 Okay, so my name is Valerio De Angelis. I'm originally from Italy, but I've been in the US since 94. So a little bit of my history and how I got here. So after my PhD, I started a software company called Mindflash Technologies, which was then sold. So I spent 10 years doing software as a service systems. Then after that, I moved on to the City College of New York, New York City, where I started the Energy Storage Company, where basically we made our own batteries, developed from the university work, and then my job was to go from the battery to the system. And that's where I kind of understood that there are so many layers, and all these layers are built by different people in a way which is just not compatible. And so, and I think that's one of the reasons the industry is moving so slowly. It's simply the lack of standardization and also how companies still see their technology and their software base as a competitive advantage, and they're just not willing to share. So I'm now with Sandia National Labs. I'm lucky to manage a fairly large group of really smart, talented people. So whatever work you're gonna see here has been mostly done by folks in my power electronics group, energy storage, and software. So there's a work of a lot of people and what we are trying to do is to slowly bring up software and hardware layers up that we can open source. So not only they need to work, but they need to be robust enough that people can actually use them to further development or in applications. So for this audience, I'm just gonna have just a few slides to introduce what is the problem that we are trying to solve as an industry. So as we, traditionally, electricity has been produced by power plants that either burn natural gas or nuclear fuel or coal, some diesel plants. But over the last, I would say 10 to 15 years, the price of solar panels and wind has dropped to the point that we are seeing more and more photovoltaic. So that's happening like in area like Southern California, but everywhere, throughout the country, Texas is a lot. And in addition to that, we are seeing growth of electric vehicle in the same residential areas in which solar is put. So this starting to create some imbalance on the electric grid where during peak hours, which are typically when the sun starts going down, people go home, charge their cars, solar doesn't produce, air conditioning comes on, and that's where in some distribution network, the grid, it's starting to suffer from some imbalances. The duck curve, which when people first talked about it 10 years ago seemed impossible, but it's happening. So, and one of the problems is that when the first solar systems were connected to the grid, they were connected using grid following technology. So, to the point that if the grid is down, your solar farm in your house won't produce any electricity unless you have your own battery, right? So, but this is not sustainable because if more and more of the generators disappear and more and more of the resources are grid following, who is he gonna set the frequency and keep the grid stable? That's making now power electronics and power converters as a more important resource. And the idea is like a lot of these converters will start appearing from on the electric grid, but they will all be made by different vendors. They will all be made at different times and they will all need to depend on software. And that's where one has to stop and think. A transformer is this massive machine of windings and oil and magnetic cores, which is immutable. And then when is the last time any of you are sold a PC or a Mac for more than 10 years, right? But how are we gonna put an asset on the grid that depends on software and last 30 years? I mean, that's kind of, that's a question. So, and you know, just to continue on the introduction on what power converters are, typically, when you look at one of the system, you start with your solar or a battery. And solar or a battery, they provide, you know, like current, which is kind of like DC. And you have to find way through IGBTs and other power electronics, silicon-based components to convert it to AC, which looks like the figure at the top. So, three phases, shifted by 120 degrees. So, that's all power electronics does. Takes AC, converts it to DC, and back. And now, as the electric grid changes, in addition to have the grid following inverters, which basically they will sense a frequency and give you what they can. You know, like the solar might give you 100 kilowatt one day or 50 another day, right? As long as it can sync with the frequency, the solar plan will give you what it can. You have to switch to a situation in which the grid will get formed by the inverter. That's a much harder problem. Because the inverter will have to sense what is the load that it needs to serve and provide enough pulse power to maintain that 60 hertz frequency and that voltage. Right, so that's where things become harder because the grid-forming inverters have to be able to communicate with one another. So, they don't just shut down each other. And, yep, and at the very core, there are two ingredients that will make all these systems. One is lithium ion batteries, and the other one is these IGBTs, these tiny little MOSFETs. So, the idea is, okay, if we have these very simple building blocks, batteries and MOSFETs, why is this so hard? The problem starts when you start looking at the complexity of controlling these devices. So, this is IGBT. You can buy them, just they cost a few dollars. Now, they make some that will work up to 100 amps, and this is the simplest possible circuit that one can make. So, the way the circuit is operated, you can think of it as a pump. You have a certain voltage on the battery side, and you're gonna either build the voltage up on the other side, or you're gonna make more current flow, right? So, in order to do that, you activate switches. So, in this particular case, you need two MOSFETs. So, you turn one on, and then you turn it off, because you turn the other one on, and you keep going. A device might have thousands of these. If you turn one on before you've turned the other one off, you're gonna short the circuits and burn the MOSFET. So, that's how easy is to destroy these devices, right? So, that's one of the reasons why writing software to control power converters is hard. There are hundreds of these, sometimes thousands of these. There may be thousands of offsets, one wrong sign, the whole