No items found.
min read

Episode #23 reaction: ASC's potential

October 14, 2020

Happy Thursday!

Welcome to this week’s deep dive exclusively for Nexus Pro members. It’s an honor to have you here. This deep dive is a follow up to our recent panel discussion on advanced supervisory control (ASC) at the September Nexus Pro Member Gathering. I learned a lot from this conversation and want to share my takeaways and the full transcript with you below.

In case you missed it in your inbox, you can find the audio or video here:

Nexus site | Apple Podcasts | Spotify | YouTube | Add to other podcast apps

Enjoy!

—James


Outline

  • My reaction, including highlights
  • Full transcript

My reaction

I obviously biased, but I thought this was a great panel discussion! I'm proud of how exploratory our group is. The Nexus Pro culture is beginning to form and I like it.

That said, I have two reactions to this conversation:

  • This discussion was really limited to the bread and butter of ASC, which is in HVAC control and doing that better than the BAS/BMS does. However, I think what I failed to take us into in the time we had is the potential for ASC outside of the HVAC silo. The simplest example is what if the whole building (all silos) ran on the same schedule? What if every silo had the same view of which occupants and in which spaces? That enables all-new use cases. That’s where we need to get to.
  • One thing I want to dig deeper on at some point is the disagreement between Tyson and Terry on one side and Keith on the other. The 2 T’s say they would never tune PID loops from the cloud and Keith says he’s doing it all day long. I love these sorts of gaps! Maybe they can add their thoughts in the comments…

My highlights:

  • Defining ASC (2:42)
  • Origins of the concept from Keith (5:00)
  • Difficulties implementing closed-loop optimization from Tyson (10:03)
  • Terry on national lab resources and potential for deployment of ASC as ‘the next big thing’ (14:15)
  • Jean-Simon on what’s changed (data storage, computational capacity, neural networks), and why now is the time to start pushing ASC at scale (18:47)
  • Keith introduces facil.ai, and describes the social engineering aspect of the transition to ASC (22:16)
  • Confronting perceived controversies: robot vs. operator (24:59), the control should be done by the control system (29:42)
  • The potential for closed-loop optimization as a tool for an analytics or fault detection service offering (32:03)
  • The value proposition for building owners (38:02)

What did you think?

Leave a comment


Full transcript

Note: transcript was created using an imperfect machine learning tool and lightly edited by a human (so you can get the gist). Please forgive errors!

James Dice: [00:00:00] Hello, friends. Welcome to Nexus, a smart buildings technology podcast for smart humans. I'm your host, James Dice. If we haven't met before, I write a weekly newsletter on the same topic. It's also called Nexus. Each week I share what I've learned, my opinions, and what I'm excited about in the quickly evolving world of intelligent buildings. Readers have called Nexus the best way to stay up to date on the future of this industry without all the marketing fluff. You can check it out and subscribe at nexus.substack.com or click the link in the show notes.

Since starting the Nexus newsletter, many of you have reached out to me wanting to talk shop, and we have. After a few weeks of those wonderful conversations, I realized I needed to record and share them with our growing community. So here we are. The Nexus podcast is born. This is our chance to explore and learn with the brightest in our industry together.

episode 23 is a very special episode of the podcast. This week is both a sneak peek into what nexus pro member gatherings are like. And it's also our first podcast panel, which we'll be doing more of in the future. I started out the episode by introducing the topic. Advanced supervisory controls and summarizing why this is important for smart buildings and climate change.

And much more. Then we heard from our panelists, three of whom are past podcast guests. We had Keith Gibson of facil.ai. We had Tyson Suder as Siemens. We had Terry her of in television and John Simone van of brain box AI. The five of us unpack this topic, which I think is the most misunderstood topic and smart buildings right now.

This episode of the podcast is directly funded by listeners like you who have joined the nexus pro membership community. You can find info on how to join and support the podcast at  dot com. You'll also find show notes there, which has links to our panelists LinkedIn page. Oh, and by the way, if you take a look at your feed and you're missing some episodes, that's because those episodes are exclusive to members of nexus pro sign up for a membership to get your personal podcast feed with access to all the episodes that further ado, please enjoy nexus podcast, episode 23.

all right. So let's kind of get started. So this is going to be a fun conversation. so what we'll do is I have some preset questions for all the panelists. We'll kind of run through all those and I'll do my best to leave some time.

at the end for questions let me say, only before we jumped in that these are four of the experts I know of on this topic. I know community is full of experts on this topic. So this is more of just a conversation starter at this point.

So I thought I'd start off conversation with, just defining what ASC is, what are advanced supervisory controls? I think that the most simple way to say it for me is that building control is coming from somewhere other than. Your traditional building automation system or building management system.

So it could be an overlay. It could be some other type of system. I call it an overlay, but somewhere other than your traditional already installed system, it's the umbrella acronym that I didn't mean to create. never intended to create this acronym. I don't think we should really try to make it more pervasive throughout the industry.

I think I just meant to like, Gather in all these other acronyms. So there's, closed loop optimization. a Terry calls. It simply optimization there's, automated system optimization, which is a term that came out of LBNL on the national labs. I'm trying to group all of those into one. but also to say that ASC is not just, you know, sexy optimization algorithms, it could be even just simple scheduling that comes from some other system other than the BAS.

So why is ASC important? Um, number one, it's a hot new tech area everyone's really interested in AI, right? that's one reason. The other reason I think it's important and I wanted to bring it up today is there's the biggest discrepancy of opinions I've seen in our industry today. You know, some say this is, not even a thing, it's not going to be a thing for 15 years.

And then we have startups that are seemingly gaining like a lot, a lot of traction and have really cool offerings. how can both of those things be true at once? I think is the reason why I wanted to bring this up. finally I just wanted to point out again, like I always do the buildings aren't really controlled that well today.

And so any sort of innovative solution on controlling them better, helps us in problem we're trying to solve. so let's kind of jump right into it. I want to start to bring our panelists in one by one here. So first let's bring in Keith.  So Keith is the CEO and CTO of a new company called facil.ai. Keith and I are good friends, but he has a long career in smart buildings. some of it, even before smart buildings was even a word, right Keith. So he was a technician at Honeywell. he went to JCI. He founded several companies, including Phoenix energy technologies.

And now this new one. So I'll let you take the stage here. I want you to tell us as an introduction to yourself. you talked about this fault detection, diagnostics and optimization like 15 years ago, right? So this is kind of the first, mention I know of in the history of our industry where this concept of ASC started popping up.

So what was your earliest experience with this concept and why didn't it take off? immediately.

Keith: [00:05:22] Okay. Yeah. great to be here. Thank you, James, for inviting me to this round table. So, I had a company called Silicon energy corporation back in 1997. We were the first, web based EEM enterprise energy management company.

so it started off as, a way to look at all your assets, your energy across all your, your portfolio. end. It was a read only system, but it was, it was great because it was web based. you know, got the United States patent on it and it was, it was really good, but in those days we weren't thinking about actually doing two way control back into the, I mean, it was just enough just to have a web based overlay, you know?

So that was the start of the overlays. actually Tridium was  found like a month before I found Silicon energy. So we were kind of going like Nick and Nick kind of developed a personal web based platform because before that  there weren't any web based building control systems.

I think we ended up kind of edging them out by like a month or two or something like that. but even back then, there was market confusion that I had to deal with in those early days of enterprise energy management, because people thought that. Silicon energy and then subsequently Phoenix was like, yeah, a control system.

And so I know you can appreciate this James, because will go to these bids and they pull me in for RFPs or different things, Crowell systems. I'm like, I'm not supposed to be here. I'm not a control system. It's an overlay. You know, it goes on top of it and addresses it. Yeah. Oh, and then we would get, we would get compared to treating them all the time.

And that would be like, you know, well Tridium is just another BAS it's just a proprietary system with Johnson, Siemens, they're all BAS systems to me, this is one step up, I would say, in this, an overlay. So. Lots of confusion, lots of arrows in the back, but I really see that this, advanced supervisory control market really is an extension of what I used to call D dos.

So FDD used to be called FDLS back in the early days, like 2004. And really the emphasis was on that fault detection, diagnostics, and optimization. We kind of forgot about the optimization piece. We shortened the acronym. We chopped off the old, then it became FTDs and it became FD fault diagnostics.

It just became diagnostics or analytics. So we went from, from four, four letter acronym to like one letter, you know, Right. Why do you think they chopped off? Or I guess I said they, but we as a collective industry, why did the Ole get chopped up? I don't think the technology was there. like I said, it was difficult enough to integrate across all these multiple systems, not just open ones, but proprietary  systems.

I remember in the early days got funding from one of the local utilities in, California. And we developed this FTD so we were generating all these kits of anomalies. We called them back then, you know, economize are not working. And I remember the program manager at the utility.

He was like, you know, Keith, this is great. But all you're doing is generating gigantic report that is stacked up on the engineer's desk. And what we're hearing from our customers is even back as early as 2003, Um, this is nice, but we really don't want to, you know, how messed up my building is a report every month, it's not actionable.

I mean, some of it you can do like, causes and effect and root cause and all that stuff. And what he told me back in 2005, before he pulled the plug on it, I mean, we actually lost our funding. He said, not going to fund this because until you can actually go back in and fix stuff, it's not really valuable to me.

