This paper proposes a probabilistic load flow method considering random branch outages as well as uncertainties of nodal power injections. Branch outages are simulated by fictitious power injections at the corresponding nodes. A unified procedure is given to deal with random branch outages, generating unit outages, and load uncertainties by their moments and cumulants. The variations of nodal voltages and line flows produced by normally and discretely distributed input variables are handled separately. The method proposed by Von Mises is employed to solve the discrete distribution part of each state and output variable. The final distribution of a desired variable is obtained by simply convoluting its continuous and discrete distribution part. Results of 24-bus IEEE Reliability Test System are analyzed and compared to those obtained by Monte Carlo simulation. A numerical test on a real power system shows the effectiveness of the proposed method.