 Okay. Well, thank you, everyone, for attending our third New Mexico Smart Grid Center webinar series. My name is Anne Jekyll. I am the Associate Director of New Mexico EPSCORE. And for our webinar today, we have a really great lineup from Los Alamos National Laboratory and Sandia National Laboratories. My name is Lisa Barnes from Lannell and Jay Johnson from Sandia and they'll be presenting on microgrid related research that is relevant to the New Mexico Smart Grid Center project. All of these webinars are recorded and archived so we will put them up on our website afterwards and you can access this information. I also want to point you toward the question and answer box on our webinar feature so via your zoom interface if you have any questions for our panelists today. You can type them into the Q&A box and we'll have a little bit of time for Q&A so please look for that. Usually at these webinars we announce the next webinar that we'll be having but we will take a break for December due to the holiday break and start up again in late January. And right now we are planning out our spring semester webinar topic. So if you have an area that you would like to hear about, if you would like to give a webinar yourself, please contact me and we can look at fitting that in. And in addition we will be sending out a poll to our entire project team to find out which day of the week might be the best day. So we might move it off of Fridays. We'll probably keep it at noon so we want to hear from you about the best day for you. So with that, I am going to cede control of the interface to Arthur Barnes from Lannell. He's a power systems engineer in the information systems and modeling group at Lannell. His current interests include geometrically induced currents and electrical transmission networks, analysis of protective devices and design of microgrid systems. And he has a PhD, MS and BS in electrical engineering from the universities of Arkansas, Florida and Colorado respectively. I've heard a lot about the work that he's doing up at Lannell and pleased to hear more about it today so thanks Art for joining us. Go ahead. All right. Afternoon everyone. So I've already been introduced. So, I don't have to do that. So what I want to talk about today is a look at protection of inverter interface microgrids. And specifically, I'm going to be looking at admins protection. And how does this perform in the absence of high fault currents that you would see an inverter interface microgrids. And, you know, are we able to detect the fault. So, the first thing I'd like to go over with everyone is. Microgrid protection. Difficult. So, there's a number of reasons for this. The first one is a, you're, you know, when we typically think of microgrids we think of them being supplied by inverter interface generation, you'd be at connection engines, micro turbines, PV. Also microgrids will operate in both islanded and grid connected mode so we can get very different fault currents on different operating regimes. In addition to this, we may have micro grids that have a mesh structure as opposed to a radial structure. Both of these are difficult for traditional distribution system protection. Traditional distribution system protection relies around overcurrent. It also relies on assuming that the structure of the grid is radial and microgrids break this. Currently, current industry practices and microgrids is actually kind of embarrassingly simple right now. They generally still rely on overcurrent protection, either zero sequence or negative sequence. This is usually focused on protection for the purpose of protecting the generation, and it also relies on the assumption that all the elements of the microgrid are pretty close to each other. So, if protection operates, generally the generation is going offline along with the entire microgrid. So important in the future, this may not be the case. We're looking at the idea of microgrids which are not worked across distribution systems, so generation can be spread over an area of miles. But when we get to this stage, being able to have protection that operates selectively and minimizes outage areas is fairly important. The first question that came to me is why don't they use differential protection? And the answer to this is for microgrids which have a large number of load taps if they're supplying an industrial facility or residential area, we wind up with radio lines coming off of them, where there's a sufficient number of nodes that we can't cost effectively provide differential protection on every line. So, this scared me in the direction of admittance protection. However, it's certainly not the only possibility. One solution that has been proposed is providing fault current on microgrids, say with traditional generation or synchronous condensers or induction motors. However, this is a bit of a step backwards. One of the nice things about microgrids and inverter interface generation is they don't provide fault current. Fault current is dangerous. It can damage equipment and we'd like to find a solution which takes advantage of the lack of it. Steered me in the direction of transmission system protection, which, you know, although operates in the presence of fault current, you know, it doesn't strictly rely on the ability of generation to provide overcurrent as is the case of traditional distribution system protection. So, there's a number of options which are listed in the slide of which admittance or distance protection is one of them. In this presentation, I prefer to use admittance protection because as I'll show further on. In