Smart home technology can be used for a lot more than just asking Amazon Alexa to turn off the lights or play your favorite album. It also has the potential to lower your electricity bill — and help the environment in the process.
As a doctoral student in 2012, Aditya Mishra, now an associate professor at Northeastern University in the Khoury College of Computer Sciences, helped demonstrate that when developed and published research on a battery-based system called SmartCharge that relied on AI algorithms to help consumers do just that.
More than a decade later, Mishra's SmartCharge research continues to be influential, so much so that he was awarded a Test of Time award by the Association for Computing Machinery (ACM) for the project.
He received the honor this summer at the ACM International Conference on Future and Sustainable Energy Systems (ACM e-Energy 2024) held in Singapore.
“I see it as concrete feedback from the research community,” he says. “Knowing that our work has been noticed and seen is a great feeling, and it motivates me to continue working on more amazing projects here at Northeastern University.”
To understand how the SmartCharge system works, it's important to understand the context surrounding its development, Mishra explains.
In the early 2010s, Mishra and his colleagues began researching how they could help utilities reduce the amount of energy homeowners use during “peak hours,” which are periods of the day during which use a lot of electricity in the grid. (Think around 5 p.m. when residents use stoves, microwaves, and other appliances to prepare dinner.)
During peak hours, power grids are under heavy load and utilities must activate special peak generators to meet demand. These generators are expensive to run and consume large amounts of energy. Being able to reduce demand would not just reduce costs, but could help reduce carbon footprints, says Mishra.
At the time, a select number of utilities were experimenting with time-of-use pricing plans aimed at incentivizing people to rely less on energy during peak hours, Mishra says. They did this by creating a payment structure that would charge users more for energy use during these peak times and less during off-peak times.
“Unfortunately, the problem is that electricity demand tends to be quite inflexible for the end user,” he says. “For example, it's usually not possible for all of us to start cooking dinner at midnight just because that's when electricity is cheap.”
With an interest in green computing, Mishra aimed to develop a better solution that would allow users to take advantage of the low energy costs that come with time-of-use plans without being inconvenienced.
To address this, they developed SmartCharge, a battery storage system that exploits machine learning technologies “to intelligently switch” between electricity consumption from the electricity grid and the battery storage system.
“Based on these various machine learning and optimization algorithms, it can decide when it's ideal for the battery to charge from the grid and when it's ideal to use the power stored in the battery instead of drawing power from the grid.” He says.
For the research project, the team developed an original SmartCharge prototype, which was essentially a modified lead-acid battery that the team purchased and modified in a lab with specialized equipment. It was trained on data they had collected on real homes in which they had installed sensors.
“Our sensors were used to collect the total consumption of the houses every minute and also the consumption from various different outlets in the houses,” he says. “We collected data from these real homes where people were actually living their lives.”
“Every day, machine algorithms continued to learn and update their models. He learned how much energy the house uses,” he adds. “Based on that learning, every day, at the beginning of the day, it made predictions about how much electricity that house is going to use.”
With this information, Mishra and his team were able to develop optimization algorithms that would decide when the battery should be used for storage and when the battery should be used to draw power.
“At a high level for a general audience, we first found that it is possible to significantly reduce a customer's electricity bills with smart battery storage in the presence of time-of-use electricity pricing plans without any involvement from the end user. Second, we found that batteries with time-of-use pricing can significantly reduce grid demand during peak hours.”
The final key takeaway from the paper was a warning, Mishra explains. While batteries can be a useful way to help keep costs down, they must be used judiciously when it comes to changing demand.
“Suppose everyone starts charging their batteries at midnight because electricity is cheap, that could actually in a larger case create a new peak in off-peak periods.”
In the decade since Mishra's research, SmartCharge has been cited more than 150 times by other researchers, Mishra notes. He also did a follow-up research project adding SmartCharge to a solar panel project called GreenCharge that allowed users to take advantage of the extra energy coming from their solar panels to power their homes.
There have been many positive developments in the industry that Mishra said he and his colleagues envisioned all those years ago.
More and more energy utilities have begun offering time-of-use plans for residential customers. In addition, consumers can now buy home energy storage systems off the shelf. Shown Tesla's Powerwall as an example. Finally, governments are offering more tax incentives for consumers to use clean energy.
“I continue to work on these problems here in my position at Northeastern,” he says. “We can use this as an indication that we should continue to pursue this direction.”
