key: cord-0784327-0yszou8f
authors: Bento, P.M.R.; Mariano, S.J.P.S.; Calado, M.R.A.; Pombo, J.A.N.
title: Impacts of the COVID-19 pandemic on electric energy load and pricing in the Iberian electricity market
date: 2021-11-30
journal: Energy Reports
DOI: 10.1016/j.egyr.2021.06.058
sha: f1451686c7132d3a693a07e00ba1174fe316e964
doc_id: 784327
cord_uid: 0yszou8f

The ongoing COVID-19 pandemic has established itself has one of the biggest health crises facing humanity. Countries all around the world were forced to adopt unprecedented restrictive measures in order to halt the spread of the virus and safeguard public health. These measures have profoundly changed the way of life and severely affected practically all sectors of activity, with major demand and supply shocks, leading to one of the largest recessions in world history. An essential pillar to the proper functioning of modern societies is energy security, particularly electricity security, which guarantees a reliable and efficient supply of electricity. Energy distributors and utility companies remained operational during mandatory stay-at-home orders, to ensure an uninterrupted power supply. Given the relevant role of energy in society, this work will study the consequences of the economic shutdown on the Iberian electricity market, and discuss the timeline of events, the macroeconomic outlook, the financial status of the major electric utility companies (prior to being hit by the COVID-19 health pandemic), the changes in load profile, the generation mix and, finally, the electricity market spot prices.

When on January 4th, the World Health Organization (WHO) acknowledged a cluster of pneumonia cases in Wuhan, Hubei Province in the central China region, it was far-fetched to imagine that, by June, the COVID-19 pandemic would established itself as one of the biggest public health crises, certainly so-far in this century (Sarkis et al., 2020) . Since that initial report, the novel COVID-19 proved to be not only fast spreading but also particularly dangerous in high-risk groups. The absence of effective therapeutics or vaccines has led governments worldwide to adopt full or targeted lockdowns and social distancing policies to halt the spread of COVID-19 and consequently help ''flatten the curve''.

As we look back, by early March, it was becoming apparent that the outbreak was fully installed in Europe. For instance, Spain, where the first case for SARS-CoV-2 was reported on January 31st, was already reporting dozens of daily new infections by early March. Unsurprisingly, Spain enforced stay-at-home orders on March 14th, three days after the WHO declared COVID-19 a pandemic. Due to the sharp increase of cases, the lockdown * was further tightened to all non-essential workers for 2 weeks, starting on March 28th. In close coordination with its Portuguese neighbor, land border control was implemented from 11pm on March 16th, at nine authorized crossing points. These border restrictions remained in place until the end of June. Portugal had until then experienced an inferior level of infection and death rates compared to neighboring Spain, but declared a state of alert on March 13th, affecting schools, coffee shops, restaurants and nightclubs, and to some degree introduced stay-at-home directives. The extent of the confinement measures was broadened on March 18th, with the declaration of a state of emergency. The Portuguese government imposed additional travel restrictions between municipalities, during the Easter and Labor Day weekends. Finally, in the beginning of May, with a continuous drop in new infections and deaths, both countries began easing lockdown restrictions by phases and periodically reassessing public health indicators, in order to slowly change gears towards the recovery of their ''shattered'' economies.

To properly address the pandemic effects on the Iberian energy market, it is important to first identify the key dates when government restrictions were put in place in an effort to mitigate the spread. The relevant events and government responses, illustrated in Fig. 1 , ultimately induced significant supply and demand shocks and consequently cutbacks in the economic output.

Unfortunately, the ongoing consequences of this pandemic go far beyond the widespread disease and subsequent morbidity and death toll. As countries slowly started to reopen, the vast trail of economic devastation became ever clearer. In fact, there is a known positive correlation between health and economic growth. Even with smaller outbreaks, such as the SARS epidemic in 2003, GDP contractions were significant and not limited to the countries affected directly (McKibbin and Sidorenko, 2006) . The current outbreak proved that no country is an island in this integrated global economy. Even before March, the ultra-dependence upon China inputs led to a contraction in production for companies around the world, irrespective of size (McKibbin and Fernando, 2020) , with China's exports plunging close to 20% in the first two months of 2020.

Furthermore, the economic output, usually measured by gross domestic product (GDP), historically tends to be coupled with increases in electricity demand, as populations grow, and more services and commodities are produced and consumed. These findings are well documented in the literature. In the Portuguese case, electricity consumption is one of the major drivers for economic growth, with an asymmetric effect, meaning the negative change of electricity consumption appears to not exert the opposite influence (Shahbaz et al., 2017) . Nonetheless, especially since the mid-1990s and particularly for developed countries, i.e., countries with less energy-intensive industries and more service based economies, this link between economic output and electric energy consumption has weakened (Hirsh and Koomey, 2015) and is much more visible for short-term trends (E.I.A. (EIA) Energy Information Administration, 2013; Peterson, 2017) . Authors in Dogan and Aslan (2017) analyzed the relationship among CO2 emissions, real GDP, energy consumption and tourism for a number of EU and candidate countries over the period 1995-2011 (including Spain and Portugal) and revealed a unidirectional causality from economic growth to energy consumption and a bidirectional causality between CO2 emissions and energy consumption. In fact, early 2020 reports revealed declines of up to 30% in industrial production and energy consumption in some countries. In the specific case of the EU 27 & UK, the lockdowns led to a variation in CO2 emissions of −6.5% YoY in March. In turn, power sector emissions alone contributed with a decline of 3% YoY in the first quarter of 2020 (Liu et al., 2020) .

A deep dive into the literature revealed many related worksof varying depth and using different variables -quantifying the broad range of impacts during the first wave of the pandemic. One paper (Bahmanyar et al., 2020) briefly analyzed the demand variation index (DVI), on a weekly basis, for a set of EU countries that followed divergent containment measures. The results placed mainland Spain on one end of the spectrum, as the most affected country in terms of DVI, and Sweden on the opposite end, with (even) some growth in DVI.

