PRONÓSTICO DE DEMANDA DE ENERGÍA ELÉCTRICA USANDO PROCESOS GAUSSIANOS: UN ANÁLISIS COMPARATIVO

Abstract

Abstract—Load demand forecasting is an essential component for planning power systems, and it is an invaluable tool to grid operators or customers. Many methods have been proposed to provide reliable estimates of electric load demand, but few methods can address the problem of predicting energy demand from a probabilistic point of view. One of them is the Gaussian processes (GP) that considering an adequate covariance function are suitable tools to carry out this load forecasting task. In this article, we show how to use Gaussian processes to predict elec- trical energy demand. Additionally, we thoroughly test various covariance functions and provide a new one. The performance of the proposed methodology was tested on two real data sets, showing that GPs are competitive alternatives for short-term load demand forecasting compared to other state-of-the-art methods