And I say, well, the technology doesn't exist. You know, this is 2005. And actually I just called, her, my old buddy at the utility a week ago. And I said, okay, the technology finally exists. We can now punch controls. It's not answered. I can do, uh, the building control system. And not only that one more quick, point is not just the, a matter of commanding.

If you're going to optimize a building, you have to be able to optimize the schedules. So that means you have to read and write schedules. That also means you have to fine tune the PID loop revolt. It might not even mean that you have to reset a panel. So this is a deep integration. This is not just lightweight reading, but you're doing the deep integration in to proprietary, open and legacy control systems, which is very difficult.

So I think that's, what's enabled facility to come out as the technology and the ability to do it is finally here.

James Dice: [00:09:33] Okay, cool. We'll get into facil in a minute. I want to bring in Tyson now. So Tyson suitor. our Australian superstar here coming to us from Switzerland. Uh, I got a lot of great feedback from Tyson's podcast, a couple of weeks back as a Tyson's global business develop manager for digitalization, the most budget priority title, and all of Siemens.

so thanks for being here with us Tyson. so when you were on the podcast, you talked about how. you saw firsthand that there was no Australian startup that started a long time ago. I think I know what startup you were talking about, but you mentioned on the podcast, how this closed loop optimization concept was difficult to implement, from when the projects you saw.

Can you give us a little bit more background on those experiences?

Tyson: [00:10:12] Yeah, definitely. So I think, yeah, I won't mention any names because the technology can, can exist and work and work well. The difficulty comes when you start implementing it and then engaging with the operations teams, engaging with existing processes.

So what we saw was the savings were occurring, but it was creating some operational problems. Our industry is like very closely linked to the hardware in a building. So even if you have a beautiful software solution, there is some relationship you have to take into a mind who's owning this hardware and who has to maintain it.

And I don't want to justify this as like defending that. I don't think that's the right thing, but that's just the reality of it. So we need a way of being able to just say, okay, even if this solution, but what's working and it was, it was causing too much of a problem operationally where they were unable to.

Determine who's responsible for maintaining this equipment and what was causing a problem. So once I started having service calls come in around temperature complaints or abnormal behavior, it doesn't mean it was the technology causing it, but it led to such a difficulty in understanding who's responsible.

I actually ended up getting removed out of the system.

James Dice: [00:11:21] Got it. Got it. Cool. So I think we'll dig into a little bit of that a little bit later on, you mentioned on the podcast also where you've seen this same concept is already taking off worldwide, across Siemens with chilled water plan optimization.

So how do you see that playing out?

Tyson: [00:11:38] Yeah. And the types of offerings around this it's been around for quite awhile. the first time I saw this coming, I was a bit skeptical of these solutions initially. Um, especially as a controls background, I'm like my control algorithms. Perfect. Why trying to replace it.

Um, but it wasn't, and it definitely wasn't. And so I saw these solutions come in and the way they approached it. I think it's a good, realistic first step. So when you come in and you want to take advanced servers, advanced controls, um, they basically replaced the sensors.

They put their own controller in. You can call it a black box if you like, but they, reduced the risk of. What senses am I getting? How much can I write back to these controllers? What priorities do I have? How much access do I have? And who's responsible. who's going to take responsibility when something goes wrong, especially in a critical asset, like a chiller.

Um, so I think the approach is, yeah, this can definitely work. And I like how you can limit it down to specific equipment or systems. I'm saying that you can just tune or change that points up your day loops without thinking about what system you're connecting to. I think this is a very difficult problem.

I'm not saying it can't be done. but I think focusing on what kind of system am I controlling learning the behavior of that system and then taking those learnings into other similar systems and other buildings, but focusing on like, you know, crack units or package units or chillers and boilers. This makes a lot of sense to me.

And I think that's where you'll see like huge, advances in the technology, because it's going to be able to really learn from that system. Especially if I've got a closed system, I can package units perfect. Or chill is fantastic as well.

James Dice: [00:13:16] Yeah. This is one of the areas where I feel like there's some discrepancies here that I'm just like trying to point out to our industry in general, where we're saying no, this whole entire like AI based control.

That's no good. But if we say just the chiller plant, Totally great. Like, you know, I think that's, that's a little bit of a gap that's growing. That's interesting to me,

Tyson: [00:13:38] just one point on that, is this a, it's not that it can't be done. I think it can be done, but it's just a matter of like, how do we approach it?

And then how do we get the confidence that this is working? And it comes back to that change management of how do we make sure the technical people involved in the process have faith? That this is working and it's done in a way that they can understand it and then allow it to be integrated into their process because otherwise it's going to be difficult to push the topic and I think people are doing it and I've seen good results of it too.

James Dice: [00:14:07] Yeah. So speaking of people are doing it. I want to bring in Terry her. So Terry, is president and television. I think he did a podcast that was like episode nine, something like that.  Terry's doing this. Sort of concept, on the ground, on a day to day basis. So I wanted to bring him in for sort of the practical day to day experience.

But first, before we get into that, Terry, I want to ask you about your involvement with the national labs and, you've been involved with the efforts at PNNL and LBNL for a while. So can you give us just a little bit of the history on what they've done for the industry and how they're promoting this ASO concept?

Terry: [00:14:42] So a lot of our investment, it is with PNNL and, and early on what they call retuning, building retuning what we might call  RCX. But if you want to me the best sort of, approach to, optimizing a building, I would say their retuning is a great start. And we do a lot of our work based around the research that they've done in doing that.

now that's, I mean, definitely talk about optimization and I'll just say, I think not only does optimization have to happen. I mean, the PAs is, are trying to do it now, and they're not doing it terribly successfully. So to me, there's no question that we can do it better with, with ASO and I, you know, so I'm, I'm a big believer.

we do optimization every building that we touch, we do it a couple of different ways. It depends. I do think it will get done in the cloud. Well, algorithms will get run in the cloud right out to set points. I don't think we're going to be retuning PIDs from the cloud. I think the pitch to stay at the controller level.

but anyhow,  peanut now has re tuning. we use Vultron, they, they develop that. So we use that as our middleware data acquisition layer. They also have some applications built on both Tron. Um, one is called air CX the developed some years ago, I think kgs was involved as a matter of fact, in that it stands for automatic indication of retuning measures.

So that basically finds when you're not optimized, which I think is quite frankly the easy part. you know, are you doing a pressure reset? but the better part is that it will, be able to do the optimization. So, and then LBL, um, they had a really cool podcast here a couple of weeks ago where they talked about machine learning, and how that's coming.

I think that is coming. I don't know that we need it right now, to be honest with you. I think we have a lot of. optimization get to without having to get too complicated with AI and machine learning yet.

James Dice: [00:16:29] Yeah, I just kind of wanted to point out that this is not just like, only controls nerds pointing out that the control system sucks.

It's also a bigger ecosystem of people saying, like, if we start to move this way from a technology standpoint, we can start to check other boxes beyond just a simple control sequences. So Tara, you said on the, on a podcast that. optimization is the next big thing. And in the context of the interview that we did, we were talking about the history of analytics, how it went from, you know, you were the, one of the first ones to start doing fault detection, diagnostics, monitoring based commissioning.

And, what you said was the next big thing after the faults are detected is to start to close that loop. So can you give us a little, is that on, on your processes?

Terry: [00:17:13] Hi, when I look at sort of the software stack, you know, you have BAS is that have been here for 30, about years. EIS. I think fairly developed.

And if you look at the adoption curves, right, the BAS is certainly, you know, well, well out there, the EIS, I think there's a lot of those. And then that's deployed in most buildings, you know, dealing with meter data FDD. Now. I mean, I stopped tracking it. 30 something. I know James, you gotta, you gotta list that's well beyond that.

and unfortunately I think we're only, although there's a lot of product in the space, the implementation, I mean, I would say the saturation rates probably only five or 7%, but that's still got a ways to go, but optimization from our perspective, when we do a building, I tell people we get probably half our savings from FDD.

And the other half from optimization. So to me, an optimization implementation is, further behind the curve yet. there's less companies doing it. There's less products in the space. So I just think that in order to get all the energy savings, it takes FDD and optimization. And so I think, and because it's further behind the adoption curve, um, you know, we see it as, you know, both of them have a lot of deployment capabilities here.

I mean, deploying both of those across. the other 95% of the buildings that are out there is, is what we all should be focusing on for the next decade about,

James Dice: [00:18:31] yeah, I agree. Uh, so cool. Let's let's bring on John Simone now. So John Sherman CTO at brainbox AI, one of the startups I was talking about earlier, so Johnson one, I want you to kind of paint us a picture pretty quickly on what's changed.

So we would just went through the whole history of this. So what's changed and why to brainbox. AI come about. And why, why is it a good time now to start pushing this technology at scale?

Jean-Simon: [00:18:57] Yeah. So, um, thank you, uh, Jane, so a lot of things are changing in that too. The, you know, um, what we're doing at green roads, we're just like, basically surfing on that change.

so look at, what's changing right now on the, on the computer side, storage storage was like a big issue, not too long ago. I mean, you want to, 10 years ago you want a storage,  a couple of terabytes. there was no CFO that would have signed up PO it's way too expensive. Why do you want to do it at that data?

so certainly an hour terabyte, like really,  do you have a credit card? It's not that expensive, so, but we have the capacity to accumulate a lot of the data, which was not really that easy not long ago. Um, then as the CPU, We, forget that  the calculation computational capacity is still increasing, not as fast as it was according to the moral law, but it's still increasing quite rapidly.