Following this line of thought, authors in Ghiani et al. (2020) studied electricity consumption, market pricing and ancillary services during the COVID-19 pandemic in Italy, one of the most affected countries. They uncovered a reduction in consumption of up to 37%, and a decrease in wholesale energy prices of around 30% between late March and early April. As a result, CO2 emissions decreased, the energy share of renewables increased (above 40%), and there was a substantial increase in ancillary market costs. The prevalence of renewables during the coronavirus outbreak is also outlined in Norouzi et al. (2020) , where the authors make the argument for a more decentralized generation. In another instance (Qarnain et al., 2020) , industrial and commercial energy consumption in New York City decreased around 7% during the current crisis, with a shift towards domestic household consumption, which grew by 23% and 10%, in March and April, respectively. A similar shift occurred for the entire US, where after analyzing energy consumption from late March to June 7, the authors found that overall electricity demand and coal-fired electricity declined by less than 10%, and the increase in residential electricity offset a percentage of this decline (Gillingham et al., 2020) . In China, due to the drop in energy demand since the start of the outbreak, there was a considerable drop in the daily coal consumption of six main energy companies (Eroğlu, 2020) . Finally, authors in Santiago et al. (2021) thoroughly studied the changes in electricity demand in Spain during this pandemic, including case studies of a typical household and hospital loads, and contributed a very interesting analysis of policy implications and market behavior.

Considering these studies of the pandemic scenario, one would expect to find a short-term dip in electricity demand, electricity spot prices and CO2 emissions in the Iberian Peninsula. The present work aims to: provide a timeline of relevant government responses in the Iberian Peninsula; reflect upon its enduring economic harms and outline the concerns/expectations of the largest electricity retailers in these markets; and quantify the ongoing impacts of the COVID-19 pandemic on the Iberian electricity market. The core aim of this work is to analyze the data in terms of load profile and generation mix, with particular emphasis on the behavior of the wholesale electricity market and its indirect consequences. Hopefully this analysis will be a valid contribution for all market participants and policy makers, highlighting the relevance of different impacts, so they can better anticipate the consequences of te, this precarious situation is expected to worsen (Graff and Carley, 2020) . In response, governments, regulators, and retailers approved or extended a set of further pandemic ''waves'' or future virus outbreaks. The structure of this paper is as follows: Section 2 describes the economic harms from the outbreak, as well as the foreseeable economic environment; Section 3 presents and discusses the changes in the electric energy loads; Section 4 investigates the fluctuations in the Iberian wholesale market; and finally, Section 5 draws the main conclusions obtained from this study.

The extensive consequences of the COVID-19 pandemic can be seen in every sector of the economy, leading to one of the biggest recessions in history, only comparable to the major financial crisis of 2008 or even the great depression (Sharif et al., 2020) . In terms of the energy sector, we can see a plunge in total energy demand, driven mainly by a decline in commercial and industrial activities. In contrast, confinement measures increased domestic demand for energy due to a larger occupancy (Mastropietro et al., 2020) . If before the pandemic, energy insecurity (the uncertainty that a household can pay its energy bills) was already an issue faced by low-income households, now, with a skyrocketing unemployment rate, this precarious situation is expected to worsen (Graff and Carley, 2020) . In response, governments, regulators, and retailers approved or extended a set of important emergency measures to protect vulnerable consumers in both residential and non-residential segments, as acknowledged in Qarnain et al. (2020) , Graff and Carley (2020) . These included disconnection bans, energy bill deferrals and discounts, payment extension plans, energy assistance programs. For instance, in Spain, self-employed workers, as well as small and medium enterprises, were given the option to defer bills until confinement measures were lifted, and then pay in six monthly installments. In addition, a new beneficiary category for the electricity social tariff was introduced (Mastropietro et al., 2020) .

We found a similar adverse outlook in the Iberian Peninsula, with both countries already experiencing the dreadful impacts of the pandemic in the first quarter of 2020. In this period, data shows a year-on-year GDP growth rate of −2.3% and −5.2% in volume (de Estatística, 2020a; de Estadística, 2020b), for Portugal and Spain, respectively, thereby interrupting a positive run in macroeconomic indicators over the last few years, as shown in Table 1 (IMF, 2020). More importantly, the IMF estimated a very adverse outlook for both economies in the remainder of the year. This predicted outcome is partly related to the weight of the tourism sector in Spain and Portugal, the greatest among the 28 countries of the European Union in terms of GDP percentage (World Tourism Organization, 2018).

If we focus the analysis on the ''big'' electric power players in Iberia (those highlighted as offset ''slices'' in Fig. 2) , and consider the available first quarter results, the overall picture shows a pretty robust performance with relatively unaffected results. However, confinement measures only began in mid-March, meaning that the second quarter results could provide a more real picture of the COVID-19 pandemic's effect on the largest electricity retailers in these two countries.

Understanding the framework (financial status) and considerations (outlook) of these key companies is relevant. For instance, in Portugal, EDP -Energias de Portugal group, which in Portugal has over 3/4 of the liberalized electricity market share (as of May 2020) (ERSE, 2020a), as shown in Fig. 2a , plus close to 1M clients in the regulated market. In the first quarter, EDP reported a quarterly net profit of 146Me, an increase of 45% year-over-year (YoY). This is explained by a strong recovery of hydro production (up 87% in comparison with the 1Q19), a rebound of energy management results, and a financial deleverage. In contrast, EDP registered a 35% YoY decrease in average sell price in Iberia, from 65e/MWh to 42e/MWh (EDP -Energias de Portugal, 2020). Note that Fig. 2a shows the liberalized electricity market share, whereas Fig. 2b shows the electricity customer's share. The reason is that a separate entity of the EDP group operates exclusively in the Portuguese regulated electricity market, unlike the Spanish regulated market, so in this case the market customer's share is more representative.

The Iberdrola group experienced a 5.3% YoY increase in net profit, in the first quarter. As for its operation in the Spanish market, there was a YoY drop of 2.8% in distributed electricity. According to the company, this slump in demand can already be explained by the stay-at-home orders that started in mid-March. In contrast, the net energy production registered an increment of 3.2% YoY, with renewables, particularly hydroelectric output, increasing 15.8% and 43.8% YoY, respectively. Nonetheless, the revenues associated with renewable generation in Spain dropped 8.2% YoY due to a reduction in the pool price (Iberdrola, 2020) . ENDESA-Empresa Nacional de Electricidad -the largest electric utility company in Spain, as seen in Fig. 2b , which combined with Iberdrola represents about 70% of the customers (1Q 2020 across all segments and for both the liberalized and regulated markets) (CNMC, 2020) -followed the same trend. This retailer reported 133% YoY growth in net profits (reaching 844 Me), in the first quarter of 2020. The total energy gross output fell by 11%, with a decrease in demand of around 3%, mainly due to industry and services. In terms of generation, renewables, led by hydro production, recovered 44% and 52%, contrasting with the plunge in coal generation of over 90%. Regarding the average electricity wholesale prices in Spain, the company's prices dropped 37%, from 55e/MWh to 34.9e/MWh. The risks of the lockdown policies are also perceived in the company's reports, with a foreseeable collapse in demand, commodities and electricity pool prices (Endesa, 2020). In a related study about the Indian power system, authors in Madurai Elavarasan et al. (2020) recognized the importance of risk management in the power and energy sector when dealing with reduction and shifting electricity demand.

