BrainBox AI Grid Integration Lead and AI expert, Arthur Montes, recently attended the AEE WORLD 2024 conference in Nashville. With over five days of technical sessions covering everything from nuclear power plants to cloud-based AI solutions, the conference brought together the brightest minds in energy management.
We sat down with Arthur to discuss his key takeaways, including the evolution of AI in the energy industry and the role of a diversified mix of energy sources for a more sustainable future.
Arthur, tell us about your experience at AEE. Were there any key themes that stood out to you?
AEE was eye-opening. It was filled with industry experts deeply committed to real solutions. I left with a much deeper understanding of where the industry is heading.
In terms of themes, there were a few clear focal points. Emission reduction was a major topic, especially in balancing grid demands while pushing toward net-zero targets. Another recurring theme was grid flexibility and balancing energy production with consumption.
Let’s start with emission reduction. What made it such a topic of interest at AEE?
Reducing emissions, particularly in the built environment, is critical as it directly aligns with our urgent need for sustainable energy practices and enhanced energy efficiency. As global energy demands continue to grow—fueled by the expansion of data centers and the electrification of transport—the composition of our electricity mix becomes even more important. This mix, which still relies heavily on fossil fuels but increasingly incorporates renewables like solar and wind, plays a significant role in shaping a building’s overall carbon footprint.
To reduce this footprint effectively, a dual approach is necessary: improving building energy efficiency while simultaneously decarbonizing the electrical grid. This combination ensures that not only are buildings optimized for energy use, but also that the energy powering them is as clean as possible.
Interestingly, the conversation around sustainable energy has changed somewhat—something I noticed at AEE. For example, nuclear energy, once considered outdated, is experiencing a resurgence thanks to advancements like small modular reactors (SMRs) and enhanced safety protocols. This shift highlights how innovations in legacy technologies can play a key role in achieving our climate goals.
Were there any talks on GHG reduction at AEE that you specifically enjoyed?
There were many. One of my favorite sessions was “Navigating the Green Transition: A Strategic Roadmap for Decarbonization in Commercial Real Estate” presented by Ali Hoss, CSO of Triovest in Canada. His presentation offered a strategic approach to decarbonization tailored for the commercial real estate sector, emphasizing the urgent need for sustainability as a competitive, compliance-driven, and investor-satisfying strategy. The session broke down a comprehensive roadmap for decarbonizing real estate assets, focusing on energy efficiency, renewable integration, electrification, and deep retrofits.
Another standout for me was Kelley Whalen’s session, “Decarbonization: Integrating It into Strategic Energy Management”. In her talk, she highlighted the importance of embedding decarbonization into Strategic Energy Management (SEM), where planning, monitoring, and implementing energy-saving measures play a critical role in reducing operational costs and environmental impact. Whalen’s insights underscored how SEM supports broader ESG goals by driving energy efficiency and carbon reduction through renewable energy, advanced technology adoption, and data-driven decisions. This session emphasized that, to achieve a sustainable future, energy management must go beyond cost savings and foster a culture of environmental stewardship, social responsibility, and governance.
What I liked was that both these sessions offered actionable insights and underscored the growing momentum in the industry to align real estate practices with sustainability and decarbonization goals.
You mentioned that grid flexibility was also a key theme at AEE. Could you tell us a bit about that?
Sure. So, one key focus was around harmonizing energy demand with production – essentially grid flexibility. What really stood out to me during the conference was the sense that the industry is shifting from simply trying to increase renewable energy production to making the best use of our existing infrastructure.
So, you see a shift away from discussing greener energy sources and toward grid flexibility? Is this a “yes, and…” moment, where we assume more renewables will be implemented and now, we have address how we can prep our grids for that?
Exactly. For years, the focus was on scaling up renewable energy sources like solar and wind. But now, the conversation has evolved. While increasing renewables is still crucial, it’s become equally important to optimize how we use the energy we already have. This shift emphasizes preparing our grids to be more flexible and efficient.
"The conversation has evolved. While increasing renewables is still crucial, it’s become equally important to optimize how we use the energy we already have."
In this context, achieving a balance is essential. It’s like trying to catch a runaway train: you must do two things at once—accelerate your pace (enhancing energy production) while also applying the brakes (reducing unnecessary energy demand). Only by harmonizing these efforts—optimizing production while minimizing waste—can we close the gap between what we generate and what we truly need. This alignment is crucial if we want to make real progress in the energy transition.