microgrids, it's generally fairly difficult to estimate the distance of a fault, especially in ones that are relatively spatially compact. Let's go back to what's happening in industry these days. They're generally using still overcurrent protection. They rely on a negative sequence and zero sequence protection, which can work if loads are close to balanced. However, if we have a lot of loads that are single phase circuits, you know, such as residential or office loads, if we have a fault on one of the single phase circuits, we can lose that increase the unbalance of the system. And this could trip the negative sequence or zero sequence protection erroneously. So, it's not a ideal solution. So, now let's look at admittance protection. The first thing to consider with admittance protection in a inverter and phrase microgrid scenario is that inverters behave very differently from synchronous generation. And what dominates the behavior of inverters is their control systems. Current state of the art in microgrid controllers is a move away from controllers in a rotating reference frame, such as DQ zero, which is commonly used in industrial motor drives to static reference frames, whether the Clark or the just the raw ABC coordinates, the Clark coordinates, if zero sequence is not supplied by the inverter, it will only require two two controllers for the alpha data coordinates, although zero sequence is also included and that requires the second third gamut coordinate. So, I think current practice is favoring using ABC coordinates, but with a proportional resonant controller, which is able to provide zero steady state error at 60 Hertz. Now, what happens during during faults. In the inverter, the thermal time constants of the semiconductor switching devices is they're very low compared to the copper in a synchronous generator, so they can overheat in seconds or even less than a second. So the strategy is to limit their current during faults to within twice that of rated current. There's a number of ways to do this one is instantaneous saturation. This isn't ideal because it introduces very high harmonics into the system during the faults. A preferred solution is a set reset solution where when the inverter detects that it's reaching an overcurrent condition. The inverter which is operating in voltage regulation mode will then switch to current regulating mode. So where the affected phase will supply rated current and the other two phases will attempt to regulate voltage. Now, now that we know what's going on with the inverter. Let's go into a little review of admins protection. So I'm considering ground faults for this presentation. And in this case, I'm using admins protection with current compensation. So the top equation that we see this is equivalent to measuring the positive sequence impedance of the line. In the case of line line to line faults, ZLL is also equivalent to measuring the positive sequence of a line during a fault. And this is beneficial because of the question that I'm asking in this first bullet point, which is the ratio of positive negative sequence voltage to current. So we'll see in the following slide that this first quantity will actually give us a distance to the source or generator as opposed to the fault. So that will completely screw up protection coordination. So here's a look at a small case study system. So this is a three bus system with two lines and a terminal load, and I'm looking at a line to ground faults between these two lines. When I break out the system one line into its individual conductors, we have a look at what what the line to ground fault looks like. And what this results in is neglecting the load, we're going to get a fault current along phase A and phase B and phase C will be approximately zero current. If I break up the three phase circuit into its equivalent sequence networks. And then I use the inverse sequence transform. What I see is the sequence component of the fault current. They're all equal. And this gives me an interconnection of the sequence networks where they're all in series, doing a little bit of simplification. I got the A Thevenin equivalent circuit. And I'm able to calculate what the current is measured by my relay. Now, what we saw in the previous interconnection of sequence networks was, there's only a positive sequence voltage. Now, for the case of an inverter which is operating in current limiting mode as I described. I now have the presence of negative and zero sequence voltages on my equivalent sequence networks. So, the voltages and currents that I'm going to be measuring are now going to be somewhat different. And this gives me a new simplified network where I got a second source, which is equal to the sum of the zero and negative sequence voltages produced by the inverter. Fortunately, as I show analytically and in simulation. Current compensation for distance protection is robust against unbalanced operating conditions. So it's able to correctly calculate the impedance between the relay and the fault under both analytical and simulated cases. Now, there's still potential issues with this protection method. In particular, the most significant issue is in a relatively small microgrid the line impedances are going to be negligible compared to the source impedance of the inverter. So this means that what I had to do was set the operating region for the admins protection to be very large. So that's why I'm using the term advanced protection as opposed to distance protection because I can't reliably estimate where on the line the fault is. And this means that in order to maintain