In this episode of the Pipeliners Podcast, host Russel Treat welcomes Vicki Knott, CEO and co-founder of Crux OCM, to discuss model predictive control (MPC) and its application to the pipeline industry.
They explore the concepts of advanced process control, model predictive control, and their relevance in optimizing pipeline operations. Vicki explains the difference between AI, machine learning, and model predictive control, emphasizing the focus on achieving optimal operation rather than creating average models. Vicki provides insights into the application of model predictive control and its benefits in increasing efficiency and revenue for pipeline operators.
Model Predictive Control Show Notes, Links, and Insider Terms
- Vicki Knott is the CEO and co-founder of Crux OCM. Connect with Vicki on LinkedIn.
- Crux OCM enables the autonomous control room of tomorrow, operating within the safety constraints of today. Combining advanced physics-based methodologies with machine learning, CRUX software helps clients increase throughput production and energy efficiency (up to 10%), improve safety, and ensure operators stay safe while contributing to a seamless, continuous operation. Find out more about their technology at CruxOCM.com.
- PHMSA (Pipeline and Hazardous Materials Safety Administration) is responsible for providing pipeline safety oversight through regulatory rule-making, NTSB recommendations, and other important functions to protect people and the environment through the safe transportation of energy and other hazardous materials.
- Model Predictive Control (MPC): A control strategy that uses mathematical models to predict the future behavior of a system and optimize its performance.
- SCADA (Supervisory Control and Data Acquisition) is a system of software and technology that allows pipeliners to control processes locally or at remote locations.
- PLCs (Programmable Logic Controllers) are programmable devices placed in the field that take action when certain conditions are met in a pipeline program.
- IEEE Triangle of Control: A classification system in process control that includes basic regulatory control, advanced process control (APC), and real-time optimization.
Model Predictive Control Full Episode Transcript:
Russel Treat: Welcome to the “Pipeliners Podcast,” “Episode 310,” sponsored by EnerSys Corporation, providers of POEMS, the Pipeline Operations Excellence Management System. Compliance and operations software for the pipeline control center to address control room management, SCADA, and audit readiness. Find out more about POEMS at enersyscorp.com.
Announcer: The Pipeliners Podcast, where professionals, Bubba geeks, and industry insiders share their knowledge and experience about technology, projects, and pipeline operations. Now, your host, Russell Treat.
Russel: Thanks for listening to the Pipeliners Podcast. I appreciate you taking the time. To show that appreciation, we are giving away a customized Yeti tumbler to one listener every episode.
This week, our winner is Tom Correll with Northern Natural Gas. Congratulations, Tom, your Yeti’s on its way. To learn how you can win this signature prize, stick around till the end of the episode.
This week, we speak to Vicki Knott with Crux OCM about model predictive control and its application to pipelining.
Hey, Vicki, welcome back to the Pipeliners Podcast.
Vicki Knott: Thanks for having me back. It’s always fun.
Russel: I’ve asked Vicki back to talk about model predictive control adapted to pipelining. As a way to build background, Vicki, for the listeners, let’s start. If you would tell us, to remind the listeners, who you are and what you do, if you don’t mind.
Vicki: Sure. My name is Vicki Knott. I’m CEO and co‑founder of the company, Crux OCM. My background is in chemical engineering, mostly process control. Then, I trained as a control room operator on the biggest batch pipeline in the world.
I met my co‑founder, Roger Shirt, who’s done lots of work up north. We came to the realization and understanding of how we operate pipelines today as a control room operator. It’s busy, it’s a lot of work, it’s quite challenging.
We made the comparison to pilots and planes of autopilot software. That our pipeline operators are pilots for our critical energy infrastructure. How do we help them be as successful as possible?
Russel: You’re doing autopilot for pipeline control?
Russel: It’s a good way to summarize it. There are a lot of things you use autopilot for. There are a lot of things you don’t use autopilot for.