Electric load is significantly correlated with a heterogeneous set of parameters, including temperature recordings, rainfall levels, hour of the day, season of the year, massive social events, industrial and services constraints, and the number of holidays, among many others. Thus, it is important to consider the changes in these exogenous external variables, to properly understand the different patterns in the load time-series. Yet, these parameters of recurrent and seasonal impact are by nature quite distinct from those originated by the current health crisis, which caused a sudden demand shock with a subsequent impact on generation. Hence, in this section, with the aim of evaluating the impact of the COVID-19 pandemic on the electric load (system level loads) in the Iberian Electricity Market, we present a graphical and statistical analysis of the Spanish and Portuguese electric load during the first five months of the year between 2017 and 2020. The 2020 monthly data is then compared in a year-over-year base against 2019, as well as the previous 3-year average, which in many ways provides a fairer assessment of the recent load trend. Moreover, two common distribution shape measures are included, where skewness evaluates the distribution's symmetry and kurtosis its tail-heaviness. These measures are a very handy tool to understand how a time-series changes, complementing scale-related measures, and to quantify changes in the shape of well-known distributions.

The first case study considered the Spanish peninsular monthly load time-series (10-min intervals) for the past 4 years (2017-2020), from January 1st to April 30th, as provided by the Spanish transmission system operator Red Eléctrica de España (REE) and available in REE (2020). In this subsection, we present a general graphical analysis, complemented with details gathered from a more refined statistical analysis of the different periods, available in Appendix, namely in Table A .1.

Although the first two months of 2020 were still lived in ''normality'', we chose to include these months, to confirm that changes in the subsequent months were not the result of some preceding trend in the load data. While some changes were seen, these were well in line with the usual electric load seasonality and working patterns and were not yet effects directly linked to the COVID-19 outbreak, which at the time were still very circumscribed to Asian countries. The same is acknowledged by REE, in its February report. After factoring in the influence of multiple parameters, this report concluded that the demand for electricity remained unaltered with respect to the 2019 figures (REE, 2020a). In March, there was a downward trend in almost all descriptive statistics, explained by the imposition of restrictive measures to slow spread in Spain, restricting the country's economic activity and consequently affecting the load over the second half of March 2020 (REE, 2020b). April registered the most abrupt variations, coinciding with the most restrictive measures. Finally, in May, despite the very gradual reopening of the economy, the plunge in energy consumption was still quite severe (REE, 2020d), with a higher than 12% drop in average consumption.

In sum, after the second half of March, the load profile never reached the levels recorded over the previous three years. Nevertheless, and despite the drop in terms of mean, median and standard deviation, the typical daily pattern persisted, namely a morning ramp (albeit more gradual), followed by a mild decline during the day, reverted around dinner time when another peak occurred (again less sharply than previous years), followed finally by a steep decline in consumption during the night.

The peninsular Spanish load data was separated by weekdays and weekends to create Fig. 4 , displaying the aforementioned electric load fluctuations with monthly boxplots for weekdays and weekends, comparing the 2020 against 2017-2019.

There were generalized decreases in all three months and for all inter-quartiles ranges, with special emphasis for the months of April and May. In addition, and as anticipated, the decrease was more pronounced during weekdays, where the statistics revealed a more uniformly distributed load, since a large portion of the population was at home, with many economic activities strongly conditioned or even closed, namely the services (largest fraction) and industry sectors, hence, partially curtailing the common morning and early night peaks.

The second case study considered the monthly Portuguese load time-series (a quarter-hourly data set) for the past 4 years (2017-2020), from January 1st to April 30th, as provided by the Portuguese transmission system operator (REN) and available in REN (2020).

As for the first case study, the variations in the peninsular Portuguese load time-series during the COVID-19 pandemic were analyzed and compared with the aggregated three-year period between 2017 and 2019. In addition, a more detailed study including the monthly descriptive statistics is provided in Appendix, namely in Table A. 2.

The first two months of 2020 exhibited small variations in the time-series, in line with seasonal and specific factors, as described in the beginning of Section 3. Then, and as seen for the Spanish load, the direct consequences of the COVID-19 pandemic on the Portuguese load can be traced back to the last fortnight of March 2020, coinciding with the adoption of stay-at-home orders. In this regard, EDP acknowledged week eleven, i.e., from March 9th onwards, as the date after which there was a significant decline in electricity demand from the non-residential segments. The main drivers of this decline were small and medium-sized enterprises (special low voltage) and industry (medium and high voltage). Although there was a slight growth in the residential segment (normal low voltage), it was insufficient to counter the decline registered in the other segments mentioned (EDP -Energias de Portugal, 2020). This analysis is confirmed by a visual inspection of Fig. 5 illustrating the load profile of the Portuguese peninsular system for the months of March, April and May 2020, compared with the load profile (area) defined by the monthly distributions of the past 3 years, 2017-2019.

As in the Spanish case study, we can clearly see the impact of the confinement measures resulting in a substantial decline in electricity demand, although perceptually smaller when compared to Spain. Interestingly, May revealed the greatest effect, although some of the restrictions were being lifted, whereas in the Spanish case, April exhibited greatest variation. The Portuguese load was also split between weekdays and weekends, and the electric load fluctuations were analyzed with monthly boxplots for weekdays and weekends, comparing 2020 against the 2017-2019 Portuguese mainland load distributions (Fig. 6) . As in the first case study, the downward trend was more significant in the months of April and May, with the average value tumbling more than 10%, and more intensely on weekdays, since many of the economic activities were strongly conditioned or even closed, namely the services (largest share) and industry sectors. With the end of the COVID-19 restrictions, and without any substantial changes in the economic activity, self-consumption and demand response patterns, the system loads are expected to fully return to pre-pandemic levels. 

