Google, Facebook, Amazon, IBM, Intel, Apple and other behemoths are pouring billions of dollars into artificial intelligence. In the energy sector, one UK start-up thinks it can take them on. Brendan Coyne reports.
By connecting the best human minds and then pitting the algorithms they create against each other, Upside Energy founder Graham Oakes thinks UK academia can outsmart Silicon Valley. His goal is to enable a more intelligent, cost-effective energy system based around millions of small, connected loads optimised via artificial intelligence. A grand plan. But how big a role can artificial intelligence play in a smarter energy system?
“I think it will be fundamental,” says Oakes. “If you look at demand-side response right now, people are doing fairly simple things – single services from a dedicated asset. But as the energy system transforms to a marketplace that buys flexibility as well as energy, you need to be doing multiple things to get the most from these assets.”
Sometimes that will mean providing services to the distribution network operator, sometimes to National Grid, sometimes an energy supplier. At other times it might simply be arbitraging price differentials, says Oakes.
“So now you’ve got a question: How do you make intelligent decisions to best use the individual asset and the portfolio of assets?” He thinks machine learning will come up with the most efficient answer – and a best value energy system.
Food for artificial thought
Great in theory. But, if the energy system is already struggling to adapt to technology shifts, what is a realistic timetable to embed AI?
Oakes says the firm’s algorithms will make significant improvements in areas such as battery life management “quite quickly, within 12-24 months”.
However, “AI is dependent upon data – and the challenge for start ups is getting enough”. Additionally, the broader energy market is not entirely sure what it wants.
“The market to deliver the full range of value services to all parts of the energy system will take five to 10 years to really develop,” says Oakes.
“Every time we talk about flexibility my question is ‘what do you actually mean by flexibility?’ Because we haven’t actually defined what flexibility is. How does it add value to people, what types of flexibility add value? It will take time to learn that stuff. So the intelligence will really come into its own five years from now.”
But Oakes says the company – and any other market participants aiming to develop smart grid solutions – must focus on that horizon now.
“If you have five years’ worth of data you can immediately start to add value. If you only start to try this when we define flexibility, you are going to need to go out and gather years’ worth of data.”
He uses Wayne Gretzky’s ice hockey analogy: “The best players don’t follow the puck, they go where the puck is going to be. And that is what we are doing.”
The firm, which was funded by the now defunct Department of Energy & Climate Change, as well as Innovate UK, has managed about 400kW of load from about 20 different sites for almost two years.
Last November it signed an agreement to provide frequency response to National Grid and is now qualifying some 2.5MW (including a megawatt-scale battery being managed by the University of Sheffield) for delivery from April. That will enable Upside to start earning some money, with National Grid paying a premium to parties that can react within two seconds to keep the power grid stable.
In January this year, Upside was also granted another £100,000 from Innovate UK to fund AI-driven demand response research at Heriot Watt University. While that adds to Upside’s academic firepower, it is probably not worrying the titans of Silicon Valley just yet.
But Oakes thinks the collective brain of UK academia – the firm is also working with universities including Oxford, Imperial, Manchester and Lancaster – offers the start-up a fighting chance against their might. And if the universities come up with a better algorithm than Upside’s, they earn a revenue share.
“There is an enormous amount of academic firepower out there doing really interesting stuff with machine learning and algorithm development and fantastic modeling of the energy system,” says Oakes.
“Working with them gives us enormous intellectual weight. We think we can leverage the expertise within the UK – and ultimately within the global academic base. We think that gives us a way to compete with Google.”
Sum of all parts
Most traditional demand-side response aggregators target larger firms due to the cost of amalgamating hundreds or even thousands of small sites, which analysts suggest represents a barrier to genuine scale. Meanwhile larger corporates, they say, will be reluctant to cede control to those without sufficient brand credibility.
But Oakes thinks that creates a vacuum for companies such as Upside to fill. He also believes a more diverse range of assets will unlock higher value.
“Managing small, diverse loads is our core. We think we are unique in terms of the range and type of those assets – and the benefit of doing this in the cloud is that we can talk to almost anything – whereas other people are operating in narrow niches.”
Oakes says that approach is shedding light on how managing combinations of batteries, heat pumps and hot water tanks, for example, adds more value than the equivalent capacity of a single asset, such as a battery.
“Using a battery to manage frequency response, for example, you need to have capacity to go in both directions; to add load to the grid when frequency is high and take load off when frequency is low. That means you need to manage your battery in about a 50% state of charge. So you go out and buy this very expensive asset and you just use half of it,” says Oakes.
“What we are starting to see in our modeling is that batteries are wonderful for [frequency balancing] but you want to keep them fully charged so you can use their full capacity,” he says.
“Hot water tanks are the reverse. You can’t export to the grid; you can’t do very much when frequency goes low. But when frequency goes high, you can always put a bit more heat into them. So this combined thermal and battery store means we can optimise in a way that a single portfolio of batteries or thermal stores cannot deliver.”
Oakes thinks that will become a “very powerful model”. But it will require National Grid to give further thought to procuring flexibility.
“National Grid has started to get its head around aggregating smaller assets and to get its head around batteries,” he says. “But now when you start to aggregate diverse asset classes, that is yet another shift of mindset to be taken.”
So how does Upside, without a sales force, overcome the cost of sales barrier mooted by some traditional aggregators? Oakes says the simple answer is “by being lazy”.
“I don’t want to build a sales force, the innovation in our business model is to find people to sell it for us. The PV and battery manufacturers have installer channels. We give them something that makes the economics work for them and let them sell it for us.”
Upside has also signed an agreement with one of the three large UPS manufacturers; where Schneider, Eaton and Vertiv (formerly Emerson) collectively hold about 85% of the UK market. Although Upside’s website cites Emerson as an advisory partner on the Innovate UK project, Oakes says the firm is under non-disclosure, so cannot reveal which of the three will incorporate Upside’s technology.
Nevertheless, Oakes says UPS firms have recently begun to convey a “more nuanced” message than selling kit based on fear of failure.
He thinks that will help break down customer barriers to demand-side response participation, which, according to research by MCP sister publication The Energyst, are principally around unsuitability of equipment and fear of disruption to core business.
“We have seen a significant change in mindset over the last six months,” he says.
Oakes thinks domestic demand-side response could also scale more rapidly than many market participants believe. However, he says it may be hampered by consumer versus commercial silos within the large industrial conglomerates involved in batteries and UPS. Whether grid operators grasp the nettle is another key factor.
“Behind the meter storage isn’t challenged in quite the same way [as some of the larger scale markets]. Talking to people at Ofgem and BEIS, they are all very supportive,” says Oakes.
“[But] National Grid and the DNOs are still very much getting their heads around the possibility of domestic DSR [they think] ‘great idea but we’ll address it in five or ten years’ time’.
“Actually, I think it is just waiting to be tapped into. It is going to take time to kick off, but I think if we start now, we can make it happen in the next two or three years.”