Abstract: With the integration of large-scale wind power generation, photovoltaic generation and electric vehicles, the operational state of the smart distribution network becomes gradually uncertain. By analyzing the probability distribution characteristics of the uncertain variables, this paper proposes a probabilistic load flow (PLF) method for the distribution network based on the Gauss-quadrature-based point estimate method (GPEM). The Gauss-Hermite quadrature and the normal transformation are used to select the estimated nodes and the corresponding weights of the input variables. In the multiple random variable response function, the statistical moments of the output variables are estimated through polynomial approximation. Taking the unbalanced parameters of the distribution network into consideration, a three-phase power flow model is constructed. Then the whole process of the three-phase probabilistic load flow based on GPEM is directly given. Simulation results of a modified IEEE-33 bus system show that this method has a higher precision compared with the traditional point estimate method, and that it can solve the non-normal distribution and correlation of new sources and loads.