1. Introduction
Intraday (ID) electricity markets allow market participants to adjust their market positions, taking as a starting point the day-ahead (DA) ones to better match their real-time generation or consumption levels. The need for flexibility in the intraday timeframe emerges from the increasing forecast accuracy of weather-dependent generation and load profiles closer to real time, operational schedule modifications, or additional plant outage information available in this time frame with respect to DA [
1,
2,
3]. ID markets can also be used to manage the technoeconomic constraints of the plants, particularly when certain formats of complex bids are not permitted in the DA markets [
3]. By using the ID markets to adjust the DA market positions, market participants can reduce the imbalance costs associated with the difference between their real-time output and market commitments.
Intraday markets play a key role in integrating high shares of renewable energy, especially wind and solar. Weber [
4] claims that liquidity in intraday markets is a major factor that reduces the societal costs of wind integration. However, as observed by Weber and Chaves-Ávila, liquidity is not uniformly defined throughout the literature [
4,
5]. The metrics used for measuring ID liquidity include traded volume and number of trades [
2,
6,
7], transaction costs [
4], bid–ask spreads and market depth [
8], and liquidity costs [
9]. Despite the differences in the approach, several studies indicate that the ID market liquidity in many European countries falls short of theoretical benchmarks [
3,
4]. The examination of ‘successful ID market designs’ [
2,
3,
4] reveals internal factors that drive the use of ID markets. An alternative to ID market participation is for agents to self-balance within their portfolios or to manage large imbalances bilaterally. In Spain, market participants are obliged to make adjustments through ID markets, increasing the trade volumes [
2,
10]. In Italy, the lack of complex DA bids requires the participants to use the ID market to manage their technical constraints [
10]. Additionally, the existence of multiple bidding zones in Italy advises the involvement of agents in ID markets for cross-zonal capacity allocation for portfolios with assets spread throughout the country. Furthermore, external factors such as the type of imbalance pricing applied (single or dual), the market share of agents, and the share of intermittent resources, among others, influence ID liquidity [
2,
11].
Assuming all the external factors remain equal, ID liquidity is closely tied to the market model. In Europe, the DA market is organized as a unified pan-European auction market, called the Single Day-Ahead Coupling (SDAC) market [
12]. However, despite the fact that there is also a Single Intraday Coupling (SIDC) market at the European level, in some countries, there are additional regional and national intraday markets coexisting alongside SIDC. The SIDC market is organized as a continuous trading market where the market participants can submit bids as soon as they have new information that can affect their operational schedule [
13]. In contrast, in those countries with additional regional and national intraday markets, these markets are structured as auction markets [
14]. The participants can submit their bids till an established gate closure time (GCT) [
15]. After the GCT, the market operator (MO) clears the market by minimizing the total procurement cost.
The selection of a continuous trading model for SIDC is primarily driven by historic reasons [
16]. At the time of SIDC implementation, Nordic countries had already obtained significant experience operating a continuous ID market. On the other hand, Spain, Italy, and Portugal had discrete auction-based ID markets in place. These countries opted for a hybrid design, maintaining the status-quo ID auction markets but complementing them with SIDC [
14].
The organization of the market, whether through an auction or continuous trading-based, can determine the overall efficiency of the intraday markets and condition the behavior of the market players. Hence, many academic works have studied the differences between auction-based and continuous-trading-based market structures. The main advantage of the auctions is related to their ability to accumulate offers and efficiently allocate the scarce transmission capacity [
10,
16,
17]. In contrast, continuous trading models allocate the cross-border transmission capacity on a first come first serve basis. As a result, the generators with the lowest bids or demand with the highest bids may not be allocated the capacity, reducing the allocational efficiency of the market [
16,
17]. Auction systems employ a merit-order-based clearing mechanism that prioritizes the most efficient trades and ensures the creation of transparent congestion rents. In a study comparing different ID markets in Europe, Weber [
4] states that market players may also develop a preference for auction designs due to their planning processes. Inflexible power plants may have technical restrictions that prevent instantaneous rescheduling in response to trades occurring in continuous intraday (c-ID) markets.
On the other side of the argument, in a renewable-dominated power grid, there is a value in increasing the speed at which new information is integrated into the market schedule [
6,
16]. The speed advantage serves as an incentive for the participants to invest in more accurate forecasting methods and improve their planning procedures. In this aspect, continuous trading systems are advantageous, as they incentivize the participants to use the new information as fast as possible [
16,
18]. Henriot [
19] argues that restricting trading to certain hours, as occurs in an auction design, can lead to market inefficiencies and opportunity losses. The evolution of the accuracy of forecasts varies across market players, making it extremely challenging for market designers to select optimal timings for ID auctions (ID-a). A similar argument is made by Scharff and Amelin in [
6], where the authors state that inflexible generators, such as thermal power plants, need to offer their flexibility much earlier than flexible generators like hydro plants.
