 All right, everyone, welcome to today's seminar. Okay, we have two speakers today. They actually have been working together for 10 years. More than 10 years. Almost 20 years. Oh, perfect. Yeah. George Lee from PowerFlex and you know, from KELTAP. Let me remind everyone that the next seminar is in two weeks, round three. George was the co-founder and CEO of this KELTAP startup company, PowerFlex. It has been acquired by BDF in New York for years ago and is now the chief product of it. George invested in a few of the company's original holy program for level two UV charge. So we have that in concentration and control and was responsible for, takes a few info, analogy and a few good strides of PowerFlex. George was a better degree in the GCS, MIT and master in the building process of KELTAP. Stephen is a human professor of the Department of Computing and language science, the Department of Educational Engineering at KELTAP and has been a honorary professor at the University of Vermont on the street. He has received numerous awards that including the IEEE Vocal Achievement Award and the ECM Sygnatrix Test or Prime Award and he's a fellow of IEEE Vocal Achievement Award. And he received the inspector from Cornell University and he received from the University of Berkeley. And that's where I met Stephen. Okay, so without further delay, I'll end here. Okay. Well, thank you very much, Timu. So I will start and then, but then the real tree is to hear from George. Thanks again, Timu. So we met at Berkeley when I was a grad student. So a long time. There was a long, long time ago, before time. So I'll talk about identity charging network and how we got from research to something real because of George. That's the EV charging startup house less that was acquired by EDF three years ago, 2019, co-family. So we started, we used to work, George and I actually both used to work on the internet research about 10 years ago. We started to really learn about how to do it from scratch. So we did quite a bit of work on a theory side and trying to understand how systems, especially as we carbonized the brief, what are the emerging issues and how to address them and so on and so forth. So there are a lot of interesting theoretical research that's done. And as we all know, not every piece of good research may have immediate applications and most of them don't. But it was not because some of the research that we've done really can be applied to address some of the issues and George invented this smart EV charger, which was the industry first, that really allow us to do real-time communication, computing and control. But that started the whole effort to really try to deploy this technology first at Caltech and then eventually through the startup and then now as the business cell through EDF. So it's basically the story that we're going to talk about. Okay. So I just want to acknowledge early contributors for George, of course, the team that founded Power Flats from scratch, really from his garage at home and eventually became Power Flats. And then we have early contributors from various people, a lot of people at Caltech, students, research staff and all the visitor students from Europe. It's amazing students. So I talk about the Caltech testbed that George built first at Caltech. And then when you transition into a commercial operation, then it's no longer a testbed that we can play with and break and review a research for those. So that was done by the PhD student at Caltech, Zach Lee, who's now at Power Flats. So I talk about the testbed and also the work that Zach Lee has done building the research order on top of the physical system. Then George can talk about how he built a business case, built a company and some of the challenges that we're working on. And then if we have time and we can skip this, I'll tell you a little bit about a theory where that is really motivated by this kind of work. And if you really have to optimize not just the grid, but also the end distributed energy resources, such as the V-Charging or the smart emerging or smart buildings, then you really have to think about modeling the sweet-faced network into it differently from the traditional method. So, and then hopefully we'll leave more times for Q&A and ask all kinds of questions about how to build clean-test startups and all the challenges to get that off the ground. So the testbed, the two references contains more of our details. And I guess we are all familiar why we're interested in this. So California, for example, has made a commitment to have a 50% original need of renewables by 2030. We are over-studying the target, and it's great when it becomes 60% now and 100% at 145. On the transportation side, we're going to a five million zero emission vehicles, which mostly will probably be electric vehicles by 2030. So if you look at the US, the electricity generation and transportation concern about two-thirds of all primary energies, and they meet between 55% to 60% of all greenhouse gas emissions. And if you really want to drastically reduce greenhouse gas emissions, you have to electrify transportation and generate electricity from clean renewable sources. So another thing that for those who are familiar with the dark curve, if you say a lot of fuel, the value of the dark curve, then you can charge about 13 million yen. So you can solve both the transportation department and also the dark curve problem if you have a lot of workplace degree charges. That's what we did. So it is quite as likely for someone to get an EV if there is workplace charging according to some survey by EDS. So this is the system that George Calder, the first system that he built in 2016. So this is a garage at Caltech from the