On October 1st, Agoria launched its latest digital course on Energy & AI. Not coincidentally two topics close to my heart. I joined the discussion about the role of AI in Energy applications. For those who missed it, here’s a quick recap.
AI and energy? What’s the deal?
Electricity systems are responsible for about a quarter of human-caused greenhouse gas emissions each year. The urgent need to create a more sustainable way of producing electricity pushes the demand for low-carbon electricity high on the agenda. And that means renewable energy sources.
There’s a hiccup, however. Renewable energy sources, like solar and wind, are subject to various circumstances. This variety creates a strong need for better forecasting of energy demand and production, greater coordination between all actors, and more flexible consumption to ensure that power grids can be operated safely and reliably. Digitalisation and machine learning have a critical role to play in fulfilling these needs.
What’s happening already?
Priya L Donti, researcher at Carnegie Melon University and co-founder of Climate Change AI, took Agoria’s stage and gave us a broad overview of AI & Energy applications that people are already working on across the world. Fascinating and promising projects are happening. Some examples:
- Forecasting energy supply & demand. Weather affects both variable source generation (wind & solar energy) and demand (domestic cooling & heating). This leads to a growing importance of weather forecasting.
- Improving scheduling and flexible demand to balance the electric grid in real time.
- Accelerating materials science for new batteries or synthetic fuels.
- Battery farm management to help control batteries located at solar and wind farms to increase these farms’ profits, for instance by storing their electricity when prices are low and then selling it when prices are high; prior work has used ML to forecast electricity prices.
- Accelerating Fusion Science by guiding experimental design, monitoring physical processes
- Improving clean energy access by managing microgrids, or off-grid methods to displace diesel generators, wood-burning stoves, and other carbon-emitting energy sources.
- Data-efficient ML techniques while ML methods have often been applied to grids with widespread sensors, system operators in many countries do not collect or share system data.
Things are certainly moving. All the synergies between energy applications and AI she touched, prove the role AI has to play in this upcoming Shift. But what about possible pitfalls or hurdles?
During the Q&A session with Priya, I asked about the main blockers in bringing AI & Energy solutions to production environments.
Priya first mentioned that testing AI & Energy use cases is really hard. When building digital products, we’re used to extensive test scenarios and rolling out features gradually, and rolling them back in seconds if things go wrong.
These design patterns are impossible when managing an energy grid. It’s all or nothing, and it cannot go wrong. We don’t see a clear solution yet, but building new and proper testing tools would be a clear step in the right direction.
The second blocker was about sharing experimental data. Many grid operators are busy building non-interoperable solutions on themselves, which will lead to suboptimal coordination and incompatibilities in the future. We need more collaboration and an open culture where successes and failures are shared so we can all learn faster together.
Shift towards consumer-centric energy services
What about the consumer side of things? Energy consumers want reliable energy, use new and clean technology, and minimise costs. This creates a clear opportunity to build digital products for consumers to interact with the energy system.
We’re entering a paradigm shift in how users interact with the electricity system. Today, energy production mostly follows consumer demand. We use energy when we need it. Yet as we’re shifting to a growing proportion of variable energy sources, consumers will see more incentives to shift energy consumption to the time slots when variable sources are producing energy. Energy will be free to use in some periods, and we’ll be paying a costly fee when we cross peak usage thresholds.
Users are more likely to embrace this shift when interacting with the energy system through a reliable, transparent and intuitive digital experience that resolves a clear set of consumer needs. A challenge? Most definitely. But it’s the kind of challenge we like to get their heads around.