I mean, by 2015, we did pass the, basically the brain power of a mouse. and now by 2023, we're gonna equal the brain power of a human. So that's good that have capacity to calculation capacity. Certainly we have in our hand, so, Oh my God, what do you do with that? I know you're so we're starting for the first time to have enough power, actually too much power for what we need to do.

And that's completely new because I remember in the eighties, when I was in university, I mean, running a program on the mainframe. you keep being chopped down. I mean, the guy was saying, no, no, no. You only have 15 minutes on the mainframe. You have to reduce, reduce, reduce your equation, because you only have 15 minutes of computing.

Shouldn't tiring so, I mean, do you get frustrated because do you want to resolve that operational research problem and you're not allowed the computational power to resolve it? so what are you doing really? So now for the first time in history, we have more power than we need to resolve the problem.

So we're kind of limited by our imagination, what to do with it. so that's really what change. And then there's this, you know, AI exists for the last four. I mean, there's nothing new about AI. You're doing regression, relented, regression. You're doing AI machine learning has been around for 30 years.

I mean, come on. I mean, we all play chest, winter computer, even if it was like 25 years ago and you know what it was already winning on us. Um, so really. What's new is these deep learning network. These neural network kind of mimicking our brain, which are exploding in the last like eight years it's exploding.

And, um, and these techniques are giving us the capability to learn to self-learn to basically. Put a lot of data into it and they will learn from it and they will tell you what's going to be happening. And that's, what's really new. So we're still like, you know, kind of like a kid with a new toy. He was still playing with that Nutanix.

Like what, what can you do with it? And we're still trying to figure out what kind of business model can I, how can I make money with that? You know, um, how can I use it to create value and then sell it to somebody? so that's what we're still, uh, we're still in that stage and it's, it's quite a fascinating stage.

And of course there's people, you know, hurting themselves and there's other people which are, finding the right combination and moving forward. Um,

James Dice: [00:21:58] Great. Great. Thanks. Yeah. And you and I, and the product, that's when deep into the brain box AI approach. So anyone can go kind of deep dive on that if they want to.

Keith, before we go the perceived or my perceived controversies here, I want to definitely want to hit those pretty hard here for the next 10, 15 minutes. The Keith  the chance to introduce your new company, which launched. Yesterday to kind of solve this the same problem. So can you give us a 32nd elevator pitched on, facil.ai?

Keith: [00:22:28] Sure. Thank you. it's very interesting. it's a little bit controversial, you know, about what, what we're trying to solve here and what the problems are. I want to put  kind of a different spin on it. Um, so, you know, facil.ai isn't is an AI driven system is an optimization system.

But I wanna, talk for a few minutes about  why we're doing this and what we're doing. So I said a few minutes ago, the technology wasn't there. Right? Okay. So obviously we have to have the technology we've got, you know, increased memory and all that stuff.

But there's also a social engineering aspect to this. We've got to change as an industry and stop doing things that are not profitable when we're running buildings. Okay. So really, I really got it. Well, yeah, we know the dog. Okay. Yeah. So let me just like radon entree into that. So, there's a quote.

One of my favorite quotes, is, by Warren Bennis and he said the factory of the future will have two employees, a man and a dog. The man is there to feed the dog and the dog is there to keep the man from touching the equipment. That's what we have here. So as an industry, I saw Moneyball, I saw Moneyball a few months ago and that just sealed it.

And I'm like, I'm like Billy Bean. I mean, I remember  one of the things he was saying, well, if he's a good hitter, why doesn't he hit it? Good. You know, Tyson mentioned. Yeah. That, um, you know, my control system is perfect. No, it's not.  if it's controlling perfectly, why do you have so many hot calls and cold calls?

So this is more a matter of like, stop doing stupid things that we do as an entry, like resetting the chilled water set point up to 53 degrees when it was a design for 45. Um, stop running unoccupied set points in death Valley, California in the middle of July of 85 degrees. You're never going to be able to make that occupied set point.

So this is like I kinda turned facil or this advanced supervisor control is like the virtual plastic box on top of the thermostat. we're not going to do any of this stuff because humans are biased. We think we're trying to save energy. We're trying to push the envelope. If one pill is good, two is better.

that puts us into all this stuff. And machine is relentless. It's not biased.  It's not gonna make a logical decision. So  part of It's a social re-engineering of the industry and really getting back down to basics. Cause I remember when I was controlling DDC systems in 1987 and I had a little Commodore 64 and it can, you know, it was like a Nintendo, it had like 64 K a memory.

And I don't remember it being all. Messed up. I remember my buildings were tuned pretty well.

James Dice: [00:24:59] So this, this actually gets me to the first perceived controversy. So I think I'm getting just started to kick it off right here. So I think there's a controversy around it's kind of two levels. So the robot versus the operator.

So will the operators allow this. So Terry talked in the podcast about how we need to get the operator job into less of an operator and more of a maintenance versus maintaining equipment. So do you guys see operators allowing this and how do we train them? educate them on, well, this take their jobs and those types of fears.

Terry, you want to kind of fill us in on your thoughts on that.

Terry: [00:25:38] Yeah, well, so like every, a lot of jobs, right. Technology changes things. And I do think the operators, I don't think there's a choice. I think the guy above them, the facility manager, if he thinks that some software can do a better job. And I agree, I there's no question to me.

I believe the software will get. Good enough to do a better job. Um, in fact, honestly, a lot of the problems. We find it in a building, our operator caused right. Overrides for instance, or they don't really understand the sequence of op.  so I don't think it's an option. Yes. They, they might find it.

we hope we can train them to just do their job differently. you know, train them to understand, you know, they'll, they'll still be at the controls. They can probably still stop the optimization. Let's say if they want to, for some reason. but again, keeping the systems running and letting  the machine and the software, I think do it, and do it better.

We're finding that with self driving cars, right? I mean, there's no doubt in my mind that, you know, I'm, I'm, I'm waiting for myself, right. Driving park. Cause I'm sure it can do better than I can. right. That doesn't get sleepy, tired, distracted. So I think, I think we're going to have the same thing in buildings.

It'll be a little while. I mean, I'm not convinced it's around the corner. I'm convinced it's five, 10 years out. Maybe that level of automation, I mean, there's lots we can do right now, but using, complete, machine learning or AI, I definitely think is, is coming, but  Yeah.

James Dice: [00:27:05] Is there an aspect of, I headed to you Johnson, Mona about your, your perspective here, but is there an aspect to this of there's the pride of the operator, but then there's also the responsibility that a lot of controls vendors have taken.

So  a lot of times you have a service contract that contractor basically owns. The responsibility for these control sequences. And now we're coming in and saying, Oh, we're just going to override a few things here. No worries. Right. So there's a, there's a liability aspect. And there's the finger pointing aspects of how are you guys thinking about that and managing that with your clients?

Jean-Simon: [00:27:38] I mean, that's exactly the, uh, the number one bottleneck to, to deploy these types of technologies. Exactly. That there's like, how do you make it fit in the actual settings? Uh, that is in the building? so yes, there's a parade or in to Terry's point. I mean, absolutely. I mean, the, the, the fact that you're starting to automate and optimize.

We're we're bringing it to a new level, but, but also you're, you're fixing a lot of things which were supposedly they should have been fixed. Right. And that are the low hanging fruit are being there. And it's not everybody that has this top notch building operator in front of the BMS. A lot of building, they have this guy, which is rotating coming one day per week.

And you know, he's not there most of the time. So who's driving the building when he's not there. So to answer your question, Jane is it's really, Usually like a SSI will start to freak out when we get into a building and they go, wait a sec, this is my control sequence. This thing is going to very quickly start to override thing.

Then my control sequence is not really doing what it's supposed to do. and we have literally to take responsibility like contractually speaking, we're saying we're going to be much relating. We're taking responsibility and this SSI is now off the hook. so, but that's a discussion that needs to take place.

And some OSI are basically saying no way. I'm not going there. And some other are kind of embracing the fact that, you know what it's coming, we better get on the train. Uh, and we always start with the air side. I, we never start to, to Tyson point, we never, never start on the water side because we're scared ourselves , our, where the neural networks are going to do.

Um, and, also we have a legal people telling us be shirt or what you're going to be doing before you do anything. So yeah, I have to open that Pat, but it's, uh, it's like, yeah. Autonomous car. There's so many question about who's responsible for what if the algo take the wrong decision and kill the poor lady that was crossing at the sidewalk.

I mean, who's responsible. Right. And there's still a lot of a question to be answered on that, on that path.

James Dice: [00:29:37] Yeah. I think it's cool that you guys are just on this discovery path. It's really interesting to see. So Tyson, give me your thoughts on this. I think the main barrier here is like there's a perception that the control should be done by the control system.

So being someone from one of the traditional controls companies, kind of, how are you thinking about this? I'm not saying represent Siemens.

Tyson: [00:29:58] One of the, one of the enemies that is, that what you're saying right now. I fully agree with the points, right? It's it always comes down. Always say like any digital sips offering has to engage with the people and become part of the process.

