Should we convert the power grid to artificial intelligence?
Should we convert the power grid to artificial intelligence?
By Ross Pomeroy | Published: 2024-12-14 16:00:00 | Source: The Future – Big Think
Sign up for Big Think on Substack
The most surprising and impactful new stories delivered to your inbox every week for free.
California, August 2048. Much of the state has been experiencing an intense two-week heat wave, with temperatures consistently reaching 100 degrees Fahrenheit. The air is dry. The ground is dry. Plants kindle. In this box, suddenly there is a spark…
Within 15 minutes, the power grid’s artificial intelligence (AI) system detects signs of smoke and fire in satellite images and alerts human operators during service. They sent a drone to the area to confirm the emerging forest fires. The flying robot feeds live, high-resolution video and images to the AI system to ascertain any potential danger to the electrical infrastructure. According to the rapid assessment conducted by artificial intelligence, the escalating fire could threaten major substations and electrical poles, threatening power outages to tens of thousands of customers.
The AI then presents mitigation options to grid operators. They quickly chose to reroute power in the area and bring an energy storage facility online, as well as a virtual power plant consisting of thousands of electric vehicle and home batteries for customers. Although a substation is soon lost, there are no power outages. The power continues to flow. The network remains stable.
The electricity grid today
America’s power grid is a giant, complex, mixed network. As described by researchers at the Department of Energy, Keith J. Bennis, and Joshua E. Porterfield, and Charles Yang in a 2024 expanded study a report“It consists of tens of thousands of power generators that deliver electricity across more than 600,000 miles of transmission lines, 70,000 substations, 5.5 million miles of distribution lines, and 180 million utility poles. This system has evolved organically over a century of incremental additions, and now operates at the heart of America’s $28 trillion economy.”
In this tangled web, old wires, hybrid transformers, and crumbling poles intertwine with modern superconducting materials, smart grid meters, sensors, and massive lithium-ion battery installations. Somehow, it all works out. The US power grid has an astonishing 99.95% overall reliability. And this He has To be this good. Americans’ lives and jobs depend on electricity 24 hours a day, 7 days a week.
“Even a power grid that is 99% operational will leave people and businesses without power for 3.5 days per year,” the authors point out.
Now, our highly complex electricity grid is about to undergo perhaps the most disruptive renovation ever. To mitigate catastrophic climate change, operators must build and integrate disparate, often intermittent, carbon-free electricity sources. Wind, solar, battery storage, hydroelectric and nuclear power are needed in large quantities. Moreover, thousands of miles of high-voltage wires must be laid to deliver the generated electricity to the places where it is actually used. But is it possible to make these power systems work well together, while maintaining impeccable grid reliability and affordability? AI may be the key to making it all work.
Control room of the future
The scenario presented at the beginning of this article could not happen today, but it could happen sooner than one might think. In May of this year, a team of researchers from the National Renewable Energy Laboratory (NREL) published a report Technical report Describing “eGridGPT”, their generative AI model is designed to virtually support power grid control room operators by aiding decision-making processes and interpreting data and models.
As the power grid grows more complex — with increasingly variable power generation, electricity flowing to and from customers, and piles of data from wide-ranging sensors — NREL scientists imagine it may become more difficult for human grid operators to comfortably manage it on their own. So they developed the AI assistant “to act as an interface between the screen in front of the operator and the coordinator for the comprehensive processing of large amounts of data, scenarios and coupled digital simulations.”
In the future network, NREL researchers envision humans as the decision makers, but artificial intelligence will inform and perhaps enforce those decisions.
Seung Choi He is a principal engineer at the National Renewable Energy Laboratory’s (NREL) Power Systems Engineering Center and lead author of the eGridGPT report.
“The primary goal of eGridGPT is to help operators make decisions in near real-time by analyzing large data sets, recognizing patterns, simulating scenarios, and suggesting mitigation strategies,” he explained to Freethink.
Some network operators have already shown interest in testing it, Choi says.
Artificial intelligence and power supply
With the emergence of large language models and Power-hungry data centers needed to operate them, AI has recently been seen as an obstacle to decarbonising the grid. But many scientists and network experts say that artificial intelligence systems are being used Deep learning And machine learning is exactly what we need for a future grid powered by renewable energy sources.