protection coordination, I'm dependent on using pilot relaying. And if the pilot relaying fails, then I'll end up with an outage area that's greater than necessary. Now, given the requirement for communicating this channel. This does open the door to more sophisticated protection schemes. So one particular set scheme that I've been looking at recently is that of state estimation based protection which has been proposed. Both on the transmission distribution system. And this is a generalization of differential protection. And it's able to operate with the failure of one or more sensors, while still being able to discriminate between in zone and out of zone bolts. So that brings me to my conclusion. Thanks Art and I would encourage everybody if you have any questions on his presentation to type them into the Q&A box in the zoom interface. I don't see any as of yet. And so maybe we can have a joint Q&A at the end of both of the presentations. So with that, I would like to introduce Jay Johnson from Sandia National Laboratories. And he is a principal member of the technical staff at Sandia and where he leads several multidisciplinary research projects focused on power systems control electric vehicle charging distributed energy resource cybersecurity and renewable energy integration. He has seven patents and is authored over 100 technical publications and will be presenting today on a recent project that they undertook at Sandia. So take it away Jay. Thank you. Thanks and I assume I'm sharing my screen is it showing up on your end. Yes, I see it. Okay, fantastic. Good afternoon, everyone. Today I'm going to be talking about distribution voltage regulation using DER grid support functions. This was a component of a larger project. A solar energy technologies office funded project under the energized solicitation. And it was a pretty big project between the DOE contributions and our cost share it ended up being a $5 million program with a number of partners that you can see here. A piece of this effort was looking at different methods of doing voltage regulation on distribution systems. And we took a lot of this work from kind of a conceptual level to an actual field deployment. So that's what I'd like to walk through today and operate my screen. So just taking a few steps back and talking about the context here. As you know PV is growing very quickly as the cost is dropping you see more and more penetrations of both distribution connected devices as well as we're starting to see some higher voltage level connected equipment. And that could include energy storage systems as well which have a lot of the same capabilities and functionalities as the other PV systems. Problem is that as we install all these devices we run into different technical challenges at the distribution level and the transmission level around voltage and frequency regulation protection which we just heard a lot about. And we need to come up with solutions to those challenges in order to continue installing renewable energy systems on our grid. And so one of the obvious solutions here is advanced inverts which include a suite of grid support capabilities that actively support the great voltage and frequency by changing their output characteristics. And they interact with tolerance to grid disturbances that's like a voltage and frequency ride through and they interact with grid operators and or aggregators using communications. So that's the interoperability capabilities of these devices. And so we have some research questions based around that what's the best way or technique for providing this voltage regulation. How can we evaluate these techniques prior to actually doing a field demonstration. So that's what we're trying to address with this project which is very briefly conceptually distribution voltage regulation is kind of depicted here traditional power systems from distribution substation down to the secondary. You can see a dropping and I don't know if you can see my cursor but I'll kind of sketch it here. You see a voltage dropping with distance from the substation down to the end load. So if you look at your voltage, you can get lower and lower voltages. Typically what will happen as you install PV systems is especially towards solar noon, you're injecting a lot of current and this is increasing the voltage there. There are a couple of limits that utilities have to abide by. These are the ANSI range A and B limits associated with that voltage band and they need to operate within this to not damage equipment and to provide, you know, good service to their to their customers. So the question is, well, okay, we can potentially add some additional equipment to this system as we install PV to correct for this problem. Or the more cost effective technique would be to use the capabilities of the inverters themselves to help bring down that voltage and control the voltage. By communicating or using autonomous grid functions programmed in these devices. And so that's, that's what we're trying to do here. So that brings us to the project, right? The project was called Protromos, which is Programmable Distribution Resource Open Management Optimization System. But the idea was shown on the right that we'd take voltage regulation power simulations probably in OpenDSS is typically what we use. Convert those to real time simulations in an OPAL-RT environment. And then we can run power hardware in the loop simulations where we actually connect physical equipment to the power simulation prior to doing the field demonstrations, which we did with National Grid. And so by stepping through each of these