Like we talked about value in the past. If you can engage and give confidence into a process. It will have a lasting impact. If you come in and try to take control, or maybe not, control's the wrong word for this situation, let's say take transparency and ownership of the system away from somebody. I think that's when you start running into problems, I think you have to not just replace a process, but like data enrich the process.

So I think there is a solution there that works. But it has to be able to engage with people on site. Can we talk about, I love that. I love the point about let's change people from operators to maintain us. That's a hundred percent what needs to happen. The gap in the market between skills and people and availability is getting bigger and bigger and the consistency across portfolios.

It doesn't exist. It doesn't exist. As soon as you lose one person, you lose performance. So I think there's a place, but it has to be part of the process. It can't just be a software solution. It has to be saying, we're going to integrate ourselves as part of your operations team, a tool. And that's what it is. It's a tool to be able to improve that process.

James Dice: [00:31:19] Yeah. I think a lot of the, Issues here because not enough stakeholders are being brought into the solution. So we have these new algorithms, these new technologies that, really will help us, but it's not one person's algorithm. It's everyone, you know, engaging together and to, to use the new technology for the betterment of the building.

So as we kind of close, it was out here. I don't see a lot of questions in the chat here. Does anyone have any questions for these guys?

Tyson: [00:31:49] I might have a question for Terry actually. Yeah, shoot. So I liked your point about how we've got a long way to go in terms of improvement and a lot of your savings function, fault detection, because I had the exact same experience.

Um, do you see closed loop optimization as being. I tool for a analytics or pulse detection service offering.

James Dice: [00:32:11] Oh, so this gets at another one of my questions actually. I'll piggyback on top of that, Terry. So when you talk, when you talk about this, will you talk about your 70, 30 concept or your 80 20 concept where you're saying, you know, most of the savings are from traditional control sequences?

Terry: [00:32:27] Um, both. So, the first, I guess the first comment is that, Uh, every time I talked to an FTD company, when I asked them about their plan, on optimization, I've been doing this for years and, you know, it used to say, Oh, they don't want us to do anything. They want us to be passive.

And I was like, well, why, you know, there's something controlling the building and it's not doing a great job. I think that's all changed. I know kgs and copper tree. everyone I've spoken to recently that, you know, optimization you know, writing back  is on the radar screen or you know, it's in the development.

Backing that really helps with that too. Right? You can write priority arrays. So it's not very hard to just write at the next level priority. and then if the optimization dies, the BAS takes over. So it's not hard to do. It's the way most of the planet optimization actually works.

and James I think you have this thing where you say, uh, you know, , walk, run, or, you know what I mean? Like yeah. So I do think, some buildings and this varies by building, but to me, if you go into a building and you do the basics, remove the faults, the overrides, and set schedules and set points and dead bands correctly, you're going to get savings, right.

Five or 10%. And that's, you can do that with a BAS and well, The fault detection part is better than FTD. Yeah. And then, so we always do that first. That's all, that's our sort of basics that gets a 10 to 30%. We think of the savings. Next are the, kind of already known algorithm standard, optimal start temperature, pressure resets.

This is stuff that BS is could be doing and some are all right, like, pressure reset on VAV. Air, air handling has been known for years. How to do that. Guideline 36 requires it now, but I would tell you, it's probably still only done 40% of the time. Optimal start probably 10% of the time.

So those algorithms doesn't require machine learning, AI. Any of that to me gets another 30 to 50% incrementally. and I think what's coming is that final 20%, right? That's the model predictive control or data-driven control, AI driven, right? That's sort of predictive adaptive stuff.

I think that's, what's coming and it's pretty exciting. We would like that, but for us right now. We're getting  all that 80% part. That's easy to get known technology, somewhat easy to do. and the other 20%, we'll just wait until the software gets better.

James Dice: [00:34:46] Got it. Keith Johnson, what do you think about that? 80, 20 or 70 30 rule or whatever, 50, 50,  whatever we want to call it.

Keith: [00:34:54] I completely agree. And I think it's 80 20. I mean, so I've been doing the 80% for the last couple of decades. I mean, that's what I built a career on, you know, optimizing buildings taking overrides off, resetting set points back to reasonable set points.

Addressing the schedule alignment. We've all been doing that. We all, I know that that works. The other 20% is where there's real opportunity. So, this time around already know I can do all the other stuff.  So the first project I'm working on is with a university.

That has a, um, they've been using these believable, smart valve, smart energy valves, and they're, they're changing the flow right at the coil to optimize for Delta T and temperature. So we're no longer optimizing for bowel position or open-loop control  it's optimizing it right there at the business end of the coil.

So it's got its own little ultrasonic flow meter. It's got his own temperature sensor, so it can calculate Delta T. But now what that's done is that caused a disturbance in the hydronic loop because you know,  you used to have a fixed coil valve position, and now you've got all these changing valves, lastly changing, and that's wrecking habit with the central plant.

So it's like the tail wagging the dog. So. We're starting out on, that part of it with the 20%. And we know that  if we can have this ASC system balance out of the hydraulic tool of Waterloo with, you know, four 1100 ton chillers, it can certainly change the set point on a, little RTU.

So we're starting out with the 20% this time around, and then we're going to scale back down to what all of us already knew how to do.

James Dice: [00:36:27] Got it. What do you think Johnson?

Jean-Simon: [00:36:28] Yeah, I mean, just because, you know, I want, I want to create some debate here. I'm not going to support Keith and Terry position.

Um, I think it's more 50, 50. and the reason being, yes, I agree with all of the low hanging fruit and sometime I'm completely in state of shock when we connect to a building and I'm looking at the way it's been managed, I go like really. You don't need AI. You just need somebody that knows what to do.

intelligence. I mean, yeah. And we look like magician, right? Okay. Well, no, really. That's not us. But there's this other 50, the reason why it's not 20 it's because if you start to look at the billing and okay, this is fine, let's optimize all of the classical things that we know how to optimize, but then you want to marry the air side and the Waterside to get, or so they work as a team and read their, you get extra.

And at the same time, you want to optimize that peak load. You want to start doing these shifts of peak when at different time of day, at the right time. And then suddenly that 20 become more cause the pie has just grown bigger. So you're kind of playing in a bigger sandbox. So, you're changing that percentage.

and open for debate on this, of course.

James Dice: [00:37:40] Right, right. Cool. Well, let's kind of wrap this. I think we can probably debate about a lot of these things all day. I want to say thank you to the four of you. Thanks for doing what you do every day, but also come in here. And, uh, and let's have your brand For a little bit. Um,  maybe we can do one more question. One question from the crowd to close us off.

So I liked John's question here and so to a building owner, how did they view. Supervisory controls value proposition. When the BAS companies say that they do that. And I think what I want to also say is like, what I'm hearing is that building owners don't want this.

And so do you guys say?

Keith: [00:38:18] I literally deal with this idea with this question all the time. So when, building owner says that, I say, okay, so you would let the IRS do your taxes for you. It's a perverse incentive. Okay. So, I'm and I'm not, you know, I'm a controls guy, Honeywell Johnson worked with Siemens, all this stuff.

I'm saying that there has to be, an independent party. That's not biased for any reason. Okay. So. another way we say it is like the Fox guarding the hen house, you've got to have some sort of ability to do not be biased in the situation.

And so that's the way we kind of address that because you know, the control system, we're not talking about replacing the control system here. We're not talking about ditching the control sequence. this chill water example. I mean, am I going to ditch the sequence of operation probably program?

When I was a Johnson 10 years ago, I'm not going to ditch the sequence of operation on the contrary, I'm going to embrace the sequence of operation and fine tune it so that it works better. So it's not about teaching the controls that I don't, I don't think it adversarial at all. I think these two, the control system and ASC can peacefully coexist.

both of them are the better because of,

James Dice: [00:39:27] Hmm. Okay, cool. Any other quick, final, thoughts from anyone

Tyson: [00:39:32] Just one point on that, I think of course, a building owner is going to want a system that reduces their cost. So I think if the value proposition is there, that they can control their systems cheaper than they do today, they'll buy it every single time.

So I think that's, what's yet to be proven at scale. And once the competence of that solution is done, there's no more arguments. I'll do it every time. So I think, we're early in the piece because people are new. It's a new topic, but once the value starts being shown, you know, day in, day out and brainbox have some great case studies ready.

Um, once that value is done at scale, building owner will buy it every time.

James Dice: [00:40:08] Totally love it. Um, cool. So anyone that wants to go deeper on this topic, several podcasts and the Texas podcast. I talked about this and the guy written the four or five essays on it now. So I'm always happy to talk about it.

These guys are too. all right. With that, thanks for everyone and, um, see you soon. Alright, friends. Thanks for listening to this episode of the Nexus podcast. For more episodes like this and to get the weekly Nexus newsletter, please subscribe at nexus.substack.com. You can find the show notes for this conversation there as well. As always, please reach out on LinkedIn with any thoughts on this episode.

I'd love to hear from you. Have a great day.

Upgrade to Nexus Pro to continue reading

Upgrade

Happy Thursday!

Welcome to this week’s deep dive exclusively for Nexus Pro members. It’s an honor to have you here. This deep dive is a follow up to our recent panel discussion on advanced supervisory control (ASC) at the September Nexus Pro Member Gathering. I learned a lot from this conversation and want to share my takeaways and the full transcript with you below.