However, if the Spanish auction market is taken as an example, one can argue that such inflexibilities can be accommodated by ID auction markets with a sufficient number of discrete auctions [
2]. The thermal generators typically participate in the first few auction sessions, giving them sufficient time for planning and operation. Also, it is worth highlighting that many studies comparing the two ID market models, discrete auction and continuous trading, often compare one model implemented in one country to the second model implemented in another country. While doing so, some factors external to the ID market, such as national regulation and the balancing market design, play a major role in determining the performance of the corresponding ID market. In this regard, the hybrid ID market structures in Spain, Portugal, and Italy may appear as excellent case studies to compare the performance of the two ID market designs, as the external factors remain the same within a country. However, closer examination reveals that even within a country, certain factors prevent the direct comparison of two different market designs.
The first is related to the differences in market architecture between the two models. Despite trading the same product within the same timeframe, there can be subtle nuances that distinguish one market type from another, such as the differences in the bidding formats and geographic scope. The second factor is related to the behavior of market participants. Profit-maximizing firms will strategize their bidding decisions to leverage any market arbitrage opportunities that may arise in the market. These arbitrage attempts may not necessarily indicate flaws in the market model but rather highlight potential challenges stemming from ill-designed market regulation.
Considering these factors, this paper aims to provide insights into the market-architecture-related and market-behavior-related factors that pose challenges when directly comparing ID auctions and continuous ID market models. We use a qualitative assessment centered on a review of the operational rules for both models to examine the main differences in their market architecture. Later, for the market behavior analysis, we conduct an empirical analysis of the Spanish market data to try to identify clear cases of strategic bidding. We further discuss the implications of implementing each of the two market designs, and both in parallel, as well as those of implementing efficient alternative market models.
To our knowledge, this is the first time a detailed comparative study of the Spanish intraday market architecture and the gaming opportunities corresponding to the two main ID market models has been performed in the literature. This is particularly relevant considering the future plans outlined by ACER (Agency for the Cooperation of Energy Regulators) to combine the continuous intraday trading model with complementary intraday auctions [
20]. By examining the Spanish market in detail, this study can offer valuable insights into the potential challenges associated with such a hybrid model, which will soon be applied in many European countries. Also, the comparison of intraday market models is extremely relevant for countries where intraday markets are still not in place or under development. It is especially important for countries that aim to integrate a high share of intermittent renewables into their energy systems. Hence, understanding the differences between the performances of the two markets is important in designing efficient, fair, and nondiscriminatory markets.
This paper also addresses the coordination between national markets (Spanish markets in this case) and higher-level regional markets (European coupled markets). The issues discussed in this paper provide valuable examples and insights for market coordination projects in other regions, such as the Central America Power Market, South African Power Pool, Nile basin initiative, and others [
21]. The lessons from the interaction of Spanish intraday auction markets with the European continuous intraday trading markets will provide insights into potential challenges and pitfalls in the coordination mechanisms.
Furthermore, the ongoing discussions focusing on the introduction of new products into the intraday markets, such as 15 min and 30 min products, will add another layer of complexity to these studies [
22,
23]. Exploring the implications of these developments will contribute to a better understanding of the intraday market dynamics and the participant behavior in each of them.
This paper is structured as follows.
Section 2 provides a brief background of the evolution of the Spanish intraday markets, particularly focusing on the developments that led to the adoption of a hybrid market model. In
Section 3, the methodology used for our study is described.
Section 4 presents a detailed discussion of the differences in the market architecture between the two ID models.
Section 5 provides the main results of the empirical analysis, along with a discussion of its significance.
Section 6 concludes the work and provides some policy recommendations.
2. The Evolution of Spanish Intraday Markets
The intraday auction market has been a part of the Spanish electricity market ever since the beginning of an open wholesale market in 1998 [
24]. The main motivation behind the implementation of the six-session ID auction market was to allow the market players to make adjustments to their schedules and to deal with unexpected events, such as the unavailability and outage of plants. In 2004, the Santiago International Agreement was signed by Spain and Portugal to create the Iberian Electricity Market (MIBEL) as a regional integration project [
25,
26]. The spot markets of MIBEL (day-ahead and intraday) became functional in July 2007. Meanwhile, at the European level, the development of pan-European day-ahead and intraday projects was under discussion. In 2014, SDAC was successfully implemented, shifting the focus to the SIDC project [
12].
However, unlike DA markets, ID markets were not developed according to a single market design (such as double-blind auctions in the case of DA). As mentioned in
Section 1, the use of ID continuous trading in continental Europe and the Nordics greatly drove the adoption of a continuous trading model for SIDC. Although the continuous trading model allows the fast integration of the latest available formation within the procurement schedule, it inherently suffers from inefficient cross-zonal capacity allocation. Identifying this potential inadequacy, the Capacity Allocation and Congestion Management Guidelines (CACM) directed the regulators to develop a single pricing methodology for efficiently pricing the cross-zonal capacities [
27].
