A Novel Algorithm for the Fast and Accurate Quantile Computation in Probabilistic Power Flow

摘要

AC power flow computations are applied to random inputs in the probabilistic power flow setting. We present a novel algorithm for computing quantiles of the resulting distributions. The approach is inspired by the principles of the Common Rank Approximation. It combines the high accuracy, that could be achieved through massive sampling with full nonlinear power flow computations, with the fast processing of large samples via the simplified linear power flow proxy. Compared to the common rank approximation, the accuracy of the computed quantiles is strongly improved, since ranking index mismatches are suppressed through a well chosen local regression. We demonstrate the efficiency of this approach with an illustrative toy example and a second test case based on the 118 bus IEEE test grid, with correlated input variables.