What happens to the electricity system when 67 million French people “reste chez eux” (stay at home)? Like in many European countries, the spread of COVID-19 through France has been quick, and aggressive. On March 16th, the French President Emmanuel Macron declared a "sanitary war" ordering 67 million French people to stay at home in order to combat the exponential spread of the virus. The confinement started the following day, March 17th at noon. Anybody leaving their home during the confinement would be required to carry a signed statement declaring for which, of the few permissible reasons, they were leaving their home. This change in behaviour overnight begs the question: What happens to a nation’s electricity system when its entire population is abruptly confined to their homes with only one day’s notice?
Electricity demand is the first change that comes to mind. It is clear that a country whose citizens are confined to their homes will not use electricity in the same way as when they are going about their daily business. Leaving March 17th aside, we can start by looking at the first full day of lockdown (March 18th). So how was the demand different from last year? Figure 1 answers that question (you should note that March 18th was a weekday in both this year, a Wednesday, and last year, a Monday).
Looking at Figure 1, a significant drop in demand from 2019 to 2020 is striking, almost 15 GW at peak load! However, if you know something about electricity demand, you probably know that this graph would be better placed in a chapter of the classic “How to Lie with Statistics” than an article for energy engineers on electricity demand. The problem is that a quarantine is not the only difference between March 18th 2019, and March 18th 2020. In fact, there are many things that are different between these two dates. However, to explain the difference in electricity demand we want to focus on the differences between what are called regressors, explanatory variables or features, depending on who you ask.
Sticking to the term regressors, a regressor is a variable that has a significant correlation with electricity demand. This means that there is some sort of relationship between the two variables, thus knowing the value of one of these variables gives you information about the other variable. The greater the relationship between the variables, the more information about the other you get from knowing one - this is called correlation. In this case, we can assume that demand is positively correlated to temperature, which means an increase in one of these variables will cause an increase in the other. However, correlation alone does not indicate a relationship. In general, you should conduct a sanity check on the two variables and ask yourself, is it reasonable for these variables to have a relationship? This is because many variables will be correlated, without sharing a relationship. These are sometimes called spurious correlations, and there are many examples of them.
From several regressors of electricity demand, like humidity or other environmental variables, ambient temperature has a historically significant position in France due to electrical domestic heating and cheap nuclear power. Furthermore, temperature data is often readily available. In this case, a simple daily average temperature is taken for the country of France as a whole. As shown in Figure 2, plotting electricity demand versus this average temperature yields a clear correlation, best fit by a parabolic curve using linear regression shown in red. Recognizing this correlation, we can correct for it, subtracting the variation in the data that we attribute to the temperature variation. Nevertheless, a parabolic fit may not be the best match for this data, but for the purposes of this analysis it will do as we are mostly interested in a small range of values in the 5-15 °C range (spring weather). Note that the vertical lines presented in Figure 2 are a consequence of the using daily temperature data, i.e. multiple demand values correspond to the same temperature.
Having made this correction for the difference in temperature, we return to the energy demand from March 18th, shown in Figure 3 below, and observe how the data is better aligned now. In fact, we find that during the night, the demand is almost the same, something fairly reasonable given that the quarantine would not have a particularly large effect then. It is during the day that the two demand profiles start to diverge, with the quarantine having generally reduced demand, yielding a slower ramp to the slightly later morning peak. The overall reduced demand is what is expected, because of the stop of almost all industrial activities during the period of confinement.
However, with almost two months of quarantine behind us, we can do better than looking at the changes of a single day. Assuming that the demand is reasonably similar on weekdays, and weekends, Figure 4 shows how the demand has decreased. Comparing the demand change at every hour of the day, the demand has dropped on average by approximately 10%. Despite this decrease, during the weekdays there is not a significant change in the consumption pattern, with demand peaks similarly placed. On the other hand, on the weekend there is a significant difference in the consumption pattern, where peak demand has shifted almost 3 whole hours later. People seem to be taking advantage of having nowhere to go by sleeping in!
Finally, with this change in demand, we could ask how did the supply of electricity change? Figure 5 below is a stacked area plot including most of the major sources of electricity in France, starting from mid-February. Though it is hard to make a detailed analysis from looking at only a stacked area plot, a few trends can be observed in the following. Overall, it seems clear that gas turbine and hydroelectric generators have both decreased their generation since the time period started, and especially recently. This is contrasted by the ever strong nuclear generation, which appears largely unchanged.
In summary, the confinement due to COVID-19 has resulted in the reduction of electricity demand in France, by approximately 10%. There is also some indication of behaviour change, with regards to electricity usage in response to the confinement. Something that we might expect, and maybe also find relatable. Finally, with the advent of technology, all of the data used in this analysis is readily available for free online (RTE and Open Data Soft). The data processing and visualisation is also available, and was conducted in python (script on GitHub).