Abstract
This thesis comprises three papers, whose common denominator is the use of quantitative metrics and approaches to appraise electricity markets. The reliance of econometric techniques and modelling has been widely adopted in recent years, due to the dramatic changes affecting the power systems since the process of liberalization and unbundling in the late 90s. The need to assess system fundamentals in a dynamic and comprehensive way, by considering energy inputs and nearby systems alike, paved the way to an ever-improving set of methodologies. Fundamental approaches may be paired with statistical inference, as highlighted in recent literature, to deliver forecasting of quantities (e.g. prices) or to assess new policies and market equilibria. In this work we contribute in both ways: the first paper focuses on the assessment of the new capacity market in the Integrated Single Electricity Market in Ireland, the other two are about electricity prices forecasting in Northern Italy. The first paper assesses the outcomes of the Reliability Options scheme for Capacity Mechanism, as implemented in the Integrated Single Electricity Market in Ireland (I-SEM). We provide evidence of the effectiveness of this scheme in terms of system reliability, as well as understanding of the best practices by early adopters of a technically complex and sophisticated market solution. The originality of this contribution is given by the regional focus, as well as the inclusion of the time dimension, by modelling three different datasets (Full, Ante I-SEM and Post I-SEM), and by adding specifications of fundamental variables related to the electricity system. The second paper assesses the robustness and predictive power of a model comprising a complete set of variables affecting the electricity prices in Northern Italy. To the best of our knowledge, no other work focused on a specific regional dimension by factoring in all variables such as the import from bordering countries and market zones, the impact of non-programmable renewables, the load factor, the weather data and the prices of underlying commodities have. The econometric modelling of an ARMAX process for Northern Italian prices results in an underperformance compared to a standard ARMA, with potential implications regarding the reliance on timely adjustments on market information at trading floor level. The third paper builds up on the former, by expanding the analysis and performing a more precise and sophisticated model selection. The iterative procedure is based on the automatisation of the forecasting process, by allowing for an adaptive switching of form, both in linear and nonlinear specifications, ranging in the class of ARFIMA GARCH models. Predictions indicate a strong predictive power from forecast demand at any hour and from renewables mainly at peak hours, as well as a non-diminishing role of natural gas and C02 prices, and a high level of significance of electricity weighted inflows, especially during the morning hours.