thing is gone. It's kinda, you know, and they're gone with a little puff of fire, actually, and that's it. Batteries are kinda like the same kind of deal. So, this is a 1816-50 kind of battery. This is the initial Tesla Roadster mounted these batteries. The initial Tesla Powerwall was using this battery, like literally this small. And internally, they have anode and a cathode without getting into the chemistry of it. Lithium ions will go into this structure in and out, and then they will bind on the other side with a very complex metal. But lots of things happen over time. The carbon structure cracks. The cathode also cracks. They get detached from the current collectors. So, when you look at one of these batteries, you see that they will produce less and less energy for you as they age. So, not only some of them will break in unpredictable ways, but some of them will age faster than others, right? So, and imagine you have a hundred thousands of them in an installation, how are you gonna manage them? So, this is what makes management of battery systems difficult. What we are trying to do at Sandia with the help of a lot of people in hopefully the industry, if any of you has some spare time to work on opens or soft, is to bring systems up that can be used as grid forming inverters and depend on batteries to do that. So, in order to do that, we have to build electrically mechanical systems. We have to have control software, and then we have to have data-driven modeling and analysis tools, so that if we have hundreds of these devices, thousands of millions, however many we're gonna need, somehow they can interact with each other in a standardized way. And so, this is what we have been doing for the last few years. Some of these ideas started in 2018, some of the work, but over the last four or five years, say. Traditionally, what you do, you connect the battery to the grid using a single power converter. So, this gives very little flexibility to the grid. If one battery fails on this side, the entire system is down, you have to keep the voltage stable so you can operate this piece of power electronics. We switched to a different topology, which basically has some intermediate steps between the battery pack and when you go to alternating current. So, we are developing a cascade of DC to DC converters, which basically allow to take any battery on this side and build a common voltage on that side. So, this basically allows to kind of plug and play. So, vendor A as a battery, we connect it to one of these devices. As long as we know how to charge that battery, that device can become part of this network. 10 years from now, this battery dies. We get another one, reprogram this converter and the battery can be part of the network. So, this is a way to kind of make system more scalable and standard test of time a little bit. And this is the first version of the system we built. Again, we have to build the mechanical components. They have to be robust enough so that our mechanical components don't catch on fire. We don't wanna be those people. Then we need to have a battery management system. For now, we are using off-the-shelf BMS, but we are transitioning to open architecture BMS, starting from a reference design from a company called Fox BMS. There is Fox BMS is an open source BMS for automotive industry that we're gonna repurpose it. And then, we have two different models of the DC to DC converters. This is the block. So, basically you can see in one box, we have the BMS, the converter, and the batteries. And so that now allows to basically decouple the problem somehow. And as long as we release the designs of the DC to DC converters, people can take the converter, they can take the embedded software, and they can run systems. So, that's the idea. When you look at the software a little bit more, the green is kind of like the core of the software itself, is energy management system. Decides when to charge and discharge the battery to meet whatever demands of the application. And then there is what's called battery management system. Battery management system job is to make sure batteries stay within safe boundaries, otherwise they may either catch on fire or be destroyed. And we are also starting to work on supporting the communication protocols that all these devices, all these external devices will use. So, we have libraries for serial, Modbus and Canbus communication. And we are trying to write software in such a way, and this is where the lack of standard really slows us down. We are trying to write the software in such a way in which we have configuration files in which in principle, you could load the characteristic of a device, whether it is a BMS or a certain type of battery, and then from there drive the software. And we have been working with IEEE and other standard entities to kind of come up with a consensus on at least what the format has to be. And we are slowly trying to understand what are the cybersecurity risk that such a system will present because a lot of these protocols like Modbus and Canbus were designed before the internet was even a thing. So they are not inherently secure. And then the other problem that we're gonna find, and we're trying to figure out how to work around, is that currently at the fire department or under spreading laboratory will not certify just the hardware or just the software. They will certify the hardware and software together. So how do you have a configurable system and get it certified? And that's another open issue. But the good news is the system works. So this is some data that we generated on the left, this is a lithium ion battery pack for lithium ion cells. On the right is six lead acid cells and we were able to, we can operate them freely. We can discharge charge. We have correct charging protocols for both of them. And now we are expanding from this initial system that totally had four battery packs to one that is a complete rack that will run at the same voltage of the batteries that you guys installed in your campus, which is a thousand volt DC. So as we do that, we find that cells quickly start to deviate. So like the top shows like the voltages of the four lithium ion cells and the bottom shows the voltage of the six lead acid cells. You can see that they don't track each other. So over time the cells will deviate even more and that will impact how much energy you can actually take out of the system. And that's why we are developing databases