electric room. He built this set of transformers and breakers and controllers and everything and then connect to all the smart chargers that he built from his garage. So that was the physical system that supplied power from the electric room in the garage to each of those smart chargers. The cyber system to go with that is that when the driver puts in a EV, there's a mobile app, through the app, the driver can tell the system how much energy she or he needs. So I need 30 miles and also the expected departure time. And if the system will have that information for every EV in the garage, how much energy they need and by what deadline? So that information is sent to the public cloud and where a lot of the computation is done. And the computation is very simple. Every two minutes or so, you form a combat quadratic constraint, quadratic program. You solve that problem to be from us and then they will determine the charging rate for every EV in the garage for the next two minutes. And then the next two minutes comes round, you repeat. When you look at the state, you form a new QQP, you solve that and then you get your optimal charging rate for every EV in the garage. That's how it works. The, this simple QQP is very flexible. The, for example, the objective function and call different preferences of that sideholds. So the sideholds may say, I think you want to charge a positive powerful for every EV in the garage. You can say, I want to maximally utilize my own size of the generation or I want to charge in the way, charge my constant garage in the way, it will minimize my electricity bill, for example, a combination of those. So all of those depending on the sideholds, you can encode into the objective function. The constraints takes care of the energy requirement before the deadline, the capacities, safety constraints, and others. So they will determine again in real time the charging rates of all the EVs in the garage. So that's how the cyber system works. That's George, thank you for that. At that night before you got energized, meet February 2016. So that was the research prototype. It was the first generation smart chargers that George built. And then now it's all standardized commodity. And there was the garage and all that. Okay, so that was 2016. By July 2020, so his garage has delivered one gigawatt of energy corresponding to about more than three million miles, a thousand tons of oil as your tool. So you would, so the company, and then George built a company and deployed it outside Caltech, a Slack as well here. So he was a choir and he would tell you a lot more details on that story. And I encourage you to ask him all the questions, all the tough questions at the end. So there's some interesting services which actually is pretty consistent with what people see around the world. So roughly, for example, each in the, every day is about, they will take about 10 to 11 kilowatt hours. That's the typical average driving distance at least. Caltech, for example, people spend on every six hours, but you can typically finish charging in less than two hours, which means that there's a lot of flexibility, inherent prosperity that's extremely valuable to help innovate renewables, which currently is not taxed. And therefore, the system that we built really tried to untapped, to really extract its untapped value, which would be important. Let me sort of start more pictures. So this is the charging rate, it ends over 24 hours. So each curve corresponds to one unit. So you can, you can monitor the charging rates and so on to achieve different objectives subject to various constraints. So this is early demonstration at JPL, this was 2016. So the idea is that we can demonstrate that you can in real time control the charging rate, aggregate charging rate of the graph, for example, to say track a signal. So in this case, it was the real time PV signal, actually, I think out of George's loophole. And then you can see that using the JPL installation, you can track the signal pretty well. Except you don't launch hours. At the time when you have eight charges at JPL, so if you don't launch for hours, you simply won't have enough to track. But otherwise, you do have the ability to track in real time. So this is a installation at NREL in 2018. And the idea is we want to reduce demand charge for those who know, basically want to control the P. So this is the building load over the weekend. You can see the gap curve over the weekend. There's not much in the building load. During weekday, you don't see this building load that such ways. You can still see a bit of a gap curve. And this is the EV charging. So this is the net load from the building minus the onsite PV charging load. But the point is that you can control the EV charging so that the P doesn't exceed whatever you specify. That's how you can reduce the charging demand and therefore reduce the operating cost. This is the impact of COVID for March, 2020. Okay, so that's the physical infrastructure that was built at Caltech in the end of 2018. So before November, 2018, is a test bed where we offer free charging and we encourage people even outside Caltech to charge them, get the data, give them the system and all of that. After 2018 November, you transition to commercial operation and therefore drivers have to pay now so we can no longer just break the system. And the idea here is to say, we want to build a software layer on top of the physical system that will make it, everything's open source, that will make it a research facility that's available worldwide. And you provide three things. So that's in your references. It provides, oh, so that's good because November, 2018, we started charge payment, the number of charging back in