Cold winter days are the hardest to endure for people suffering energy poverty. “I was getting lethargic sitting still to keep warm,” says one. “Physically, I wasn’t even taking a bath. I was saving up the money. Mentally, I was losing my health, cutting down on so many things,” says another. “I was holding off with the laundry, even holding off going out looking for a job because you need clean clothes.”
These are cries for help from people who have contacted the Fuel Bank Foundation over the past year. The charity provides emergency credit to those struggling to pay their energy bills. Requests for support have increased by 23 per cent since the start of the coronavirus pandemic. Worse still, the foundation says self-disconnection, where households switch off their power supply completely, is a growing problem.
Choosing between heating and eating, or between having power or going into debt, are decisions increasing numbers of people are having to make. It has been a long, tough winter. Unemployment currently stands at 5.1 per cent, the highest level since 2016. And for many of us, as our homes have become our workplace and school, domestic energy consumption has jumped as a result.
Artificial intelligence (AI) innovations and smart technology are typically thought of as preserves of the wealthy. Charities and academics, however, believe they can provide vital solutions to fuel poverty by tracking and managing energy usage, enabling cheaper power consumption and providing short-term solutions when there isn’t enough money for the bills.
One size does not fit all
How? At a community level, installing smart local energy systems can make electricity cheaper by deploying low-carbon, local power. This can cut generation and distribution costs, and deliver the savings back to users.
Within homes, smart prepayment meters can now add instant emergency credit when funds run low, tiding users over until they can afford to top up. In turn, the consumption data smart meters generate can be used to give suppliers insights into their customers’ energy usage patterns.
Yet this data has to be used carefully and thoughtfully. People struggling to power their homes can often slip through the net, simply because they don’t behave according to what suppliers and the government expect. Applying blanket assumptions about who needs help, and how they can be helped, doesn’t work, especially as the pandemic draws more people into fuel poverty.
Dr Aidan O’Sullivan, associate professor at University College London, where he leads energy and AI research, says the government and suppliers spend money “trying to identify people who should be getting the winter fuel allowance and support for fuel poverty, and still get it quite badly wrong”.
“What AI can really help with is correctly identifying the customers who should be receiving support,” he says. The signals that suggest a household is starting to struggle with their energy use are complex.
“For example, someone might repeatedly reduce their consumption at the end of the month as they run out of money, which is a subtle signal that can get lost when averaging data,” says O’Sullivan.
His work includes building neural networks, which can use signals from smart devices to detect whether a home is falling into fuel poverty. Energy suppliers can then help at-risk customers with targeted interventions, such as giving personalised guidance via an AI agent.
Starting earlier will be key
Dr Rose Chard, who leads consumer insights at Energy Systems Catapult, a state-backed not-for-profit which works to accelerate new energy technologies, stresses that fuel poverty is diverse. It affects different people in varied circumstances, which is often overlooked when trying to use technology to solve fuel poverty.
“An elderly woman living on her own, in a property that she owns, on a very low state pension with no mortgage, might be fuel poor. But we also have working families, on zero-hours contracts, living in the private rental sector, who are fuel poor. So there isn’t going to be one solution that’s going to work for all households,” she says.
Someone fit and healthy may enjoy tracking their energy consumption via their smart meter, turning their thermostat down by a couple of degrees and getting cheaper bills as a result. While for others, who are in a damp or draughty home or living with a serious health condition, this could be bad for them. But they may benefit from using a smart system to heat individual rooms in their home to higher temperatures. Therefore, imposing one-size-fits-all AI and smart technology is unlikely to be the answer.
Including the vulnerable in the design process
Instead, Chard suggests tech innovators can do better by considering exactly who stands to benefit from the algorithms, products and systems they create right at the start of the research and development process. Actively involving vulnerable householders and wider services, such as the NHS, earlier in research can ensure innovations and support schemes are fit for purpose.
Her team last year successfully trialled provision of “heating on prescription” in combination with smart meters. She says: “We found with smart controls, people were able to heat their home to healthy, warm temperatures in a way they weren’t before. And what if a GP or a healthcare visitor could prescribe a warm home for six months at times in your life when you might be most vulnerable to living in a cold home?”
Providing resources to help manage, rather than minimise, energy use is perhaps the most effective application of AI and smart technology for those experiencing fuel poverty. Combining valuable data and technological advances with human sensitivity and insight may be the best way to meet that most basic of needs: keeping warm and dry.