of battery data. So we, this is probably, no definitely, this is the largest database of public lithium ion cell data. It started as a small project, COVID project, where I wrote the code and one of my colleagues at the data. And now we're getting data from the major universities, national labs in the US and internationally. And I can tell you that all the energy storage companies, EVIS companies and a fairly well known rocket company, they all get data from there, which tells you how messed up was the industry, right? If some guy and two, three of us can put together a database just like a few thousand lithium ion cells and we become the most used repository of lithium ion data, it tells you how much work we need to do as an industry. But, so batteryarchive.org is available and basically allows people to browse the data. Look at some plot that makes sense in the battery world and then download the data. And the underneath batteryarchive uses open source software. So we started with a tool called Redash, which was then was acquired by Databricks. And I was just talking to the Databricks folks, they're not gonna maintain Redash anymore, which is kinda like one of the risks I guess of using open source software that these companies then stop supporting it. But the idea is you take data from big machines that you use to test batteries, you convert them to some standard format, put them in a database and once it's there, you can either visualize it or you can query it. Then at that point the data becomes available to any other program or any other data hub. So, and the project is available on GitHub. Let's see if I can get this done. Let me see, let me try. If I can find my mouse. So if you go to GitHub, there. You would find multiple repository, the agent is the one that contains the database definition, the queries and contains everything that is battery specific. The service contains the SQL alchemy part and all the plumbing necessary for the JSON API. And the battery archive Redash is a fork that we made of Redash. So we are starting to customize Redash to make it a little bit more friendly for people to use. And then for those who are interested, there are also tutorial files that we used at the talk we gave last year. So one of the things we did to Redash, for example, was to, I don't know, as any of you use Redash, you know what that is? Okay, so Redash is a dashboarding tool. It allows you to make database connections easily to most of the MySQL, Postgres, Synflux, whatever. And then you can import the data query, automatically generate graphs, and then put those graph into dashboards. So, but one of the things they did, they wouldn't allow in a table to select multiple cells or the filters weren't linked. So we made some changes to the Redash interface. Let's see if I can get this done. Ah, so to give you an idea of how all this works, for example, you can select a couple of cells and then instantly kind of look at some of the battery data. And this was beautiful, right? Because this is what the open source did for us, right? We would not have been able to build all this machinery if it wasn't for the open source itself. So thanks to the open source community and too bad data brick is not gonna maintain it. So, and then again, after the data is all in the same systems, then is all in the same format, you can use an API to kind of extract the data and do more work on it, right? So like for example, one of the things you can do is you can pull the data into a Jupyter notebook interface. We give some basic functions to query the API. You can get the metadata, which means basically is whatever characteristic the cell has and you can get the time series data. And then once is in Jupyter notebook, then you can basically do whatever you like with it. So you can plot it and you can do further analysis, right? So now for developers of machine learning programs and modeling, it means that we did all the work of taking all the battery data and putting it in one system. And now the only, what they can do, they can take this data and just analyze it, you know, build models, solve problems, look at degradations and just bring the industry forward a little bit more. Another thing that we are trying to do, we are trying to promote groups that are writing models of batteries. So there is a group in the UK that has introduced a package called PiBam. PiBam allows you to, by solving differential equations to model the behavior of a battery. So now that you have the data, you can directly compare battery data with results from the model and you can use this information to start seeing where are the degradation coming from. So for example, in this particular case, you can see that from the model that this particular cell mostly gets degraded during charging, right? So this is information that you wouldn't be able to extract quickly if you couldn't quickly compare the experimental data with the model and use the intelligence in the model to then figure out why certain degradations are happening. And I can actually, it's the same idea, right? You can put the data, PiBam is itself open source software and you can set up the experiment, run the solver and generate the data, right? And once the data is there, again, you can compare it with the battery data, right? And so we are having more and more researchers now coming to battery archive and taking the data just so that they can validate their own models. And this brings me to the last few slides. We are also building, sorry, we are also building a toolkit that allows now people to directly use machine learning methodologies with the battery data. One of the biggest problem of using machine learning methodology with scientific data is the difficulty of cleaning up the data for analysis. The data, you know, some data needs to be discarded, some other quantities need to be calculated. So we are releasing a toolkit that allows people to calculate quantities, like for example, in the battery industry, we use this delta Q, just some battery stats that people use to understand battery behavior. We are gonna give people libraries to calculate those. So they can then use things like ElasticNet, which are part of standard machine learning toolkit like Scikit-learn or TensorFlow. And then from there, find out ways to estimate battery life using information that they can collect early in the history of the battery. So, like, remember, like in one of the previous slides, I was