the jobs and now you have to pay in that code. Okay, so the idea is that we want to build this layer, software layer on top of the physical system that will provide real-time, fine-grained charging data that people can use it for all kinds of purposes. Also provide a realistic simulator environment for people who are interested in, say, coming up with new charging algorithms or even pricing algorithms. Then they don't have to simulate all the infrastructure because it's focused on simulating their own algorithm and running this similar environment. And in the third piece, you say, again, life. And the idea is that you, once you have the control algorithm that you're happy with, you test with a simulator, then we can allow you to charge the actual car, the Caltech. So I'll tell you a little bit about that. Okay, so ATM data. March, 2021, about 80,000 fine-grained, even charging sessions is publicly available. Everything is open source. If you want to use it and test it, if you want to contribute to a set, even better. So you can use it for all kinds of things. You can use it for modeling the behavior. Maybe you're interested in machine learning algorithms. You can use it to evaluate your charging algorithm or pricing algorithm. You can use it to evaluate what if the facility is the charging flexibility of your facility, for example. Or what is, if you have a lot of smart charging colleges and a theater, what is the impact, for example? So this is just more pictures. So this shows the flexibility at Caltech during the weekend, at JPRs and so on. And the conclusion is that, compared with if you do uncontrolled charging, you need about almost four times fewer capacity in your system that translates into a lower capital cost. And George can tell you what that means on the business side, the trial. You can use, again, use the data to look at the user behavior. So you can look at the arrival times and so on. And one thing you'll see is that whole COVID, the variance figure, because the hybrid of people's arrival departure time is different from the pre-COVID. You can use God's admission models, try to model the arrival, the departure, the energies and all of that. Once you build a model, you can use it to simulate all kinds of different, for different purposes, say machine learning, control optimizations and so on. You can use the model to do those. So that's the way you can get all the data. Again, everything is open source, extensible. Okay, so that's the data piece. So that means, who was a PhD student, also built a realistic example, that would simulate this batch of behavior. You can be as detailed as you want. And so you chose a batch of model, which is reasonably simple, yet relatively not for his purpose. Includes simulation of ED behavior, the charge of behavior, constraints that depends on, that is your conformance and breakers and the infrastructure in the garage. And also there's some constraints in the networks. You can also take ACM data that would drive the simulation. You can also take other signals, such as utility chargers, if you have fuel use, and you want to optimize the charging to minimize your electricity bill, the use you can use that. You can look at the solar generation that can drive the simulators and so on. And therefore the simulator is probably listed so that as long as you're in the space with that, you don't have to recreate this simulation model. You can just focus on the algorithm that you are interested in and just running the simulation. So that's your algorithm. And ACM life, the idea is the same interface, but underneath that is given, underneath that is the real garage, that's how it works. So the idea is that nothing, if you want to try a new control charging algorithm, you can easily charge and command. It will be a layer that will catch your command and check it is safe. If it is not safe, we'll just use the full power flex charging algorithm. Otherwise, we'll use the algorithm to charge the car, measure the state, send it back to the algorithm and then you can continue. So this piece is not entirely done because of good reasons. So anyone interested in that, you should let Z know. But you've done a power flex. Working with George. Okay, so... This is an exciting part. Thank you, Stephen. Good afternoon, everybody. So when Stephen and I were at Caltech, we were looking at building out all the EV chargers for the campus, right? And so, you know, we did a bunch of models and we looked at cost, we got some quotes and the cost that came back was like a million dollars which was basically impossible. So there's something like 15,000 per charger and we realized at that point that this is not just the Caltech problem, this is a global issue. We can't be spending, that's not cost effective, it's never gonna, we're never gonna get to the goal that we need to get to California if that was the case. And so, you know, we went on this journey to innovate improvements in anything from design to deployment, to instruction, all these things. And we were able to get it down to $3,000 per first call. And that was a pretty big end changer we thought, a five X difference. And so that's when we decided we should probably commercialize and do something. Thank you very much. I think that's why this extra time that we're likely to be using. It's the decrease generally. So there's this thing called load management and that allows you to over-subscribe circuits and that in return that makes you have to do less infrastructure. And so it's a cascading effect where by doing things more efficiently, it's a big savings, yeah. But I'll get into that later with a few more slides. So good question though. And so we looked at other things and once you've