Vicki: Exactly, like adaptive cruise control. It’s like, “When you don’t have it anymore, you miss it, but you might not want it.” You need to be in control. You’re still driving the car. We’re not there yet, but it’s handy when those things kick in.
Russel: Absolutely. To talk about model predictive control, it’s probably a good idea to build. Most pipeliners are familiar with the idea of automated control. You’ve got the SCADA system. I’ve got PLCs in the field. I’ve got pressure control. I’ve got flow control. Most of us are familiar with that.
There’s a term that’s out there that’s called advanced process control. What is it? Why have we not historically done it in pipelining?
Vicki: What is it? We’ll start with that. What is it? I bet you, almost everyone who’s listening to this will be like, “Wait, I know what that is. I’ve heard of it.” They probably work at midstream companies that own refining or refineries.
Inside of the refineries, a unit like a hydrocracker or a distillation column will typically have advanced process control capabilities on it inside of the DCS system, and it’ll be configured.
What it’s doing is it’s targeting an optimal operation. It’s a series of complex mathematical equations ‑‑ differential and integral. What it’s effectively doing is it’s constantly driving to optimal.
If you think of refining and our chemical industries, and if you were to do a quick Google, advanced process control in refining, what is the improvement? It’s about one to three or one to five percent depending on the source you’re looking at. It’s like, “This is proven in pulp and paper refining, chemical, gas plants.”
Russel: It makes sense to me that in these process industries, we would be focused on optimizing. They’re commodity industries, and they operate on razor‑thin margins. Any savings to cost directly goes to the bottom line. There’s a lot of opportunity to optimize at the incremental functional level.
Vicki: Not just cost, too. If you think about how these assets are financed and built, there’s an assumption on the performance. When you get above and beyond that, it’s really high‑margin revenue for the organizations, which is quite nice.
Russel: When you start talking about its algorithms and equations and differential, some of those are fairly complex. They’re modeling what’s occurring chemically, or flow, or whatever, in those units. The reason it tends to be focused on a unit is each of those units is running a different process. They’re optimizing every process in a chain.
That’s kind of artificial intelligence. It’s kind of machine learning. There’s a lot of similarity. How is that different than artificial intelligence or machine learning?
Vicki: Excellent. Yes, artificial intelligence, whether those be neural nets, however, you’re training the massive AI model, when you’re training an AI model, you’re typically looking at trying to automate or create a way to eliminate a human task.
If we think about the process industry and those operations, and we think about how humans operate them, we can look up many studies on the inconsistencies of control room operators. If we were to put all of that information into AI, we would have created a fantastic average performer. That’s great, that’s fine.
Then, if you’re spending significant capital on that investment to create those models, you want to see hard ROI returns. You want to see that revenue uplift. In order to do that, you need to be driving towards optimal operation, not average. That’s where advanced process control comes in, because that’s what it does. It’s not just recreating an average of humans.
Russel: Another way to think about that is what you’re doing is you’re returning to the base science.
Russel: You’re starting with the base science, and then going from there. What does the science say the best to do is? Then, you’re running that math real fast over and over and over again.
Vicki: Way faster than a human can. People can’t run it that fast. Because of that, you always have this time lag. The time lag is where the value is lost. The tighter you can make that, which all the folks from refining know, that’s where the opportunities are, those margins.
Russel: Exactly. What is model predictive control? How is it distinct or different than advanced process control?
Vicki: It’s a sub‑classification of. Advanced process control is considered a group of methodologies. Within the group of methodologies of using more advanced types of process control, which is common in plants, like in chemical plants and refining, you’ll see the IEEE triangle of control.
Inside that triangle, a subset of advanced process control is model predictive. It’s the most commonly used in refining. Effectively, it’s internally taking first principles, driving towards an optimal operation. Then, continually reiterating that as the process is changing around it. That’s how it’s able to…
You execute more commands at a higher frequency than what a human can do. That’s how you’re getting that uplift.