Consequently, ACER decided to complement continuous ID trading with an auction-based cross-zonal capacity pricing methodology. This was based on the principle that each capacity recalculation in the intraday timeframe must be priced first in an intraday auction [
28]. As the adoption of such ID complementary auctions is a long-term process, a medium-term option was to have complementary regional ID auctions (CRIDA) [
14,
16]. According to this approach, relevant Nominated Electricity Market Operators (NEMOs) and Transmission System Operators (TSOs) jointly develop these markets. Once the complementary intraday auctions are adopted at the European level, the existing CRIDAs, such as those between Spain and Portugal and Italy and Greece, will gradually transition towards a pan-European auction market. Considering these factors, when ID continuous trading was introduced in June 2018, it did not replace the existing ID auction markets but rather created a hybrid market combining both ID auction and continuous trading.
In the case of Germany, the introduction of intraday auctions for 15 min products complementing the continuous trading markets was associated with a significant increase in traded volumes, as demonstrated in [
10]. Within just four months of implementation, the German ID auction volumes nearly matched the status quo continuous trading market volume. However, in the Iberian markets, the traded volume in continuous ID trading remained between 30% and 60% of the ID auction volumes, even one year after its implementation, as shown in
Figure 1 [
29]. This could be attributed to path dependencies, driving the participants to the more familiar ID auction market. Also, compared with the German case where a new product was introduced, the Iberian continuous intraday markets offered an already existing ID product (1-h blocks) [
13,
30]. However, it is tough to conclude on the factors contributing to the existing differences in the traded volume between the two market types without assessing the differences in market architecture between them. Hence, in
Section 4, we discuss the main differences between the market architecture of the two models.
3. Methodology
In this study, we aim to understand the factors contributing to the significant differences in the traded volumes between ID auctions and continuous trading, as depicted in
Figure 1, while pointing out the challenges that prevent a clear comparison from being carried out. We take a two-step approach to analyze the market architecture and participant behavior for both types of markets.
In the first step, we conduct an analysis of the market architecture to identify the factors driving the differences in the traded volume between the two market models. We specifically investigate four key aspects of market architecture: market timing, pricing schemes, available bidding formats, and geographic scope. The main source of data for this part of the study is the market regulation published by the Iberian market NEMO, OMIE and the Spanish TSO, Red Eléctrica de España (REE) [
29,
31]. By examining the regulations governing the two markets, we can understand the structural and operational factors affecting the traded volume. Additionally, to validate the assumptions, we supplement our analysis with empirical evidence from the published market data available in the transparency portals of OMIE and REE.
In the second step, we focus on analyzing the behavior of the market participants. We base our analysis on the hypothesis that the behavior of market players is notably driven by the monetary incentives associated with participating in each market. Specifically, we investigate how participants adapt their bidding strategies to the observed price differences between the ID auction and continuous markets. To demonstrate this point, we select a particular period in 2022 where a substantial price difference between ID auctions and the ID continuous market was observed. By examining the bidding patterns of agents in this period, we aim to provide evidence of market participants’ responses to price differences and their impact on traded volumes.
In June 2022, the Spanish government introduced a mechanism to adjust the method applied to compute the price in the wholesale markets, a measure widely known as the ‘Iberian exception’, to drive down the high electricity prices resulting from the increase in gas prices [
32]. As a result, in the following months, the wholesale market prices in Spain and Portugal remained much below the prices in continental Europe (see
Figure A1 in the
Appendix A). In addition to this, at the same time, there was a disruption in the supply of low-cost French nuclear energy due to the simultaneous occurrence of power plant maintenance and repair works affecting a large number of units [
33]. As a result of these, in the period of minimum coupling (16:00 18 August 2022), the price differential between Spain (99 EUR/MWh) and France (535.4 EUR/MWh) DA markets was as high as 436.4 EUR/MWh [
34]. This price differential was clearly reflected in the prices of intraday markets, as shown in
Figure 2. The prices in the Spanish continuous intraday market, associated with those in continental Europe, rose high above the DA and ID auction market prices in MIBEL. The correlation between these prices is tabulated in
Table 1.
From January to May and in December, the correlation between continuous intraday (c-ID) and day-ahead (DA) markets is consistently above 0.9. However, between June and November, these correlation values decrease and fall below 0.9, reaching their lowest point in July. A similar pattern can be observed in the correlation values between intraday auctions (ID-a) and continuous intraday (c-ID) markets. Based on these correlation values, we divide the selected period of study, January 2022–November 2022, into a convergence period (January 2022–May 2022) and a divergence period (June 2022–November 2022). We use the market data available in the transparency portals of OMIE [
29] and Red Eléctrica de España (REE) [
31] to analyze how the two markets interact with each other and how the behavior of market players changes during this period.