In case you missed it in your inbox, you can find the audio or video here:

Nexus site | Apple Podcasts | Spotify | YouTube | Add to other podcast apps

Enjoy!

—James


Outline

  • My reaction, including highlights
  • Full transcript

My reaction

I obviously biased, but I thought this was a great panel discussion! I'm proud of how exploratory our group is. The Nexus Pro culture is beginning to form and I like it.

That said, I have two reactions to this conversation:

  • This discussion was really limited to the bread and butter of ASC, which is in HVAC control and doing that better than the BAS/BMS does. However, I think what I failed to take us into in the time we had is the potential for ASC outside of the HVAC silo. The simplest example is what if the whole building (all silos) ran on the same schedule? What if every silo had the same view of which occupants and in which spaces? That enables all-new use cases. That’s where we need to get to.
  • One thing I want to dig deeper on at some point is the disagreement between Tyson and Terry on one side and Keith on the other. The 2 T’s say they would never tune PID loops from the cloud and Keith says he’s doing it all day long. I love these sorts of gaps! Maybe they can add their thoughts in the comments…

My highlights:

  • Defining ASC (2:42)
  • Origins of the concept from Keith (5:00)
  • Difficulties implementing closed-loop optimization from Tyson (10:03)
  • Terry on national lab resources and potential for deployment of ASC as ‘the next big thing’ (14:15)
  • Jean-Simon on what’s changed (data storage, computational capacity, neural networks), and why now is the time to start pushing ASC at scale (18:47)
  • Keith introduces facil.ai, and describes the social engineering aspect of the transition to ASC (22:16)
  • Confronting perceived controversies: robot vs. operator (24:59), the control should be done by the control system (29:42)
  • The potential for closed-loop optimization as a tool for an analytics or fault detection service offering (32:03)
  • The value proposition for building owners (38:02)

What did you think?

Leave a comment


Full transcript

Note: transcript was created using an imperfect machine learning tool and lightly edited by a human (so you can get the gist). Please forgive errors!

James Dice: [00:00:00] Hello, friends. Welcome to Nexus, a smart buildings technology podcast for smart humans. I'm your host, James Dice. If we haven't met before, I write a weekly newsletter on the same topic. It's also called Nexus. Each week I share what I've learned, my opinions, and what I'm excited about in the quickly evolving world of intelligent buildings. Readers have called Nexus the best way to stay up to date on the future of this industry without all the marketing fluff. You can check it out and subscribe at nexus.substack.com or click the link in the show notes.

Since starting the Nexus newsletter, many of you have reached out to me wanting to talk shop, and we have. After a few weeks of those wonderful conversations, I realized I needed to record and share them with our growing community. So here we are. The Nexus podcast is born. This is our chance to explore and learn with the brightest in our industry together.

episode 23 is a very special episode of the podcast. This week is both a sneak peek into what nexus pro member gatherings are like. And it's also our first podcast panel, which we'll be doing more of in the future. I started out the episode by introducing the topic. Advanced supervisory controls and summarizing why this is important for smart buildings and climate change.

And much more. Then we heard from our panelists, three of whom are past podcast guests. We had Keith Gibson of facil.ai. We had Tyson Suder as Siemens. We had Terry her of in television and John Simone van of brain box AI. The five of us unpack this topic, which I think is the most misunderstood topic and smart buildings right now.

This episode of the podcast is directly funded by listeners like you who have joined the nexus pro membership community. You can find info on how to join and support the podcast at  dot com. You'll also find show notes there, which has links to our panelists LinkedIn page. Oh, and by the way, if you take a look at your feed and you're missing some episodes, that's because those episodes are exclusive to members of nexus pro sign up for a membership to get your personal podcast feed with access to all the episodes that further ado, please enjoy nexus podcast, episode 23.

all right. So let's kind of get started. So this is going to be a fun conversation. so what we'll do is I have some preset questions for all the panelists. We'll kind of run through all those and I'll do my best to leave some time.

at the end for questions let me say, only before we jumped in that these are four of the experts I know of on this topic. I know community is full of experts on this topic. So this is more of just a conversation starter at this point.

So I thought I'd start off conversation with, just defining what ASC is, what are advanced supervisory controls? I think that the most simple way to say it for me is that building control is coming from somewhere other than. Your traditional building automation system or building management system.

So it could be an overlay. It could be some other type of system. I call it an overlay, but somewhere other than your traditional already installed system, it's the umbrella acronym that I didn't mean to create. never intended to create this acronym. I don't think we should really try to make it more pervasive throughout the industry.

I think I just meant to like, Gather in all these other acronyms. So there's, closed loop optimization. a Terry calls. It simply optimization there's, automated system optimization, which is a term that came out of LBNL on the national labs. I'm trying to group all of those into one. but also to say that ASC is not just, you know, sexy optimization algorithms, it could be even just simple scheduling that comes from some other system other than the BAS.

So why is ASC important? Um, number one, it's a hot new tech area everyone's really interested in AI, right? that's one reason. The other reason I think it's important and I wanted to bring it up today is there's the biggest discrepancy of opinions I've seen in our industry today. You know, some say this is, not even a thing, it's not going to be a thing for 15 years.

And then we have startups that are seemingly gaining like a lot, a lot of traction and have really cool offerings. how can both of those things be true at once? I think is the reason why I wanted to bring this up. finally I just wanted to point out again, like I always do the buildings aren't really controlled that well today.

And so any sort of innovative solution on controlling them better, helps us in problem we're trying to solve. so let's kind of jump right into it. I want to start to bring our panelists in one by one here. So first let's bring in Keith.  So Keith is the CEO and CTO of a new company called facil.ai. Keith and I are good friends, but he has a long career in smart buildings. some of it, even before smart buildings was even a word, right Keith. So he was a technician at Honeywell. he went to JCI. He founded several companies, including Phoenix energy technologies.

And now this new one. So I'll let you take the stage here. I want you to tell us as an introduction to yourself. you talked about this fault detection, diagnostics and optimization like 15 years ago, right? So this is kind of the first, mention I know of in the history of our industry where this concept of ASC started popping up.

So what was your earliest experience with this concept and why didn't it take off? immediately.

Keith: [00:05:22] Okay. Yeah. great to be here. Thank you, James, for inviting me to this round table. So, I had a company called Silicon energy corporation back in 1997. We were the first, web based EEM enterprise energy management company.

so it started off as, a way to look at all your assets, your energy across all your, your portfolio. end. It was a read only system, but it was, it was great because it was web based. you know, got the United States patent on it and it was, it was really good, but in those days we weren't thinking about actually doing two way control back into the, I mean, it was just enough just to have a web based overlay, you know?

So that was the start of the overlays. actually Tridium was  found like a month before I found Silicon energy. So we were kind of going like Nick and Nick kind of developed a personal web based platform because before that  there weren't any web based building control systems.

I think we ended up kind of edging them out by like a month or two or something like that. but even back then, there was market confusion that I had to deal with in those early days of enterprise energy management, because people thought that. Silicon energy and then subsequently Phoenix was like, yeah, a control system.

And so I know you can appreciate this James, because will go to these bids and they pull me in for RFPs or different things, Crowell systems. I'm like, I'm not supposed to be here. I'm not a control system. It's an overlay. You know, it goes on top of it and addresses it. Yeah. Oh, and then we would get, we would get compared to treating them all the time.

And that would be like, you know, well Tridium is just another BAS it's just a proprietary system with Johnson, Siemens, they're all BAS systems to me, this is one step up, I would say, in this, an overlay. So. Lots of confusion, lots of arrows in the back, but I really see that this, advanced supervisory control market really is an extension of what I used to call D dos.

So FDD used to be called FDLS back in the early days, like 2004. And really the emphasis was on that fault detection, diagnostics, and optimization. We kind of forgot about the optimization piece. We shortened the acronym. We chopped off the old, then it became FTDs and it became FD fault diagnostics.

It just became diagnostics or analytics. So we went from, from four, four letter acronym to like one letter, you know, Right. Why do you think they chopped off? Or I guess I said they, but we as a collective industry, why did the Ole get chopped up? I don't think the technology was there. like I said, it was difficult enough to integrate across all these multiple systems, not just open ones, but proprietary  systems.

I remember in the early days got funding from one of the local utilities in, California. And we developed this FTD so we were generating all these kits of anomalies. We called them back then, you know, economize are not working. And I remember the program manager at the utility.

He was like, you know, Keith, this is great. But all you're doing is generating gigantic report that is stacked up on the engineer's desk. And what we're hearing from our customers is even back as early as 2003, Um, this is nice, but we really don't want to, you know, how messed up my building is a report every month, it's not actionable.

I mean, some of it you can do like, causes and effect and root cause and all that stuff. And what he told me back in 2005, before he pulled the plug on it, I mean, we actually lost our funding. He said, not going to fund this because until you can actually go back in and fix stuff, it's not really valuable to me.

And I say, well, the technology doesn't exist. You know, this is 2005. And actually I just called, her, my old buddy at the utility a week ago. And I said, okay, the technology finally exists. We can now punch controls. It's not answered. I can do, uh, the building control system. And not only that one more quick, point is not just the, a matter of commanding.

