Abstract: The high proportion of distributed generation（DG） access to distribution network presents high uncertainty and strong dynamic characteristic, which brings new challenges to operational reliability. Considering multiple uncertainties in distribution system operation, a flexible distribution network reliability assessment method based on Markov chain and improved minimum path method is proposed. Firstly, based on analyzing the operation and control mode of soft open point（SOP）, the impact of different types of SOP on the operational reliability of distribution network is studied. Then, considering the source and load uncertainty,mean-drift clustering is used to generate the source and load operating status, and obtain the multi-status transfer probabilities based on Markov chain. Secondly, based on the diversity of island formation after high proportion of DGs and SOPs are connected,Dijkstra algorithm is used to investigate the construction principles of single DG and multi-DG island. Considering the effects of aging, accident failure and weather factors, Bayesian network is used to obtain the time-varying failure rate of component, and the operational reliability model of the single DG and multi-DG island is established based on the improved minimum path method,which obtains the operational reliability of system. Finally, the accuracy and validity of the proposed method is verified by taking the example of improving the F4 feeder system.