In the current framework, Electricity day-ahead markets are an integral part of the electrical energy production market and are where most of the consumed energy is traded. This liberalized and horizontal environment fosters competitiveness between market players in a clear contrast with the previous vertical and monopolistic structure. Nevertheless, in a more liberalized and competitive market environment, electricity traders are generally more exposed to extreme events including different levels of renewable generation output (mostly benefit from a feedin tariff), transmission constraints and demand disruptions (as during the current pandemic), among several others.

In the specific case of the Iberian electricity wholesale market (MIBEL), the regulatory structure impedes price gouging, i.e., there is a maximum bid price of 180.3e/MWh for the dayahead electricity market, and at the opposite end, the lower bound for bid prices is set to 0e/MWh, in both the day-ahead market and local intra-day market (Andrade et al., 2017; Juan Weik -News Argus Media, 2020) . In fact, even in the intraday continuous market, where market participants manage their energy imbalances up to one hour before delivery and have access to liquidity beyond the local level (from other European markets), negative prices are an extremely rare instance. In the past two years, average negative prices were recorded in only two occasions in the Spanish market. One occurred on December 19th (hour 5) of 2019, when the average price reached -0.66e/MWh and the second instance took place on April 5th (Sunday) of 2020, with negative prices recorded for eight hours and with average negative prices for three of them. On the same day, in the Portuguese Market, the number of hours with average negative prices was even higher, totaling 7 out of 24 h. Moreover, the lowest contract prices were -9.25e/MWh and -10.36e/MWh during hour ten in the Spanish and Portuguese markets, respectively (OMIE, 2020). Both these circumstances were attributed to a higher renewable generation, particularly wind output -which in turn led Spanish nuclear plants to reduce their output -and also to reduced demand, as that day fell within the most acute period of the pandemic in Europe, when the most restrictive confinement measures were in place (Juan Weik -News Argus Media, 2020).

These rare events in MIBEL, with negative prices seen in the intra-day market, are a direct consequence of the abundance of RES coupled with lower demand, and since in a foreseeable future the RES penetration is even higher, the regulator and policy makers need to address the current market distortion, where feed-in tariffs or other support mechanisms support the RES costs, therefore, detaching the dependence upon regular market pricing (where only the close to zero marginal cost are reflected) (Keay, 2016) . This behavior has been seen in some European markets, particularly in Germany where a large-share of RES was added to the mix in a few-years, thus, frequently leading to the occurrence of null or negative prices (even in the day-ahead market). Consequently, the financial health of thermal utilities is under significant stress, with smaller margins to pay fixed costs and reduced output times, threatening the survivability of even newer and cleaner gas plants. This scenario is forcing some local utilities to close and large utilities like RWE (Germany's largest energy power producer) are asking for a compensation (Eckert, 2019) . So, to sum-up, although its viable this path also presents risks and immediate challenges, and thus, it is critical to preserve a backbone of conventional thermal generation technologies, to provide the missing flexibility and energy security to a future system with high levels of RES. Moreover, this pandemic framework created additional ''stress'' for load forecasting, which then affects how generation is scheduled -with potentially greater imbalances levels addressed in the shorter-term intraday sessions -, leading to penalties and price adjustments (Santiago et al., 2021) .

In another note, the COVID-19 health crisis in Portugal has also affected the number of consumers switching to the liberalized electricity market during those three months, the lowest on record since 2012, especially in April. Nonetheless, it still represented a gain of 2.7% YoY, with an 88% share of all domestic consumers that account for 98.8% of all clients in the liberalized electricity market (ERSE, 2020c). However, by May 2020, with the country gradually reopening, the regulator reported a recovery in the number of costumers switching to the liberalized market, around 20k, which was a twofold increase (close to 11k) in comparison with April (ERSE, 2020a).

With this in mind, the impact of the COVID-19 pandemic on the Iberian Peninsula's day-ahead electricity market was quantified with the same approach used for the electric load, for both the Spanish and Portuguese markets.

The first case study considered the hourly Spanish day-ahead electricity prices for the past 4 years (2017-2020), from January 1st to April 30th, as provided by the electricity market operator (OMIE) (OMIE, 2020). Table 2 highlights the changes in the Spanish day-ahead electricity price time-series during the COVID-19 pandemic, indicating the same monthly descriptive statistics as in the previous section, for 2019 and 2020 separately, as well as the aggregated statistics for the three-year period between 2017 and 2019.

One can see right away that every month, including January and February when the consequences of the pandemic on demand were still negligible (as we saw above), revealed a substantial decrease in almost all statistics, in comparison with the 2017-2019 aggregated period and the year of 2019. For the first two months of the year, there was an average drop in prices of around 1/3, when compared with previous periods. While the minimum value was within the common range (because the abnormal minimum value in January 2018 distorted the % change vs 2017-2019), the maximum value stumbled 38.7% and 16.4% compared with the 2017-2019 aggregated period and 2019, respectively. Both months revealed a negatively skewed distribution, even reaching moderately to highly skewed distributions. There was a pronounced drop (>90%) in skewness, compared with both past periods, especially in January. There was a similar decline in kurtosis, which contrary to the load analysis, revealed higher kurtosis levels (leptokurtic distributions are common). These variations in the day-ahead electricity prices are in line with the observed dips in consumption and with an increased share of renewables in the generation mix. Until March 2020 (in a year-to-date comparison), the renewable share was up 13.1%, with a special emphasis for hydro and solar power, which gained 43.3% and 11%, respectively (IEA, 2020).

As we saw in the Introduction, by the second quarter of March most EU countries had already implemented confinement measures to slow the spread of the virus. Therefore, the severe impacts of these measures are reflected in the statistics for the market prices in March 2020, with considerable drops of 43.2%, 46.2% and 21.4% in the average, median and maximum values, respectively, on a YoY basis. A similar picture emerges when comparing the months of March between 2017 and 2019, (except the minimum value, which in March 2018 plunged towards 2.3e/MWh). The fall in skewness was so significant that the distribution shifted from being highly (left) skewed to an approximately symmetric distribution. Kurtosis followed a similar pattern, with drops of 46.6% and 65.9%, in comparison with the 2017-2019 aggregated period and 2019, respectively. Because of these dips, the distribution in March 2020 was mesokurtic instead of leptokurtic. The following month of April revealed a steep downhill trend, both in comparison with the 2017-2019 aggregated period and 2019, with plunges above 60% in terms of average, median and minimum values, as well as falls over 50% and 30% regarding the maximum and standard deviation values, respectively. These statistics are particularly revealing of the huge impacts of the comprehensive stay-at-home orders on the day-ahead electricity market (MIBEL). Furthermore, in April 2020, the standard deviation represented almost 36% of the average, meaning an increase of 17% YoY, and energy prices reached the abnormal minimum price of 1.95e/MWh. These results were expected since it was the first complete month under state of emergency policies, which led to a substantial decrease in consumption (16.5%), as can be seen in Table A .1.