If you're going to optimize a building, you have to be able to optimize the schedules. So that means you have to read and write schedules. That also means you have to fine tune the PID loop revolt. It might not even mean that you have to reset a panel. So this is a deep integration. This is not just lightweight reading, but you're doing the deep integration in to proprietary, open and legacy control systems, which is very difficult.

So I think that's, what's enabled facility to come out as the technology and the ability to do it is finally here.

James Dice: [00:09:33] Okay, cool. We'll get into facil in a minute. I want to bring in Tyson now. So Tyson suitor. our Australian superstar here coming to us from Switzerland. Uh, I got a lot of great feedback from Tyson's podcast, a couple of weeks back as a Tyson's global business develop manager for digitalization, the most budget priority title, and all of Siemens.

so thanks for being here with us Tyson. so when you were on the podcast, you talked about how. you saw firsthand that there was no Australian startup that started a long time ago. I think I know what startup you were talking about, but you mentioned on the podcast, how this closed loop optimization concept was difficult to implement, from when the projects you saw.

Can you give us a little bit more background on those experiences?

Tyson: [00:10:12] Yeah, definitely. So I think, yeah, I won't mention any names because the technology can, can exist and work and work well. The difficulty comes when you start implementing it and then engaging with the operations teams, engaging with existing processes.

So what we saw was the savings were occurring, but it was creating some operational problems. Our industry is like very closely linked to the hardware in a building. So even if you have a beautiful software solution, there is some relationship you have to take into a mind who's owning this hardware and who has to maintain it.

And I don't want to justify this as like defending that. I don't think that's the right thing, but that's just the reality of it. So we need a way of being able to just say, okay, even if this solution, but what's working and it was, it was causing too much of a problem operationally where they were unable to.

Determine who's responsible for maintaining this equipment and what was causing a problem. So once I started having service calls come in around temperature complaints or abnormal behavior, it doesn't mean it was the technology causing it, but it led to such a difficulty in understanding who's responsible.

I actually ended up getting removed out of the system.

James Dice: [00:11:21] Got it. Got it. Cool. So I think we'll dig into a little bit of that a little bit later on, you mentioned on the podcast also where you've seen this same concept is already taking off worldwide, across Siemens with chilled water plan optimization.

So how do you see that playing out?

Tyson: [00:11:38] Yeah. And the types of offerings around this it's been around for quite awhile. the first time I saw this coming, I was a bit skeptical of these solutions initially. Um, especially as a controls background, I'm like my control algorithms. Perfect. Why trying to replace it.

Um, but it wasn't, and it definitely wasn't. And so I saw these solutions come in and the way they approached it. I think it's a good, realistic first step. So when you come in and you want to take advanced servers, advanced controls, um, they basically replaced the sensors.

They put their own controller in. You can call it a black box if you like, but they, reduced the risk of. What senses am I getting? How much can I write back to these controllers? What priorities do I have? How much access do I have? And who's responsible. who's going to take responsibility when something goes wrong, especially in a critical asset, like a chiller.

Um, so I think the approach is, yeah, this can definitely work. And I like how you can limit it down to specific equipment or systems. I'm saying that you can just tune or change that points up your day loops without thinking about what system you're connecting to. I think this is a very difficult problem.

I'm not saying it can't be done. but I think focusing on what kind of system am I controlling learning the behavior of that system and then taking those learnings into other similar systems and other buildings, but focusing on like, you know, crack units or package units or chillers and boilers. This makes a lot of sense to me.

And I think that's where you'll see like huge, advances in the technology, because it's going to be able to really learn from that system. Especially if I've got a closed system, I can package units perfect. Or chill is fantastic as well.

James Dice: [00:13:16] Yeah. This is one of the areas where I feel like there's some discrepancies here that I'm just like trying to point out to our industry in general, where we're saying no, this whole entire like AI based control.

That's no good. But if we say just the chiller plant, Totally great. Like, you know, I think that's, that's a little bit of a gap that's growing. That's interesting to me,

Tyson: [00:13:38] just one point on that, is this a, it's not that it can't be done. I think it can be done, but it's just a matter of like, how do we approach it?

And then how do we get the confidence that this is working? And it comes back to that change management of how do we make sure the technical people involved in the process have faith? That this is working and it's done in a way that they can understand it and then allow it to be integrated into their process because otherwise it's going to be difficult to push the topic and I think people are doing it and I've seen good results of it too.

James Dice: [00:14:07] Yeah. So speaking of people are doing it. I want to bring in Terry her. So Terry, is president and television. I think he did a podcast that was like episode nine, something like that.  Terry's doing this. Sort of concept, on the ground, on a day to day basis. So I wanted to bring him in for sort of the practical day to day experience.

But first, before we get into that, Terry, I want to ask you about your involvement with the national labs and, you've been involved with the efforts at PNNL and LBNL for a while. So can you give us just a little bit of the history on what they've done for the industry and how they're promoting this ASO concept?

Terry: [00:14:42] So a lot of our investment, it is with PNNL and, and early on what they call retuning, building retuning what we might call  RCX. But if you want to me the best sort of, approach to, optimizing a building, I would say their retuning is a great start. And we do a lot of our work based around the research that they've done in doing that.

now that's, I mean, definitely talk about optimization and I'll just say, I think not only does optimization have to happen. I mean, the PAs is, are trying to do it now, and they're not doing it terribly successfully. So to me, there's no question that we can do it better with, with ASO and I, you know, so I'm, I'm a big believer.

we do optimization every building that we touch, we do it a couple of different ways. It depends. I do think it will get done in the cloud. Well, algorithms will get run in the cloud right out to set points. I don't think we're going to be retuning PIDs from the cloud. I think the pitch to stay at the controller level.

but anyhow,  peanut now has re tuning. we use Vultron, they, they develop that. So we use that as our middleware data acquisition layer. They also have some applications built on both Tron. Um, one is called air CX the developed some years ago, I think kgs was involved as a matter of fact, in that it stands for automatic indication of retuning measures.

So that basically finds when you're not optimized, which I think is quite frankly the easy part. you know, are you doing a pressure reset? but the better part is that it will, be able to do the optimization. So, and then LBL, um, they had a really cool podcast here a couple of weeks ago where they talked about machine learning, and how that's coming.

I think that is coming. I don't know that we need it right now, to be honest with you. I think we have a lot of. optimization get to without having to get too complicated with AI and machine learning yet.

James Dice: [00:16:29] Yeah, I just kind of wanted to point out that this is not just like, only controls nerds pointing out that the control system sucks.

It's also a bigger ecosystem of people saying, like, if we start to move this way from a technology standpoint, we can start to check other boxes beyond just a simple control sequences. So Tara, you said on the, on a podcast that. optimization is the next big thing. And in the context of the interview that we did, we were talking about the history of analytics, how it went from, you know, you were the, one of the first ones to start doing fault detection, diagnostics, monitoring based commissioning.

And, what you said was the next big thing after the faults are detected is to start to close that loop. So can you give us a little, is that on, on your processes?

Terry: [00:17:13] Hi, when I look at sort of the software stack, you know, you have BAS is that have been here for 30, about years. EIS. I think fairly developed.

And if you look at the adoption curves, right, the BAS is certainly, you know, well, well out there, the EIS, I think there's a lot of those. And then that's deployed in most buildings, you know, dealing with meter data FDD. Now. I mean, I stopped tracking it. 30 something. I know James, you gotta, you gotta list that's well beyond that.

and unfortunately I think we're only, although there's a lot of product in the space, the implementation, I mean, I would say the saturation rates probably only five or 7%, but that's still got a ways to go, but optimization from our perspective, when we do a building, I tell people we get probably half our savings from FDD.

And the other half from optimization. So to me, an optimization implementation is, further behind the curve yet. there's less companies doing it. There's less products in the space. So I just think that in order to get all the energy savings, it takes FDD and optimization. And so I think, and because it's further behind the adoption curve, um, you know, we see it as, you know, both of them have a lot of deployment capabilities here.

I mean, deploying both of those across. the other 95% of the buildings that are out there is, is what we all should be focusing on for the next decade about,

James Dice: [00:18:31] yeah, I agree. Uh, so cool. Let's let's bring on John Simone now. So John Sherman CTO at brainbox AI, one of the startups I was talking about earlier, so Johnson one, I want you to kind of paint us a picture pretty quickly on what's changed.

So we would just went through the whole history of this. So what's changed and why to brainbox. AI come about. And why, why is it a good time now to start pushing this technology at scale?

Jean-Simon: [00:18:57] Yeah. So, um, thank you, uh, Jane, so a lot of things are changing in that too. The, you know, um, what we're doing at green roads, we're just like, basically surfing on that change.

so look at, what's changing right now on the, on the computer side, storage storage was like a big issue, not too long ago. I mean, you want to, 10 years ago you want a storage,  a couple of terabytes. there was no CFO that would have signed up PO it's way too expensive. Why do you want to do it at that data?

so certainly an hour terabyte, like really,  do you have a credit card? It's not that expensive, so, but we have the capacity to accumulate a lot of the data, which was not really that easy not long ago. Um, then as the CPU, We, forget that  the calculation computational capacity is still increasing, not as fast as it was according to the moral law, but it's still increasing quite rapidly.