So, it’s all about achieving balance through grid flexibility - matching energy production precisely with consumption?
Yes, but it’s also important to recognize that grid flexibility plays a different role depending on the time of day and the nature of demand. To use our example from above, nuclear power plants operate 24/7, even when demand is low at night. But grid flexibility alone won’t solve that specific inefficiency since nighttime demand is naturally low.
Where grid flexibility really shines is during periods of peak consumption—like on a hot summer day when air conditioning use is at its highest. In these situations, the existing production capacity often isn’t enough to meet the spike in demand, forcing utilities to turn on marginal power plants, which usually have a much higher carbon output. By leveraging grid flexibility, we can shift energy consumption away from those peak hours. This prevents the need to activate those high-emission backup plants, reducing overall carbon impact.
In essence, it’s about using smarter, more responsive systems to align demand with sustainable production, ensuring we can meet our energy needs without resorting to less efficient, carbon-heavy solutions. That’s where the true value of grid flexibility lies—not just in balancing production and consumption, but in optimizing when and how we use the energy that’s already being produced.
Can AI help us optimize energy consumption?
Absolutely, AI can play a crucial role in optimizing energy consumption, especially when it comes to leveraging grid flexibility. Let’s take buildings as an example. During peak demand periods, like a hot summer afternoon when everyone’s cranking up their air conditioning, the grid often struggles to meet the surge in consumption. This is where AI can make a significant difference.
In a typical scenario, if demand spikes beyond the grid’s capacity, utilities have to turn on marginal power plants that emit a lot of carbon. But AI-powered systems can predict these peaks ahead of time and adjust a building’s energy usage accordingly, pre-cooling a building in the early morning hours when demand on the grid is lower and electricity is cheaper. Then, during peak hours, we can reduce the HVAC load, shifting the building’s energy consumption away from those critical times.
By optimizing how and when a building uses energy, AI helps alleviate pressure on the grid, preventing the need to activate those high-emission backup plants.
Was AI a big topic of conversation at AEE?
Oh yes. I actually presented on that very topic, though it was mostly educative and focused on the fundamentals. Basically, I broke AI down into its core components: machine learning (ML), deep learning (DL), and generative AI (Gen AI), citing use cases for each.
There was definitely a lot of interest in the subject. One thing that struck me, though, was just how much large language models dominated the conversation throughout the conference – and how many missed opportunities there are for employing more traditional machine learning methods.
Can you explain where traditional machine learning might be more effective than generative AI?
Sure. So, generative AI definitely has its strengths (especially when it comes to handling complex, open-ended tasks like language processing or making sense of unstructured data). However, we need to recognize that different challenges require different tools.
For instance, in grid management, when it comes to load forecasting, a Long Short-Term Memory (LSTM) model can do an excellent job. LSTM models are designed to handle time-series data, making them ideal for predicting energy loads based on historical patterns. Using a generative AI model in this context could introduce unnecessary complexity and computational overhead. Similarly, for real-time decision-making systems where response time is critical—like optimizing HVAC systems in buildings —generative AI by itself might introduce latency.
That's why tech like
ARIA (BrainBox AI's virtual building engineer) is so powerful. It has the best of both worlds: combining traditional machine learning techniques, like reinforcement learning, with generative AI capabilities. That way, it has a narrower, more targeted base of data to pull from, which results in more accurate outputs and less energy consumed.
The way I see it is that GenAI tools like ChatGPT are incredibly powerful, but they can be overkill for some specific use cases – a bit like using a rocket launcher to swat a fly. Sometimes, a straightforward machine learning model would do the job just as effectively, using less computational power and resources. And sometimes you need a combination of both. It’s all about choosing the right tool for the task.
"GenAI tools like ChatGPT are incredibly powerful, but they can be overkill for some specific use cases – a bit like using a rocket launcher to swat a fly. Sometimes, a straightforward machine learning model would do the job just as effectively, using less computational power and resources. And sometimes you need a combination of both. It’s all about choosing the right tool for the task."
What advice would you give to companies that are just starting to scope out and integrate these AI tools?
I’d say: Start with knowing what problem you’re trying to solve. Understand your problem will naturally help you choose the solution that fits best – it’s like cooking a nutritious meal instead of grabbing fast food. It takes a little more effort but, in the long run, it’s ultimately more beneficial.
Discover how BrainBox AI uses traditional machine learning and GenAI to create powerful energy-optimizing tools
