 That's known everyone and Welcome to the spring quarter for the smart group seminar. So in the in the past few years for smart group seminar we always invite jazz speakers to summer and the new parts of different topics and So this quarter we are going to do slightly different and some of our researchers and even senior PhD students to talk about the awesome work they have done and they have been doing here and One main purpose of this type of presentation is to really try to unlock So open source open innovation we have done at the different parts of Stanford related to any system So in the next Two months we're going to appear five to six different products open source products produced Through Stanford research and the students in the past four or five years. So Yeah, I started with the first Product we're going to first product. I would see this way not just the one hits at least two they're going to talk about from Jack Dean Shenner and Jack was the first fellow graduate student fellowship awarding for the bits and months and Working with a professor Sally Benson and and Peter in the last four or five years and Love interesting work and both globally and also hands on the Stanford energy system so Now he is an interprofessor at Stanford and also work at the power managers So the topic today for jazz talk are going to share with us Two interesting work one is on the sample energy system. Another one is It's a platform and he created a call the real-time emission tracking system So without further ado, I will handle this over to Jeff so thanks, so actually Actually, I'm going to try to focus on one of the two products and see how you know how long that takes me and You know, I did by a time they go a little bit more about the second And I guess for so for so I gave a talk on Monday on Summer stuff around the campus energy system. So I won't talk too much about that and I'll talk mostly about this work on tracking emissions in the US electricity system which started Actually, when we were thinking about the campus energy system, so it's all very very very linked and so I don't know how you usually do Questions in this seminar, but I guess I feel free to direct me to the crime I think you know questions are too long. Maybe I'll keep that for the end. Otherwise, I'm going to begin And actually one last thing before I start And I guess to all the students whether in the room or online I'm always looking for students, you know, so if anything here is interesting to you, please Please contact me as much as you research. We're looking for people And so okay, so today Okay, so kind of the starting point for some of this or I guess the content which I'll talk a tiny bit at the beginning, but it's really sort of motivation for this is Sanford campus, which is roughly equivalent to a 30,000 people city And in California, the reason I say 30,000 people I'm looking at the campus is total electricity consumption, which is the really one of the things I care about When I look at Stanford campus and and it's a it's a great place it's a great place to do research and experiments whether real life or thought experiments because Well, the Stanford system has it's sort of an island in the sense that there's one Distribution system one substation for the campus to transformers off of it One that goes to the central energy plant or to produce the heating and cooling for the campus One that goes to the hundred plus buildings That that are on the campus. It's also an island in the sense that there's Central management, you know, we have this almost Totalitarian regime, you know with one president making decisions for everything that happens on campus and that means you can do You know, you can do experiments you test things out in a way that might not be so easy In a different project system, you know, so that's one of the things I like about The Stanford system and this is my sort of, you know, my rough schematic for what's in on this campus on the bottom middle there of campus buildings So depending on how you count 100 150 of those and the bottom right the central energy plant, which is almost fully electric There was a major overhaul starting 2014 in 2015 To go from what was before which was a co-generation system using gas to an almost fully electric system using heat pumps big industrial heat pumps to produce the heating and cooling that sent through pipes Ten miles of pipes around the campus buildings and then to the left on the bottom there. We also have these different these different systems that are either consuming electricity that are either active Actively they managed a or for that we're thinking of actively in the near future and then from the outside we get gas we get electricity we get water The one I'm going to think, you know, really think about today is the power and one of the key Associated with the power of coming over that line What are the to try to get access for? What are the operating emissions if you like of Stanford's electricity? It's a why why does this matter and this is sort of the base We started with the thing about this which is How do you schedule, you know, if I had to go in every day and schedule The electricity going to the machines and the central energy plant and so there are different types of machines Some to cruise just cooling some just heating some we use both the heating and the cooling But apart from the power generators that run mostly on gas But the tiny bit of electricity for the fans the other types of machines are all heat pumps Different types of industrial heat pumps. So almost only electricity is the key word, you know That building uses up to 15 megawatts of power. So Up to 45 percent of the campus's power consumption is from the CEO the central energy plant and And I think it's roughly 20 to 25 percent depending on the year of our energy in terms of electricity is a big part of our of our electricity use and it's fairly flexible because there's There's centralized thermal storage in that building that can store for a couple hours today depending on the season they work though Cooling water and cold water. So this big problem is how do I schedule? How do I schedule the different