Russel: That’s interesting. When I think about where we are as an industry, and how we think about things like AI and machine learning, and what we’re hearing about what we ought to be doing, where I go is we’re probably not thinking deeply enough.
For example, if I have a SCADA system and I’ve got automation in the field, and I’m doing the things that I need to do, then I can apply AI to the pressure control algorithm, and have it work more smoothly, more effectively.
If I take what a human does, and I try to automate what a human does and I don’t go back to first principles, then I’m missing something. What would be your commentary about the whole mess of stuff I said there?
Vicki: The commentary, or the easiest way for pipeliners to think about it is, “We’ve always done things this way. We think of it as a logistic system, not a process.” Let’s look at our friends that are probably in our organization that are in refining, and like, “Huh, interesting.” The way that they treat it is a process, not a logistics problem.
If you take that thinking, and you think of a pipeline network as a process problem and not a logistics problem, it becomes very different in terms of how you grasp what would be useful with regards to AI versus algorithms that try to achieve optimals. Does that summarize a little? Where would you like me to…?
Russel: No, it’s good. It makes me think about how I would take and break down a pipeline. When you think about a pipeline as a process, that’s a different way to think about it than I’m trying to get product from Point A to Point B.
What I’m trying to do is I’m trying to move the product from Point A to Point B. I’m trying to do that without exceeding any of my safety constraints. I’m trying to do that by minimizing my power consumption to support that movement, and to minimize the labor hours required to support that movement.
Vicki: You want confidence that what you’re selling to your shippers, you’re able to sell them more. You want the measurable confidence that that volume is actually there.
Russel: That’s so true. If I have a pipeline, and I’m running at 80 percent of optimal.
Vicki: You think you’re at optimal, but you could run it at 83.
Russel: Then, I go back to the first principles, and I re‑evaluate what I’m doing. I’m like, “Oh, no, you’re not running. You could be moving another 10 percent of your product. You could reduce your electric consumption by 20 percent.”
Vicki: You can do it simultaneously.
Russel: There’s a whole value that goes right to the bottom line.
Vicki: In our production installations of this, it’s exactly that. We are driving hydraulic max recalculated every 10 seconds. At the same time, operating at a lean profile, which is showing about a 10‑percent power cost savings.
It’s like it can be done. It’s just that humans, like right now, we have a human in the loop constantly computing the best that they can in their head. We can’t compute that fast. Let’s use a tried and true algorithm that’s in industries that we’re comfortable with.
Russel: I think we can. My next question is, what’s the value of taking a proven refinery technology, and implementing it in the pipeline? We’ve kind of addressed that, but have we missed anything? What is that value beyond moving to optimal operations?
Vicki: Moving to optimal, it’s also a big reduction or definitely, a reduction in control room operator workload. We see those reports. We see Mackenzie, PHMSA. We’ve got to do what we can to help these folks who operate our critical energy assets. We have to help them do that.
If they’re overloaded, and they’ve got too much work and too many displays telling them to do too many things, they’re not able to achieve and get that value that commercially needs to be sold to clients.
By introducing and getting them those tools that are tried and true in refining where refinery operators are used to seeing them, I don’t think it’s a big shift to the pipeline industry if we think about it that way. We are a risk‑averse industry as we should be. If we think of it like, “Hey, this is tried and true in just this risk‑averse industry,” let’s adapt it.
Russel: I would think the challenge is getting people to think about a pipeline as a process.
Vicki: Yeah. The control room operators will.
Russel: Can you elaborate on that a bit? I think I know what that means, but I don’t know that I could speak it very well. In my head, I’m like, “That makes sense,” but I don’t know how to articulate it. What is it about process that’s missing in the way that pipeliners typically think about things?
Vicki: For me, it’s dynamics. When you’re sitting in the control room, you feel the dynamics of the liquids. As you’re operating, you can feel this batch has a different density, because these pumps behave slightly differently. All of that is, as you said, Russel, first principles. To get us going, of course, we can use better tools like AI and ML to augment.