built the site, you have to operate it as well. So you have to pay the electrical bill, right? And so if you look at this curve here, you can see this is the over 24 hours, 42 hours. So you really need to reduce the peak and that's where the load management comes in. You can actually still satisfy all the requirements of your drivers, but make sure that your peaks are less. And so what we saw was like, most people showed up at Caltech at like 10 a.m. and they might be this crazy, crazy peak. And so if you needed to satisfy this, you'd have to build your infrastructure to above the slide. And that's where it gets really expensive because you're wasting all that capacity here and here. So that load management really comes in handy. When you work with other schools, UC San Diego with another one, they have a interest of lowering the carbon intensity. So this blue line is the real time carbon intensity before we have peak. And I said, hey, you can charge as much as you want, but keep it under the signal, right? And so we built the algorithm so that we can take in any signal and then just shake the curve underneath and still satisfy all the cars before they need to be. Okay. And then grid services, if you think about a really large scale, let's say you have a million chargers out there and then they're all in different places and instability in grid is really important to address. And so programs like the man-response, you might get an alert from PGD or whoever and say, hey, things reduce 50% of your load immediately. And so our system can still do that maybe for an hour and then we can still satisfy all the requirements for each one of these drivers. So a lot of value you can extract indirectly. Oh, we're hiring. So our headquarters are in Los Altos less than 10 minutes away. You could ride your bike there, but we have lots of positions open. We're about 200 people now. We have offices in San Diego here, Manhattan, and a lot of people will work remotely as well. Okay, cool. Let me, how do I jump to the other, other freezer? Okay. I think I got it. Okay. Well, let's play from the start. All right, cool. So I got a couple of slides here on PowerFlex post academic, academic, yeah. So we have about 10,000 stations deployed, mostly in California now. We also do battery storage or 500 megawatts, so solar, 500 megawatts there. We're actually the nation's second largest commercial solar installers that we do solar for like target warehouses on top and a lot of really big projects. And then a lot of commercial storage. So the batteries, stationary storage batteries that can help with these assets. We do everything from hardware only to turnkey. A lot of times the customer wants one firm to be able to address all the energy needs. And also when you have all these resources, you need to optimize them jointly. Because if you have separate vendors for each one of these, they may have different, you know, directives and that's not optimal. So it's very important to jointly optimize. Here's the vision. The long range vision is VPP, virtual power plant. So the idea is that if you have enough of these assets around the country, then you can basically coordinate them in such a way where it's like having a power plant. So you see these like a building like this, you might have a lot of solar up top, a bunch of charging stations, battery storage. And this is really the future that we're all building towards. In terms of customers and verticals, we're actually in a lot of places. So different cities like we're the provider for Palo Alto here, San Francisco, San Jose, a lot of real estate, private sector things, a lot of universities, a lot of OEMs, different cars here. We do the charging for their factories and things like that. A lot of non-profit, a lot of museums, a lot of fun stuff there. Research labs like Slack, GPL, workplaces and fleets. So these are the things that we've been focusing more on, larger scale. So each one of these sites generally has 30 plus chargers at one time, because that's really when the load management is applicable. If you only have like two chargers, you can just do basic algebra and figure that out. Not hard, not hard. So different cars, different needs. This is really important too, because you might have a plug-in hybrid that can run on gasoline. So they don't need a charger every day. Their batteries are really small, but they do take up a stall. So it is something you have to manage. Then you have like small battery EVs, like the ENWI-3. This guy has to charge every day or else they ain't getting home. So that's really important. And then you've got the larger long range EVs which don't have the charge every day, but when they do charge, it's a lot of energy at one time, right? So the algorithm's got to be able to satisfy all these different classes of vehicles at the same time. One slide about load management. So the National Electric Codes looks at EV charging as a continuous load. And what that means is that you have to have capacity available 24 hours a day for those chargers. So on the standard electric panel, you might only get 18 chargers before you could hit that limit. So I think in 2014, the Electric Code was amended to allow load management and that's kind of what we've pioneered is the ability to use software and math to make sure that even if you install five, six, seven, 10 times more dispensers that you're keeping everything safe on the panel board here. So the software and the math is making sure that you're not popping breakers and causing an unsafe environment. One slide here on the charge curves, very similar to what Steven showed earlier, but if you do uncontrolled charging, then you're going to have these spikes, right? And so each color is a car and as long as you cap them all together and solve it before they need to leave, then it's actually fine. So a lot of cool things here. Here is a graph of a standard 24-hour building, right? So the dotted, the thicker dotted line is the base load or the usage over the day. Solar is the curve and then you can see if you add that stationary storage and you add it all together that really you can really reduce the needs of your connection to greater. 