Another factor contributing to this sizeable drop were the changes in the generation mix, with a renewables' share of almost 50% in April, as reiterated in REE (2020c). The Spanish monthly generation mix's from March, April and May of 2020, are compared against their 2019 counterparts, as illustrated in Fig. 8 . Highlighting the mentioned increased share of renewables, with a particular emphasis on Hydro and Solar (PV and Thermal), in contrast with Combined Cycle and Nuclear technologies that saw its output fall. This trend was also reported in Italy (Ghiani et al., 2020) , where there was also an increased share of renewables. The rapid evolution of the renewable power installed in the Spanish peninsula is a major contributing factor to achieve this outcome, and in March of 2020, REE reported the integration of 6528 MW of new renewable capacity, which meant that, for the first time, it exceeds the installed capacity of the other energy sources (REE, 2020f).

A negatively skewed distribution gave place to an approximately symmetric one, and a distribution with a longer and fatter tails gave place to one with shorter and thinner tails. Lastly, despite the gradual reopening of the economy in May, the market recovery compared to April was still tenuous, i.e., the median value of 18e/MWh in April 2020 increased slightly to 22.56e/MWh in May 2020 (+24%). However, when comparing with previous periods, there were significant drops in the average and maximum values, explained by a plunge in energy consumption above 12% when compared with both the 2017-2019 aggregated period and 2019 (Table A .1) and confirmed by REE in REE (2020d). The most clear-cut effect can be seen in the minimum recorded value, with a staggering decline in wholesale day-ahead prices, bottoming around 1e/MWh. Despite variations, the shape measures revealed a more stable picture, with the distribution remaining moderately left skewed and platykurtic. to contain the COVID-19 outbreak, clearly indicating how the percentage drops in prices, tied with a slight growth tendency of renewable penetration in 2020 (YoY basis), drove the market sharply lower. A recurrent feature during these months was that for large periods the maximum hourly prices did not even reach the minimum values observed in the previous three years. Nevertheless, with the gradual reopening of the economy, this accentuated decline in the average price can be reversed. This recovery will be crucial for many of the market players, already under stress from the current framework that compromises operating margins and can originate liquidity problems in the short to medium term, as recognized in Mastropietro et al. (2020) .

The peninsular Spanish day-ahead electricity price time-series, separated by weekdays and weekends, are presented, in Tables A.3 and A.4, respectively ( Appendix). As expected, the descriptive statistics paint a similar picture with sweeping decreases in practically all the months and statistics, with special emphasis for April and May. Contrary to the energy load, the plunge in prices on weekdays compared to weekends was not statistically relevant. Nevertheless, we can conclude that weekdays were more likely to have a higher average and median, as well as higher kurtosis levels, a trend that continued even during the pandemic.

The price fluctuations over the weekdays and weekends are highlighted in Fig. 9 , with monthly boxplots for the most impacted months, comparing monthly distributions in 2020 and 2017-2019 aggregated period.

At the end of 2019, the regulated electricity market in Spain still accounted for more than 1/3 of the universal supply points (CNMC, 2020), which emphasizes the need to also look at the price structures surrounding customers in this market. Interestingly, the sharp tumble in the retail electricity market prices (MIBEL) benefited more consumers in the regulated market (at least in short-term) -who contracted the PVPC tariff (Voluntary Price for the Small Consumer) -, compared to the majority of consumers in the liberalized market. This dynamic tariff for customers in the regulated market was introduced by the Spanish Government in 2014, replacing the old Last Resort Tariff (TUR), and is limited to contracts with active powers below 10 kW (low voltage), replacing the old Last Resort Tariff (TUR). The competitiveness of this pricing system during this pandemic crisis results from the hourly and daily price updates that reflect the daily evolution of the wholesale electricity market (translating the balance of supply and demand), marking a clear contrast with the old TUR where prices were normally set quarterly. That is, contracts on the liberalized market and a fringe of consumers with a fixed price regulated tariff -in which the prices are fixed for a certain period with a frequent yearly rate update, i.e., in a similar fashion to the old TUR -, penalized consumers during the current events. The other two electricity bill components, namely access fees and taxes, are the same for both markets. Also noteworthy regarding the pricing system, customers with PVPC type tariffs had their increased domestic consumption compensated or even offset by how the price is usually structured, i.e., the price is higher during the morning ramp and early night, which penalizes customers during these periods. However, since a significant portion of the population was confined at home, the domestic load profile was diluted during the day, almost in the form of a peak shaving. This unintended pattern marked an unusual shift to a more elastic price response, as acknowledged in Santiago et al. (2021) .

This benefit can be seen in Fig. 10 when comparing the default tariff of active energy invoicing price (PVPC) for the months of March to May 2020 versus 2019. The plot is quite illustrative of the lower 2020 prices, with an average decrease of 26.7% compared to 2019.

On a positive note, for March and April 2020, the REE reported the lowest figures regarding CO2 equivalent emissions coupled with electricity production since 2007, linked to the sharp thermal power generation decline. This fact is illustrated in Fig. 11 with the data analysis revealing an approximate 25% YoY reduction in power generation-related CO2 emissions. However, it is important to acknowledge that this crisis-induced CO2 reduction is not an isolated event in the Spanish Power System, since in the last 5 years CO2 equivalent emissions fell dramatically by 30 million tons (REE, 2020e). Worldwide, we can see a similar pattern, and the IEA recorded the largest relative and absolute fall on associated CO2 emissions. Besides, the reduced electricity demand, this fall is primarily explained by the rapid integration of renewable sources, which in 2020, saw the biggest annual expansion on record, representing a share of 29% of the global electricity generation (IEA, 2021).

The second case study considered the hourly Portuguese dayahead electricity prices for the past 4 years (2017-2020), from January 1st to April 30th, data is also provided by the electricity market operator (OMIE) (OMIE, 2020).

