In the report for Italy we observed a steep drop in energy load following the national lockdown. This drop severely impacted both the generation mix and the energy balance with neighbouring countries. Surely, Italy was not only the first but also one of the most severely COVID-hit European countries. Its strict confinement measures had a huge impact on the electricity grid and market, but how about other countries?
In this report we seek to understand how the quarantine period affected consumption and provision in 12 European countries: Italy, France, the UK, Germany, Spain, Switzerland, Austria, Belgium, the Netherlands, Portugal, Sweden and Norway.
The energy data is obtained from the Transparency Platform of the European Network of Transmission System Operators for Electricity (ENTSOE). For reference, we plot data from 2016 until 2020, although earlier years present more missing data (especially about the types of generation source). In general, missing data are imputed interpolating with the mean. We observed differences in data submissions between countries (for pumped storage for instance) which we unified. For earlier years, occasionally, we noticed inconsistent data (0 values or very low values) which we excluded from the analysis.
We consider the lockdown phase (main intervention phase) as the time interval between the country’s official start of the confinement (or the strongest intervention taken) and its end, in accordance with this table. As we could not find a clear ending date for some countries (either the confinement end or the strongest intervention end) we arbitrarily set the 5th of June (see Sweden, for which no lockdown was imposed, but some interventions were still on by such day).
Here is the table of the dates considered:
When comparing the national evolution of electricity load, all considered countries showed a drop in consumption from before to after the lockdown period. A load reduction also occurred in the previous years, due to seasonal changes. Still, the drop in 2020 was more significant for most countries and was probably due to the lockdown restrictions causing a slowdown of industrial and commercial activities.
For reference, we plotted daily consumption for 2020 (and previous years) from each country, alone and aggregated, together with lockdown dates. The blue line shows the actual consumption while the orange line (day-ahead) shows the energy bought on the day-ahead market.
We observe that most countries experienced a serious decrease in consumption that goes beyond seasonal drops and can be associated with their respective lockdowns. In some countries, such as Italy, we can observe that the consumption is raising again as industrial and commercial activities picking up with the loosening of confinement measures.
To understand the exceptionality for the year 2020, we compare the percentage change from before to after the lockdown for the years ranging from 2016 to 2020, for each country, calculated with the formula
where and are the mean load for the period before and after the lockdown, respectively, in year . We use the lockdown phase as described in the introduction. As for the pre-lockdown phase we consider the 8 weeks before the lockdown (roughly corresponding to the first or second week of the year. Percentage change enables the comparison between countries in different years and with different baseline consumptions, effectively putting smaller and larger countries on the same scale.
In 2020, most of the countries experienced a drop in average load that largely deviates from the changes in the previous years, considering that a systemic load decrease occurs with the transition from colder to warmer season in the intervals of time considered. Three countries display a different trend: Switzerland and Sweden, which showed no particular deviation; Norway, for which the load decreased less than previous years.
In order to give a more detailed account of the load variation that occurred during the COVID-19 emergency, we also performed a different kind of analysis. First, we addressed the daily load for the countries considered. In each country, we calculated daily running averages of one week (to remove weekends effects) and compared 2020 data with the 4 previous years averaged, for each day. For example, the 15th of January 2020 has been compared with 15th of January 2019, 2018, 2017 and 2016. This has been done in two ways: we compared the relative difference with the average of the four previous years, obtaining the percentage of variation observed in 2020 respect to the mean of the four years before; we compared the 2020 value with the mean and the standard deviation of the four previous years, obtaining a Z score, to take into the account the different levels of variability of different countries. In the following graph we show the results for these two comparisons.
For each day from the beginning of 2020, the percentage of variation in the electric load of 2020 with respect to the average of the 4 previous years. The size of the dots is proportional to the Z-score, so larger dots do represent variations that are particularly meaningful. For each country, the phase of most severe non-pharmaceutical intervention phase has been highlighted changing the color of the dots border in that phase.
We can observe that there is, already in normal condition, a certain amount of variation in the electric load. For the first months of the year, in fact, we observe variation in the range between -10% and +10%. In correspondence with the beginning of lockdows we do observe a meaningful fall for various countries, in some cases particularly deep (Italy, almost -30% at the peak) or particularly meaningful (Belgium, -20% at peak). Even though after the lockdowns the loads do increase, consumption for the affected countries does not seem to go back to normal levels, at least not for every country. It is also very interesting to notice that some countries (Denmark, Norway, Sweden and Switzerland) did not experience any decrease (Norway in particular had a meaningful increase). Mainly, governments from these countries decided not to adopt extreme restrictive measures (like lockdown) but chose different approaches.
Another way to look at the variation in the load is to analyse the hourly load pattern. We performed a similar analysis but by hourly aggregation rather than daily. For example, we compared 8 a.m. of 25th of january 2020 with 8 a.m. of the same day of the previous 4 years. Again, we compared the percentage of variation as before. In this case, instead of the Z-score, for practical reasons, we show the standard deviation of the average of the 4 previous years (always same day, same hour). The result is reported in the following graph.