4 to 9 p.m. is the period down in California. So you want to really avoid the need to do that period. You can see here that during that time, this load is all being taken care of by storage. I should go over here. Couple examples, yeah, so it's not like city of Palo Alto. We have over 50 chargers there. There's another hundred being built. We've done like all the downtown garages, the libraries, the junior zoo, all these things. So a lot of fun there. We're in County of Los Angeles. So this is over a hundred different sites like Walt Disney Concert Hall, like just the jails, all those things there. So another pretty big customer being used daily. LAX International Airport. We installed 1,200 chargers there, I think a couple of months ago. This was the biggest project at any airport in the world, I believe right now. So you can find us in any of the short term or long term parking lots. We were able to do all these chargers without upgrading any of their infrastructure because there was just no more power to give out anyway. So having load management is a key differentiator here. Fleets, this is like DHL. So we're their provider of the charging for all their depots. So the packages come in on their plane and then it goes on a bigger truck to like a depot. They might have a depot or the people around here in Sunnyvale, but and then they have all these last mile electric trucks that actually deliver the package to your doorstep. So this is a mission critical, like because if they don't charge the vehicle, then they can't deliver the packages. It's really important to make sure that this is always working for them. And then last slide on non-profits. So this is like Getty Center in Los Angeles. So world class facility. We want to make sure that we provide charging for all their guests and staff there. So we're hiring. In summary, so do you want to go back to yours? Sure. So funny story, funny you asked that. We actually came to Thoughtstand for, I think three, four years ago. And I guess maybe I talked to the wrong person. It was the director of transportation at the time. And he said, we actually don't want more people driving here. We want you on bicycles. So that was the response that I got a couple of years ago. And that's why we went over to Slack. And then we installed like 60 chargers at Slack. But I think the world has changed. And I think that there's a lot more EVs on the road. So we would be more than happy to come here and install a couple of hundred for you guys. You would like. I think it's more of a policy management. Do you want to share with us your? Okay, great. Sure. So let me maybe, I want to leave more time for Q&A. So maybe I will just skip some of the slides but just tell you five minutes. Let me see. Yeah, five minutes or so. Tell you, so the work really motivated, we have to think about three-phase modeling, powerful analysis and optimization slightly differently from how we do this traditionally. And let me just tell you the problem in the current way of thinking. And then the details, shameless advertisement. I'm writing a test for the details all here. So you can take a look. All right, so the motivation. So most papers and research papers are self included. We really think of the powerful analysis in a single-phase environment. And that is great for transmission system, a lot of transmission system applications. Or even if I have an algorithm, so I'm not that interested in the details of my infrastructure, but I want to demonstrate my algorithm works. Typically a single-phase model is good enough for that purpose. However, when you're actually controlling optimize on optimizing the actual system, like the rechargeable smart buildings and so on, you are not controlling the right variables. You'll see that the next slide, you can model your system using a single-phase model. You really have to look at the unbalanced three-phase model much more carefully. And if you don't do that, you may get into trouble when you see the slide of that when you actually implement system, which is three-phase. So that's the motivation. So let me get into a bit of what I mean by that. So if you look at a network, which basically is a bunch of devices, these can be multi-sources, current sources, differences or current power sources. They're connected by transmission lines or performance. So this is one statement where you have one device connected to a transmission line. This is the high model for those two my thinking power systems. So give more of a transmission line or a transformer and connect to another device or another node and so forth. So that's how you do our network. Now, if your system is single-phase or if you're sitting with three-phase, the difference at this level is triggered. For example, if you look at the line cones in your three-phase system, how they relate to the voltages at this terminal, VJ faces ATC, and at this terminal, it should be VK, ATC. Then the equation looks exactly the same, whether in the single-phase or three-phase. In the single-phase case, these are failure. These are complex numbers, failure. In the three-phase case, these are metering and these are three-by-three complex matrices. That's the only difference. But only structurally, they are exactly the