Power Load Probability Density Forecasting Method Based on Gaussian Mixture Model and Variational Message Passing

Abstract

The mathematic model of power load forecasting can be expressed as a regression problem. The applicationof technologies such as renewable energy generation and power demand side management has madepower load characteristics more complicated, and a single distribution model has been difficult tocompletely represent the changing characteristics of power load. With the rise of concepts suchas energy interconnection and regional integrated energy system, smart grids put forward higherrequirements for the prediction of load in the grids, based on regression algorithms such as artificial intelligence,nonlinear regressions, and correlation vector machines. Good performance has been achieved,but the model generally relies on the training of historical data and cannot provide analyticalsolutions and intuitive explanations. In response to the above problems, this paper uses a Gaussianmixture model to represent the power datadistribution, based on the variational message passingalgorithm. Data analysis realizes the corresponding parameter estimation of the mixed Gaussiandistribution, and uses the estimated data to perform posterior estimation and prediction without relyingon training historical data. This article elaborates on the theoretica