Article 6GJMJ Four ways AI is making the power grid faster and more resilient

Four ways AI is making the power grid faster and more resilient

by
June Kim
from MIT Technology Review on (#6GJMJ)
Story Image

The power grid is growing increasingly complex as more renewable energy sources come online. Where once a small number of large power plants supplied most homes at a consistent flow, now millions of solar panels generate variable electricity. Increasingly unpredictable weather adds to the challenge of balancing demand with supply. To manage the chaos, grid operators are increasingly turning to artificial intelligence.

AI's ability to learn from large amounts of data and respond to complex scenarios makes it particularly well suited to the task of keeping the grid stable, and a growing number of software companies are bringing AI products to the notoriously slow-moving energy industry.

The US Department of Energy has recognized this trend, recently awarding $3 billion in grants to various smart grid" projects that include AI-related initiatives.

The excitement about AI in the energy sector is palpable. Some are already speculating about the possibility of a fully automated grid where, in theory, no humans would be needed to make everyday decisions.

But that prospect remains far off; for now, the promise lies in the potential for AI to help humans, providing real-time insights for better grid management. Here are four of the ways that AI is already changing how grid operators do their work.

1. Faster and better decision-making

The power grid system is often described as the most complex machine ever built. Because the grid is so vast, it is impossible for any one person to fully grasp everything happening within it at a given moment, let alone predict what will happen later.

Feng Qiu, a scientist at Argonne National Laboratory, a federally funded research institute, explains that AI aids the grid in three key ways: by helping operators to understand current conditions, make better decisions, and predict potential problems.

Qiu has spent years researching how machine learning can improve grid operations. In 2019, his team partnered with Midcontinent Independent System Operator (MISO), a grid operator serving 15 US states and parts of Canada, to test a machine-learning model meant to optimize the daily planning for a grid comparable in scale to MISO's expansive network.

Every day, grid system operators like MISO run complex mathematical calculations that predict how much electricity will be needed the next day and try to come up with the most cost-effective way to dispatch that energy.

The machine-learning model from Qiu's team showed that this calculation can be done 12 times faster than is possible without AI, reducing the time required from nearly 10 minutes to 60 seconds. Considering that system operators do these calculations multiple times a day, the time savings could be significant.

Currently, Qiu's team is developing a model to forecast power outages by incorporating factors like weather, geography, and even income levels of different neighborhoods. With this data, the model can highlight patterns such as the likelihood of longer and more frequent power outages in low-income areas with poor infrastructure. Better predictions can help prevent outages, expedite disaster response, and minimize suffering when such problems do happen.

2. Tailored approach for every home

AI integration efforts are not limited to research labs. Lunar Energy, a battery and grid-technology startup, uses AI software to help its customers optimize their energy usage and save money.

You have this web of millions of devices, and you have to create a system that can take in all the data and make the right decision not only for each individual customer but also for the grid," says Sam Wevers, Lunar Energy's head of software. That's where the power of AI and machine learning comes in."

Lunar Energy's Gridshare software gathers data from tens of thousands of homes, collecting information on energy used to charge electric vehicles, run dishwashers and air conditioners, and more. Combined with weather data, this information feeds a model that creates personalized predictions of individual homes' energy needs.

As an example, Wevers describes a scenario where two homes on a street have identically sized solar panels but one home has a tall backyard tree that creates afternoon shade, so its panels generate slightly less energy. This kind of detail would be impossible for any utility company to manually keep track of on a household level, but AI enables these kinds of calculations to be made automatically on a vast scale.

Services like Gridshare are mainly designed to help individual customers save money and energy. But in the aggregate, it also provides utility companies with clearer behavioral patterns that help them improve energy planning. Capturing such nuances is vital for grid responsiveness.

3. Making EVs work with the grid

While critical for the clean-energy transition, electric vehicles pose a real challenge for the grid.

John Taggart, cofounder and CTO of WeaveGrid, says EV adoption adds significant energy demand. The last time they [utility companies] had to handle this kind of growth was when air conditioners first took off," he says.

EV adoption also tends to cluster around certain cities and neighborhoods, which can overwhelm the local grid. To relieve this burden, San Francisco-based WeaveGrid collaborates with utility companies, automakers, and charging companies to collect and analyze EV charging data.

By studying charging patterns and duration, WeaveGrid identifies optimal charging times and makes recommendations to customers via text message or app notification about when to charge their vehicles. In some cases, customers grant companies full control to automatically charge or discharge batteries based on grid needs, in exchange for financial incentives like vouchers. This turns the cars themselves into a valuable source of energy storage for the grid. Major utility companies like PG&E, DTE, and Xcel Energy have partnered on the program.

DTE Energy, a Detroit-based utility company that serves southern Michigan, has worked with WeaveGrid to help improve grid planning. The company says it was able to identify 20,000 homes with EVs in its service region and is using this data to calculate long-term load forecasts.

4. Spotting disasters before they hit

Several utility companies have already begun integrating AI into critical operations, particularly inspecting and managing physical infrastructure such as transmission lines and transformers.

For example, overgrown trees are a leading cause of blackouts, because branches can fall on electric wires or spark fires. Traditionally, manual inspection has been the norm, but given the extensive span of transmission lines, this can take several months.

PG&E, covering Northern and Central California, has been using machine learning to accelerate those inspections. By analyzing photographs captured by drones and helicopters, machine-learning models identify areas requiring tree trimming or pinpoint faulty equipment that needs repairs.

Some companies are going even further, and using AI to assess general climate risks.

Last month Rhizome, a startup based in Washington, DC, launched an AI system that takes utility companies' historical data on the performance of energy equipment and combines it with global climate models to predict the probability of grid failures resulting from extreme weather events, such as snowstorms or wildfires.

There are dozens of improvements a utility company can make to improve resiliency, but it doesn't have the time or funding to tackle all of them at once, says Rhizome's cofounder and CEO, Mish Thadani. With software like this, utility companies can now make smarter decisions on which projects to prioritize.

What's next for grid operators?

If AI is capable of swiftly making all these decisions, is it possible to simply let it run the grid and send human operators home? Experts say no.

Several major hurdles remain before we can fully automate the grid. Security poses the greatest concern.

Qiu explains that right now, there are strict protocols and checks in place to prevent mistakes in critical decisions about issues like how to respond to potential outages or equipment failures.

The power grid has to follow a very rigorous physical law," says Qiu. While great at enhancing controlled mathematical calculations, AI is not yet foolproof at incorporating the operating constraints and edge cases that come up in the real world. That poses too big a risk for grid operators, whose primary focus is reliability. One wrong decision at the wrong time could cascade into massive blackouts.

Data privacy isanother issue. Jeremy Renshaw, a senior technical executive at the Electric Power Research Institute, says it's crucial to anonymize customer data so that sensitive information, like what times of day people are staying home, is protected.

AI models also risk perpetuating biases that could disadvantage vulnerable communities. Historically, poor neighborhoods were often the last to get their power restored after blackouts, says Renshaw. Models trained on this data might continue assigning them a lower priority when utilities work to turn the power back on.

To address these potential biases, Renshaw emphasizes the importance of workforce training as companies adopt AI, so staff understand which tasks are and aren't appropriate for the technology to handle.

You could probably pound in a screw with a hammer, but if you use the screwdriver, it would probably work a lot better," he says.

External Content
Source RSS or Atom Feed
Feed Location https://www.technologyreview.com/stories.rss
Feed Title MIT Technology Review
Feed Link https://www.technologyreview.com/
Reply 0 comments