same. And if you're marking your system at this level, three-phase is trivial. It's a trivial extension of the three-phase. However, if you are dealing with the kind of system that George described earlier, where you have to optimize and control the actual charges. So you're not controlling these variables. You're controlling these variables inside your system devices. Then that's the difference. So these internal variables are typically these kind of applications where you can actually control it. What interests on the network, over the network, are these terminal variables that are externally observable. The matching between the terminal variables that you actually control and the terminal variables that are observed externally, that matching may not be invertible. That's the issue. That's really the cross of the three-phase model. Let me just look at the time. Okay, four more minutes. Whether your three-phase device is the current source which the model, which is exactly the model of the charger, or it can be a quantum power source or impedance and so on. It's the same story. You are internal variables which may be the only thing that you control. They will induce terminal variables which are observed externally and which interact with the network. So the connection between the terminal variables and the terminal variables is really the key if you have to look at unbalance to the system. So that's the issue. Okay, so one last time, the motivation. If you don't do that, so this is actually the charging, that if you don't do the proper three-phase analysis for modeling, then you think you are controlling or charging so that all the safety constraints like in reality, they do not. So you can have basis EMC that violates the safety constraint. Whereas if you do the proper three-phase modeling, then you see to make sure that every phase conforms to the safety constraint. Okay, I still have three minutes. So let me just say how do you think about this? So if you think about how to model a system, where in a single phase or three-phase, you model the device, whether you can model the device as multi-source data generator as a current source that they know or what the impedance or constant output. So that's the device model. And then you model your lines and transmissions. You put them together, you write down specifically, only two things you can write down, the nodal balance equations or nodal tau flow equations. That's combined to give you the overall model. In a single-phase case, the device model is trivial. In a three-phase case, that's the key to this. Again, the transmission model and therefore the network equations are trivial between single phase and three phase. All the complicated subtleties appear at the end of the device model. Whether you need to do that. You don't need to do that. You're controlling at this level. You don't need to do that. You're controlling at the end of the devices. So that's the key. Oh, that's good. Okay. All right, so let's see. What else do I want to say? Okay, so the key question is really how do we think about the conversion between internal variables that you control and internal variables that you travel the network? And that mapping is, without a configuration, it's not invertible. And that's where a lot of subtleties comes in. So you have to look at that basically. So I want to be primed for Q&A and EFOR. Let me skip the details of this. Let me just put the advertisement. Details are here. So we can go to Q&A. Okay, thank you very much. All right, there's a question here from Craig Lewis. How much flexibility does PowerFlex EV chargers have with respect to set pricing that discourages EV charging during the peak time of use, four to 9 p.m. daily when the grid is most stressed and energy is most expensive? Yeah, so I can answer that. So we have a discussion with the site host and really to figure out what they want to do during this period. So some customers will say, hey, track the real-time pricing from PGD. And so we'll actually, let's say you start your session at 3 p.m. and it goes to 10 p.m. or when it crosses the 4 to 9 p.m. then our rate will actually go up to match that. And for example, in another case at NREL, they wanted something a little bit more exotic. And that was like the more energy you use, then the price changes. I think it gets cheaper as you get more. So it's encouraging people to hard as much as you can in your vehicle. So our algorithm has the ability to address either of those situations. Okay, next question from David Cole. Are you working on V2G? So vehicle to grid? Absolutely. So vehicle to grid, I think it's gonna have a really big year. There's been lots of announcements from various manufacturers, including Ford. Ford's new F-150 has some backup power solution. Lucid air has something as well. So I think it's really that if you look at the global supply chain for batteries, the automotive sector has really taken a dominant piece of that production. And so the stationary storage market has a very little percentage of those batteries. So it's gonna be way more cost effective to literally buy a EV that's a battery on wheels than it is to do a battery without wheels. So we are actively working on this and it's, I think it has a very bright future. A lot of the issues you're looking on to a lot of, or there's a lot to be in this adoption. For example, are you looking at mass driving or are you looking at the engine type? Yeah, we do a lot of fast charging today. There's a lot of challenges there. Takes a lot more power to bring in. The construction is very expensive. The chargers themselves are $50,000 apiece. So they're pretty expensive. Also you get the cost of energy because of the high spike. That's