layers of fidelity, we were able to improve or provide some sort of confidence that our approach was going to work in that field environment by adding more and more fidelity to the operations. And so the three different grid support or I guess the voltage regulation techniques that we used, which use different grid support functions were distributed as autonomous control, which is the volt bar function in our case. And that's a technique called extremum seeking control where you inject a reactive power ripple by multiple DVR devices and pull out their influence on a global objective function based on that ripple. And then a more, I would say a more traditional optimization approach where we use the state estimator to calculate the optimal power factor set points for the devices. The volt bar is pretty straightforward. We understand what that is based on the local grid voltage we inject or absorb reactive power at the device. The two other probably less well known capabilities are extremum seeking control, which is depicted here and the optimization, which I'll talk about on the next slide. So in this one, what we have is shown on the right here a few different inverters that are all injecting a probing signal. That's a reactive power ripple. So the relatively low, you know, think maybe, I don't know, a period every, I don't know, 30 seconds or so something like that are relatively low periodicity in terms of the ripple, each of them have different frequencies. And because of that, you can extract their influence on a globally transmitted objective function. And since we didn't have the ability to broadcast that objective function, what we did was we pulled in data from the field and then calculated the objective functions locally and then submit it and send each of the inverters the reactive power set points that they needed using a power factor set point. So while this is rippling like this really it'll trend in one way or another based on the gradient it sees based on this ripple. So a little bit complicated, but hopefully that makes some sense. The last technique is based on a particle swarm optimization technique using an OPF optimal power flow, where we take and measure from the power system and a bunch of intelligent electronic devices like PMUs, micro PMUs, the inverters themselves, different data that is passed to a Georgia Tech state estimation tool called WinIGS, it calculates the current states. We export that into a particle swarm optimization routine that instantiates a bunch of power factors for the different DER devices, and then runs the optimization on that to calculate the optimal power settings based on an open DSS simulation running here. And then those power factor settings are then communicated down to the end devices. And the objective function that we're trying to minimize is shown here. Essentially, we wanted to minimize cases where we're exceeding ANSI limits. We don't, you know, you minimize that essentially when you get closest to the nominal voltage and a power factor of one. And the power factor component is some crude method of considering the economics of it. So if you're operating off a unity power factor, you're not producing potentially not producing as much real power and therefore generating as much money for the customers. Okay, so in terms of methodology, we've got a few steps in this process. We collected distribution system models from our two utility partners, PNM and National Grid, because we're in open DSS and SIME. We converted those to open DSS models. And then in order to run these in real time using eMegasim in Opal RT, we did a circuit reduction technique. And this is some work that Matt Reno has done here at Sandia for many years. And so you can see we went from 9,300 lines here on the left down to just 32 lines, which could be modeled in Simulink and run in real time. And then we put that in a real time simulation environment using RT Labs and RT Lab. And then we connect the real time simulation to a power DER simulator that presents the active and reactive power set points of multiple DER devices to the simulation. And in the final stage here, we connected physical equipment and replaced some of the simulated DER pieces of equipment with actual physical inverters. We conducted the experiment at Dettel, the Distributed Energy Technology Laboratory at Sandia here. Dettel is a highly reconfigurable laboratory with about 150 kilowatts of PV coming into, I'd say probably a dozen different inverters. And we can hook up different things to either PV simulators, grid simulators, different load banks, create single phase and three phase microgrids, and do a whole host of other experiments. And in this case, we were using a programmable DC power supply, which is our PV simulator to create a highly variable voltage or highly variable irradiance profile. And our AC grid simulator and Amatek simulator that was connected to the OPAL RT to provide that hardware and the loop feedback. Okay, so looking at how this works in the bigger scheme. The bottom of the image here is Dettel, our lab. Inside the OPAL RT real-time simulation, we have a distribution system running. We've got PMUs connected to each of the buses there. And each of those PMUs are exporting C37118 data streams to a physical SEL3373 phasor data concentrator. Data is then exported over the public internet to our connected energy software vendor, which was running the Georgia Tech WinIGS data estimator. So it completes the state estimation and exports that of another C37118 stream to the PSO. We calculate the set points for the devices and then feed