AI efficiency gains come at a high energy cost

{"text":[[{"start":6.1,"text":"Former US energy secretary Steven Chu once described energy efficiency as “not just low-hanging fruit” but “fruit that is lying on the ground”. Yet despite being cost-effective, this critical method of decarbonisation is widely underused."}],[{"start":21.75,"text":"The process of identifying where and how energy is wasted is complex and time-consuming, which is why technologies and management practices in this area are not adopted as widely as they could be. Artificial intelligence could now be changing that."}],[{"start":36,"text":"“AI can understand where the problems are and where the waste is — and you don’t need five people with PhDs looking at monitors all day,” says Brian Motherway, head of energy efficiency at the International Energy Agency."}],[{"start":49.35,"text":"In large factories with thousands of motors, valves and other pieces of equipment, for instance, AI can analyse everything and make efficiency adjustments. “It’s beyond the power of anyone to handle all that data,” he says."}],[{"start":62.75,"text":"A boost is badly needed. IEA data shows that the rate of progress towards global efficiency targets is slowing, particularly in industry. “In some ways that’s surprising because it’s where the financial case is clearest,” says Motherway."}],[{"start":78.25,"text":"However, he says, barriers to energy efficiency adoption should not be underestimated: “Maybe the CEO loves the idea of cutting ribbons on solar-panelled roofs but it’s hard to get excited about a boiler.”"}],[{"start":null,"text":""}],[{"start":90.75,"text":"Energy efficiency investments produce results less rapidly and less visibly than shifting to renewable energy sources, says César Quilodrán-Casas, an advanced research fellow in machine learning at the Grantham Institute for Climate Change and the Environment at Imperial College London. “It’s not such a big story to tell.”"}],[{"start":109.75,"text":"Meanwhile, increasing energy efficiency often requires equipment upgrades that can be more complex than building a wind farm or solar installation."}],[{"start":118.05,"text":"“Companies need to go factory by factory investing in bits of kit that are often bespoke,” says Sam Kimmins, director of energy at the Climate Group, a non-profit that works with companies and governments on meeting global net zero goals."}],[{"start":132.75,"text":"But if complexity is hampering progress on energy efficiency, it is something AI is effective at managing. “AI is really good when you’ve got complex and highly unpredictable data,” says Kimmins."}],[{"start":145.45,"text":"In factories and other large industrial operations, he explains, sensors attached to every piece of equipment generate large volumes of data that AI can use to identify inefficiencies in equipment or processes."}],[{"start":158.75,"text":"When it comes to physical infrastructure and the built environment, AI can be combined with digital twin technology, which creates digital representations of the real world."}],[{"start":168.15,"text":"Developed in the 1960s for Nasa’s spacecraft simulations, digital twin technology simulates the workings of an entire facility, enabling potential changes to be tested virtually before making physical adjustments."}],[{"start":182.05,"text":"AI expands its possibilities, as it means efficiency adjustments can be made to new building designs or to the maintenance and operation of existing structures at a faster rate."}],[{"start":192.4,"text":"“AI does a lot of the thinking for the team, makes suggestions and takes some of the guesswork and human interaction out of it,” says Kimmins."}],[{"start":200.45000000000002,"text":"One study found that applying AI-powered digital twin models to renewable energy facilities could cut unplanned downtime by 35 per cent and increase energy production by 8.5 per cent. It also improved fault detection accuracy by 98 per cent while reducing energy costs by 26 per cent."}],[{"start":220.25000000000003,"text":"Other studies indicate similar efficiencies could be made in other sectors. In agriculture, for example, researchers found that AI could potentially enable large reductions in the power used to heat greenhouses, and predict energy consumption patterns in irrigation based on factors such as light and heat, allowing farmers to plan watering to minimise energy use."}],[{"start":240.80000000000004,"text":"Of course, set against these types of gains is the power consumed by AI itself. “Given the phenomenal amount of energy AI uses, it’s going to have to do a lot of heavy lifting on energy efficiency to wash its face,” says Kimmins."}],[{"start":255.15000000000003,"text":"Nor will AI alone provide all the answers. “A lot of people want to throw AI at everything, but systems need to be ready for it,” says Quilodrán-Casas."}],[{"start":263.70000000000005,"text":"Stephen Horrax, UK and Ireland director of energy at engineering consultancy Ramboll, agrees. “AI is unquestionably a powerful enabler, but it works best when combined with electrification, infrastructure investment and supportive policy,” he says. “Without those, its impact will be incremental.”"}],[{"start":283.25000000000006,"text":"However, some believe the buzz surrounding AI could in itself help direct more resources towards energy efficiency. “If it makes it more exciting to engage with energy efficiency, that’s great,” says Kimmins. “Because we have a lot of really simple solutions that we need to implement to save a lot of money and energy.”"}],[{"start":303.00000000000006,"text":"Motherway thinks both enthusiasm for AI adoption and the technology’s data management capabilities could accelerate implementation: “We could be at a moment where we could step up efficiency progress, particularly in industry, unlocked by AI.”"}],[{"start":324.9000000000001,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1781524327_3899.mp3"}