A year of pandemic effects on residential and small commercial electricity use: Examining evidence from the SDG&E service territory

As the COVID-19 pandemic progresses into its second year, we are eager to understand how COVID-19 is affecting customers’ energy usage in various sectors over the short and long term

As the COVID-19 pandemic progresses into its second year, we are eager to understand how COVID-19 is affecting customers’ energy usage in various sectors over the short and long term. We know that many residential customers are at home (working or unemployed, with or without children attending school remotely), and that many small businesses had to close or operate at partial capacity. To understand what this means for customers’ energy usage requires data analysis.  

To begin exploring how the pandemic changed the energy usage and patterns of residential and small commercial customers, DNV conducted an analysis using publicly available data including San Diego Gas & Electric (SDG&E) Dynamic Load Profiles,1 actual weather data,2 normal weather data,3 and Google Mobility Statistics.4 We compared customers’ usage patterns from 2019 (“pre-COVID”) and 2020 (“COVID”), controlling for the different weather experienced in the two years so that we could estimate the direct effects of COVID on energy usage. SDG&E did not review or comment on this analysis.    

Our analysis found that in the residential sector, the pandemic caused a short-term increase in electricity use, but that in subsequent months, usage leveled off to pre-pandemic levels. By contrast, the small commercial sector experienced visible and consistent use reductions throughout 2020 and the first quarter of 2021.   

Figures 1 and 2 show the hourly usage patterns for March 2019 and March 2020 beginning on March 16—the date when California instituted a shelter-in-place order. To allow for easy visual comparison, we aligned the day of the week in these figures by shifting the pre-COVID usage back by two days. In these figures, the 2019 “pre-COVID” dynamic load profile is shown in orange and the 2020 “COVID” one is shown in red. To show how the day of the week impacts energy use, weekends are highlighted in green.  

Figure 1: Residential dynamic load profile 

20210609_Examining Evidence from the SDGE Service Territory - blog - figure 1

Figure 2: Small commercial dynamic load profile 

20210609_Examining Evidence from the SDGE Service Territory - blog - figure 2

These figures show pronounced differences in the energy use patterns the year before COVID and at the start of COVID. However, we needed to determine whether and how much these differences are attributable to weather. Thus, we quantified the change in energy use for these two customer groups after controlling for outdoor temperature, which is the strongest predictor of electricity use for most customer types.  

The process of making energy consumption comparable for different periods by removing the effect of differences in outdoor temperature is called weather-normalization. It’s a regression-based method that is widely utilized in energy analysis. We created one weather-normalization model for residential customers and another for small commercial customers. The models describe daily energy use before and during COVID as a function of outdoor temperature and day of the week.  

California electric IOUs have published dynamic load profiles through their respective websites since 1997. It is important to note that SDG&E’s dynamic load profiles reflect net usage, not actual usage. Actual usage is the amount of electricity that customers used.  Net usage is the amount of electricity that customers take from the grid minus the amount of electricity that they put back into the grid. For non-solar customers, actual and net usage are the same. For net-metered solar customers, who are a sizable percentage of customers in the SDG&E service territory, net usage is not reflective of actual energy use.   

We selected SDG&E for this study due to the compact nature of the service territory, which allows a single weather station to be reflective of the weather experienced by most customers in the service territory. This simplifies the analysis and makes the model more accurate, as dynamic load profiles reflect usage for all customers in the service territory. We selected the Montgomery-Gibb weather station for the models because it had more explanatory power than other weather stations in the service territory. 

We ran the usage models5 with all combinations of heating degree days base 45 to 70⁰ F, and cooling degree days base 60 to 90⁰ F, and an indicator for weekends/holidays as explanatory variables. We selected the models with the highest adjusted R-squared. The temperature bases in the selected models are the temperatures at which the DLP shows evidence of heating and cooling electricity use. The optimal base temperatures are presented in Table 1. 

Table 1: Heating and cooling base temperatures 

20210609_Examining Evidence from the SDGE Service Territory - blog - table 1


Normalized outdoor temperatures are maintained by CALMAC for selected weather stations throughout California. Figures 3 and 4 show the weather-normalized model results for residential and small commercial customers, respectively. Figure 3 shows that for residential customers, a large amount of the variation in usage before and during COVID can be attributed to differences in outdoor temperature, and not the effect of the pandemic. This contrasts with Figure 4, which shows significant changes in weather-normalized usage before and after the start of the pandemic for small commercial customers. This change is reflective of large numbers of businesses that were closed or limiting operations after the start of the pandemic.  

Figure 3: Residential weather-normalized modeled usage 

20210609_Examining Evidence from the SDGE Service Territory - blog - figure 3

Figure 4: Small commercial weather-normalized modeled usage 

20210609_Examining Evidence from the SDGE Service Territory - blog - figure 4


To gauge change in electricity use, we summarized weather-normalized consumption in temperature bins, before and during COVID6. The changes in usage aggregated by temperature bin are shown in figures 5 and 6 for the residential sector, and figures 7 and 8 for the small commercial sector. Residential usage is similar in both periods and shows no systematic differences. The small commercial usage differences are visible and systematic. They show a reduction in usage ranging from 3% for the 50-degree temperature bin all the way up to 15% for the 80-degree temperature bin. The small commercial class had larger usage reductions at higher temperatures, indicating reductions in air conditioning load. This matches our understanding that many small businesses were operating at partial capacity and experienced a sharp reduction of in-person customers, therefore reducing their need for cooling. 

