But it’s a “tension-filled discussion” inside utilities, said Elizabeth Cook of Pittsburgh’s Duquesne Light Company, “because there is an urgency to keep the power on”
Image credit: Anne Bailey / Duquesne Light Company
What is it like to evaluate artificial intelligence inside a utility? It’s not straightforward.
At the Pittsburgh utility Duquesne Light Company, director of advanced grid systems and grid modernization Elizabeth Cook is examining a wide range of applications and pitfalls. In a conversation earlier this year, she told me that the discourse about integrating AI into the grid is “very split” within utilities, because new technology “makes everyone uncomfortable.”
However, as grid edge technologies proliferate, she added, the need for AI to improve data analytics and modeling is becoming increasingly clear. It’s just a question of demonstrating its ultimate value to the customer — and to other leaders inside the company.
Our conversation has been edited for brevity and clarity.
Lisa Martine Jenkins: How have AI and machine learning evolved as a part of the grid modernization conversation at Duquesne in your years there? Has it picked up recently?
Elizabeth Cook: When we say artificial intelligence, I feel like it’s a buzzword without a clear definition, so we have to really peel back the onion. AI means embracing the data that’s now becoming available, putting it into a back-end or a cloud, and running analytics that provide insight that we've never had before.
We historically have had a one-way power grid, where electricity travels one way through the transmission, down to a lower voltage distribution, into the homes of our passive customers. Utilities didn’t need to know much beyond the predictability of humans — when we wake up, go to work, come home, go to bed. But now with all the grid-edge technologies, we must see the changing flow of electrons. We need to model it. We need the consumption data to then study the trends and to start actually understanding the true utilization of the assets that we built over the last 140 years.
These new analytics start changing the way we think about the grid of today, and then allow us to alter how we manage and drive the grid of tomorrow. AI will enable the management of the edge in a way that allows this reality.
LMJ: How would you characterize the tenor of conversation about AI’s use that is happening inside utilities?
EC: The conversation is very split. Over the last 60 years, we built a grid to do one thing, and everyone that was part of the utility space, that's what they did. Their job was to ensure that the electrons continue to flow through the assets. Their job was to make sure things work as they were designed to work.
Then you bring in this new technology, which makes everyone uncomfortable. Within the utility’s walls, the challenge is that we have to have some people maintain our core business, but then we have to have others that are visionary, saying “While you're doing that, we must transform you.” It’s a tension-filled discussion because there is an urgency to keep the power on. It’s really hard to defocus those who are in charge of doing that, and those who will need to drive this whole new way of thinking.
LMJ: It sounds like there’s some wariness inherent to the dynamic of utilities, given that power provision is the priority.
EC: Right. Enthusiasm about AI hinges entirely on how this benefits our customers. There’s a continuous need — from an external stakeholder perspective, from the regulatory environment — to show that we’re prudently spending our money.
From transportation electrification to building electrification, all of the drivers are saying that we can do this, and the technology is here. But we need the grid owners to adopt the technology. And even if the grid owners are enthusiastic, there’s that regulatory element. Who’s paying for it? How much are they paying? Can the rates sustain it? Are you even allowed to, based on your regulatory environment that you sit in within the state? Educating the regulators is key, as well as the consumer advocacy groups, so that they all understand where we're going.
There are a lot of utilities that have done pilots. There are proof-of-concepts. We’re far past that. We now know the technology is available and we could drive towards electrification and ensuring the grid is prepared for this transformation.
LMJ: How do you evaluate AI against the many other data science tools at a utility’s disposal?
EC: If we could build models where we’re running all of the analytics across all the individual unique assets in totality, I know that we would start seeing commonalities that we've never even thought to connect — and that's where the artificial intelligence part comes in. It allows us to take the things we know well, and start to connect the dots to gain knowledge that wasn’t threaded until we had the tools and the data to do so. It starts breaking down the silos and bringing all the parts together, and then looks at them from a higher level, from a systems perspective instead of individual parts.
Think about a car mechanic. Their job is to open up the hood, see the broken part, take it out, and fix it. That’s kind of where we are with the grid. You’re reactive, and you’re assessing problems after an event occurs. We only have the tools to be the mechanic today. But with artificial intelligence, we’ll be able to look at the whole system and how each individual part works together.
LMJ: How does AI differ from other "hyped" technologies like blockchain that were criticized as a solution looking for a problem?
EC: I don’t think AI in the utility space is going to create solutions looking for a problem. I think there are enough problems where AI can come in and start helping immediately.
That’s been my tagline: this is no-regrets technology. Having the data and having the analytics to drive the insight — which is the foundation of AI — allows us to not only prepare for this grid of tomorrow, but also make us better at serving safe, reliable, affordable power today. I feel like that gets lost in translation because artificial intelligence sounds like it’s designed for a world that is futuristic and intangible to those engineers that are keeping the lights on today.
LMJ: What are the big unknowns you are trying to figure out around the risk and consequences of sophisticated AI systems?
EC: Our biggest hurdle is getting the buy-in from individual workstream business units responsible for keeping the power on. That's the biggest unknown: getting enough to believe in the reality of where we're headed, and that this new way of thinking will set us up not only for that future, but also for our everyday business.
I’ve had a few engagements where I've been freely able to provide the new data that we have — which may not be the best, which may be dirty, but which will get better — and have seen data scientists use this powerful approach to analytics to give immediate value within a very short period of time. We need to get enough of those wins internally to start sharing externally, especially to regulators and consumer advocates.That’s my challenge right now — getting the buy-in to invest in the people, the processing changes, and then ultimately, the deployment of technology to enable all of this good that AI can bring us.
This interview was originally published on the Post Script Media website on June 7, 2023