As in the previous case study, the fluctuations in Portugal in the first five months of 2020 were analyzed using the same set of monthly descriptive statistics for the years of 2019 and 2020 separately, as well as the aggregated statistics for the three-year period between 2017 and 2019 (see Table 3 ). As market participants (consumers and producers) operate in the same (Iberian) electricity whole-sale market spot (MIBEL), the clearing prices are composed by a set of transnational purchase and sale offers, and these take into account the available interconnection capacity (data from 2020 reveals a maximum exchange capacity around 3.5 GW between Portugal to Spain versus a maximum exchange around 3.8 GW in the opposite direction (REE, 2012) ). Consequently, if sufficient interconnection capacity is available for the flow of negotiated electric energy, the price of electricity will be equal in both countries at a specific time. However, this 1:1 relationship breaks down if the interconnection capacity is fully used in that period, leading to divergent spot prices in each country, i.e., market splitting. The higher price is attributed to the area/country that is importing electricity at the time when the congestion occurs. More often than not, when the market is separated, the prices in Portugal tend to be higher than those recorded in Spain. This pattern is typically explained by the greater annual average marginal production costs in Portugal compared to Spain, which translates into higher bidding prices (Borges, 2008) .

For instance, there was a significant difference in the month of February, when the minimum price was 77% lower in Spain than in Portugal. Moreover, compared with the 2017-2019 aggregated period, the minimum price in February grew by 13% in Portugal, but fell 36.3% in Spain. Comparing Tables 2 and 3, we note that the differences in both the recorded value and the respective % change are sporadic occurrences, resulting from market separation. Similarly and for the same reasons, the Portuguese day-ahead price evolution, illustrated in Fig. 12 , almost matches the outcomes exhibited in Fig. 7 ; and Fig. 13 presenting a comparison of the Portuguese day-ahead prices monthly distributions for weekdays and weekends also closely resembles Fig. 9 . The same inferences would be valid if we divided the day-ahead Portuguese electricity prices between weekdays and weekends occurrences and compared them with Tables A.3 and A.4 ( Appendix). Likewise, Fig. 14 is introduced to illustrate the changes in the Portuguese generation mix in a YoY basis, and as we saw in the Spanish case, the share of renewables has increased, especially hydro-energy, which in its ordinary regime grew its share by 54%, 64% and 90% in the months of March, April and May, respectively, meaning an accumulated YoY 88% growth. This coupled with a minor recovery of PV generation has compensated the wind generation drop. On the opposite side, coal generation has almost ceased during this period, a similar decline can also be seen for combined cycle production. Another consequence of this hydroelectric grow in a scenario with reduced load, was the increase of pumped storage. As a result of this (continued) environment with slim profitability of coal power plants, the EDP group anticipated the closure of coal plants in the Iberian Peninsula to the beginning of 2021, specifically in Sines (largest Portuguese thermoelectric utility) and unit 3 in Soto de Ribera, Spain EDP (2020). With this move, the company reinforced its commitment with the group's strategy of decarbonization.

Therefore, the COVID-19 pandemic has had the virtue to increase the percentage share of renewables in the generation mix, which already accounts for more than 50% of the generation mix for a couple of years. Revealing the viability of these sources to achieve the ambitious targets of the European Commission's ''Energy roadmap 2050''. Particularly, in a country where the total installed capacity of the two leading RES, hydropower (7193 MW) and wind power (5322 MW), already exceed by 2-fold the highest recorded peak loads, together with a growing adoption of solar PV farms (ANDRITZ, 0000; APREN, 2020a).

However, with the projected integration of higher levels of renewables, policy makers need to properly address the variability trait of these sources, i.e., days where RES far-exceeds the demand, contrasting with periods of low RES production, due to unfavorable weather conditions (reduced wind and hydroelectric production indexes), typically during the summer. Besides, a distorted regulatory and policy system, conventional thermal utilities are being ''squeezed'' to remain competitive market players. Overall, these are serious issues, in a framework with a limited storage (about 1 3 of the installed hydropower capacity is pumped storage) and import/export capacity, together with growing concerns about grid congestion problems and a compromised flexibility due to market distortions caused by renewable subsidies (Höckner et al., 2020) .

To avoid repeating the same detailed inferences drawn from the previous case study with a slightly different scale, this case study will focus on the market splitting, displayed in Table 4 , analyzing whether there was any notable disparity between the 2017-2019 period (average terms) and the year of 2019.

First, in January, the number of market splitting occurrences decreased more than 50%, with a greater expression over the weekends (decrease of approximately 97%). There was also an inversion in the pattern of price separation observed in previous years: in only 3 market splitting occurrences (out of 41) was the price for Portugal higher than the price for Spain, a nearly 100% fall compared with the average number of occurrences between 2017 and 2019 and the year of 2019. Second, in February, the downward trend in the number of market splitting occurrences not only continued but worsened (negative variation of all indicators), reaching a reduction of 60% compared to the aggregate period between 2017-2019 and a 71% YoY drop. Third, the downward trend in the number of market splitting occurrences continued in March, although less abruptly when compared to the previous month.

In all (41) market splitting occurrences the price for the Portuguese market was higher than its Spanish counterpart. This was also the second consecutive year in which the Spanish dayahead prices in March never exceeded the prices in Portugal. In addition, there was a clear inversion of occurrences by type of day, with nearly twice as many occurrences on the weekend than weekdays. Fourth, April 2020 was the most impacted month since the start of the pandemic, both in terms of consumption and prices. The monthly analysis of market separations exposed this impact, including a 65.7% plunge in average value of occurrences compared with the 2017-2019 period, with a special emphasis on weekends, as well as a 100% negative variation in market splitting occurrences. For the second consecutive month, the Spanish dayahead prices were higher than the Portuguese prices in all market splitting occurrences. This was due not only to the sharp decline in consumption in Iberia in April compared to previous years (>16% in Spain and >11% in Portugal), but also to changes in generation, since in this month there was a significant fall in wind production (REE, 2020c) in contrast to the continued recovery of hydro production. Fifth, the month of May showed mixed signals, since the number of market splitting occurrences did not vary significantly, decreasing when compared to 2019, but increasing when compared to the average of the last 3 years. In turn, there was a 200% growth in occurrences when day-ahead prices were higher in Spain than in Portugal, as the result of a very low basis for comparison (only 1 occurrence). Still, there was an inversion of the trend both in comparison with the average number of occurrences between 2017 and 2019 and the year of 2019, and the majority of market splitting occurrences took place over the weekend.

