There are giant, complex machines out there that we all rely on. Without them, civilisation as we know it would collapse.But these machines – power stations – are often pretty dumb, according to people.
“Power plants,are just robots that don’t have a brain yet.”
For years, NeuCo had been developing optimisation technologies – a form of artificial intelligence or AI – that can make power plants more efficient.
The idea is to get a computer to monitor the hundreds of fine-grained controls that may be altered in, for example, a coal-fired power plant, and learn how to adjust them in a more effective way.
Human operators in such facilities are tasked with overseeing all kinds of minutiae, such as the level of oxygen in the furnace, the frequency of the soot blowers that keep tubes in the system clean, or the build-up of slag that, if left unchecked, can grow into huge boulders ready to break off and wreck the equipment.
“There’s too much data and it overwhelms the human ability to respond,” explains Mr Kirk , MD.
Instead, a computer can take over. Machine learning allows software to identify small changes that improve the efficiency and stability of the coal-firing system. The result, Mr Kirk says, is sometimes an efficiency improvement of about 1%.
Software is used to improve a power station’s efficiency and stability
That might not sound like much, but coal power plants are massive carbon emitters.
“I mean, that’s 1,000 cars coming off the road,” says source
GE Power plans to develop this technology, which has already been used in many plants around the world. It has set up a new development centre at the Birchwood coal plant in Virginia.
AI, of course, is booming. The sophistication of machines that can recognise patterns or rules and automate a response to them continues to evolve at many firms – from those in retail to financial services.
But the tech is also cropping up in energy, not always the quickest sector to adapt to new technologies.
GE has another example of how AI can help – with wind turbines. The idea is to better predict the likely output from turbines, based on weather patterns, so that maintenance days can be more accurately scheduled for times when they are less likely to be operational.
More from Future Energy
Will buses be run on coffee?
China leads world in solar power production
LA crystals turn cars into energy source
The challenge is in getting AI to respond to situations with the kind of nous we expect from it, of course.
“We as humans understand that New Year’s [Eve] is a time when we use more power, but artificial intelligence doesn’t know that, it just sees patterns,” he says.
Computer brains can also be used to improve control at the other end of the energy supply chain – demand.
Google reported last year that it had been able to cut the energy used by cooling and support systems at its data centres by 15%.
AI from the firm’s artificial intelligence division, DeepMind, was able to predict more accurately when cooling equipment – essential to keep hot servers running – should be switched on.
This was achieved by carefully analysing when people were more likely to access Google services like YouTube, and thereby increase the load on servers.
Engineers realised that it was more efficient at some times to spread the cooling load across lots of devices, rather than running fewer fans more intensively, according to Jim Gao, Google’s data center engineer.
But shutting cooling off entirely was not an option.
“If you shut off all the cooling you could probably go for a few minutes at most before your servers would melt or exceed temperature thresholds,” he told the BBC.
Data centres consume gargantuan amounts of electricity, largely because of those cooling systems. It is no surprise, then, that other technology firms are trying to achieve similar results.
If smarter power plants and electrical appliances mean reduced emissions, and lower costs, for both supplier and customer, there’s good reason to believe we’ll see the trend continuing.