Forecasting Electricity Demand for Small Colombian Populations.
.
Pronóstico de la demanda de electricidad para pequeñas poblaciones colombianas.AbstractThe socioeconomic and cultural behavior of a population may be reflected in the consumption of electrical energy. Due to the foregoing, researchers and academics have developed models to predict electricity demand in the short, medium and long term. This paper presents an Artificial Neural Network (ANN) for the prediction of daily electricity demand (GWh) in small Colombian populations. The methodology proposed by Kaastra and Boyd is used for the construction, training and validation of the network and the development of the model in the statistical software SPSS. This paper conclude that the predicted values with models constructed with Artificial Neural... Ver más
2027-8101
2619-5232
7
2016-07-19
111
120
CUADERNO ACTIVA - 2016
info:eu-repo/semantics/openAccess
http://purl.org/coar/access_right/c_abf2
| Sumario: | Pronóstico de la demanda de electricidad para pequeñas poblaciones colombianas.AbstractThe socioeconomic and cultural behavior of a population may be reflected in the consumption of electrical energy. Due to the foregoing, researchers and academics have developed models to predict electricity demand in the short, medium and long term. This paper presents an Artificial Neural Network (ANN) for the prediction of daily electricity demand (GWh) in small Colombian populations. The methodology proposed by Kaastra and Boyd is used for the construction, training and validation of the network and the development of the model in the statistical software SPSS. This paper conclude that the predicted values with models constructed with Artificial Neural Networks (ANN) present a greater degree of approach with the real values of electricity demand (GWh). Also it indicates that the values obtained using models developed with other forecasting techniques (game theory, time series, simulation models, and others) allow to include variables and external factors that are difficult to quantify with simple equations.Keywords: Electricity demand, forecasting models, multi-layer perceptron, artificial neural networks, seasonal time series.Resumen El comportamiento socio económico y cultural de una población puede verse reflejado en el consumo de energía eléctrica. Debido a lo anterior, investigadores y académicos han desarrollado modelos que permitan pronosticar la demanda de la misma en el corto, mediano y largo plazo. Este trabajo presenta una red neuronal artificial (RNA) para el pronóstico de la demanda diaria de electricidad (GWh) en pequeñas poblaciones colombianas. Para la construcción, entrenamiento y validación de la red se empleó la metodología propuesta por Kaastra y Boyd en el software estadístico SPSS. Con el desarrollo de este trabajo se concluye que los valores pronosticados con modelos construidos con redes neuronales artificiales (RNA),
presentan un mayor grado de acercamiento a los valores reales de la demanda de electricidad (GWh),
que los valores obtenidos con modelos desarrollados con otras técnicas de pronóstico (teoría de juegos, series de tiempo, modelos de simulación, entre otros),
ya que permiten incluir variables y factores externos que son difíciles de cuantificar por medio de simples ecuaciones. Palabras clave: Demanda de electricidad, modelo de pronóstico, perceptron multicapa, redes neuronales artificiales, series de tiempo estacionales.
|
|---|---|
| ISSN: | 2027-8101 |