Resources / Blog

PowerSecure’s Marshall Worth on Auditing AI for the Microgrid

Join Infinia ML’s Ongoing Conversation about AI

Episode Highlights from Machine Meets World

This week’s guest is Marshall Worth, Senior Product Manager at PowerSecure.

Highlights from our conversation include:

“We developed some years ago some AI software that was predicting a customer’s load and dispatching batteries based on that prediction. . . .

“And had we not been monitoring that data to really look at saying, ‘Hey, I think we were wrong in what we were presupposing here,’ then it would just be happily doing the wrong thing forever.


We need to have regular cadence with the audit team or someone like that . . . Sitting down with somebody else that’s saying, ‘Hey, we’re noticing a trend that maybe we saw from somewhere else. We want to talk about that.’

“And of course that’s only possible because we monitor and collect data.”


“With the pandemic, Tuesdays don’t really look like Tuesdays anymore. Tuesdays might look like Sunday. They might look like Christmas Eve.

Had we not been monitoring that, then we wouldn’t see that, but that’s a bias that was in there because we had the historical logs that we said, ‘Okay, we’ll get started with this.’”

Photo by Anders Jacobsen on Unsplash

Audio + Transcript

Marshall Worth:
With the pandemic, Tuesdays don’t really look like Tuesdays anymore. Tuesdays might look like Sunday. They might look like Christmas Eve. Had we not been monitoring that, then we wouldn’t see that, but that’s a bias that was in there because we had the historical logs that we said, “Okay, we’ll get started with this.”

James Kotecki:
This is Machine Meets World, Infinia ML’s ongoing conversation about artificial intelligence. I am James Kotecki and my guest today is Marshall Worth, Senior Product Manager at PowerSecure. Marshall, welcome to the show.

Marshall Worth:
Thanks. I appreciate it. It’s great being here. Look, I’m honored to be here as well. I’m looking at some of your videos y’all have done in the past and you’ve got quite the roster there, so I’m glad to be here with you.

James Kotecki:
Well, thanks so much. We have very high expectations for this interview and we know you will exceed them. So PowerSecure deals with something called the microgrid, which as I understand it, is putting power supplies, power generating supplies, next to things like big box stores or other institutions that need to have their power on when the rest of the grid, I guess the macrogrid, goes out. When you look at what happened in Texas and the deep freeze and massive power outages that occurred there, is that a good example of why PowerSecure exists?

Marshall Worth:
Oh, definitely. But in order to really be able to do anything to keep reliable power, we need intelligence. We absolutely have to have intelligence that coordinates how these services come on. So PowerSecure, as you said, a microgrid, we install these resources. PowerSecure is going to install a Tier 4 final generator, a natural gas generator, a battery storage system, solar — on and on. You can’t just put those things down behind a customer’s service and just turn them on. How do we create reliable power — lights don’t go out for the customer — how do we create reliable power that we can coordinate together?

James Kotecki:
So tell me about a world without AI. Maybe this is the world of five, ten, however many years ago. How did it used to work?

Marshall Worth:
There is an incredible amount of intelligence into the core components of the microgrid and it runs on control algorithms to bring in generators so that they deliver power effectively. The question is, which one do I turn on? Which one is the best for me to turn on right now? And there are a multitude of different variables that come into that. How much does power cost? How much does a generator cost? Now, what we’re really getting into is, “Okay, let’s look at how we optimize that.” How do we create a lower energy cost for the customer that looks at 100 different variables? How do we achieve carbon objectives for the customer? And there’s ways to do that. You try and analyze how much carbon is coming from the local grid, how much you’re producing on site, and you try to even those out.

Marshall Worth:
So really that base level intelligence is established. We’re kind of like going up the ladder a little bit and adding another rung there, adding some higher-level intelligence to make more sophisticated decisions for the rest of the group.

James Kotecki:
Was this level of complexity a long sought-after dream that only when you had AI that you realized you could achieve it, or was it the development of AI that made people look differently at the technology that they had and realized they could do things with it that they had actually never thought about before?

Marshall Worth:
I think this level of intelligence has been in progress for a long time now. Back in the nineties and the eighties, they looked at different optimization because the math was there. The math has been here — if you look at neural networks, the math has been here for decades. So the math was there to make some macro-level decisions, say for the utility as a whole. Personally, what I see now is that when we add in these assets at the customer level, and then you have to bring them and let them harmoniously operate with the utility, that’s where those decisions get a little more difficult.