As you’re sitting there, that’s what you’re feeling. You’re like, “I’ve got to tweak this here because I know that this pump has this impeller. From my pattern matching of operating for the last 10 years, I know that this density with this batch is a little less.”
You create model predictive control in your head as a control room operator to manage all of those dynamics along the line to shoot for that higher operation.
Russel: That’s a very liquid‑centric kind of [inaudible 15:05] which is fine. It helps illustrate. That makes sense. If you break down what’s going on with the pipeline and get to the details, then you start seeing where the opportunities are.
Vicki: It’s a process problem.
Russel: A particular pump with a particular impeller type is probably optimal for a particular density of fluid.
Vicki: If you’re not switching to it, then you lose the opportunity.
Russel: As a human controller running that pipeline, I’m not going to be watching for every one of those switches. It’s too much workload, but a model could do that.
Vicki: Software can do it.
Russel: Modeling every pump. Then, it becomes more like the advanced process control that’s implemented in refining. Now, I’m putting that on every pump. I’m putting that on every separation process. I’m putting that on every intake process, and so forth.
Vicki: The feel when you’re a control room operator in pipeline, and an advanced process control we call pipeBOT is working for you. The feeling of that as a control room operator is like, “OK, cool. It’s doing it for me. It’s doing what I want it to do.”
It feels, too, like you feel the pumps are talking to each other like the pump stations can talk to each other. You can see how this change resulted in this change over there. I didn’t have to go do it, but the displays are the same…
Russel: It means that the pipeline controller is going to become a lot more familiar with the hydraulics and details of the process versus just keep it running.
Vicki: Versus just keep it running, because they’re going to watch how the advanced control is operating it. They’re going to be like, “That’s cool.”
Then, in their next training cycle, which is ‑‑ I’m going to go back to aviation ‑‑ extremely important from a control room management standpoint. They may have learned some cool, new tricks for the next time they’re going in annually.
Russel: How is a technology like this applied if somebody wants to evaluate this type of thing? When you start talking about modeling every pump…
Vicki: [laughs] Sounds scary.
Russel: That’s scary. That’s exactly right. It’s like, “Oh, my God. That’s crazy. I don’t want to do that.”
Vicki: Thankfully, they’re all the same algorithm. [laughs] Sorry.
Russel: …and get the values.
Vicki: Yes. To get the value right, essentially, it’s also looking at, “How is value measured in an industry that we’re all familiar and comfortable with?” In refining, it’s been measured. “Here was your baseline before, historical average. Here’s your baseline now, historical new.” It’s that simple.
With one of our clients who has some new stuff that will be coming out public with us, we took two years of their historical data before implementing. We took the average, but we also made sure to comb the data and pull out if the pipeline wasn’t a portion that month. We wanted to make sure we’re getting a good apples‑to‑apples comparison.
We thought, “That was the historical flow rate.” Reviewed it with the client, and made sure everybody’s happy with it. We put on pipeBOT, and, in this case, maxOPT, and we’re running it. We ran it for about three months. We do it like, “What’s it looking like?”
We ended up with about 100 data points or something. We did two methodologies because this was our first time vetting this. We had our internal client doing their methodology. We did ours. It looked, to us, after about four or five months of runtime, the values would converge.
It was interesting to see in terms of a project and an ROI, you can put this thing in, and measure exactly how much volume you’re going to get. We’ve seen it now in production. What that means is that your commercial team can go take that, and can confidently sell it.
This isn’t something that engineering tools can achieve, because you don’t know that that volume is moving. With an implemented closed‑loop system, you know that it’s there. Does that make sense for how it’s measured? It’s actually simple.
Russel: Yeah. Then, as you build competency and capability, you start applying it in other places. You start dividing, and say, “What can I do with this pump station? This pump station is a substantial asset. How could I apply this model predictive control there?”