machines inside that building? subject to a set of Different strengths and what I'm trying to minimize by default is the campus is total electricity which to To give you maybe a sense of scale is something like 20 25 million dollars per year So, you know, so, you know, if you do 100% on that electricity bill, you know, it's I guess bigger savings than that at least Michael and So, you know, so it's worth it in other words is what I'm trying to say and that that's something that That the people who manage the central energy plant are reviewing, you know, they're already looking at how can we optimize the operations One of the questions I'm going to be asking here is well, what if we wanted that's that system to be used for other things And the sorts of things that, you know, they might want, you know, we might want to use it for our things like reducing costs So that I've talked a bit about Providing services to grid that I talked a bit about on Monday and I won't talk about again today Reducing carbon footprints, which is really what I'll talk about Today and we're really thinking about what's the carbon footprint not so much producing the carbon footprint Is what I'll talk about today and then things like so integrating electric vehicles There's also some research doing that on campus and I've done a bit of that as well and and things like adapting to heat and also something I've worked on that I'm actively working on but I won't talk about today just to give you a sense of Different things you can think about with that system And so so the what I really want to focus on today and these are some results kind of at the end To give you a sense of where this is headed One of the so one of the things we did back in so this was in 2018 2019 What 2018 2019 was to think well, what if Stanford were to self-impose the carbon price? And let's say we have a proxy for what the carbon content of the electricity we're purchasing is How much difference would it make to change the operations at the central energy facility? To follow that carbon price and so what the picture on the left there is sign you there's sort of two Two ways you reduce emissions one is well now that samford is almost fully electric Just if the grid becomes cleaner. Well, naturally, you know, sort of automatically when you look at the carbon emissions associated with Samford's operations that just goes down because The issues you were getting is cleaner and the second is This idea of carbon we're scheduling if we start moving and there the idea is really if we start moving towards a grid in california where There's a strong difference in the carbon intensity of the grid between the daytime and the nighttime Then it will have value to schedule operations in that way in other words to make You know to make heating cooling while the sun shines and and so the picture on the on the right was A sort of first attempt back then and saying what's the sort of for california What's the carbon intensity of the grid? So I won't talk about this picture too much because I'll I'll show many more pictures like these in a bit But and so so this was kind of some of the motivation here And so one more slide on this to give you a sense of you know, why why I care about the carbon intensity of the grid And so on these so these pots are basically showing you a summary for what's the total electricity consumption of the campus Over a full year. So this is and this is by solving So this is the schedule that's generated by my optimization programs And this is a sort of a perfect foresight kind of solution to What you would do in the year of 2016 where each Each column there is a day of the year in 2016 and each row is an hour of the day And so and the left is basically so we you that's business as usual what you know, what what if you're just paying for electricity the way we pay for like just today what you should do and So without going into too much detail Basically what the current structure the current tariff structure is telling us is try to be as flat as possible so when you look at The electricity consumption from those seeds of the central energy facility Basically try to do it as much as possible when the other buildings are continuing less power So that the aggregate is as flat as possible and then when the price of electricity is higher Which is in the afternoon that's the sort of blue you see around 6 p.m The bottom try to consume a bit less And whereas the the right plot is okay if I really you know, I'm just tracking Carbon and say I was trying to minimize my footprint. What would I do? Well, you know, and this is in a scenario where we have Assuming carbon it has a lot more solar. This is sort of you know, looking at it Perspectively, um, well try to consume as much just see in the middle of the day Another way to interpret the picture on the right, you know, I think is basically If Sam if Samford had onsite solar Well, we do have onsite solar But if we had much more onsite solar how much of that solar would we be able to just absorb ourselves in the middle of the day? And so all that was kind of motivation to get to what I'm really hoping talking about today I have roughly And is uh, okay If I want to be able to do this kind of thing if I want to be able to do carbon anywhere scheduling I need to have some idea of what's the carbon intensity of Of the of the electricity that I purchased So in other words, I care about what's the carbon footprint of the electricity assumption And then you know, actually when you start thinking so these are sort of more questions that come up when you start thinking about carbon intensity of the grid Things like what's going to be the impact of renewables on grid carbon intensity. So how does that change? But also, you know, as we started digging more and more into this work Questions started coming up like what's the impact of renewables on other Generating units and on these changes that this Something I'm still working on very much today. And then of course, how are these answers changing in time even just in the Especially it's a bit more than four or five years. I've actually been in San Francisco for seven and a half years