The concept of Cost-at-Risk (CaR) within ancillary services markets is introduced.
•
Several models to compute CaR are proposed and compared.
•
An application to the Italian case shows that the best performing model is the semi-parametric GAM-GARCH model.
Abstract
Measuring the risk exposure of TSOs on the dispatching market is a crucial task for the correct management of liberalized electricity markets. To fill a gap in the literature, the notion of Cost-at-Risk (CaR) is defined in the context of the dispatching market. Moreover, we propose a set of semi-parametric and non-parametric models for the estimation of the Cost at Risk (CaR) for the Italian TSO (Terna) and evaluate the corresponding out-of-sample forecasting performance. The empirical analysis relies on a rich hourly dataset provided by Terna, including several costs’ drivers. The results, in terms of 1-day and 30-day ahead predictions, suggest that the model with the globally best performance is the semi-parametric GAM-GARCH model.