Abstract: Distributed generation possesses intermittence and fluctuation as well as its load has the characteristics of randomness. At present, the probabilistic statistical theory based study on the power grid planning, operation and control becomes current research hotspots. In recent years the distributed generation represented by wind power generation and photovoltaic (PV) generation is connected into flexible DC distribution network increasingly, so it is necessary to research the probabilistic load flow. By means of Nataf transform combining with Cholesky decomposition, the correlative processing of uncertain variables was performed, and a method to calculate the probabilistic load flow in flexible DC distribution network by the three-point estimate method, which took the correlation of input variables into account, was put forward. Taking the first domestic five-terminal flexible DC distribution demonstration project, namely the DC distribution center, as calculation example, a three-point estimate method based unified DC probabilistic power flow calculation model was established. The effectiveness and rationality of the proposed algorithm are verified by calculation results.