now And you know, even in that Short, you know, shortage times go now it's starting to be a bit longer. The answers have been changing quite a lot You know, and and so having tools that can accurately track What's happening in the grid is going to be super important, you know And having relatively Fresh information is also going to be super important. And so the map on the top right that's sort of a teaser for what's coming next You know, we started off with a pair of my samper But you know, and then we started cooling on the grid and then to say actually we want to do this for the us And so that that map and I'll talk more about that map what it means But it's looking at What's the carbon footprint of electricity? in different parts of the united states So Why does this matter? This matters because we want to We want to manage carbon and if we want to measure carbon the managed carbon. We need to be able to measure it This is especially important since We just want to electrify Everything out there right now, you know, it's so so we're heating is heating and cooling is something that I look at a lot Electric vehicles are another big one. That's going to have Big impacts on a power sector. It also means that a bigger part if you look at entities like Stanford a bigger part of their What I call their operating emissions is going to be linked to electricity. So just measuring The carbon content of electricity, which was already important. It's just going to get even more important And this this sort of tool or you know, there's a unit that we care about here Is what we're called emissions factors And so emissions factors measure the carbon content or the carbon emissions associated with electricity consumption or production in units of kilograms of CO2 per One hour of electricity And this is this picture is just to give you a sense of I guess these are numbers that I like to have in my head To get a sense of what's what and so these are life cycle analysis estimates from the FTC and You know the bottom line for me is coal is roughly double is you know, roughly a thousand which is roughly double gas Which is roughly in order of magnitude, you know, so they take out a zero to get dollars And so the other so the other thing that we cared about there was Well, sorry, I mean I'll get that actually in just a bit in order to be able to do the sort of thing We we we want to look at We need sort of two things. We need data on emissions. So a CO2 that's historically been released By varying fossil fuels to produce electricity and then electric grid data. So those are the two sources of information we were using to do this One thing I didn't talk about yet is the difference between consumption and production based accounting and production based accounting of CO2 emissions Is basically looking at what are also called territorial emissions What are the emissions associated with electricity? Production that the idea actually didn't really come from looking at electricity looked at other things, but This work was applying this idea to electricity and then consumption based emissions would be what are the electricity Sorry, the emissions associated with electricity consumption and so I'm currently There you start thinking about imports and the reason this matters is when you are in california, whether it's running a job on the data center or running the machines in the snapper center energy facility You're using electricity, but electricity was put up, you know that grid that you're drawing power from Is being energized by supplying electricity to it But that could be you know in california. Actually, there's a mesh, you know It's like it could be much farther off inside that grid and the the u.s grid is Roughly divided in three parts. They're sort of the west the east And then texas is kind of by itself in the in the bottom there and you know And so really the whole western right the western interconnect Wek is is all connected. So so in quarts matter if you want to think about That's in california So this I kind of went through a bit of between production and consumption based emissions and so I'm going to show a succession of maps that kind of give you a sense of how I'm computing these estimates So this here is showing electricity consumption Uh in the u.s. So the circles there are So the area of the circles are representative of 2016 electricity consumption, which is the data I was working with when I made this And then the arrows here that I've added on top that's that exchanges of electricity So I just see trade and here are the width of the arrows Is representative of how much electricity we're sending. So this is net transfer over the year 2016 and then the So this is what we're trying to get consumption based carbon intensity So the color here is the carbon intensity of the grid and do I have a mouse? Yeah And then and if you look at the scale at the bottom there That's roughly from zero, you know is Less than a hundred is green all the way up to a thousand and so I said coal You know again carbon intensity of coal that would be roughly a thousand here And then gas would be basically the the white, you know, so pjm is roughly cool But uh if they were just gas and then you know the very green like so here, there's a lot of hydro Is uh less than a hundred And so I won't talk too much today about how we made this map that I guess maybe one thing I'll just say is that Somewhere you have to solve a linear system what I will so What I will talk about more is In just a bit is is how do I get it in real time? Um, but before you know, what why does it matter to get this? If you look one thing you could kind of see from the colors, but this picture is showing a bit more is If you look at the u.s. There's a pretty big difference in the carbon intensity of the grid depending on where you are and So in 2016 according to this the u.s. Average was roughly equivalent to gas, you know, so that was kind of like pjm the the balancing area and the Northeast and if you're in a place like libel power On the far left where there's a lot of uh hydro and they also have a lot of wind electricity is very clean if you're in the central Iraqis There's