6. Conclusions
The development of intraday markets in Europe points towards a possible hybridization of the market design involving the combined use of auction and continuous trading markets. Spanish intraday markets are a good example of this. However, even within a country, it remains challenging to conclude how market players develop preferences for one market type over the other. This seems to be primarily due to the nuances of the specific market architecture and the profit-maximizing behavior of the market participants.
The qualitative assessment of the architecture of each market type reveals notable differences between both market models in terms of available bidding formats, pricing mechanisms, market timing, and geographic scope, even if they trade energy products within the same timeframe. The complex conditions available to be set on bids in the ID auction markets allow market players to represent their technoeconomic constraints within the bids, such as the minimum income or the ramping rate. Hence, thermal units like CCGT and coal ones have a preference for bidding in auction markets, especially for the first few sessions. Continuous intraday markets, on the other hand, allow players to set execution conditions affecting the matching of bids rather than explicitly specifying the corresponding unit’s technoeconomic constraints. Given that similarly complex conditions in day-ahead markets may be replaced with scalable complex orders, the available bid formats in ID auction markets can also be expected to undergo similar changes soon. This move may probably impact the participation of thermal units in the ID markets and may encourage additional market players (like those with thermal units) to consider the continuous ID markets as an alternative. Nevertheless, existing bidding formats in both intraday markets are not well-suited for nonconventional technologies like demand response and batteries. Introducing more sophisticated bidding formats, such as loop orders, will be instrumental in attracting liquidity across technological groups.
The choice between ID auction and continuous trading markets is also closely related to the pay-as-clear vs. pay-as-bid choice. The pay-as-bid pricing combined with the first come first serve principle of ID continuous trading markets can result in inefficient allocation of scarce transmission capacities. Additionally, in a pay-as-clear system, the participant has an incentive to bid their marginal cost, whereas, in a pay-as-bid system, the market profits it makes depend on its ability to snatch the most attractive orders in the Order Book as soon as they become available or predict the prices set for the orders yet to be made available. Hence, a participant in a pay-as-bid continuous trading system may require sophisticated market monitoring algorithms to snatch competitive bids in time. This may act as a barrier to the participation of small market players. The effect of algorithm trading is also relevant from a market timing point of view, as continuous trading markets can incentivize speed over cost efficiency.
In the Spanish case, the differences in the geographic scope and regulations affecting the interaction between ID auction and ID continuous trading markets led the participants of continuous trading markets to arbitrage between the markets of both types. We have demonstrated this arbitrage behavior by market participants making use of data available on the continuous trading market results for the year 2022. The price difference between the ID auction and ID continuous markets during the period June 2022–November 2022 prompted certain participants to strategically submit to the continuous market multiple orders of a small size precisely at 15:10, the market opening time for continuous intraday trading. While market participants may opt to submit a large number of small-sized bids as an optimal way to offer their production/consumption capacities, in this case, this behavior was part of a potential platform destabilization attempt. Some preliminary ideas on the exact motivation behind the need for risk platforms are provided here, though this behavior needs to be further investigated. In any case, this is clearly not a desired behavior. This behavior in the continuous ID markets could be identified in the Spanish market and is associated with the fact that the differences in prices between the ID markets of both types in the aforementioned period deviated substantially from the usual ones. However, in a market solely reliant on ID continuous trading, such behavior by agents may be mistakenly interpreted as increased liquidity.
Additionally, the rise in the use of automated trading platforms in continuous trading markets requires careful consideration. In the end, intraday markets should be rewarding investments in low-cost technologies and the application of accurate forecasting techniques rather than solely favoring investments in advanced trading platforms. As the Spanish case study revealed, the participants engaged in this strategic behavior did not individually possess large enough market shares to influence the price formation. However, when multiple small firms engage in the same strategy, the overall market performance may be affected. Auction market design also has its own challenges. The performance of an auction market is largely dependent on the number of intraday auction sessions and their timing. Many studies have discussed the complexity of determining the optimal number of sessions or their timing.
A popular alternative to both continuous trading and auction models is frequent batch auctions [
10,
51,
53]. This approach involves conducting discrete high-frequency auctions, allowing them to mimic the availability of a continuous trading market while increasing the tolerance for minor differences in the timing of bid submissions [
51]. Such a design also eliminates the need for multiple rules governing the interaction between the two types of markets. This becomes particularly important as countries like Spain, where only hourly products are currently offered in markets, plan to expand their product offerings in the future, making the management of the interactions among markets even more challenging. The cross-matching of products, which is currently considered in SIDC, can also be considered as an option in frequent batch auctions. Future studies should look further into alternate options, exploring their feasibility and comparing them with the existing ID market models.