Predicciones de la demanda de la energía eléctrica con datos de la actual crisis económica y financiera. Aplicación a la región Canaria

Abstract

The economic development is the most influential factor on the power consumption of each country and each region, in long term estimation. In years of economic and financial crisis like the current one, a great variability of Gross Domestic Product (GDP) and Consumer Price Index (CPI) is observed. Particularly, CPI is sensitive to changes in the price of energy and the establishment of monetary policy. Therefore, the improvement of including CPI, in addition to GDP and population, as an explanatory variable to forecast the electricity consumption is investigated. For electricity companies it is important to have efficient prediction techniques to reduce uncertainty in the energy demand and obtain an optimal and realistic scheduling of the production of electricity. In pursuit of more objective conclusions, estimates are made using prediction methods of different nature, such as Multiple Linear Regression and Multiple Logarithmic Regression, which are classical statistical techniques, Support Vector Machine, which is a statistical learning technique, a Genetic Algorithm, which is an evolutionary computation techniques and an Artificial Neural Network, which is a machine learning technique. As a case study, the prediction of electricity demand in the Canary Islands is considered. It is of great interest for being an insulated electric system. The best prediction results are obtained with techniques which posses a greater capability to emulate nonlinear dependencies of the electricity demand in relation to population, GDP and CPI.