a lot more coal and so the carbon intensity of the electricity is much sure here and so the on each of these the These are summary statistics for hourly data. So the circle is the median and then the length of the line there is the tenth to 90th percentile over hourly data for a year So what so the and i'm going to have one Technical part in in this talking just a second because i wanted to give a sense today of you know, how do you How do you do this? You know beyond just the the great box in the picture is how do you get there? And then i'll i'll get back to something that's a bit to some more results, but um We wanted to be able to get this uh in close real time So the plots i was showing earlier that was so that was data that we published in 2019 But was using data from 2016 because that was actually Sort of the given what we were doing the latest set of data that were available at the time but One thing i said earlier is by the time you're in 2019 actually that data is already a bit outdated So of course it's it's interesting because you know, you can kind of look back As sort of before 2016 all the years were to have data at how are things changing But we also care about getting things that's much more Recent and you know, and and there are a lot of reasons why we we care about that Roughly what you know, this is giving you a sense of how How to get these data. So there's some data on the left there whether it's from the most of it from the it's from the epa of energy Sorry, the What does it use that for again? The Environmental Environmental Protection Agency and EIA is energy information administration. So two u.s um agencies So the epa gives me data on emissions through the sems systems that are live sensors installed on I think it's all power plants above 50 megawatts and then the EIA i'm using to get data on the electric grid production consumption exchanges And e-grid so so these are data sets that come out every couple years that do a lot of consolidation But a pretty aggregate level and so there are two steps and i'll talk about the one on the top one step is When i'm getting data, that's much more recent. So data from an hour ago There are often issues with the data. So I need to find a way to reconcile the data that cleaned up And then two I compute the the consumption based emissions, which I said Involves solving a linear system that's roughly the size of the network Yeah, sorry just a quick question. How do you reconcile um The production from coal power plants with the specific discourse that apply because I mean it's all one big powerful one So how do you know that the electrons that come from my specific core plant are going to the specific disco Is that just based on sort of the The purchase agreements between the two? Oh, yeah, so That's a very good question. Um, let me just share this while I talk and so the What I'm doing here is I'm trying to look at the physical grid. So I'm not looking at power purchase agreements I'm just looking at how much power is injected at different nodes So I'm just looking at the grid as an electric network And saying how much power is injected here? And what's the carbon content associated with that? You know, so if you take like the uh, the midwest Inevent system operator here I'd say what are all the power plants that we're in here How much co2 are they're emitting and how much energy are they putting on that grid? And I also have between the nodes how much I just use exchange And then basically so I'm not so in so this is a Non-market based way of this of attributing emissions if you like here. I'm just looking physically, you know If you're taking power from a grid, you know, you're getting everything better. You can't say In this in this framework. I'm using you can't say I'm taking california from the grid But I'm you know, I'm just buying the wind power That's on the other side here. This is saying whatever comes off that grid you're Well, I'm not really, you know, this is this is more of a technical Work, you know, so I'm not I'm not really thinking about responsibility, but basically, well, you know the way I am Really what I care about is What's the power on that grid? So So that's how I do the matching but actually um, so the first and actually thanks for the plug So the first set of maps I was showing I was doing exactly what we said just there I was looking at all the power plants in a given balancing area How much are they putting onto the grid and saying, okay? That's the electricity generated and the emissions and then I was looking at Exchanges of electricity to compute carbon consumption based carbon intensity Here so so this map is Is live and so how I get to this map is what I wanted to talk about next And in this map, I'm actually so the first one I was looking at what you would call a direct emissions estimate for the consumption based intensity of the Avocacy here, this would be a life cycle analysis based estimate And the difference here is that I'm not looking at individual power plants to make this one Here what I'm looking at is how much power from different generation sources was produced in each In each balancing area and then I have an emissions factor associated with each of the generation technologies to say what are the emissions in that in that time and you If we have time so there are ways to do a bit of both at the same time that we're exploring and so what I want to show this okay, so so so this This what yeah, so so this is kind of the end result what I wanted to get at and this thing So this thing if you update basically even see was updated A couple of hours ago, you know, and so it's an update sort of Basically, it's updated every hour and there's a small timeline But so how I how I get there is what I wanted to talk about next Uh, so I'm going to talk about regions. So so basically what I what I the way I think about the power system here I guess to come back to that the question just before I'm thinking about a network So we're thinking graph if you're if the graph is something that you like thinking about, you know So I have a set of nodes set of regions Then I have a set of generation sources in each region and