Published in


Tracking the US electricity demand

COVID-19 has shaken the world! Countries are in different phases of struggle against the pandemic. While some have just begun the lockdown, others are in the recovery phase. With economic activity plumbing across the globe, data that quantitatively captures this macroeconomic trend has become more important than ever, specifically for investors.

Along with a host of datasets that Synaptic tracks, we investigate the US electricity demand data, given its relevance to the extent of economic activity. These electricity demand datasets are near real-time and could be pretty good indicators of social and economic activities returning to their normal state after the pandemic.

US Energy Information Administration publishes several useful datasets, one of which gives per hour electricity demand for different regions in the country. Here are our findings…

The overall daily demand has decreased, duh!!

In the figure below, we show how electricity demand varied over the last three years for the same period of time in the New York region. As expected, with reduced industrial and commercial activities in the lockdown period there is a slump in the overall electricity demand. We found similar trends for other states viz. California and Texas.

Mean daily electricity demand for the New York region

Peak demand is shifted and is low

Normalized hourly electricity demand (in MWH) in the New York region. [A] Day on Day data, [B] Mean of the data for each hour

We see some interesting trends in the hourly demand patterns. Typically electricity demand rises up steeply between 6–10 am, as people wake up and arrive at work. Normalized (Appendix-2) day on day data during the lockdown period reveals that this ramp is less steep now, i.e. the morning peak demand is both low as well as delayed. This hints that probably people are starting their day later than they used to, and of course, offices and factories being closed are the factors here. Data also shows that the evening peak demand is shifted by approximately an hour. It's worth noting that we only consider weekdays for this analysis since weekends wouldn’t capture the routine so consistently.

In order to rule out factors such as changing weather and seasonality effects, we compare the data during the lockdown period with the same period in 2019. The figure below shows this data. From both these analyses, the delayed and low morning peak can be clearly attributed to the slowdown caused by the COVID-19 pandemic. The same observation with a little variation is made for other regions in the US.

Afternoon demand bumps in Texas

In addition to the observations made for the morning peak demand, for the Texas region specifically, interestingly, we also observe a bump in the afternoon demand. We are not sure how much of this can be attributed to HVACs given the weather is not extreme, yet, though there have been some analyses around this.

These patterns can be tracked during the recovery phase to quantitatively understand the extent of economic activities restored in the US.

Synaptic harnesses a number of alternative data sources to help investors get actionable insights from vast amounts of data. Get in touch for a demo.

View more from Synaptic around COVID-19.


[1]. Hourly demand pattern without normalization is shown below.

[2]. Normalization strategy: We normalize the daily data by subtracting the minimum demand value for each day. Typically this minimum is observed early in the morning (around 4–5 am) for most of the regions. We tried several normalization strategies, of which mean subtraction highlighted the slowdown pattern most clearly.

[3]. We also analyzed detailed demand distribution for each hour of the day in the New York region. These detailed distributions also support the same observation made above. In the graphs below, the x-axis denotes the normalized power (in MWH) and the y-axis is normalized KDE. We clearly observe the shift in power distribution in the morning and evening times, wherein they both are lower than usual.

[Red] indicates the distribution for the lockdown period and [Blue] for the rest of the days in 2020




Creating opportunity through better insights

Recommended from Medium

The Promises and Perils of “Automatic Data Visualization Recommendation”

A gladiator stands over a defeated foe. The bloodthirsty crowd is signaling “thumbs down.” I’ve been told that, contrary to popular belief, thumbs down was actually the “spare him” sign, but then I told that this is also a myth, and so we’re at nested layers of wrong, here. I’d advise just keeping your hands to yourself if you end up in ancient Rome, just in case.

Delete rows and columns from a DataFrame using Pandas drop()

One Feature to Supercharge Your Excel in 2020

Predicting NBA Post All-Star Break Wins

Top 5 Business Intelligence Tools for Embedded Map Visualizations (Plus 1 Bonus Option)

Yurbi - Embedded Analytics Software

Community Safety Health and Environment :Second Fortnight Progress Report of Mega Project

Patterns of data institution that interact with people and their rights over data

LGM VIP Internship Experience

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Sankalp Gulati

Sankalp Gulati

More from Medium

💊Our 12th Investment: OK Capsule — Enabling Personalized Nutrition At Scale. 💊

7 Lessons Learned from the Pittsburgh Technology Landscape

Neuralert’s First-in-Class Wearable Detects Stroke Quickly So Brain Function Can Be Saved

SEC Proposes New Cybersecurity Risk Management Rules