Day Ahead Electricity Price Forecasting in Coupled Markets: An Application in the Italian Market

Introduction

In the last two decades, the liberalization and deregulation of the electricity markets have completely reshaped the landscape of the traditionally monopolistic and government-controlled power sectors. European Union (EU) has passed several laws trying to achieve a single internal electricity market through the implementation of a target model for the operation of the EU electricity markets (Daskalou, 2019). At the same time, electricity displays a set of characteristics that are uncommon to other markets; it is non-storable and requires a constant balance between consumption and production, in order to ensure the stability of the power system (Carrion et al., 2007). Moreover, the electricity consumption is influenced by the weather conditions and the intensity of business of everyday activities, which, on their turn, result in prices that exhibit seasonality at a daily, weekly and annual basis (Weron, 2014). As a result, accurate electricity price forecasting, especially for the day ahead, is one of the biggest challenges for all market participants (generators, power suppliers, investors, traders) in order to maximize their profitability.

The day ahead electricity price forecasting problem is not a new one, and several statistical, intelligence-based and other techniques have been proposed in literature to address it, at all required levels that is daily, weekly and annual (see Weron (2014) and Nowotarski and Weron (2018) for relevant comprehensive reviews). What is, however, still largely lacking, are methods that take into account the forthcoming market integration. Given that complete integration has not achieved yet, both relevant data and studies of forecasting considering integration are still rather sparse. Thus, most relevant studies focus mainly in methodological issues in an effort to develop and/or refine forecasting models and techniques, while little effort (Daskalou, 2019) has been placed in investigating the effects that integration may have, and whether it is indeed useful and/or necessary to include in the forecasting model of a country inputs from the electricity markets of the countries with which it is already or will eventually be integrated with.

It is the aim of this chapter to focus on and study the integration effects in day ahead electricity price forecasting in an effort to contribute in filling the aforementioned gap. To this end, an Artificial Neural Network (ANN) is developed, and several, with respect to inputs, forecasting scenarios are used considering the Italian electricity market. Italy has been chosen for this study because its electricity market is already coupled (Gestore Mercati Energetici, n.d.) with some of its neighboring countries, thus data are available. This was not, however, the only reason. Another equally important reason is that the electricity system and therefore the market of Italy is divided into zones (Gestore Mercati Energetici, n.d.), the behavior of which resembles the behavior of coupled markets. More specifically, the zones correspond to geographical areas, which, for reasons of system security, present limited electricity exchange between them. Based on its own supply and demand, an electricity price is determined daily for each of these zones, and a weighted, by zonal-purchased electricity quantity, average is then used to determine the national single price (PUN-Prezzo Unico Nazionale). This means that more insight can be gained on the integration effects, by studying not only the behavior of the PUN price, but the behavior of the zonal prices as well.