Estimating Density Functions for Probabilistic Power Flow Using Invertible Neural Networks

Abstract

Probabilistic power flow (PPF) is essential for quantifying operational uncertainty in modern power systems with high penetrations of renewable generation and flexible loads. Conventional PPF methods primarily rely on Monte Carlo (MC)- based power flow (PF) simulations or simplified approximations of voltage probability density functions. Although MC methods provide high accuracy, they incur substantial computational and data-storage costs, whereas simplified approximations often sacrifice accuracy. In this paper, we propose a novel PPF density approximation framework that avoids repeated PF simulations during inference and can, in principle, approximate complex voltage distributions without restrictive distributional assumptions. The core idea is to learn an explicit invertible mapping between stochastic power injections and system voltages using invertible neural networks (INNs). By combining this mapping with the change-of-variables theorem, the proposed framework directly evaluates voltage probability densities without repeatedly solving the PF equations. Extensive numerical studies demonstrate that the proposed framework achieves state-of-the-art performance both as an accurate PF surrogate and as an efficient PPF density estimator.