With the wide application of microgrid system in island area, the reliability evaluation of microgrid system becomes more complex and different. Firstly, K-means clustering algorithm is available for collecting typical scenes of microgrid operation with big data, and the “supply-load” which reflects the time correlation is treated as the extracted feature quantity. Secondly, Latin hypercube sampling method which put with scene probability distribution is proposed, and in this process the sampling frequency is identical to the scene occurrence probability. Finally, according to the state time series of each cell in each scenario, the reliability index of each scenario is computed. Then the multiple reliability index of microgrid can be obtained by using the full probability formula. A microgrid example is used to be analyzed, and its results make clear that this modified model method proposed can quickly and correctly evaluate the reliability of the microgrid.