was presented under the given confidence probability. The method is reasonable and effective by RTBS verification.