Vicki: We don’t think of it that way. We think of it differently. If you think about optimizing something, if you optimize that one pump but the pump next to it isn’t optimized or they don’t talk, then all your work here at this one point doesn’t matter. That’s how I like to think of it.
When it comes to implementing advanced process control on a pipeline network, you’re implementing it from the human’s perspective. You want to implement it in SCADA, and have it talk to every single pump station.
So that, then, when you go to start up the pump station and start up a pipeline, instead of executing all of your pressure set points or pump starts, you select your target flow rate or your max setting, and you select on. Then, the software will execute every single one of those intermediaries on your behalf.
Russel: What I’m driving at is with more time and more competency, you have the opportunity to implement this on smaller segments of where you want to get optimal. I mean optimizing the system. I don’t mean optimizing the pump, but optimizing the system. That’s a different way of thinking about things.
If you go back to per diagramming, or something like that, and you diagram this out, there’s all these different places in the process where I have the opportunity to optimize. I’m not trying to optimize. I’m trying to make the whole system optimal. That’s different than trying to optimize a point.
Vicki: Good point. I had to…
Russel: I know, it’s a good clarification. It’s a good clarification.
Vicki: [laughs] I’d like to know.
Russel: Where do you think this tech is headed? You guys are on the leading edge of control room technology. Particularly, when you start talking about this predictive control type stuff, autopilot for pipeline controllers, autocontrol.
Vicki: We call it pipeBOT.
Russel: I know.
Russel: I’m making up other names for it. Where do you think this is headed?
Vicki: As of right now, it’s fun. Honestly, when control room operators get their hands on it, folks really love it. Think about, as a human, all of a sudden, you buy a car with adaptive cruise control. It’s cool, this is fun. I have this new tool that makes my life more enjoyable and easier.
In terms of where it’s going, I think that remembering fundamentally that making these folks’ life easier, they’re the ones on the sharp end. They’re pushing the buttons that generate the revenue. Anything we can do to make their lives easier as employees and shareholders of all of these wonderful energy companies, that is extremely important.
What I’m hoping in terms of where the technology is going is that there becomes a better understanding of what the actual capabilities are of the technology, and what the point is. The point is moving oil and gas safely and efficiently. How do we do that? It’s not about the AI. It’s not about the ML. It’s about moving oil and gas safely and efficiently.
Russel: I’m with you. I want to segue a bit. Might be a little clumsy segue, but we’ll see. We were talking before we got on about, how do you apply? How do you get to optimal? You were talking about the first thing you need to do is you need to look at your system and understand what the safety constraints are.
Vicki: Yeah, absolutely.
Russel: Then, once you have the safety constraints like, “Don’t get outside of this,” then you can start looking at applying artificial intelligence, machine learning to do some of these highly‑computational, highly‑repetitive tasks.
You’ve got to start with getting clear about what are the safety constraints, and make sure the overall system model doesn’t go outside the safety constraints.
Russel: Again, I’m going to ask you to elaborate on that. That idea of how to build a pipeline control system is radically different than how we have historically done it.
Vicki: You mean in terms of APC?
Russel: What I mean by that, being a guy that’s done a ton of SCADA projects when you go to do a SCADA project, basically, that’s all about understanding all the I/O, and all the buttons I can push. Making all those buttons reliably available to a person. When I start trying to run on autopilot, the first thing I got to do is make sure…
We rely on the person to keep it inside the safety constraints. If I’m going to optimal, I’m going to run the system in an optimal way and optimize the human input, then I’ve got to first build my model around, “What are my safety constraints?”
Now, if I’m building a control system, the first thing I got to know is I got to know, “What are the safety constraints? What are the pressures and flow rates that I can’t get outside of? What are the RPMs and the horsepower and the current draws on these pumps that I can’t get outside of?”
That’s where I have to start. That’s a different way of thinking about how to build a control system.
Vicki: Some of it’s the same in terms of how we do it. For us, specifically, we’re very aware of all the critical safety constraints that are at the PLC level and the SCADA level.
