Forecasting the hourly electricity spot price plays a crucial role for agents involved in energy day-ahead markets. However, traditional time series processes used for this issue model each hour separately not taking into account the intraday energy market microstructure information. In this paper, we appeal to a Functional Data Analysis (FDA) viewpoint that allows modeling and forecasting the intraday electricity spot price of the Colombian Electricity Market. Specifically, we use the Hyndman-Ullah-Shang method, which relies on a functional principal component decomposition of the nonparametric smoothed price curves, where the short-term forecasts are obtained by using the empirical functional principal components and the univariate time series forecasts of the corresponding estimated scores. Results show that one of the main advantages of this approach is that it allows to capture the underlying intraday common structural patterns shared by the daily spot price curves, and also behaves well for one-month-ahead price predictions compared with standard benchmarks. |
