Artificial intelligence has been in the headlines a lot lately. Rapidly increasing energy demandespecially rapidly increasing Data center power usage Enables training and deployment of modern generative AI models. But it’s not all bad news. Some AI tools have the potential to reduce some types of energy consumption and enable a cleaner grid.
One of the most promising applications is the use of AI to optimize the power grid, making it more efficient, more resilient to extreme weather events, and enabling the integration of more renewable energy. To learn more, Massachusetts Institute of Technology News I talked to priya dontithe Silverman Family Career Development Professor in the Massachusetts Institute of Technology’s School of Electrical Engineering and Computer Science (EECS) and principal investigator in the Institute for Information and Decision Systems (LIDS), where his research focuses on applying machine learning to optimize the power grid.
question: Why should we optimize the power grid in the first place?
answer: We need to maintain a precise balance between the amount of power put into the grid and the amount of power output at any given moment. However, there is some uncertainty on the demand side. Utility companies do not require customers to register in advance how much energy they use, so some estimation and forecasting must be done.
Second, on the supply side, there is usually some variation in costs and fuel availability that grid managers must respond to. This problem is exacerbated by the integration of energy from time-varying renewable sources, such as solar and wind, where weather uncertainties can have a significant impact on the amount of power available. And at the same time, as power flows through the power grid, power is lost through resistive heating in the power lines. So, as a grid operator, how do you make sure everything is working all the time? That’s where optimization comes in.
question: How can AI be used in power grid optimization?
answer: One way AI can help is by using a combination of historical and real-time data to more accurately predict the amount of renewable energy available at a given point in time. This allows these resources to be processed and utilized more effectively, potentially resulting in a cleaner power grid.
AI can also help tackle the complex optimization problems that grid operators must solve to balance supply and demand in a way that also reduces costs. These optimization problems are used to determine which generators should produce power, how much to produce, when to produce it, when to charge and discharge batteries, and whether power load flexibility can be exploited. These optimization problems are computationally very expensive, so operators use approximations to help solve them in a feasible time. But these approximations are often wrong, and as we integrate more renewable energy into the grid, energy deviates even further. AI can help by providing more accurate approximations faster, which can be deployed in real-time to help grid operators manage their grids in a responsive and proactive manner.
AI could also help plan the next generation of power grids. Power grid planning requires the use of large-scale simulation models, and AI can play a big role in making these models run more efficiently. This technology also helps with predictive maintenance by detecting where abnormal behavior on the power grid is likely to occur and reducing inefficiencies due to outages. More broadly, AI can also be applied to accelerate experiments aimed at developing better batteries, which will allow more energy from renewable sources to be integrated into the grid.
question: How should we think about the pros and cons of AI from an energy sector perspective?
answer: One important thing to remember is that AI refers to a disparate set of technologies. There are many different types and sizes of models used, and different ways to use them. If you have a model trained on a small amount of data with a small number of parameters, it will consume significantly less energy than a large general-purpose model.
In the energy sector context, there are many places where the cost-benefit trade-off is favorable when using these application-specific AI models for targeted applications. In these cases, applications are realizing benefits from a sustainability perspective, such as integrating more renewable energy into the grid or supporting decarbonization strategies.
Overall, it’s important to consider whether the types of investments we’re making in AI are actually aligned with the benefits we want from AI. On a societal level, I think the answer to that question at this point is “no.” While there is a lot of development and enhancement of specific subsets of AI technologies, these are not the technologies that will provide the greatest benefits across energy and climate applications. That’s not to say these technologies aren’t useful, but they’re incredibly resource-intensive and aren’t responsible for most of the benefits felt in the energy sector.
I’m excited to develop and ensure robust deployment of AI algorithms that take into account the physical constraints of the power grid. This is a difficult problem to solve. Even if the LLM says something slightly wrong, we as humans can usually correct it in our heads. But mistakes of the same magnitude when optimizing power grids could result in large-scale power outages. Although we need to build models differently, this is also an opportunity to leverage our knowledge of how the physics of power grids work.
And more broadly, I think it’s important that those of us in the tech community work towards fostering a more democratized system of AI development and deployment, and that it’s done in a way that’s tailored to the needs of the applications on the ground.
