Short-term time series forecasting of the electricity consumption in Spain using an Evolutionary Algorithm and an Ensemble Method

Abstract

The ability to predict short-term electricity consumption provides several benefits, both
at the economic and environmental level. Indeed, it would allow for an efficient use of
resources in order to face the actual demand, reducing the costs associated with the
production as well as the emission of CO2.
The aim of this thesis is to propose two methodologies based on an Evolutionary
Algorithm for regression trees and an Ensemble Method by Stacking strategy in order to
tackle the short-term consumption forecasting problem.
The Ensemble Method uses a Stacking ensemble learning method, where the predictions
produced by three bases learning methods (Random forest, Artificial Neural
Networkand Evolutionary Algorithm for regression trees) are combined in a generalizer
(Gradient Boosted Method) in order to produce final predictions.
The two methods are applied on a dataset reporting the electric consumption in Spain
over more than nine years.

Date of Award	18 Jun 2018
Original language	English
Awarding Institution	University of Namur
Supervisor	Wim Vanhoof (Supervisor)