The Portuguese Energy Services Regulatory Authority (ERSE), following the sustained decline in MIBEL clearing prices, approved a decrease of 5e/MWh in the energy tariff applied to the regulated market. According to the regulator, estimates would translate into a reduction of around 3% for final consumers (ERSE, 2020b) . This decision was made after verifying that electricity prices in the wholesale market were ∼20% lower than the annual forecast considered for the purpose of fixing the energy tariff, for the year 2020, which was 58.45e/MWh. Lastly, mainland data from Portugal revealed a considerable reduction of CO2 emissions in the electricity sector, which in May recorded a value 48% lower on a YoY basis (111 gCO2/kWh), following a significant drop in electricity demand and an increased share of renewable generation. In cumulative values (Jan-May) and in a YoY comparison, this share grew 13.4%, with renewables accounting for 69.2% of the overall electricity generation in mainland Portugal in May (APREN, 2020b). Last but not least, it will be interesting to observe the price market behavior after the COVID-19 pandemic, in a scenario with higher load levels and after a particularly harsh period for conventional thermal utilities, which may lead to a price hiking in the pool market to compensate, at least partially, the financial losses of 2020. So, in one hand, this combination of factors highlighted the viability of a system dominated by RES but also the shortcomings in terms of market and final consumer pricing. 

The whole world is currently under severe stress from the COVID-19 pandemic, as the repercussions go far beyond the health crisis. Although most of the restrictions only came into full force in the second half of March, macroeconomic data indicate a severe economic slowdown already in the first quarter of the year, in contrast with the relatively unaffected performance reported by the major electric utility companies in the same quarter. Nevertheless, this adverse scenario will certainly be tenuous when compared with the data for the second quarter of 2020, as acknowledged by the biggest market players. While virtually all economy sectors were affected, energy-intensive businesses (industry and some services) were some of the most affected sectors. This was reflected in the load profile, with not only a relevant drop in average values but also a change in the load profile itself, with a less intense peak activity, and subsequent echoes in the energy markets, as they match the balance between supply and demand curves. These unparalleled challenges were certainly a first for market players, namely electric utilities operating in MIBEL. While in itself this situation would already be difficult under the old paradigm of centralized and vertical structures, it became particularly complex in the more packed, competitive, horizontal and liberalized framework, with potential liquidity problems down the road. This environment may lead utility companies to make provisions, since profit declines are anticipated.

The results in this study are in line with the effects observed in other countries in terms of electricity prices, generation mix and energy consumption. Average consumption decreased up to 12% and 17% in the months of April and May (under stateof-emergency rules) for the Portuguese and Spanish peninsular loads, respectively. There were significant decreases in standard deviation, as well as a sizeable drop in skewness, leading to approximately symmetric monthly distributions during the pandemic. The aftermath of the load plunge in both countries was visible not only in the day-ahead electricity market, but also on a higher scale, with the average wholesale energy prices decreasing well-above 50% and sometimes 60% in the months of April and May. The minimum day-ahead prices reached values under 2e/MWh in April and May, and there were even negative figures in the continuous energy market (which is a rare event in MIBEL).

The fact that renewables thrived during these months, with wind generation leading the charge along with a significant recovery of hydropower and PV generation in 2020, also played its part in the observed plunging prices, since this curtailed the more costlier thermoelectric generation (higher marginal costs). Additionally, with the fall in demand and the resulting generation mix, nonrenewable sources with reduced levels of operational flexibility, such as nuclear energy, were curtailed preferably, because a large share of renewables is non-dispatchable and favored by a feed-in regime that influences the pecking order in the market supply curve. This framework has exposed some weaknesses in the current market structure, i.e., the same RES support mechanisms can compromise the financial viability of the utilities that provide the required flexibility to cope with the variability of RES, and preserve energy security of the system. As a consequence, policy makers and regulators need to learn from this experience and restructure the market, in order to have more transparent and cost-driven price formation, where the important role of thermal units is acknowledged.

Due to these factors, CO2 emissions also tumbled accordingly. Lower demand was also reflected in a lower number of market splitting occurrences, i.e., a reduced use of the available Iberian interconnection capacity. Finally, last resort consumers in the regulated market, in particular under the Spanish PVPC regime, proved to be one of the most beneficial tariffs for consumers during this period, contrasting with a more conservative cut by the Portuguese regulator ERSE, of 5e/MWh. Therefore, a greater effort needs to be made in the liberalized electricity markets to better reflect/adjust market fluctuations as the main driver of spot prices. 

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. 

This work is funded by FCT/MCTES through national funds and when applicable co-funded EU funds under the project UIDB/50008/2020.

Please see the detailed load analysis and Appendix Tables A.1-A.4 below. Table A .1 complements the analysis in Section 3.1. Monthly descriptive statistics are presented for the years of 2019 and 2020 separately, as well as the aggregated statistics for the three-year period between 2017 and 2019. Starting with the month of January, statistics show relatively similar values, with a minor downward trend in practically all statistical indicators, with differences around or below 3%. However, the skewness for January 2020 did not follow the same trend, with substantial drops of 16.2% and 34.6% when considering the absolute value, in comparison with the 2017-2019 aggregated period and 2019, respectively. These drops indicate a more symmetric distribution, i.e., less negatively skewed data. The following month of February revealed a more accentuated downward trend when compared with the two previous periods, notably for the standard deviation and the maximum values. In part, this explains why skewness is again an exception, here increasing 72% in absolute value, compared to the 2017-2019 aggregated data. In the first two months of 2020, the negative variations recorded are the result of known seasonal trends. The first two months reveal a stable kurtosis, with a platykurtic distribution.

In March, there was a similar negative trend for almost all indicators, with significant drops of 8.2%, 7.0% and 9.5%, regarding the median, minimum maximum, respectively, in comparison to the 2017-2019 aggregate period. There was an inversion to distribution with minor right skew (approximately symmetric) and a 14.5% YoY increase in kurtosis. Then, the month of April exhibited the most abrupt variations and coincided with the period with stricter stay-at-home orders. In particular, there was a drop above 16% for both comparative periods, in both average and median, with a considerably smaller peak value. The staggering dip in skewness was not statistically significant, since the value was practically zero. Moreover, the standard deviation was reduced by a quarter, meaning a lower dispersion of values. Finally, in May, despite the gradual reopening of the economy, there was a sizable fall in energy consumption, with the average, median, standard deviation and minimum dropping above 12%. The shape measures revealed an approximately 100% decline (absolute value) in skewness, achieving an almost symmetric distribution, and a 20% rise in kurtosis, meaning marginally longer tails when compared with both the 2017-2019 aggregate period and 2019.

