Modelización de la demanda de energía eléctrica: más allá de la normalidad

The main characteristic that differentiates electricity markets from other markets corresponds to the need to produce energy at the same time it is consumed, to such an extent that in real time the systems must maintain a perfect balance: at each moment the demand for electrical energy is equal to its generation. This characteristic prevents, for example, intertemporal arbitrage by those who carry out transactions in this market. In this regard, when modelling demand, it is common to find econometric analyzes that consider the assumption of normality; however, this assumption may ignore, a priori, an eventual presence of bias, kurtosis or higher order moments in this variable. In this paper, the Semi-Nonparametric approach (SNP) is studied to describe the demand for electricity in Colombia and the residuals of an ARIMA process. We propose the selection of probability density functions in terms of a finite Gram-Charlier expansion adjusted by the criterion of maximum likelihood. As a case study, the demand for electrical energy in the Colombian market is considered. As a result, it is found that the SNP type distribution achieves better adjustment than the normal distribution for some transformations of the electrical energy demand where it can be required more than four moments to represent this variable.

• Autores: Rendón, Juan F.; Trespalacios, Alfredo; Cortés, Lina M.; Villada, Hernán D.
• Fecha de publicación: 2019
• Institución: Universidad EAFIT
• Repositorio: Repositorio EAFIT
• Idioma: spa
• OAI Identifier: oai:repository.eafit.edu.co:10784/13601
• Acceso en línea: http://hdl.handle.net/10784/13601
• Palabra clave: Energy demand; Semi-nonparametric modeling; Energy market; Demanda de energía; Modelización semi-noparamétrica; Mercado de energía.
• License: Acceso abierto