Review of Gaussian Mixture Model-Based Probabilistic Load Flow Calculations

Abstract

It is challenging to approximate multimodal distributions of probabilistic load flow (PLF) result variables stemming from discrete and non-standard continuous input random variables (RVs). The Gaussian mixture model (GMM) approximates the probability distribution of the above input RVs as a “K” weighted sum of Gaussian distributions. The expectation-maximization (EM) algorithm effectively estimates the mixture component parameters. Nevertheless, knowing the true number of components a priori is vital. In pursuing a pragmatic GMM-based PLF, several approaches have been suggested in the literature to determine the true number of mixture components and parameter initialization. This paper comprehensively reviews GMM-based PLF using EM. The criteria adopted in the literature for selecting the value of “K” and the initialization strategies are given special attention. This detailed review is expected to help novice readers in the area of GMM-based PLF.