In addition, the peninsular Spanish load data was separated by weekdays and weekend and, as expected, the descriptive statistics indicated a similar pattern as described in Fig. 4 , with decreases in the months of March, April and May. This drop was more noticeable for weekdays, due to the shutdown of economic activities: the median value for the weekday load recorded drops of 19% and 15% during the months of April and May, respectively. Interestingly, there was an approximately symmetric distribution for weekends, in contrast to a moderately (left) skewed distribution for weekdays in the 2017-2019 aggregate period and 2019. As a clear consequence of the broad stay-at-home orders, the skewness during the weekdays of March, April and May 2020 were noticeably curtailed in terms of nominal value, i.e., consistently plunging more than 1/3, thereby altering the shape to an approximately symmetric distribution. Finally, there was a consistent platykurtic distribution for all considered periods, both on weekends and weekdays.

Following the same procedure, Table A .2 complements Section 3.2, for the second study, using the same set of monthly descriptive statistics, i.e., for the years of 2019 and 2020 separately, as well as the aggregated statistics for the three-year period between 2017 and 2019. First, despite the YoY variations, distributions for all the months were approximately symmetric, with an overall tendency to be skewed negatively, and with shorter and thinner tails, i.e., platykurtic distributions. The few instances with large variations in skewness (in terms of absolute value) were not statistically significant since the values approach zero (symmetric distribution). Comparing the months of January and February 2020 with the two previous periods (2017-2019 and 2019 alone), indicated relatively small changes in the different statistics, explained by seasonal characteristics. Data indicate a slim growth in the peninsular Portuguese load in January, which did not exceed approximately 2% for both the median and average values. In February 2020, the opposite pattern was seen, with the median and average falling approximately 3.5%.

Furthermore, there was a 11.3% and 7.2% drop compared to the February peak during the aggregated period of 2017-2019 and 2019, respectively. In March, there was modest drop in most statistics, particularly when compared with the 2017-2019 aggregated period, as well as in the peak and off-peak loads (more extreme values). However, when compared with the first case study, the load variations in Portugal, recorded in March, were less pronounced. In April, except a rise in kurtosis when compared to April 2019, all indicators fall considerably, with the standard deviation dropping roughly 20%, and the average and median values both dropping roughly 10%. Finally, although the month of May overlapped with the lifting of some confinement restrictions, it showed the 

Monthly Descriptive statistics analysis: Spanish Day-ahead Electricity Prices (Weekends)

aggregate period and 2019 YoY, with differences around and above 11% affecting the average, median, standard deviation, minimum and maximum, as well as skewness

Probabilistic price forecasting for day-ahead and intraday markets: Beyond the statistical model

Portugal -On the wind of innovation

Evolution of the installed capacity of the different sources of electricity generation in Portugal

The impact of different COVID-19 containment measures on electricity consumption in

MIBEL and the market splitting -5th International Conference on the European Electricity Market

Quarterly National Accounts of Spain: Main Aggregates

Quarterly national accounts

Exploring the relationship among CO2 emissions, real GDP, energy consumption and tourism in the EU and candidate countries: Evidence from panel models robust to heterogeneity and cross-sectional dependence

Does renewables pioneer Germany risk running out of power? -Reuters

EDP Anticipates closure of coal plants in Portugal and Spainedp.com

1Q20 Results Presentation

EIA) Energy Information Administration, 2013. U.S. economy and electricity demand growth are linked, but relationship is changing -Today in Energy -U.S. Energy Information Administration (EIA)

Endesa Q1 2020: Consolidated Results

Effects of Covid-19 outbreak on environment and renewable energy sector

Boletim do Mercado Liberalizado de Eletricidade -maio 2020

ERSE aprova descida no preço da tarifa de energia aplicada no mercado regulado do setor elétrico

3 milhões de clientes em março

Impact on electricity consumption and market pricing of energy and ancillary services during pandemic of COVID-19 in Italy

The short-run and long-run effects of Covid-19 on energy and the environment

COVID-19 assistance needs to target energy insecurity

Electricity consumption and economic growth: A new relationship with significant consequences?

Market distortions in flexibility markets caused by renewable subsidies -the case for side payments

Results First Quarter

Global energy review: CO2 emissions in 2020 -analysis -IEA

Report for selected countries and subjects. In: World Economic Outlook Database

Spain sees negative prices in intraday market

Oxford Institute for Energy Studies

COVID-19: Impact analysis and recommendations for power sector operation

Emergency measures to protect energy consumers during the Covid-19 pandemic: Global review and critical analysis

The economic impact of COVID-19

When pandemics impact economies and climate change: Exploring the impacts of COVID-19 on oil and electricity demand in China

Acceso aos arquivos -OMIE

Link between growth in economic activity and electricity use is changing around the world -Today in Energy -U.S. Energy Information Administration (EIA). In: Today in Energy

Review on government action plans to reduce energy consumption in buildings amid COVID-19 pandemic outbreak

Electricity Interconnections: a Step Forward Towards a Single Integrated European Energy Market

Generación y consumo -ESIOS electricidad · datos · transparencia

Demand for electricity in Spain falls 1.5% in February -Red Eléctrica de España

Demand for electrical energy in Spain falls 4.6% in March -Red Eléctrica de España

Demand for electricity in Spain falls 17.3% in April -Red Eléctrica de España

Demand for electricity in Spain falls 13.1% in May -Red Eléctrica de España

Emissions from the Spanish electricity system have fallen by 30 million tonnes in the last 5 years -Red Eléctrica de España

Renewable energy already exceeds the installed power capacity of other sources of energy on the Spanish Peninsula -Red eléctrica de España

Electricity demand during pandemic times: The case of the COVID-19 in Spain

A brave new world: Lessons from the COVID-19 pandemic for transitioning to sustainable supply and production

Production function with electricity consumption and policy implications in Portugal

COVID-19 pandemic, oil prices, stock market, geopolitical risk and policy uncertainty nexus in the US economy: Fresh evidence from the wavelet-based approach

World Tourism Organization (UNWTO), Madrid