“By harnessing the capabilities of artificial intelligence, it is possible to develop intelligent systems that can adapt to dynamic environmental conditions, predict energy production, and optimize resource allocation,” said a team of engineers from the University of Johannesburg in South Africa. books In a review published earlier this year.
The model can cut overall maintenance costs by half.
Integrating AI into the energy grid can bring a range of benefits. For starters, it can monitor large amounts of data about individual assets such as solar panels, wind turbines, and inverters to determine what needs maintenance, thus minimizing downtime, maximizing production, and reducing costs for utilities and ratepayers.
Energy equipment giant GE Vernova is already using AI for predictive maintenance of wind farms. By analyzing data from sensors on wind turbines, machine learning algorithms predict potential equipment failures before they occur. It was developed by scientists at Argonne National Laboratory An AI model that predicts failure and works with a range of components, allowing facilities to schedule maintenance accordingly. In one case, where they tested it on solar inverters, they found that the model could cut total maintenance costs by half and unnecessary crew visits by two-thirds.
Another potential benefit of artificial intelligence: matching supply and demand more closely. To ensure the availability of electricity, the former must always match the latter. But this becomes more difficult because more electricity comes from intermittent wind and solar versus always-on power plants, such as nuclear power plants or natural gas plants that can run on demand. AI can better predict cloud cover for solar panels and wind patterns for turbines, helping forecast future supplies. In turn, it can also use utility data to better predict potential demand during those days. Using this data, grid operators can decide whether they need to pull more resources from other areas or activate battery storage assets.
“As more renewable energy sources become available and extreme weather events become more frequent, operators are facing uncharted territory,” Choi adds. “They often rely on educated guesses to address these new challenges, while AI can provide predictions of possible future scenarios.”
At the ground level, AI models can also help optimize the placement of solar panels to capture the most sunlight throughout the day. For wind turbines, they can adjust blade angles to improve energy capture. Using batteries, it can quickly monitor market conditions, discharging power when it is expensive and storing it when it is cheap. Tesla, one of the leading battery storage companies, is already using artificial intelligence algorithms to do this.
Disadvantages
Ultimately, the power grid is the kind of data-sensitive machine that is ideally suited to AI automation. But there are some reasons why integrating AI may be difficult.
The first is security. As control of the network becomes increasingly in the hands of software, it becomes more vulnerable to traditional cyber attacks. But machine learning models are also at risk of a more specific form of subterfuge: data poisoning. If these systems were somehow fed with garbage data, they could make decisions that lead to mass power outages.
Accuracy is another matter. There are concerns that these models could become biased over time in ways that misalign their goals with the basic need to ensure access to electricity for all.
“Accuracy remains a major challenge, especially in building operators’ trust. Building that trust is a critical goal for AI,” Choi told Freethink. “AI relies on historical data to learn and lacks awareness of current conditions, which means it may not always provide up-to-date insights.”
Second, AI is not without costs. As the University of Johannesburg researchers note, “While AI can improve renewable energy systems and reduce carbon emissions, it also comes with its own environmental footprint. Training AI models requires significant computational power, leading to higher energy consumption and associated carbon emissions.”
Essentially, the network must evolve into one supercomputer.
Third, and more difficult, to get to the point where AI can make reliable decisions at network scale, the network itself will require massive upgrades.
“While today’s grid primarily transmits electricity in one direction from large, centralized power plants to electricity customers with relatively little information exchange, the grid of the future will manage multidirectional flows of energy and information across a variety of grid-connected resources,” the DOE researchers wrote.
You will need to install a large number of sensors across the network to feed models with large amounts of data. SCADA (supervisory control and data acquisition) systems that include a range of electrical components will need to be inserted across network assets. Smart meters should become ubiquitous in homes and businesses. Essentially, the network must evolve into one supercomputer. Completing this unprecedented transformation will cost hundreds of billions of dollars.
Worth the cost
However it will likely be worth it in the long run. the The global cost of climate changeDamage to infrastructure, property, agriculture, and human health is expected to range between $1.7 trillion and $3.1 trillion annually by 2050. A massive boost to renewable energy is needed to avoid these costs, and if artificial intelligence eventually makes this carbon-free grid cheaper and more efficient, that is much better for people and the planet.
this condition Originally published by our sister site Freethink.
Sign up for Big Think on Substack
The most surprising and impactful new stories delivered to your inbox every week for free.
ــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــ
