Powering data centers: how AI is shaping energy’s next chapter
Currents: AI & Energy Insights - September 2024
Welcome back to Currents, a monthly column from Reimagine Energy dedicated to the latest news at the intersection of AI & Energy. Every last week of the month, I’m sending out an expert-curated summary of the most relevant updates from the sector. The focus is on major industry news, published scientific articles, a recap of the month’s posts from Reimagine Energy, and a dedicated job board.
1. Industry news
The race for AI supremacy is fueling an unprecedented scramble for energy. Microsoft plans to reopen Pennsylvania’s dormant Three Mile Island nuclear plant by 2028, buying all its power for 20 years. Sam Altman pitched the White House on building massive 5GW data centers—each consuming as much power as three million homes. Y Combinator startup Lumen Orbit suggests moving data centers into space to harness unlimited solar energy. Meanwhile, Meta is researching Reinforcement Learning applications to optimize data center cooling, cutting energy and water usage. There’s also increased interest in edge computing, pushing AI processing closer to where data is generated, reducing latency and easing the strain on central data centers.
What I’m thinking: OpenAI and Microsoft seem to believe that we haven’t reached the limits of transformer architectures yet, and want to exponentially scale up compute (Mark Zuckerberg seems to agree). The predictions on the energy demands of data centers are so bold that some people believe the economics of operating data centers in space might become viable. Demand response will have a massive role to play here. Edge computing might be part of the solution, with Apple apparently embracing this route as well.
San Francisco-based Voltus has unveiled AI Adjuster, a new service that boosts demand response earnings for customers. It uses AI to adjust participation in real-time, tapping into untapped load flexibility. Meanwhile in Europe, Beebop.ai raised $5.5 million seed round to integrate residential devices into the power system. Their software turns household items like storage units, heat pumps, and EV chargers into assets that support the grid.
What I’m thinking: Over the next decade demand response will become the new normal for all energy users. Companies in both the U.S. and Europe are pushing ahead, even though the market structure is in development. Different countries have different regulations, making things tricky. But those who figure out the tech now will be way ahead once the regulatory framework settles.
Schneider Electric has unveiled the Building Decarbonization Calculator, an online tool that helps building owners and operators quickly assess energy and carbon conservation measures. Drawing from nearly 500,000 building performance models, the calculator offers recommendations on construction, energy efficiency, and decarbonization strategies. It’s designed to aid compliance with strict regulations like New York City’s Local Law 97, which will impose fines on buildings exceeding emissions limits starting in 2025. The tool also helps users prioritize retrofit plans based on carbon reduction and financial return on investment.
What I’m thinking: This caught my eye since I’ve been leading development for a similar product, Ento Strategy, over the past year. If a giant like Schneider is diving into this space, it signals a huge market potential. The real game-changer for me will be the user experience. The success of these tools will depend on how they can connect with various data sources and support multiple certifications and regulations.
2. Scientific publications
Toward global rooftop PV detection with Deep Active Learning. The authors developed a Deep Active Learning method to detect photovoltaic panels in satellite and aerial images. By intelligently selecting which images to label, they reduced the required dataset by 97% compared to traditional methods. The approach uses uncertainty-based algorithms to focus on the most informative images, boosting model accuracy while saving time and resources.
What I’m thinking: At Ento we developed algorithms to detect PV systems by analysing the electricity time series data of a facility. Being able to cross-reference this with the results of an aerial image analysis could be very interesting and would definitely boost confidence in the results achieved. I’d love to test out this model.
Deep learning-based electricity price forecasting: Findings on price predictability and European electricity markets. This article introduces a deep learning-based toolkit designed for day-ahead electricity price forecasting in Europe. It uses a multilayer perceptron model that recalibrates daily and optimizes hyperparameters annually, outperforming traditional models in both speed and accuracy. Findings reveal that while electricity prices have become more volatile due to events like COVID-19 and the global energy crisis, this volatility does not necessarily equate to unpredictability when using advanced deep learning models.
What I’m thinking: Highly volatile (but easily predictable) power prices are unlocking the deployment of AI-powered demand-side flexibility solutions. Interestingly, time series forecasting remains a domain where transformer architectures haven’t yet become dominant. Instead, traditional models like multilayer perceptrons (MLPs) continue to offer simplicity and computational efficiency while delivering state-of-the-art results.
3. Reimagine Energy publications
The second part of my Python tutorial on integrating building energy consumption and rooftop solar generation data went out last week:
Can we power a building entirely with solar energy?
Solar power is on the rise globally, driven by falling installation costs and higher efficiencies. But what does it take to power an entire facility only with solar?
4. AI in Energy job board
This space is dedicated to job posts in the sector that caught my attention during the last month. I have no affiliation with any of them, I’m just looking to help readers connect with relevant jobs in the market.
Lead Data Engineer at Eneco
Head of Data Science at Yottar
Senior Data Scientist Energy and Utilities at Deloitte
Postdoctoral Scholar in Machine Learning and Data-Driven Optimization and Analysis at Future Institute Cities (University of Waterloo)
Conclusion
With so much going on in the sector it’s not easy to follow everything. If you’re aware of anything that seems relevant and should be included in Currents (job posts, scientific articles, relevant industry events, etc.) please reply to this email or reach out to me on LinkedIn and I’ll be happy to consider them for inclusion!