that back to the FPV simulator as well as our physical hardware and the loop devices here. And so that's how you close the loop on the optimization. Okay, so let's start taking a look at some of the results for the PNM use case. You can see a simulation here. It's about 1500 seconds. I think I remember that four and a half hours or so. And the simulation show a lot of variability. We wanted to use high irradiance variability to get a sense of how these voltage regulation techniques would perform. And we ran the cases for a few different runs. So running baseline, volt bar curve, extremely speaking control and particles form optimization for each of the phases that shown up here. I'll focus your attention on the lower image. And so you see a large line running down this plot, which represents the average bus voltage for the entire feeder. And then you see two patches. You see multiple patches here, but the patch represents that the upper side, the maximum bus voltage. And then it stretches down to a lower point, which is the minimum bus voltage. And so this gives you a sense of average min and max of the feeder. And so it's a nice way to represent it, although when you plot everything on top of each other, it's pretty hard to understand. But in brief, for the baseline case, which is the gray or the black one, you see we're pretty high voltages on this particular simulation. When you add volt bar curve, which wasn't a very aggressive curve, you improve that a little bit. However, if you use the extreme and seeking control, you can see the probing signals impact on the bus, average bus voltage here. But it tends to get very close to nominal voltage at one pu. And then particle swarm optimization is the blue line here. It also does a reasonable job of pulling the voltages close to that nominal value. And so we wanted to come up with some sort of metric or scoring for these different simulations. And we came up with this or just integrating over time, a scoring for a baseline test minus the voltage regulation case. And so if you, on the baseline case, moved from that point down toward nominal, you would score better. And so volt bar had an improvement in terms of this metric of about 13%. And the extreme and seeking control and particle swarm optimization did much better, like that 75% improvement. So not bad. Now we'll, now we'll go to the national grid simulations. This one turned out to be much more difficult. So we wanted to simulate just using the three phase inverters that old up 10 road, which we had access to for running the field demonstrations. In this case, on average, the voltage of this system was very close to nominal. But there is a phase imbalance. And so phase A was pretty close to nominal, but B was high and C was low, as you can see in the right plot here. And so when we ran each of the use cases, everything pretty much stayed more or less the same, right? It couldn't improve the voltage on this particular feeder, even though it was imbalanced because when you change the power factor for these devices, it changes the power factor on each phase. And so if you move it up or down, it actually gets worse overall. And so we ran into a situation that we couldn't solve for that. And so, but we wondered, well, okay, there's also some single phase devices on this in an event that we could control all these inverters. How would that improve things? And so we ran that simulation as well. And what we see compared to the top, which is just controlling the large old Upton road site compared to all the inverters, we actually do see quite a bit of improvement because each of the single phase inverters is able to pull that phase toward nominal. And so you get an improvement, in this case, Extremely Seeking Control does quite well of 15%, 30%, 17% for the different methods. Okay, so based on all that, we had some confidence that our controllers were working and we were ready to go to the actual field demonstration. And so for this, we were communicating to the old Upton road PB site that National Grid owns. We had another point on the feeder shown in the lower left that we had a feeder monitor there. And we could pull currents and voltages off of that measurement point. PB is hard to see but you can see my cursor. And so we went ahead and we ran each of our test cases. In the case of the particle swarm optimization, however, we had one major hurdle to get across. And that was that we didn't have enough field data to complete the state estimation. And so what we did was, I think a little bit clever technique where we use a digital twin. We ran the power hardware in the loop, well, the real time simulation at the same time that we were running the field demonstration. And so to populate the state estimation. What we did was we basically pulled the measurements from the simulated PMUs in the real time in the real time simulation as opposed to them from the field. And we sent that to the state estimator, which was then able to solve and then we could calculate the optimal power factor set points and then issue that to the end devices in the field. Of course, there's a lot of challenges associated with that. We're not really capturing live load data. We do know the irradiance at that PV site based on measurements. And so we can use a curtailment function to send the curtailment to the simulated inverters. And so we're getting pretty close in terms of active power contribution from each of the VR devices in the simulation. But, you know, we also don't know really any of the settings for the actual field in terms of, you know, tap changing transformer positions