Artificial intelligence has been making headlines recently rapidly growing energy demandespecially growth electricity consumption in data centers that enable training and implementation of the latest generative artificial intelligence models. But it's not all bad news – some AI tools have the potential to reduce some forms of energy consumption and enable cleaner networks.
One of the most promising applications is the use of artificial intelligence to optimize the power grid, which would improve efficiency, increase resilience to extreme weather conditions and enable the integration of more renewable energy. To learn more, MIT News talked to Priya Dontithe Silverman Family Career Development Professor in MIT's Department of Electrical Engineering and Computer Science (EECS) and a principal investigator in the Laboratory for Information and Decision Systems (LIDS), whose work focuses on applying machine learning to power grid optimization.
Q: Why should you optimize your power grid in the first place?
AND: We need to maintain a precise balance between the amount of energy we put into the grid and the amount of energy released at any given time. However, on the demand side we have some uncertainty. Energy companies do not ask customers to pre-register the amount of energy they intend to use, so some estimates and forecasts must be made.
Then, on the supply side, there are usually some differences in fuel costs and availability that network managers need to respond to. This has become even more of an issue due to the integration of energy from time-varying renewable sources such as solar and wind, where weather uncertainty can have a major impact on the amount of energy available. At the same time, depending on the energy flow in the network, some power is lost due to resistance heat in power lines. So how do you, as a network operator, make sure this all works all the time? This is where optimization comes in.
Q: How can artificial intelligence be most useful in optimizing the energy grid?
AND: One way AI can help is by using a combination of historical and real-time data to make more accurate predictions about the amount of renewable energy available at a given time. This could lead to a cleaner energy grid, enabling us to make better use of these resources.
Artificial intelligence can also help solve complex optimization problems that power grid operators must solve to balance supply and demand in a way that also reduces costs. These optimization problems are used to determine which power generators should produce power, how much power they should produce, when they should produce it, when batteries should be charged and discharged, and whether we can take advantage of flexibility in power loads. These optimization problems are so computationally expensive that operators use approximations to be able to solve them in the time possible. However, these approximations are often wrong, and as we integrate more renewable energy into the grid, they will be thrown off even further. AI can help by providing faster and more accurate approximations that can be deployed in real time to help network operators respond and proactively manage the network.
Artificial intelligence may also be useful in planning next-generation power grids. Power grid planning requires the use of huge simulation models, so artificial intelligence can play a big role in running these models more efficiently. This technology can also help with predictive maintenance by detecting where unusual network behavior is likely to occur, reducing inefficiencies resulting from downtime. More broadly, AI could also be used to speed up experiments to create better batteries that could enable more renewable energy to be integrated into the grid.
Q: How should we think about the advantages and disadvantages of artificial intelligence from the perspective of the energy sector?
AND: It is important to remember that artificial intelligence refers to a heterogeneous set of technologies. Different types and sizes of models are used and different ways of using them. If you use a model trained on less data and fewer parameters, it will use much less power than a large general-purpose model.
In the context of the energy sector, there are many places where using application-specific AI models in the applications for which they are intended provides cost-benefit advantages. In such cases, applications provide sustainability benefits, such as integrating more renewable energy sources into the grid and supporting decarbonization strategies.
Overall, it's important to consider whether the types of investments we make in AI actually match the benefits we expect from AI. On a societal level, I think the answer to this question is currently “no.” A particular subset of AI technologies is being aggressively developed and expanded, but these are not the technologies that will provide the greatest benefits for energy and climate applications. I'm not saying these technologies are useless, but they are incredibly resource-intensive and yet they don't account for the lion's share of the benefits that can be felt in the energy sector.
I'm excited about the opportunity to develop AI algorithms that take into account the physical constraints of the power grid so that we can reliably deploy them. This is a difficult problem to solve. If LLM says something that is slightly incorrect, as humans we can usually correct it in our heads. However, if you make the same mistake when optimizing your power grid, it could result in a large-scale power outage. We have to build models differently, but it also gives us the opportunity to benefit from our knowledge of power grid physics.
More broadly, I believe it is critical that members of the technical community put effort into supporting a more democratized system for AI development and deployment, and that they do so in a way that is tailored to the needs of field applications.