I mean, by 2015, we did pass the, basically the brain power of a mouse. and now by 2023, we're gonna equal the brain power of a human. So that's good that have capacity to calculation capacity. Certainly we have in our hand, so, Oh my God, what do you do with that? I know you're so we're starting for the first time to have enough power, actually too much power for what we need to do.

And that's completely new because I remember in the eighties, when I was in university, I mean, running a program on the mainframe. you keep being chopped down. I mean, the guy was saying, no, no, no. You only have 15 minutes on the mainframe. You have to reduce, reduce, reduce your equation, because you only have 15 minutes of computing.

Shouldn't tiring so, I mean, do you get frustrated because do you want to resolve that operational research problem and you're not allowed the computational power to resolve it? so what are you doing really? So now for the first time in history, we have more power than we need to resolve the problem.

So we're kind of limited by our imagination, what to do with it. so that's really what change. And then there's this, you know, AI exists for the last four. I mean, there's nothing new about AI. You're doing regression, relented, regression. You're doing AI machine learning has been around for 30 years.

I mean, come on. I mean, we all play chest, winter computer, even if it was like 25 years ago and you know what it was already winning on us. Um, so really. What's new is these deep learning network. These neural network kind of mimicking our brain, which are exploding in the last like eight years it's exploding.

And, um, and these techniques are giving us the capability to learn to self-learn to basically. Put a lot of data into it and they will learn from it and they will tell you what's going to be happening. And that's, what's really new. So we're still like, you know, kind of like a kid with a new toy. He was still playing with that Nutanix.

Like what, what can you do with it? And we're still trying to figure out what kind of business model can I, how can I make money with that? You know, um, how can I use it to create value and then sell it to somebody? so that's what we're still, uh, we're still in that stage and it's, it's quite a fascinating stage.

And of course there's people, you know, hurting themselves and there's other people which are, finding the right combination and moving forward. Um,

James Dice: [00:21:58] Great. Great. Thanks. Yeah. And you and I, and the product, that's when deep into the brain box AI approach. So anyone can go kind of deep dive on that if they want to.

Keith, before we go the perceived or my perceived controversies here, I want to definitely want to hit those pretty hard here for the next 10, 15 minutes. The Keith  the chance to introduce your new company, which launched. Yesterday to kind of solve this the same problem. So can you give us a 32nd elevator pitched on, facil.ai?

Keith: [00:22:28] Sure. Thank you. it's very interesting. it's a little bit controversial, you know, about what, what we're trying to solve here and what the problems are. I want to put  kind of a different spin on it. Um, so, you know, facil.ai isn't is an AI driven system is an optimization system.

But I wanna, talk for a few minutes about  why we're doing this and what we're doing. So I said a few minutes ago, the technology wasn't there. Right? Okay. So obviously we have to have the technology we've got, you know, increased memory and all that stuff.

But there's also a social engineering aspect to this. We've got to change as an industry and stop doing things that are not profitable when we're running buildings. Okay. So really, I really got it. Well, yeah, we know the dog. Okay. Yeah. So let me just like radon entree into that. So, there's a quote.

One of my favorite quotes, is, by Warren Bennis and he said the factory of the future will have two employees, a man and a dog. The man is there to feed the dog and the dog is there to keep the man from touching the equipment. That's what we have here. So as an industry, I saw Moneyball, I saw Moneyball a few months ago and that just sealed it.

And I'm like, I'm like Billy Bean. I mean, I remember  one of the things he was saying, well, if he's a good hitter, why doesn't he hit it? Good. You know, Tyson mentioned. Yeah. That, um, you know, my control system is perfect. No, it's not.  if it's controlling perfectly, why do you have so many hot calls and cold calls?

So this is more a matter of like, stop doing stupid things that we do as an entry, like resetting the chilled water set point up to 53 degrees when it was a design for 45. Um, stop running unoccupied set points in death Valley, California in the middle of July of 85 degrees. You're never going to be able to make that occupied set point.

So this is like I kinda turned facil or this advanced supervisor control is like the virtual plastic box on top of the thermostat. we're not going to do any of this stuff because humans are biased. We think we're trying to save energy. We're trying to push the envelope. If one pill is good, two is better.

that puts us into all this stuff. And machine is relentless. It's not biased.  It's not gonna make a logical decision. So  part of It's a social re-engineering of the industry and really getting back down to basics. Cause I remember when I was controlling DDC systems in 1987 and I had a little Commodore 64 and it can, you know, it was like a Nintendo, it had like 64 K a memory.

And I don't remember it being all. Messed up. I remember my buildings were tuned pretty well.

James Dice: [00:24:59] So this, this actually gets me to the first perceived controversy. So I think I'm getting just started to kick it off right here. So I think there's a controversy around it's kind of two levels. So the robot versus the operator.

So will the operators allow this. So Terry talked in the podcast about how we need to get the operator job into less of an operator and more of a maintenance versus maintaining equipment. So do you guys see operators allowing this and how do we train them? educate them on, well, this take their jobs and those types of fears.

Terry, you want to kind of fill us in on your thoughts on that.

Terry: [00:25:38] Yeah, well, so like every, a lot of jobs, right. Technology changes things. And I do think the operators, I don't think there's a choice. I think the guy above them, the facility manager, if he thinks that some software can do a better job. And I agree, I there's no question to me.

I believe the software will get. Good enough to do a better job. Um, in fact, honestly, a lot of the problems. We find it in a building, our operator caused right. Overrides for instance, or they don't really understand the sequence of op.  so I don't think it's an option. Yes. They, they might find it.

we hope we can train them to just do their job differently. you know, train them to understand, you know, they'll, they'll still be at the controls. They can probably still stop the optimization. Let's say if they want to, for some reason. but again, keeping the systems running and letting  the machine and the software, I think do it, and do it better.

We're finding that with self driving cars, right? I mean, there's no doubt in my mind that, you know, I'm, I'm, I'm waiting for myself, right. Driving park. Cause I'm sure it can do better than I can. right. That doesn't get sleepy, tired, distracted. So I think, I think we're going to have the same thing in buildings.

It'll be a little while. I mean, I'm not convinced it's around the corner. I'm convinced it's five, 10 years out. Maybe that level of automation, I mean, there's lots we can do right now, but using, complete, machine learning or AI, I definitely think is, is coming, but  Yeah.

James Dice: [00:27:05] Is there an aspect of, I headed to you Johnson, Mona about your, your perspective here, but is there an aspect to this of there's the pride of the operator, but then there's also the responsibility that a lot of controls vendors have taken.

So  a lot of times you have a service contract that contractor basically owns. The responsibility for these control sequences. And now we're coming in and saying, Oh, we're just going to override a few things here. No worries. Right. So there's a, there's a liability aspect. And there's the finger pointing aspects of how are you guys thinking about that and managing that with your clients?

Jean-Simon: [00:27:38] I mean, that's exactly the, uh, the number one bottleneck to, to deploy these types of technologies. Exactly. That there's like, how do you make it fit in the actual settings? Uh, that is in the building? so yes, there's a parade or in to Terry's point. I mean, absolutely. I mean, the, the, the fact that you're starting to automate and optimize.

We're we're bringing it to a new level, but, but also you're, you're fixing a lot of things which were supposedly they should have been fixed. Right. And that are the low hanging fruit are being there. And it's not everybody that has this top notch building operator in front of the BMS. A lot of building, they have this guy, which is rotating coming one day per week.

And you know, he's not there most of the time. So who's driving the building when he's not there. So to answer your question, Jane is it's really, Usually like a SSI will start to freak out when we get into a building and they go, wait a sec, this is my control sequence. This thing is going to very quickly start to override thing.

Then my control sequence is not really doing what it's supposed to do. and we have literally to take responsibility like contractually speaking, we're saying we're going to be much relating. We're taking responsibility and this SSI is now off the hook. so, but that's a discussion that needs to take place.

And some OSI are basically saying no way. I'm not going there. And some other are kind of embracing the fact that, you know what it's coming, we better get on the train. Uh, and we always start with the air side. I, we never start to, to Tyson point, we never, never start on the water side because we're scared ourselves , our, where the neural networks are going to do.

Um, and, also we have a legal people telling us be shirt or what you're going to be doing before you do anything. So yeah, I have to open that Pat, but it's, uh, it's like, yeah. Autonomous car. There's so many question about who's responsible for what if the algo take the wrong decision and kill the poor lady that was crossing at the sidewalk.

I mean, who's responsible. Right. And there's still a lot of a question to be answered on that, on that path.

James Dice: [00:29:37] Yeah. I think it's cool that you guys are just on this discovery path. It's really interesting to see. So Tyson, give me your thoughts on this. I think the main barrier here is like there's a perception that the control should be done by the control system.

So being someone from one of the traditional controls companies, kind of, how are you thinking about this? I'm not saying represent Siemens.

Tyson: [00:29:58] One of the, one of the enemies that is, that what you're saying right now. I fully agree with the points, right? It's it always comes down. Always say like any digital sips offering has to engage with the people and become part of the process.

Like we talked about value in the past. If you can engage and give confidence into a process. It will have a lasting impact. If you come in and try to take control, or maybe not, control's the wrong word for this situation, let's say take transparency and ownership of the system away from somebody. I think that's when you start running into problems, I think you have to not just replace a process, but like data enrich the process.