For various nations (separately) the hourly electric load variation observed in 2020, respect to average of the same day and the same hour of the 4 previous years. The size and the colors of the dots do represent the entity of the variation. The error bars do represent the standard deviation observed for that day and that hour in the 4 previous years. Main intervention phase is highlighted in the timescale of the animation.
Also in this case we observe a certain amount of variations even before the emergency, which represent the normal dynamics of the loads. Still, for the countries where we previously observed a drop in the load, we see the same drop in the main intervention phase. Drops emerge mainly in working time (7-12 and 15-19) with national differences. Interestingly, for several nations the main peak in the drop is observed at 7 a.m. The meaning of this peak has still to be clarified, but it could be related to industries ramping up. Another interesting aspect is that for those countries where the increase is observed (in particular in Norway) such increase is located in night hours (with peaks at 4-5 a.m.). It is possible that these peaks are due to domestic warming so it could be ultimately due to low temperatures and bad weather, but this hypothesis has to be confirmed.
As national lockdowns heavily decreased the energy consumptions, we expect this to be reflected by significant drops in generation or net imports (import - exports). Depending on the type of energy mix and interconnectivity with neighbouring countries, we suppose that each country is impacted differently. In this section, we explore how the lockdown has impacted generation and energy balance in the 12 European countries cosidered, to inform us about how flexible a given generation mix is to load changes.
For reference, the following figure shows weekly rolling mean from the generation by source for the 12 countries and 5 years examined, together with country-specific lockdown date. The first plot shows the generation of 12 countries summed up.
As the overall electricity consumption in most countries dropped significantly, so did the generation. The generation mix played an important role: in general fossil coal and gas decreased while renewables seemed to have been used to their capacity. This means the percentage of renewables increased despite a decrease on onshore wind energy which was particularly high at the beginning of 2020. In particular:
Similarly to the load, we also computed the percentage difference pre-lockdown vs lockdown for all the years, as well as all the types of generation source.
We observe the following:
Of note, Italy significantly increased its generation output (2017 we observe erroneous data on fossil gas generation with zero values for over 2 months).
Next we look into the epidemic impact on the countries’ exchange balance. According to ENTSO-E’s definitions on how losses are supposed to be dealt with, we expect this to hold:
In this section, we question how these four factors were adjusted in each country to implement such modulation, compared to previous years.
In the following graph we see this is the case for most countries (Italy, France, Germany, Spain, Austria, Belgium, Portugal, Sweden, Norway). However, some countries do not add up (UK, Switzerland, Netherlands), which could be due to: losses not included in the data, missing data, or different methods of aggregating data. We will exclude them from this analysis.
We notice that some countries are net-importers (Italy, Spain) and some are net-exporters (France, Sweden). The others have a zero balance or they change from being net-importers to net-exporters (or vice versa) depending on their needs. Some changes should be seasonal, some may be due to COVID-19, which is what we are trying to sort out.
We now investigate the percentage difference of imports and exports from before to after the lockdown, as we previously did for load and generation.
Excluding the countries whose total energy does not add up from previous section (i.e. the UK, Switzerland and the Netherlands), we observe that:
On a side note, Portugal has moved from a net exporter to a net importing country over the past 5 years, making hard to compare it with other countries.
Taking the most affected month of April (April 1 to April 30) for all 12 countries, we could observe the following energy source shares:
In particular, generation reduction was generally achieved by curtailing fossil sources (gas, oil and coal) as well as, to a much lesser extent, nuclear. This allowed solar shares to increase, while using mostly hydro energy for balancing daily fluctuations. This production of energy is extremely delicate, as sources exploitation must be balanced in accordance with their specificities:
Trading with neighbouring countries is another way to manage load changes or to balance intermittent renewables. This is especially effective in Europe, where countries often have different energy source types from one another. Indeed, again for April 2020, we observe a significant increase of inter-country exchanges:
Therefore, the European grid and markets are also a crucial way to balance a future with high penetration of intermittent renewables.
Although load drops occur seasonally and countries are generally prepared to handle them, this pandemic has fast-forwarded times to a situation only expected in several years from 2020: an exceptionally high penetration of renewables compared to the load.
For countries with a high hydro capacity it is easier to reach high renewable penetration. Biomass, waste or geothermal are of limited availability, so fossil gas could also have a role in a more renewable future. Meanwhile, coal could be avoided due to its high carbon footprint. Nuclear reduces a country’s flexibility, potentially hampering renewables penetration and resilience to severe events.
In this report we have investigated energy data only. A deeper look into the market mechanism would be needed to understand how this lockdown has impacted prices (often becoming negative during this pandemic) and how resilient such mechanism is.