or voltage regulation equipment if it's on or off, capping, etc. So it's kind of a rough estimate, but it seemed to work pretty well for us. And so here is an example run for that. I'll just, you know, there's a lot going on here, so I'll just kind of point out a couple of things. The first thing is in the image in the upper left here, you can see active power for the digital twin and the field inverter. So they match quite well. So we're monitoring the inverter in the field and then setting a curtailment level in the simulated device to match that. For comparison of possible voltages at the point of common coupling for that, you know, a little bit off this could be partly due to the other voltage regulation equipment that we're not accounting for don't have visibility into. But overall the trends match reasonably well. I would argue pretty well there. And then so for the optimization, as we saw before, there's not a whole lot that it can do. And so the voltages in average are shown here, where the feeder monitor in the blue and purple increasing, and then the point of common coupling and red, the active power is the yellow line and the forecast. Based on our forecast is the green line, which is persistence longer term like more of an energy forecast than an actual power forecast. And then the reactive power contribution is quite relatively small. It's shown here in the orange. So it seems to work reasonably well. I won't say it's perfect, but for for in order to run this optimization, it was a good technique. And so we here are the results for a baseline the bull bar control or bull bar curve and extremity control control and particle swarm optimization. It's essentially impossible to compare these because there's going to be different set points for voltage regulation equipment on the feeder. There's also obviously different radiance profiles. The PSO run has high variability, or the baseline was a nice clear day. And so it more or less just shows that in all cases were relatively close to nominal voltage. So the red lines here close to unity or sorry, one P you. And you can see in the extremity control case that that reactive power ripple and its influence at the point of common coupling there. So that, you know, if the system is sensitive to that, that's not too good. But in the case of the particle swarm optimization, again, operating relatively close to nominal and just just sitting. So in conclusion, we demonstrated that this incremental approach moving from pure simulation to real time simulation to the hardware and the loop environment and then a field demonstration was a good approach. The communications between each of these elements is a major hurdle and takes quite a long time to debug. So we could verify that through this process and it built our confidence that it would work in the in the live demonstration. The digital twin was necessary to get over the limitations of our state estimation system because it was too sparse without those simulated PMUs and volt bar did well we didn't have it too aggressive but it did improve voltage regulation, especially on the PNM system stream control is a viable means to do control over fleets of devices. We tended to bundle these into a couple of different frequency groups for the demonstrations on the larger system on the national grid system. But, you know, that's in the details. The state estimation based particles form optimization technique was was pretty good. It worked with sufficient telemetry. It's of course the hardest to get operating because you have to have so much information and be solving these state estimations, essentially in real time or near to it. It took us about I think, you know, one minute to close all the loop in terms of that cycle. And I guess in terms of open questions and observations. We're very interested to know if the old up and road PV system could have solved this problem if it had the ability to do negative and zero sequence current injection. There's some work at the Austrian Institute of Technology developing an inverter that can do that. So you can inject different reactive power levels per phase. And so that would solve that issue there. And so I think there's quite a bit of research that could be conducted in that space to solve that issue. Communications continue to be an issue in terms of getting data from both field systems. And then just like moving them between these different elements of this ADMS system, you know, larger tool is quite a pain in constructing all the PMU traffic between the components, extensive period of time. And then, yeah, I think that kind of touches on the software interoperability issue and how we're, we're still struggling with that and a lot of what we're doing. So that, that's it for me. I'll open it up to, I guess, questions at this point. Thank you, Jay. That's great and very relevant to a lot of the work we have going on and then Mexico Smart Grid Center. I encourage all attendees to type your questions into the question and answer box. And we have a couple of follow up questions for you Jay, just now. One is if you think you'll pursue the digital twin strategy again. That's a new technique that you'll use. Well, I would be very open to it if a project came along that would need that technique. Right now I don't have any work pursuing this line of research, unfortunately, but I think it, I think it can solve a lot of those issues. And so if you have, if you want to do global optimization on feeders or larger systems that you don't have full visibility into. If you have a real time simulation that could be run in parallel