During the last years, non-dispatchable generation has considerably increased its participation in electricity markets worldwide. As a result, we are seeing an increasingly common phenomenon of an extreme nature in which, in addition to price spikes, are occurring extremely low prices. Trying to predict these events is crucial for market agents in a competitive environment. This paper proposes a novel methodology to identify the appearance of extremely low electricity prices in a mediumterm scope. This methodology is applied to the Spanish market, which has experienced a large number of hours with null price in recent times. The procedure is based on the statistical identification of the process key drivers. A first empirical analysis shows that unexpected low prices are highly correlated with the collapse of thermal generation at certain moments. Logistic regression, decision trees, multilayer perceptrons and markov regime switching models are used. Proposed techniques are compared in terms of the results obtained. Examination of the model goodness of fit and interpretability is done by means of statistical and graphical tools. We give a comprehensive theoretical proof of the proposed methodology, which empirically demonstrates its effectiveness by achieving promising experimental results on real market price data set. This approach to predict extremely low prices can provide valuable information for market agents when they face the decision making and risk-management process.
Directores: Javier Reneses, Julián Barquín.
Autor: Antonio Bello
Lugar: Aula de Seminarios

Añadir a agenda electrónica: