Artificial Intelligence has recently made headlines for itself rapidly increasing energy demandand especially bounce power usage of data centers Which enables training and deployment of the latest generic AI models. But it’s not all bad news — some AI tools have the potential to reduce certain types of energy consumption and enable cleaner grids.
One of the most promising applications is using AI to optimize the power grid, which will improve efficiency, increase resilience to extreme weather and enable the integration of more renewable energy. To know more, MIT News talked to Priya DontiSilverman Family Career Development Professor in MIT’s Department of Electrical Engineering and Computer Science (EECS) and a principal investigator in the Information and Decision Systems (LIDS) Laboratory, whose work focuses on applying machine learning to optimize the power grid.
Why: Why does the power grid need to be optimized in the first place?
A: We need to maintain a precise balance between the amount of electricity fed into the grid and the amount of electricity coming out at all times. But on the demand side, we have some uncertainty. Electricity companies do not ask customers to pre-register the amount of energy they will use ahead of time, so some estimation and prediction must be done.
Then, on the supply side, there is usually some variation in cost and fuel availability that grid managers need to be responsive to. This has become an even bigger issue due to the integration of energy from time-varying renewable sources such as solar and wind, where uncertainty in the weather can have a major impact on how much electricity is available. Then, at the same time, depending on how power is flowing in the grid, some power is lost through resistive heat on the power lines. So, as a grid operator, how do you ensure that everything is working at all times? This is where customization comes in.
Why: How can AI be most useful in power grid optimization?
A: One way AI can be helpful is to use a combination of historical and real-time data to make more accurate predictions about how much renewable energy will be available at a given time. This could lead to a cleaner power grid by allowing us to handle and better utilize these resources.
AI can also help tackle the complex optimization problems that power grid operators must solve to balance supply and demand while also reducing costs. These optimization problems are used to determine which power generators should produce electricity, how much they should produce, and when they should produce them, as well as when the 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 heuristics so that they can solve them in a reasonable time. But these estimates are often wrong, and they get further off the mark as we integrate more renewable energy into the grid. AI can help by providing more accurate estimates in a faster manner, which can be deployed in real-time to help grid operators manage the grid responsibly and proactively.
AI may also be useful in planning the next generation power grid. Planning power grids requires using huge simulation models, so AI could play a big role in running those models more efficiently. This technology can also help with predictive maintenance by detecting where on the grid unusual behavior is likely to occur, reducing inefficiencies caused by outages. More broadly, AI can also be applied to accelerate experiments aimed at creating better batteries, which would allow the integration of more energy from renewable sources into the grid.
Why: How should we think about the advantages and disadvantages of AI from an energy sector perspective?
A: An important thing to remember is that AI refers to a diverse set of technologies. There are different types and sizes of models used, and different ways of using the models. If you are using a model that is trained on a small amount of data with a small number of parameters, it will consume much less energy than a larger, general-purpose model.
In the context of the energy sector, there are a number of places where, if you use these application-specific AI models for the applications they are intended for, the cost-benefit tradeoff works in your favor. In these cases, the applications are enabling benefits from a sustainability perspective – such as incorporating more renewable energy into the grid and supporting decarbonization strategies.
Overall, it is important to think about whether the type of investments we are making in AI really match the benefits we want from AI. On a societal level, I think the answer to this question right now is “no.” There has been much development and expansion of a particular subset of AI technologies, and these are not the technologies that will have the greatest benefit in energy and climate applications. I’m not saying that these technologies are useless, but they are incredibly resource-intensive, as well as not accounting for a large portion of the gains in the energy sector.
I’m excited to develop AI algorithms that respect the physical constraints of power grids so we can deploy them reliably. This is a difficult problem to solve. If an LLM says something that is slightly wrong, as humans we can usually correct it in our minds. But if you make that big a mistake while optimizing the power grid, it could lead to massive blackouts. We need to model differently, but it also provides an opportunity to benefit from our knowledge of how the physics of the power grid works.
More broadly, I think it’s important that those of us in the tech community put our efforts toward promoting a more democratic system of AI development and deployment, and do so in a way that is tailored to the needs of grassroots applications.