to pull kind of pseudo measurements off of. I think that's an absolutely valid approach. I mean, you're not going to do any worse than just guessing right so it's going to be going to be as close to as accurate as you can get, given that you don't have a lot of data. Thanks. Yeah, and another question was about coordinating across so many different entities for this project. It looks like you had a ton of collaborators and utility partners and multiple sites and if you had any lessons learned for how to best coordinate and merge models together and something that we'll be looking at in the deployment component of this project also. Yeah, I mean that's always a challenge right is, you have to balance things the way that we structured program management for this project as we had fortnightly group meetings with everyone to just bring people up to speed. But oftentimes we'd have, you know, daily or maybe even weekly calls between people running the programming side of things and trying to connect components together and solve things and so you have to analyze where all the touchpoints are and address those things independently of these large group calls it just, you can't do activity level work if you're having everybody on the call effectively. Yeah. Yeah, good advice for our team. This is for both of you and it's not research related but something that our folks are curious about. Do you host interns or work with undergrad or graduate students and what are the best ways to pursue working with Sandia and Lannell if you are a student. Yeah, we have student postings all the time on our website. I don't know the exact thing but sandia.gov will. I'm sure there's a link there for employment opportunities and you can take a look at our postings. I do know that our group in renewables and distributed integration research area is actively trying to find summer interns if you're interested, please take a look for our postings and send a resume. Okay, and then I'll kick that to art also at Lannell and here. What are the best ways for students to reach out to you or Lannell to work on these types of projects. And I believe you're on mute right now. Okay, so yes, a lot of also has open student postings. Although we're also always open to academic collaborations. One of the big things within our group at Lannell is we have a focus on trying to make sure that our work is open source and you know other people can use it really. If you look for Lannell and see on GitHub, you can see what you can look at and I encourage you to go and take a look there. Okay, thank you and a follow up to that when what are the the key skills you're looking at for students who you want to come work with you. What, what kind of background or skills do you want them to have. Yeah, start with you. Okay, sure. So, at least in the power and energy sector, people familiar with, you know, obviously a traditional power systems background is very helpful. Also, people coming from computer science and operations research. It's also quite useful because much of our power and energy work has been to be optimization focused. Also, we are branching into machine learning that that's definitely something that allows them putting quite a bit of work into. We have a lot of research topics as outage forecasting, being able to infer electrical networks based off of geospatial predictor variables. You know, these are all major areas that we're working in right now. Do you have anything to add to that for what you look for for students at Sandia. Yeah, well there's a lot of things that we're looking for but I'd say kind of the core themes and a lot of the postings revolve around strong analytical skills, programming expertise and Python or Matlab or others. A lot of work we're doing now on power systems of course, communications and cybersecurity, looking at analytical models to solve resiliency questions. We've got a lot of work just, you know, large data sets, machine learning, how do you crunch numbers, how do you manage databases, things like that. And particularly for me, I like candidates who have some lab experience. So if you've worked with hardware in the past and know how to connect physical pieces of equipment together to make them run, that's certainly a strong positive for a candidate. Great, thank you. And one final question for Jay and I think then we'll wrap up our webinar for today. You mentioned that your funding ended for the project that you overviewed. Are you going to continue threads of this research or what are the next research questions that you're going to pursue based on this work? Yeah, call up your state representatives and have them send more DOE funds our way. We are, yes, we have a large suite of research programs in this area. This actually was just one project and a much larger program of doing protection and voltage regulation at the distribution level. Matt Reno and Robert Broderick here continue to do a lot of work in that space. And we're doing a lot of hardware in the loop research now branching into wind systems and other hopefully electric vehicle chargers. And I think, yes, these core questions of voltage regulation on distribution systems come up all the time as we start changing the mix of generation on our power grids. And so there's, yeah, there's quite a bit of work, both just the modeling side and also at the power hardware in the loop level. Okay, thank you. Well, I'd like to thank both of you. Art Barnes and Jay Johnson, a tremendous presentations today. Thank you for your time and sharing with us. And that concludes our webinar today. Thanks all. Thanks. Bye.