Figure 5: Weather-normalized residential usage by temperature bin6, pre-COVID and during COVID 

20210609_Examining Evidence from the SDGE Service Territory - blog - figure 5


Figure 6: Percent of weather-normalized residential monthly usage change in 2020 compared to 2019

20210609_Examining Evidence from the SDGE Service Territory - blog - figure 6

   

Figure 7: Weather-normalized small commercial usage by temperature bin6, pre-COVID and during COVID 

20210609_Examining Evidence from the SDGE Service Territory - blog - figure 7

    

Figure 8: Percent of weather-normalized residential monthly usage change in 2020 compared to 2019 

20210609_Examining Evidence from the SDGE Service Territory - blog - figure 8


While the causes of the uneven changes in the residential sector require additional analysis, it is possible to speculate about them. They could be the result of customers initially following the shelter-in-place order and then resuming normal activities, the intentional reduction of usage by some customers who may have lost their jobs or faced other financial challenges, or the rapid increase and subsequent leveling-off of activities such as cooking and baking at home. Changes to electric vehicle charging patterns may have played a role. Additionally, as mentioned earlier, SDG&E has a very high percentage of solar customers. Changes in solar radiance that would result in increased solar output and reduced use of grid electricity were not considered in this analysis.   

It is important to note that dynamic load profiling is calculated using the utility’s open accounts. Premises with closed accounts are not reflected in these averages. In other words, these are reductions in the electricity use of open accounts, and do not reflect reductions in use from premises where the business closed and no new account opened at the same location. Landlord accounts, where the building owner opens an account for premises with no tenants, are common. To investigate whether our estimated reductions in small commercial use were understated due to closed accounts that do not figure in dynamic load profiling averages, we examined the number of customers published with SDG&E’s load profiles. Figure 9 shows that there was no decrease in the number of open small commercial accounts during the pandemic. Further, as shown in Figure 9 below, the number of open accounts continued to increase, although at a slower pace than in 2019.  

Figure 9: Number of Small Commercial Customers 

20210609_Examining Evidence from the SDGE Service Territory - blog - figure 9

To learn more about occupancy patterns, we used the Google Mobility visit data. This data includes several measures, including “residential visits” which reflect times when location-enabled phones reported their location as “at home.” Figure 10 shows the San Diego County Google Mobility residential visit   statistics for mid-February 2020 through the end of January 2021. After an initial surge of 25% above baseline through mid-April, by early June, occupancy declined gradually to 15% above baseline, and consistently maintained that level through the end of the year. Weekend residential visits (not shown) also increased consistently by 5 to 8% through year-end. 

Figure 10: Google Mobility Residential Visit Statistics 

20210609_Examining Evidence from the SDGE Service Territory - blog - figure 10

While there is certainly a correlation between certain electricity uses and occupancy, it is not perfectly proportional, so we expect that the 15% increase in occupancy shown between June and year-end will result in an increase of less than 15% in energy use. Flat energy usage and higher occupancy suggests a possible reduction in occupancy-adjusted residential energy usage.  

The workplace visits statistics have a complementary pattern, with an initial steep drop of 55-60% by mid-April, then a gradual increase to 40% below baseline by early June, which stayed consistent through the end of the year. The weekend pattern (not shown) is similar, but leveled off at 20% below baseline. 

Figure 11: Google Mobility Workplace Visit Statistics 

20210609_Examining Evidence from the SDGE Service Territory - blog - figure 11

Overall, the occupancy statistics correlate well between the workplace visits and commercial usage levels, reflecting more COVID usage impacts during the summer months to due to less cooling for mostly unoccupied spaces. Once workplaces open up, we expect that the COVID-influenced usage impacts will abate, although they may not ever fully recover, since COVID has permanently changed the way we do business. For example, office occupancy is unlikely to recover, which will in turn create less demand for services associated with offices, such as lunch restaurants and dry cleaning.  

If you would like to find out more about our work estimating COVID impacts, please contact us.




1 SDG&E Dynamic Load Profiles: https://www.sdge.com/more-information/doing-business-with-us/energy-service-providers/dynamic-load-profiles 

2 Current weather data source: https://www.ncdc.noaa.gov/cdo-web/ 

3 Normal weather data source: http://www.calmac.org/weather.asp 

4 Google Mobility Statistics: https://www.google.com/covid19/mobility 

5 This weather-normalization work was presented at the Western Load Research Association Spring 2021 Conference. The corresponding presentation has more details regarding the models. The presentation can be accessed at http://wlra.net/docs/WLRA_Spring2021.zip  

6 The temperature bins group average outdoor temperatures rounded to the closest 5 degrees: “65” includes 62.50 to 67.49, “70” includes 67.50 to 72.49, etc.

6/9/2021 5:00:00 AM