Using Artificial Intelligence (and a Little Scolding) to Reduce Energy Use

Turn your thermostat down, Dave. A new program in England looks at your energy use and then offers you suggestions on how to reduce it to save money (and emissions)

Using Artificial Intelligence (and a Little Scolding) to Reduce Energy Use Image:

By Andrew Price

Turn your thermostat down, Dave. A new program in England looks at your energy use and then offers you suggestions on how to reduce it to save money (and emissions).

Houses and other dwellings account for nearly a third of all the energy used in the United Kingdom (in the U.S. it's a bit lower). That takes a toll on the climate and on people's wallets. What's worse is that much of that energy is wasted because residents don't see an immediate connection between the thermostat, for example, and the utility bill.

Nigel Goddard, a professor at the University of Edinburgh's School of Informatics, is trying to solve that problem using cutting-edge techniques from a branch of artificial intelligence called "machine learning."

In the multi-year IDEAL project, launching in 2013, Goddard and his colleagues will outfit hundreds of British homes with relatively inexpensive sensors that monitor temperature, humidity, and light levels, as well as gas and electricity use, and wirelessly report their readings every minute. Using machine learning techniques, Goddard and his team will be able to analyze that noisy data to infer what people are actually doing--cooking or taking a shower, for example.

Then they'll use another cutting-edge technology--natural language synthesis--to generate automatic text messages that give people feedback about their energy use. A text message might read, "Last week you spent £10 on hot water for showers, if you reduced your average shower time from 15 minutes to 12 minutes you could save £100 per year." Over time, they can tweak the kinds of messages they send to make them as effective as possible.

How much energy will people end up saving with this kind of feedback? Goddard expects the results to vary based on a home's income level, among other factors, but says that "in the best groups I would estimate 20%, possibly a bit more--but its a wild guess."

The money saved with an energy reduction of 20% could pay for the roughly $800 sensor set within as little as two or three years. And that's what makes this project so interesting. If it works, it could scale quickly. "It could be offered by utilities or energy service companies, perhaps they install the kit for free and take a percentage of savings," says Goddard.