Marshall Worth:
And so now what we’re seeing is this tidal wave of AI development that we’re able to ride on. And so what I mean is Google and Facebook and Microsoft, they’re developing all these things. We’re now starting to understand, “I guess we could apply it to the energy grid now.” And coupled with that is that people are installing those assets behind their meter as we call it and we have to figure out how to do that. So we’ve got this convergence of many, many different factors coming into play right now.

James Kotecki:
You’re on an AI journey, as are many companies to continue to incorporate and integrate this into what you’re doing. What has your journey taught you that you think could be valuable to someone in another industry?

Marshall Worth:
That’s a deep question. I’m not sure what I could say about that because we’re really just getting our feet on the ground, I feel like, and understanding what we can do for our own industry. And not only that, but I really believe that we are learning from others. We are taking AI concepts developed by Google and Microsoft, and we’re taking talent that may be developed for them through university, and just trying to figure out how to apply it to our industry, our energy industry. So to turn that around, I think I just want to say thanks to everybody for all the amazing work they’re doing in the field.

James Kotecki:
Well, I think there’s a lot of companies like yours in many different industries that are, as you say, at the beginning of their journeys and there is so much to be learned from and so much to glean from other entities. I think that’s probably one of the greatest lessons of all is that you actually don’t have to build this stuff from scratch. Even if it’s new to your industry, it’s not new to the world and you can apply things from other fields.

Marshall Worth:
Yep. Definitely. We’re grateful that we’re not having to create the inertia for this. We’re riding along, and we want to contribute. I don’t want to be just feeding off of this. I want to be a part of that inertia and to come back and develop something that the rest of the industry can use.

James Kotecki:
I’m often asking myself and my guests, are we building, with this AI technology, a world that looks dramatically different from our own or are we simply building a better version of our own world where things just work the way they should more often? And it strikes me that energy is probably the extreme example of that second case where if you’re doing your job right, people will not notice any radical change in their lives at all.

Marshall Worth:
Yeah, that’s definitely the case. No one recognizes the 99.9% of the time that you’re operating well.

James Kotecki:
How do you think about monitoring the AI that you have and use and oversight of all these new technologies that you’re incorporating?

Marshall Worth:
Tremendously important and it needs to be monitored very closely. And I’ll give you an example. We developed some years ago some AI software that was predicting a customer’s load and dispatching batteries based on that prediction. It has — which is not surprising — it has bias into it, developer bias. And had we not been monitoring that data to really look at saying, “Hey, I think we were wrong in what we were presupposing here,” then it would just be happily doing the wrong thing forever.

Marshall Worth:
So I think it’s very, very important to monitor data, especially as we’re taking off right now. We always opt to collect more information as fast as absolutely possible, because right now we are still in the understanding phase of, “what do we need to improve upon?” And we need to have regular cadence with the audit team or someone like that that says, “Hey, this is…” And this could be a third-party team, so Infinia is a great example. Sitting down with somebody else that’s not energy-minded. Sitting down with somebody else that’s saying, “Hey, we’re noticing a trend that maybe we saw from somewhere else. We want to talk about that.” And of course that’s only possible because we monitor and collect data.

James Kotecki:
What were some of the biases that were built into the system that you mentioned? What are some ways that you can bias a, what was it, a battery monitoring system, you said?

Marshall Worth:
Yeah. So what we did is we developed a forecasting algorithm that did a 24 hour load prediction for a customer and we wanted to discharge the batteries during opportune times. We may presuppose that Tuesday’s power profiles, load profiles always look like Tuesdays. With the pandemic, Tuesdays don’t really look like Tuesdays anymore. Tuesdays might look like Sunday. They might look like Christmas Eve. Had we not been monitoring that, then we wouldn’t see that, but that’s a bias that was in there because we had the historical logs that we said, “Okay, we’ll get started with this.”

James Kotecki:
Marshall Worth of PowerSecure, thanks so much for being here on Machine Meets World.

Marshall Worth:
Well, thank you very much. This was a lot of fun and I greatly appreciate being asked to be here.

James Kotecki:
And thank you so much for watching and/or listening. You can always email the show. It’s MMW@infiniaml.com. Please like this, share this, give the algorithms what they want. I am James Kotecki, and that is what happens when Machine Meets World.