When thinking about that and that being important, you want to architect the system in a way that you’re effectively making it anti‑fragile. You’re taking away any opportunity that it could mess up. That also leads to you have a human in the loop. If you can minimize the human in the loop, you’re also minimizing the opportunities that things can get messed up.
If we’re thinking about that, we know that our SCADA settings are correct, we know that our PLC settings are correct, why would we mess with it?
If we’re going to implement a technology like this, some folks might think it’d be a great idea to do MQTT, and go straight to the end device. I completely disagree, because that goes against the IEEE, your typical safety standards of you’ve got your layers.
When you’re thinking about safety of implementing this type of technology, be very aware of, “Is there a possibility that you could compromise any of your existing safety layers?” If that’s a no, fantastic. If it is, you’re probably architecting it wrong, because you shouldn’t.
Then, when you’re within that, now, you know your bounds, and you’ve also set up multiple layers. Then, you’re inside of those bounds, which are defined because they’re very important. If you had an OpenAI model, it knows. It might not respect those bounds if it doesn’t know they’re there.
Now, you’ve got them, and you know that, “I can calculate theoretically what my hydraulic maximum is on this pipeline.” We’ve all done it, we know, but we know that we never get that volume. [laughs]
Russel: Right. Fair enough.
Vicki: Did I answer it, or did I start rattling there, because I might? No?
Russel: No, it’s a great answer. One of the things that you can…You can take this lesson out of aviation. I use aviation as an illustration a lot. Even though I’m not a pilot, I know a lot about it.
Vicki: That’s great.
Russel: This is one of the things that they’ve done with the flight computers in the aircraft all to minimize fuel cost. They know what altitude, what throttle setting, all that kind of stuff to minimize fuel consumption and maximize my ability to arrive on time.
It’s a similar idea. Once I have it in place the right way, then I can start looking at, “How do I make this optimal?”
Vicki: How do I make it optimal? A common misconception in pipeline is that that level of interconnectivity between pump stations to be able to have some control that behaves like the control room operator is possible with the current tools. It’s actually not.
Unfortunately, SCADA doesn’t have those libraries. It just doesn’t. When I hear folks are like, “We already do this,” the libraries don’t aren’t there, so you can’t. It doesn’t exist.
Russel: It’s fascinating. One of the things about what you guys are doing with pipeBOT, and what you’re doing in the control room space, I find it fascinating. I’d love the opportunity to play with the tech. It’s probably never going to happen, because I don’t have the bandwidth for such things these days, but, oh, my gosh, it sounds like fun.
Vicki: This could be a fun one, Russell, because we need better sales enablement tools. As a company, we’ve definitely demonstrated value. It might take us a few months, but we are working on a very nice demo. It could be a fun episode to have you try to play with it. [laughs]
Russel: Oh, man. We’d have to film that one, make sure we could get the…That would be fun. I’m up for that. Let’s do that.
Vicki: We get a Russell versus pipeBOT, and see what happens.
Russel: That would be way cool. I would dig that. All right, look, this is great. Vicki, it’s always a pleasure to have you. Thanks for coming back. Wish you all the best in your sailing adventures. [laughs]
Vicki: Oh, yes, yours as well. I’m coming to join you. [laughs]
Russel: Thank you, I appreciate it.
Russel: I hope you enjoyed this week’s episode of the Pipeliners Podcast, and our conversation with Vicki. A reminder before you go, you should register to win our Pipeliners Podcast Yeti tumbler. Simply visit pipelinepodcastnetwork.com/win, and enter yourself in the drawing.
If you’d like to support us, the best way to do that is to leave us a review. You can do that on Apple Podcasts, Google Play, Spotify, wherever you happen to listen.
Russel: If you have questions, ideas, or topics you’d be interested in, please let me know either on the Contact Us page at pipelinepodcastnetwork.com, or reach out to me on LinkedIn. Thanks for listening. I’ll talk to you next week.
Transcription by CastingWords