In reliability assessment of bulk power systems， sequential Monte-Carlo simulation（SMC） can provide not only the estimated expectation value of the reliability indices， but also the annual reliability index samples. This paper focuses on the convergence performance of SMC， analyses the probabilistic uncertainty relationship between simulation period and calculation precision， and deduces the confidence interval formulas for simulation period and variance coefficient， which achieves the tradeoff between calculation precision and cost. Moreover， based on the annual reliability index samples and the nonparametric kernel density estimation theory， the estimation of probability density distribution of reliability indices is achieved. The proposed approach overcomes the disadvantages of the traditional expectation value indices which only measure the risk of bulk power systems from the view of probabilistic average， and provides deeper understanding for bulk power system risk from the view of internal distribution laws and structural features of reliability indices.