The stochastic load flow (SLF) is extended to include non-Gaussian ‘long-term’ nodal probability-density-function (PDF) data by replacing each non-Gaussian PDF with a ‘Gaussian sum’ approximation. A series of SLFs (stochastic load flows) are then performed and the results recombined, with the correct weightings, to generate non-Gaussian PDF profiles for busbars and lines of interest. Generally less than half of the most likely convolution components need evaluating.P–ø,Q–Vdecomposition and nodal dependence is easily incorporated in the study and moment matching can be used to determine the ‘best’ lower order Gaussian sum approximation. Long-term network topological impedance changes can also be included in the proposed method.