Economic analysis of continuous intraday market in Spain from an industrial consumer's perspective

Abstract

European energy policy in the last decades has focused on developing an electricity internal market. In this context, the XBID Project is created by several European power exchanges and transmission system operators, in order to develop an integrated Intraday Market, so that market participants’ bids in one country can be matched with bids from market participants in any other country, as long as transmission capacity is available. Intraday Market in the Iberian Electricity Market was previously structured in six centralized auctions, based on marginal pricing with uniform prices. OMIE, REE and REN proposed a hybrid intraday market design in which continuous intraday market sessions are combined with the existing auctions. The Iberian Hybrid Intraday market started operating on June 12th, 2018, for the energy delivered on June 13th, 2018. Advantages and disadvantages of this new market design have been discussed. It is argued that, as continuous trading prices vary from trade to trade, a unique and clear reference price is harder to identify. Auctions, on the other hand, provide a more reliable reference price. A hybrid market design as the one proposed by Iberian operators can help overcome this disadvantage of continuous trading, providing flexibility to agents’ as well as clear price signals resulting from intraday auctions. Furthermore, the hybrid mechanism provides a playing field for small market participants, who may not have the resources to implement a 24/7 trading desk which would be required in an all-continuous intraday market. Also, the Iberian hybrid market design only considers hourly products. However, distributed resources are able to vary consumption and generation in very short periods of time. Thus, it would be beneficial to introduce shorter products as distributed resources penetration increases, especially as gate closure moves closer to real-time, in line with other European intraday markets. However, Spanish System Operator REE has already stated the operational difficulties associated to moving gate-closure up to one hour before delivery, using only hourly products, so resistance to further changes in this line can be expected. Also, this new model reduces lead time before delivery to 1 hour, increasing market agent’s opportunities to correct their imbalances. In the case of an electro-intensive industrial consumer, reducing the extra-costs associated to imbalances could yield a competitive advantage. However, expected prices in intraday continuous market are needed in order to evaluate the incentive that industrial market agents have to correct their programs in the intraday continuous market. In order to predict intraday continuous market prices, predictive models based on linear regression and multilayer perceptron methods have been used, based on variables such as hourly electricity demand, wind production or traded volume in intraday auctions. Non-linear models trained (MLP) show a better performance than linear regression models as the prediction of continuous intraday electricity prices is probably a non-linear function. Obtained continuous intraday prices over 2017 are generally higher than day-ahead prices and, in some cases, even higher than downward imbalance prices. High continuous intraday prices offer a high incentive to correct positive imbalances but a reduced incentive to correct negative imbalances. As the industrial consumer under-study generally creates positive imbalances, it may be able to strongly reduce the cost associated to imbalances by trading in the continuous intraday market. In the scenario considered, it would have been able to reduce its imbalance cost by 113%. However, these results should be taken with caution. First, the ability of an industrial consumer to actually reduce its imbalance cost in this scenario depends on its productive process, as it is only a business opportunity if imbalances are generally positive and can be forecasted at least an hour before gate-closure. Second, results are subject to the ability of the model to accurately predict continuous intraday prices. As the training dataset used is small and only takes into account data from a specific time of the year, prediction models should be updated as the amount of data available increases, in order to improve their performance and avoid overfitting.