Abstract: The traditional cumulant method uses Jacobian matrix to express the linear relationship between input and output random variables and assumes network is three-phase balanced. However,due to the high R/X of real distribution networks,its Jacobian matrix is often morbid,and distribution network is significantly unbalanced due to the high permeation of distributed photovoltaics（PVs）and the uneven distribution of loads. To solve the problems above,a probabilistic power flow（PLF）algorithm of unbalanced active distribution network based on linearized load flow equations is presented. The algorithm linearizes three-phase back/forward sweep equations with high accuracy,synthesizes the cumulant method and Gram-Charlier expansion theory to solve PLF. The proposed algorithm is validated in terms of accuracy,computational efficiency and robustness by simulating on a real Australian unbalanced LV network during 24 hours through Matlab,with two different cases selected and compared to demonstrate PVs integration into distribution network improves the operation performance of network.