Some homes with solar panel installations also have solar batteries, which store energy for later use. A Pittsburgh start-up has developed artificial intelligence software that could make those batteries more efficient.
Currently, solar batteries with decision-making abilities can only do so based on real-time information. For example, on a cloudy day when solar panels might not produce enough electricity to power a house, a charged battery would automatically kick in to make up the difference.
Matthew Maroon, CEO of Watt-Learn, said his company’s cloud-based software can look ahead at how factors like weather projections or the price of electricity will change over the next few days.
Suppose one day is hot and sunny and the next day will also be hot but cloudy, he said. That means electricity costs still be high because of the heat, but solar panels won't be able to collect as much energy.
"Our software would make a decision in the current day to store as much power in the battery as possible, not discharge the battery into the house or sell it back to the grid," said Maroon. "So tomorrow, we can discharge from the battery into the house to provide you with the lowest cost electricity for running the air conditioning or whatever whatever else you're going to be doing that given day."
The software also causes batteries to go through cycles of charge-and-discharge less frequently. He said each time a battery goes through a cycle, it degrades slightly; over time, this causes the battery's function to decline, the same way one might observe with a cell phone or laptop battery.
Maroon said that because the software is built on machine learning algorithms, it will get better at making these decisions as it gains more experience. It's also compatible with any kind of solar battery.