MODELING AND FORECASTING HOURLY SPOT ELECTRICITY PRICES WITH AUTOREGRESSIVE MODELS: AN ANALYSIS OF THE ITALIAN ELECTRICITY MARKET

This study examines Italy's electricity market dynamics by modeling and forecasting Italian hourly PUN (Prezzo Unico Nazionale) prices. Using about ten years of hourly data, the analysis focuses on in-sample model selection by assessing the performance of different autoregressive models augmented with several combinations of dummy variables. By emphasizing the role of stationarity, we find the AR14 model with day-of-the-week dummies to provide the most satisfactory in-sample performance. On an out-of-sample basis, the AR7 model with an additional coefficient at lag 14 appears to be the best trade-off between model complexity and forecasting accuracy. This work's insights into the dynamics of the Italian electricity market and the relative effectiveness of different forecasting models motivate future work investigating dynamical model selection and forecasting.