So I think there is a solution there that works. But it has to be able to engage with people on site. Can we talk about, I love that. I love the point about let's change people from operators to maintain us. That's a hundred percent what needs to happen. The gap in the market between skills and people and availability is getting bigger and bigger and the consistency across portfolios.

It doesn't exist. It doesn't exist. As soon as you lose one person, you lose performance. So I think there's a place, but it has to be part of the process. It can't just be a software solution. It has to be saying, we're going to integrate ourselves as part of your operations team, a tool. And that's what it is. It's a tool to be able to improve that process.

James Dice: [00:31:19] Yeah. I think a lot of the, Issues here because not enough stakeholders are being brought into the solution. So we have these new algorithms, these new technologies that, really will help us, but it's not one person's algorithm. It's everyone, you know, engaging together and to, to use the new technology for the betterment of the building.

So as we kind of close, it was out here. I don't see a lot of questions in the chat here. Does anyone have any questions for these guys?

Tyson: [00:31:49] I might have a question for Terry actually. Yeah, shoot. So I liked your point about how we've got a long way to go in terms of improvement and a lot of your savings function, fault detection, because I had the exact same experience.

Um, do you see closed loop optimization as being. I tool for a analytics or pulse detection service offering.

James Dice: [00:32:11] Oh, so this gets at another one of my questions actually. I'll piggyback on top of that, Terry. So when you talk, when you talk about this, will you talk about your 70, 30 concept or your 80 20 concept where you're saying, you know, most of the savings are from traditional control sequences?

Terry: [00:32:27] Um, both. So, the first, I guess the first comment is that, Uh, every time I talked to an FTD company, when I asked them about their plan, on optimization, I've been doing this for years and, you know, it used to say, Oh, they don't want us to do anything. They want us to be passive.

And I was like, well, why, you know, there's something controlling the building and it's not doing a great job. I think that's all changed. I know kgs and copper tree. everyone I've spoken to recently that, you know, optimization you know, writing back  is on the radar screen or you know, it's in the development.

Backing that really helps with that too. Right? You can write priority arrays. So it's not very hard to just write at the next level priority. and then if the optimization dies, the BAS takes over. So it's not hard to do. It's the way most of the planet optimization actually works.

and James I think you have this thing where you say, uh, you know, , walk, run, or, you know what I mean? Like yeah. So I do think, some buildings and this varies by building, but to me, if you go into a building and you do the basics, remove the faults, the overrides, and set schedules and set points and dead bands correctly, you're going to get savings, right.

Five or 10%. And that's, you can do that with a BAS and well, The fault detection part is better than FTD. Yeah. And then, so we always do that first. That's all, that's our sort of basics that gets a 10 to 30%. We think of the savings. Next are the, kind of already known algorithm standard, optimal start temperature, pressure resets.

This is stuff that BS is could be doing and some are all right, like, pressure reset on VAV. Air, air handling has been known for years. How to do that. Guideline 36 requires it now, but I would tell you, it's probably still only done 40% of the time. Optimal start probably 10% of the time.

So those algorithms doesn't require machine learning, AI. Any of that to me gets another 30 to 50% incrementally. and I think what's coming is that final 20%, right? That's the model predictive control or data-driven control, AI driven, right? That's sort of predictive adaptive stuff.

I think that's, what's coming and it's pretty exciting. We would like that, but for us right now. We're getting  all that 80% part. That's easy to get known technology, somewhat easy to do. and the other 20%, we'll just wait until the software gets better.

James Dice: [00:34:46] Got it. Keith Johnson, what do you think about that? 80, 20 or 70 30 rule or whatever, 50, 50,  whatever we want to call it.

Keith: [00:34:54] I completely agree. And I think it's 80 20. I mean, so I've been doing the 80% for the last couple of decades. I mean, that's what I built a career on, you know, optimizing buildings taking overrides off, resetting set points back to reasonable set points.

Addressing the schedule alignment. We've all been doing that. We all, I know that that works. The other 20% is where there's real opportunity. So, this time around already know I can do all the other stuff.  So the first project I'm working on is with a university.

That has a, um, they've been using these believable, smart valve, smart energy valves, and they're, they're changing the flow right at the coil to optimize for Delta T and temperature. So we're no longer optimizing for bowel position or open-loop control  it's optimizing it right there at the business end of the coil.

So it's got its own little ultrasonic flow meter. It's got his own temperature sensor, so it can calculate Delta T. But now what that's done is that caused a disturbance in the hydronic loop because you know,  you used to have a fixed coil valve position, and now you've got all these changing valves, lastly changing, and that's wrecking habit with the central plant.

So it's like the tail wagging the dog. So. We're starting out on, that part of it with the 20%. And we know that  if we can have this ASC system balance out of the hydraulic tool of Waterloo with, you know, four 1100 ton chillers, it can certainly change the set point on a, little RTU.

So we're starting out with the 20% this time around, and then we're going to scale back down to what all of us already knew how to do.

James Dice: [00:36:27] Got it. What do you think Johnson?

Jean-Simon: [00:36:28] Yeah, I mean, just because, you know, I want, I want to create some debate here. I'm not going to support Keith and Terry position.

Um, I think it's more 50, 50. and the reason being, yes, I agree with all of the low hanging fruit and sometime I'm completely in state of shock when we connect to a building and I'm looking at the way it's been managed, I go like really. You don't need AI. You just need somebody that knows what to do.

intelligence. I mean, yeah. And we look like magician, right? Okay. Well, no, really. That's not us. But there's this other 50, the reason why it's not 20 it's because if you start to look at the billing and okay, this is fine, let's optimize all of the classical things that we know how to optimize, but then you want to marry the air side and the Waterside to get, or so they work as a team and read their, you get extra.

And at the same time, you want to optimize that peak load. You want to start doing these shifts of peak when at different time of day, at the right time. And then suddenly that 20 become more cause the pie has just grown bigger. So you're kind of playing in a bigger sandbox. So, you're changing that percentage.

and open for debate on this, of course.

James Dice: [00:37:40] Right, right. Cool. Well, let's kind of wrap this. I think we can probably debate about a lot of these things all day. I want to say thank you to the four of you. Thanks for doing what you do every day, but also come in here. And, uh, and let's have your brand For a little bit. Um,  maybe we can do one more question. One question from the crowd to close us off.

So I liked John's question here and so to a building owner, how did they view. Supervisory controls value proposition. When the BAS companies say that they do that. And I think what I want to also say is like, what I'm hearing is that building owners don't want this.

And so do you guys say?

Keith: [00:38:18] I literally deal with this idea with this question all the time. So when, building owner says that, I say, okay, so you would let the IRS do your taxes for you. It's a perverse incentive. Okay. So, I'm and I'm not, you know, I'm a controls guy, Honeywell Johnson worked with Siemens, all this stuff.

I'm saying that there has to be, an independent party. That's not biased for any reason. Okay. So. another way we say it is like the Fox guarding the hen house, you've got to have some sort of ability to do not be biased in the situation.

And so that's the way we kind of address that because you know, the control system, we're not talking about replacing the control system here. We're not talking about ditching the control sequence. this chill water example. I mean, am I going to ditch the sequence of operation probably program?

When I was a Johnson 10 years ago, I'm not going to ditch the sequence of operation on the contrary, I'm going to embrace the sequence of operation and fine tune it so that it works better. So it's not about teaching the controls that I don't, I don't think it adversarial at all. I think these two, the control system and ASC can peacefully coexist.

both of them are the better because of,

James Dice: [00:39:27] Hmm. Okay, cool. Any other quick, final, thoughts from anyone

Tyson: [00:39:32] Just one point on that, I think of course, a building owner is going to want a system that reduces their cost. So I think if the value proposition is there, that they can control their systems cheaper than they do today, they'll buy it every single time.

So I think that's, what's yet to be proven at scale. And once the competence of that solution is done, there's no more arguments. I'll do it every time. So I think, we're early in the piece because people are new. It's a new topic, but once the value starts being shown, you know, day in, day out and brainbox have some great case studies ready.

Um, once that value is done at scale, building owner will buy it every time.

James Dice: [00:40:08] Totally love it. Um, cool. So anyone that wants to go deeper on this topic, several podcasts and the Texas podcast. I talked about this and the guy written the four or five essays on it now. So I'm always happy to talk about it.

These guys are too. all right. With that, thanks for everyone and, um, see you soon. Alright, friends. Thanks for listening to this episode of the Nexus podcast. For more episodes like this and to get the weekly Nexus newsletter, please subscribe at nexus.substack.com. You can find the show notes for this conversation there as well. As always, please reach out on LinkedIn with any thoughts on this episode.

I'd love to hear from you. Have a great day.

⭐️ Pro Article

This article is for Nexus Pro members only

Upgrade to Nexus Pro
⭐️ Pro Article

This article is for Nexus Pro members only

Upgrade to Nexus Pro

Get the renowned Nexus Newsletter

Access the Nexus Community

Head over to Nexus Connect and see what’s new in the community. Don’t forget to check out the latest member-only events.

Go to Nexus Connect

Upgrade to Nexus Pro

Join Nexus Pro and get full access including invite-only member gatherings, access to the community chatroom Nexus Connect, networking opportunities, and deep dive essays.

Sign Up