also very expensive. So it is pretty challenging. Now on the software side, we're always waiting for new innovations. And so like ISO 1511.8, which is this whole plug-in charge. It's basically communications between the charger and the vehicle. So right now, most public stations, you have to plug in and then use a mobile app to pay for the session and whatnot. With ISO 1511.8, it's like the station can contact your car, it can pull the VIN number, it knows, it's a bi-directional communication so that you don't have to have an app. It's just plug and go. So for you that are familiar with like Tesla and Tesla superchargers, it's a very good experience to just plug in and then the car starts charging. And that's kind of what, that's a gold standard. And that's where everyone else is trying to get to. No. Great. Let's see. Next question is here from Craig. The Clean Coalition is currently staging 30 sites in LA County for solar, microgrids and EV charging infrastructure. Each site will need between one and 20 level two EV charging ports. Who at PowerFlex should be contacted for to discuss details. Welcome to email me. It's just George at powerflex.com. We are LA County's sole provider of EV charging right now. So we're happy to help on that. We can do the solar microgrids as well. We have a lot of battery systems and software to go with them. Cool. Next question from Dirk. Have you looked into vehicle to grid at all? So I think we discussed that. Not only controlled charging but also controlled discharging. Yeah. So it's a five directional resource. That's gonna be really important. So you can charge the vehicle when it's low cost and then you can provide the grid services when the grid needs it the most. So absolutely. Next question, Craig Lewis. Do utilities like SoCal Edison embrace load management designs to allow service ratings and switch gear to be sized for the load management limit or do they simply force power calculations based on the number of EV charger ports multiplied by the level two power minimum? So this is a great question. And I think the utilities have changed their mindset over the years. So one inherent I guess problem I think with IOUs like industrial utilities is that they get paid on a port on a fraction of their deployed assets. So they're incentivized to deploy as much hardware as they can in the field. Did I get that right? That's absolutely right. So load management on the other hand is a tool that allows you to maximize your existing infrastructure. So you can see how there might be a little conflict of interest there. But I think the utilities are realizing that the adoption of EVs is coming so fast that you can't just deploy as much hardware as you possibly can. It's not financially responsible and not possible in the timeframe that you needed to do that. So we've worked with many utilities and they know who we are. We actually just got adaptive load management trademarked and in a lot of these investor IOU programs that have funding for EV charging, load management is becoming a requirement in those programs. So it's a big win for load management. And I think that they're really changing their mindset on that. Let's see, what else? All right, it's from Anonymous. Does adaptive charging network mean an optimal deployment of chargers? What's the factor that, what factors have the most impact on this problem? Is it human behavior? Yeah, I would say like, there's a lot of like oligarchies behind this too. So when somebody plugs in, a lot of times they expect full power right away until their car is done, right? Maybe it's anxiety, maybe it's just they want to make sure that their car's absolutely charged. Now, that's why in our mobile apps, it's really important to give a five hour forward projection of the charging curve to the driver. That way you are addressing their anxiety. Like when you plug in and it's only charging at past speed, you know, they're kind of worried. Like, is this really going to finish or not? So providing that forward projection, I think is one of the most important. Yeah, cool. Eduardo, is PowerFlex also interested in residential chargers? My energy comes from PCE, for Penicillic Energy, San Mateo. Will you work with a CCA like PCE to, so my EV is helping the grid? Also, have you been talking mostly about consumption? Are you PowerFlex, are you PowerFlex also looking to production, production part of DER, solar panels and batteries? So, yeah, we worked with many, many CCAs, especially around this area and they have incentive programs. So we've been deploying a lot of chargers in their service territories, mainly in condos and apartments. I think the strategy there is that if you deploy in a multi-family housing, that gives access to a lot more family at one time, right? We actually don't do any residential, like single-family home chargers at this time. We may do that in the future, but that is a totally different set of problems and optimizations. Okay, Anonymous really appreciated the discussion on the three-phase unbalanced PowerFlex. You got the name, okay. Anonymous, it says, how much more value do you think a more vertically integrated hardware system could unlock? For example, do you see value in owning the BMS or having access to the BMS and solar inverter control systems and how much value if so? Yeah, it's very important. And that's why we actually at PowerFlex, we have a certification program where we bring in a lot of different vendors and their products and we test them extensively. And what I mean by that is like we usually try to ask them for as much access into their hardware as possible, like into the BMS and getting stats, like maybe temperature, health, cycles, like all these things, as