The U.S. economy has been thrown into a tailspin by the coronavirus pandemic and the resulting lockdown, with many businesses closing their doors, countless others shifting to remote working, and universities and schools shutting down for the duration of the crisis.
That economy-wide shift is having a huge impact on U.S. energy markets, too, with commercial demand falling by more than 30%, and residential usage soaring. That brings enormous challenges for energy producers and suppliers — after all, if the switch to remote working can wreak havoc on something as apparently straightforward as our supply of toilet paper, imagine what it means for something as sprawling and complex as the U.S. electrical grid.
Fortunately, energy companies have a secret weapon: machine learning. By leveraging AI’s ability to spot patterns in large, chaotic data-sets, providers can recalibrate their forecasts in real time, and get the accurate projections they need to ensure the supply of power is always near-perfectly matched to anticipated demand.
During these turbulent times, AI-based load forecasting gives energy providers the ability to learn quickly from short-term shocks, make timely adjustments, and develop informed, data-driven strategies for future success. For example, within three days of shelter-in-place orders, Innowatts’ forecasting models — which use AI to digest readings from 34 million smart meters, processing over 2.83 billion data points per hour — were able to spot new consumption patterns and develop accurate predictions that fully accounted for the monumental shift in commercial and residential energy use stemming from COVID-19.
We quickly ascertained that residential energy consumption was trending between 6% and 9% higher than normal, and as much as 15% higher during working hours. Load patterns have also changed dramatically: mealtimes have apparently become optional, and the predictable lunchtime and dinnertime peaks we’d ordinarily see have largely evaporated as home-bound families have begun cooking and eating at whatever time suits them best.
We’re also seeing big shifts in specific industries. As you’d expect, restaurant energy use has fallen precipitously, by 51% — but food delivery services are still thriving, with pizza delivery services using 20% more energy since the crisis began. Similarly, overall consumer retail stores have almost halved their energy use, but certain categories are still doing well: liquor stores, for instance, are using 19% more energy than before the crisis.
Our data also shows the sectors where the lockdown is taking the fullest effect. School energy use is down 40%, for instance, while commercial offices are using 25% less energy — a sign that government-mandated educational closures are more comprehensive than closures in the professional community.
With utilities and energy retailers seeing a once-in-a-lifetime drop in commercial energy consumption, accurate forecasting has never been more important. Without AI tools, in fact, utilities would see their forecasts swing out of sync with reality by 20% or more — a colossal amount for the energy sector — and place an enormous strain on their operations.
By allowing operators to minimize those pressures and cope with uncertainty, AI load forecasting allows energy providers to operate efficiently and reliably, maximize profits, and ultimately keep costs low for businesses and consumers.