saying, well, different batteries will age differently. What if you had the ability to understand how those batteries will age in the first month of operation instead of having to wait years? That's where some of these machine learning methodologies come handy, but the difficulty of using them is that it's just hard to collect the raw experimental data and to clean it up for machine learning. So this is part of the work we are doing. And we are about to release a toolkit that people will be able to use to clean data and then generate this kind of machine learning results. So as you see, it's a lot, right? I mean, like, to go through it is a lot. You need the hardware, you need the software, you need the mechanical systems, you need the ability to collect data, and you need the ability to analyze this data, right? So, and you also need a lot of buying from other entities. So we've recently launched an initiative called the principle of battery data genome in collaborations with other five national labs and a bunch of universities where we are trying to at least standardize the API that battery archive and all the other tools available in national labs and universities will use. Because we think that if we can standardize at least the data communication, that will make it that much easier for them different groups to build tools that can work with one another. So yeah, and this was my talk. So any questions? Go ahead. Yeah, so, you know, that's where things, that's where some of the complexities come, right? So what we, the way we're structuring it is we are trying to add as much protection as we can in the power converters, right? So the power converters will operate with embedded software which people will have very few reasons to change. We're gonna give that code, but the code is designed to, for example, maintain all the IGPTs into safe conditions, prevent overlaps, right? It's gonna be designed in such a way that it will have some hard voltage limits which will open contactors. But then we still have to have enough flexibility for people to define what their charge and discharge protocols are gonna be, right? Because different batteries will charge in slightly different voltage ranges and you need to allow for that. So, I will say the hard limits in the converters will ensure safety, but it will still be possible for users by changing some of the soft limit in the configuration files to damage their batteries. Yeah, that is just something that I think if you wanna give the flexibility to the user, I don't think you can make a system that is 100% bulletproof. But at the same time, if you don't provide them something to use, different companies will make their own systems and some of them will explore, right? Because they just wanna have the skills to do it, right? So, there was this very sad story. There was this boat outside the coast of Catalina of one of the islands in the Channel Island in Southern California where they plug the camera in, digital camera to charge the battery pack. The battery pack went in overcharge and the entire boat caught on fire. So, we are trying to avoid situations like that by putting out there something that is guaranteed to work. Yeah, but that is the issue. That's why it's taking so long, right? So, the question is how do we... Actually, I think that... And that's where we need to work with the open source community because I feel one of the strengths of open source is that the community, through community usage, there is a certain level of testing and safety that comes from there, right? So, I will feel a lot more confident on my lithium ion-charging protocols if 100 people reported bugs rather than if I just developed one and used it, right? So, there are pro-incons and in fact, I was talking to some folks from the work on cybersecurity for the banking industry and governance of open source software is one of the issues we need to look into it. I don't know how to do it, but what I do know is that if we don't standardize some of these things, the situation will be even worse, right? So, and that's why we hope to work with folks like the Linux Foundation Energy to kind of understand how to build governance for these kind of projects, yeah, but that's the issue, right? Yeah, yeah, so... So, what we're trying to do is, by the end of this year, we're gonna finish the design of this actual system and there are several universities that want it. So, we're gonna send it to Texas Tech and other universities where they're gonna use it to develop their own predictive models. So, instead of them having to build the entire system, they will simply improve the models that operate on the system itself. And we are open to send this toolkit to battery vendors, you know, software developers, whoever, because at the end, you know, that's how we're gonna learn how to make it more useful and grow it into the industry. Yeah, so, initially, there has been some resistance in the idea of having open source as part of the portfolio of activities, but over time, especially as the battery archive site has grown and as a couple of groups started to use the open source software, now we have a pathway. So, it's more like at the end, we love to... I think there will be some process that love to be followed once we start sending these things out, but at the end of it, you know, we are a national love. So, if we don't do it, who is it gonna do it, you know? Universities are not gonna do it, vendors are not gonna do it. So, this seems like the kind of stuff the national loves have to do. Let's see, any other question? Yeah, yeah, definitely help, there are a few things. One will be help with the governance of it. You know, we wanna make sure that we put this on a solid foundation because as we put it on a solid foundation, then we love to get utilities and regulators and other folks to accept it. And yeah, that I would say is the number one part. And the other part which I think is a challenge for the community, and I don't know, maybe it's not, but how do you open source and maintain things that have both a software component and a hardware component? You know, because we saw it's hard enough to version control software. How do you version control software and hardware? So, we'll definitely wanna know how to do things like that. Yeah, yeah, so then we should, yeah, we should talk, yeah. Yeah, yeah, okay, well, yeah. Well, if there are no other questions, thanks, thank you everybody. And yeah.