AI tools could solve a lot of power industry problems — when they're ready
image credit: Photo 101370901 © Alexandersikov | Dreamstime.com
- Apr 20, 2021 10:37 am GMTApr 20, 2021 12:10 am GMT
- 222 views
Articles about the electric power industry’s use of artificial intelligence give many reasons why it’s increasing, but a recent one by Germany’s international broadcaster Deutsche Weller had an unusual one: the pandemic.
Reservations that governments and businesses had about AI “had slowed down the rollout of advanced applications in the energy sector,” the article said. But due to the pandemic, which it called “the ultimate digital catalyst … tools created by researchers over more than a decade are now hitting markets,” and changing “energy systems forever” in the process.
Despite their potential, however, AI tools have a long way to travel before they are widely adopted by the electric power industry, and some are encountering bumpy roads in the process.
For example, the Deutsche Weller article mentions DeepMind, a British subsidiary of Alphabet, that worked with its sister company Google to forecast the power output of wind farms in the U.S. The company said in early February 2019 that “our early results suggest that we can use machine learning to make wind power sufficiently more predictable."
Deep Mind also was talking to National Grid about having its AI used to balance the transmission system that National Grid runs in Great Britain, but the two never reached a deal, according to a story by Sam Shead for Forbes in 2019.
The problem, according to a source quoted in a subsequent Shead story for CNBC, was that Deep Mind wanted to charge more for its services than National Grid was willing to pay. “Most of [DeepMind’s] work is internal and they bill Google for it,” the source told Shead. “They sell the work of their AI engineers at super inflated prices, and not at the price that the market values their output.”
In that same story, Shead reported that several key climate change researchers that were part of DeepMind’s energy unit had left the company and DeepMind had shifted its focus from climate change to other areas, according to several people familiar with the matter.
One of the departing researchers, Jack Kelly, went on to help found Open Climate Fix, which is aiming to build an online solar forecasting service for the U.K. and mainland Europe, according to a recent CNBC article by Shead. The London-based nonprofit plans to use a combination of satellite images and AI software to predict cloud cover, thereby allowing U.K. and European grid operators to better forecast how much solar energy they’re going to get on a given day, in turn allowing them to reduce the amount of fossil fuel generation they have to keep operating at less than its full capacity. Kelly said Open Climate Fix is talking about its service with National Grid, as his previous employer did.
Kelly apparently didn’t burn any bridges in leaving DeepMind, as Google recently awarded Open Climate Fix with $686,350 through its philanthropic arm, Google.org. Open Climate Fix plans to use the money to expand its staff, which currently consists of three people, an indication that the nonprofit — and therefore, probably, its technology — is in its early stages.
The fact that National Grid is talking with Open Climate Fix about using its AI technology isn’t the only evidence that the British grid operator, which also owns utilities in the U.S., still thinks AI could be important to the electric power sector. At least eight of the 29 portfolio companies of its venture investment subsidiary, National Grid Partners, use AI, according to National Grid Partners’ descriptions of them and a press release National Grid Partners put out last October. They range from Audio Analytic, which uses AI to augment consumer products with sound recognition, according to a story by Will Bedingfield for Wired, to AiDash, which uses high-resolution satellite imagery coupled with AI to help energy companies perform such operations and maintenance activities as remote monitoring and vegetation and disaster management.
Products and services that use AI also are being proposed as ways to make Texas’ power grid more reliable and resilient, according to an article by Forbes contributor Jim Magill.
For example, AI’s ability to determine when preventive maintenance is needed could have helped the Electric Reliability Council of Texas to keep track of the physical condition of all the assets on its grid, from generation “all the way down to the thermostats,” during the devastating cold spell that hit Texas in February, said Tom Siebel founder and CEO of AI enterprise company C3.ai.
AI also could have helped ERCOT better balance the load coming from the generation assets on its grid, said Colin Parris, senior vice president and chief technology officer at GE Digital.
Additionally, by providing more accurate long-range weather forecasts, AI could have helped ERCOT determine in advance which units would have to go offline to protect the grid’s integrity, which in turn would have given consumers more time to prepare for planned outages, said Steve Kwan, director of grid management for AI software provider Beyond Limits.
If widely used, those capabilities could, as the Deutsche Weller article said, change “energy systems forever.” But even if the pandemic is spurring their roll out, as the article maintained, their adoption is not liable to occur overnight for a simple reason — they’re still being perfected, and, in some cases, developed.
For example, although AI could have helped ERCOT with load balancing during its recent crisis, National Grid didn’t think the load balancing capabilities of DeepMind’s AI were worth implementing.
Similarly, given that Open Climate Fix is proposing using AI to predict cloud cover for solar operators, it’s likely that the long-range weather forecasting ability that would have helped ERCOT anticipate which generation assets would have been knocked offline by the cold snap that devastated Texas isn’t readily available.
Even uses of AI to help with maintenance, such as those mentioned by Siebel, are still being developed.
Loughborough University, which is named after the English town in which it’s located, recently announced that its researchers have developed a tool that uses AI to analyze images of wind turbine blades to spot and classify defects, which could help reduce the time needed to repair them.
The researchers tested the tool and published the results of their tests in a paper that, a press release said, “also proposes a new set of measures for evaluating defect detection systems, which is much needed given AI-based defect detection and existing systems are still in their infancy.”