Prediction techniques based on non-parametric methods. Application to energy series
Autoría
Fecha
Palabra(s) clave
Editorial
ISBN
Resumen
Trabajo de Máster en Economía, Finanzas y Computación. Director/Tutor: Dr. José Manuel Bravo Caro. The prediction of the electric energy demand is a problem of great importance for the electric industry, considering that, given the results of these predictions, di erent market agents take the most appropriate decisions. This is especially relevant for power companies that generate electric energy, because this way they're able to generate the amount needed in order to supply the market without exposing themselves to overproduction, which supposes a huge saving in economic costs. This paper proposes a method to predict the electric power demand using techniques based on non-parametric regression. An empirical study focusing on the UK electricity market is presented. The main objective of this research is to obtain a solid predictor for the electric power demand that essentially captures the intrinsic particularities of the electric power demand series and serves us to obtain future predictions. The methodology followed in order to obtain said predictor is based in using a training set in order to make and empirical adjustment of the predictor. The adjustment is obtained by selecting the set of hyperparameters that minimize the prediction error. The proposed predictor is validated and compared with others predictors through a validation set. The results support the goodness of the proposal made in this work.
Trabajo de Máster en Economía, Finanzas y Computación. Director/Tutor: Dr. José Manuel Bravo Caro. The prediction of the electric energy demand is a problem of great importance for the electric industry, considering that, given the results of these predictions, di erent market agents take the most appropriate decisions. This is especially relevant for power companies that generate electric energy, because this way they're able to generate the amount needed in order to supply the market without exposing themselves to overproduction, which supposes a huge saving in economic costs. This paper proposes a method to predict the electric power demand using techniques based on non-parametric regression. An empirical study focusing on the UK electricity market is presented. The main objective of this research is to obtain a solid predictor for the electric power demand that essentially captures the intrinsic particularities of the electric power demand series and serves us to obtain future predictions. The methodology followed in order to obtain said predictor is based in using a training set in order to make and empirical adjustment of the predictor. The adjustment is obtained by selecting the set of hyperparameters that minimize the prediction error. The proposed predictor is validated and compared with others predictors through a validation set. The results support the goodness of the proposal made in this work.