much information as we can get because that only helps our algorithms treat the battery better in this case and so forth. So, yes, it does provide a lot of value if you're very feeling great. I think a good example of that is Tesla. Like they've got everything from the cars, chargers, batteries, and you can see it's a very seamless system that they've built. Okay, Eduardo again, oh, sorry, this is, sorry, repeat. Let me see, is PowerFlex business model profitable without incentives? Yeah, I think so. It depends on what vertical. We have a lot of Fortune 500 companies who paid us for chargers that have no incentives whatsoever. And what's really interesting about this that we realized early on, especially in the Silicon Valley, is that if you look at all the costs that pertain to EV chargers, like the infrastructure, the charging station itself, all the permitting, and what you find out is that the cost of the employee to go and move their vehicle is actually the most expensive thing out of the whole equation, right? So a lot of companies here in the Bay Area, like they get a lot of perks, free food, free laundry, whatever, and so free charging is one that you'll find at many, many large companies, and it's all an HR type of optimization, right? You want to make it easy for your employees to come onto campus and do what they were hired to do and then stay as long as possible. So there are certainly many situations where it is profitable without incentives. Okay, Dennis Silverman, can you use multiple phases to label power by its fossil fuel or renewable source? Is that for you or is that? Sure, so the quick answer is yes. So actually we have a project at Caltech now to look at campus decarbonization and where this is one thing that we're trying to get there. So what is important is that you need to know the circuit, the entire how the three phase service laid out so that you know which circuit, even on which phase, is connected to what kind of degradation source, whether it's the cogent plan, whether it's the grid power or whether it's sort of a panel. So with that piece of information, then indeed you will have that information. I guess the underlying question is that with this information, then you can take this into account in your EV charging optimization. If you want to let's say charge in a way, it will minimize your carbon emission, for example, then that is possible to do now. And also I assume like each generation source has a very unique signature. You can put some machine learning on it and you can probably detect it pretty easily. Yeah, good point. Okay, from Ranxin Ying, how does PowerFlex deal with the communication to the chargers, mostly open protocol like OCPP or cloud to cloud API, any challenge in talking to commercial chargers? So yeah, OCPP is open charge point protocol it's just a standard that most chargers you use nowadays. So you can tell it like start charging, stop speed up, slow down, authorize, turn off like just basic commands. So that's what we do as well. We support any chargers that are OCPP compliant. And so that works pretty well nowadays. Five years ago, that wasn't the case. We had to build custom communication chips to go into each charger and then do it that way. But nowadays it's pretty stable. From Dirk, do you have any research positions at PowerFlex or should we go to Caltech for that? Yeah, I think both, but I think for the real hardcore stuff, certainly Steven, I think is good. The use will stop then talk to coach. Who do you consider PowerFlex to be PowerFlex's top competitor? What do you think your competitive edge is? Is it the adaptive load charging? Yeah, so we're kind of doing like the work of five start-ups at one time which is challenging, but I think there's a lot of rewards to that. So we're doing the EV, storage, solar, micro grids. So you might have like in storage, you might have someone like STEM that's just doing storage. In charging, you might have ChargePoint, solar might be like Solar City or Sun Run and so forth. So we really believe that the joint optimization is a game changer because you need to have all those things talk to each other in a choreographed way. They can't just be all optimized for themselves. So I think it's really that unification. And of course, yeah, the adaptive load management is important. We're the number one provider for number of sites above 30 stations in the United States. So that's definitely because of the algorithm. And finally, it says, is the hardware part of the level two charging considered more commodity at this point? Yes, it is. The charger is pretty simple when you think about what it does, but we actually have our own charger now and that's because we've looked at a lot of people's chargers and we said, hey, like let's take the best things from every single one and then put it into one and that's what we've done. Yeah, but it is a commodity. Do you, last question from Dirk, he says, do you also optimize for battery longevity? So when we look at a new project, we do a financial model to see what that operational period is supposed to be. Maybe it's 10 years, maybe it's 15, 20. And so based on what the customer is looking to do, we will cycle the battery to meet those objectives. Yeah, so it is considered. But I would say that if you look at a meaningful battery that's like half a million dollars, like you wanna be able to save at least that much money over the operational period. So at least that much. So generally the longevity, whether it lasts 18 years versus 15 years, like it's not that big of a difference. No, cool. I think that's all from the online section. And speaking of the thing, thank you. Thank you very much. All right, so have a fun. Thank you.