so some of the data I'm getting from The EIA is I know how much electricity is consumed in each region. I know how much generation There is an each region. I know how much I'm producing from each generation source I know what are the exchanges So total, you know, so like the net exchanges with uh for A region. I know the transfers And then so w, I guess it's just a weight parameter here I'll talk about that later So when I when I'm looking at data on this network, there are certain set of strengths that I know To be true like I know that the total generation at a node should be equal to The generation from each of the different sources, I know that interchange matrix, so if I have a matrix where Index ij is the transfer from Region i to region j. I know that matrix should be antisymmetric And I know that the sum of each row should add up to The total interchange for the corresponding region I know That each node there should be energy conservation. So whatever I'm Consuming in that the service are taking out whatever I'm consuming here plus everything I'm So tr here so the total exchange a convention that's net Imports, you know, so everything I'm importing has to be Uh equal to everything I'm producing that node Sorry, just can you hold it just a second and and so the So so what I said I wanted to reconcile the data So it's getting these data where initially each of these Equations is not necessarily met by the raw data And so what I'm doing so the deltas here are variables that I'm adding on top Data adjustments to say I want to change the data The unclean data to clean them and the so and basically I'm minimizing What's called an l2 norm? I'm minimizing the square or the weighted sum of the squares of these adjustments And let me not go into the the weights maybe here And and I'm doing this so that I have at the end a set of clean data where Where I'm meeting these relationships and the so so and the reason I care about this Is to solve my linear system so I'm already said I've had two steps to compute To get to The first step was I'm getting data I need to clean it and then the second step is I want to To solve my linear system to get the color basically on the max That linear system if the data that are coming in don't need these equations I get into all sorts of numerical Issues and you know basically my code for his blog and it's not happy and you know It crashes if I solve this optimization program to get the data matched then I have something that That solves and and and so I'm happy So to give you a sense for how this is happening and so and so this is happening online. So every Every hour basically I have a server that goes and calls me the data solves this optimization program that now becomes the new reference data for that hour and then Well, then pushes it to the to the web app so you can Visualize it and you know, then the next hour it does it again and when new data comes in so that's one of the nice things here is Is actually in the data that's coming from the eia Basically, they're pulling data from the different balancing areas So and some of the balancing areas give you data with timelines and some of the some of them have holes As data is backfill as you get better data that the other than you know, just like regenerates better estimates So, you know better data just continuously improves, but at least you have a first guess With which to base yourself and so the plots are a bit small, but so this is this is showing you One week's worth of results for November of 2019 for SOCO, which is one of the balancing areas Should have been the middle south of the country And and so what you can see on the top left that's for a total generation. That's right total demand and So basically the parts where you don't have little blue dots those are parts where in the raw data that For you know for that balancing area you have missing data, you know, so you and this is Six days worth of data So there's at least two or three days if you look at the top left there are data that we're missing and so what the I didn't go through all the different steps, but The algorithm first makes the first guess you have to have the first guess to initialize the algorithm based on prior data And that's what's shown in in the small dash line You can kind of see it around November 5th on the top plot of left And then you solve the optimization Program to say how do I adjust the data so that it matches with everything that's around it? And that's the what I have in in green you can kind of see that you know Sometimes you change from the raw data you see don't agree and that means you want to match something else And so these are these different plots are showing you so Demand generation total interchange Like it's maybe just to say a tiny bit of that one. So that's the plot on the top right there And so blue is demand orange is generation green is total interchange and the red is The the energy balance constraints. So that's demand plus total interchange minus generation and we can kind of see Here is it is is once the data has been cleaned The the that's now zero and before it so if you look in the raw data One one reason I chose this period I thought it was interesting is if you look in the raw data over here From a November 2nd to 5th there was data and the energy balance constraint was verified So if you looked at demand plus total interchange plus generation that all added up to zero So good But then if you look at this so, you know, then here there was just missing data No data was supplied by the EIA So, you know, that's something I have to do something about and then over here And if you look at the raw data if you look at this This Red curve down here that was slightly negative. You know, so the constraint was not verified So that's also something that would throw off the linear system I'm going to solve And so these are two different types of issues And basically by the time you've gone through this algorithm to reconcile the data You have data that match of the constraints both for energy balance, but also constraints for interchange and the sum, you know So these are all the neighbors basically of SoCo and if you sum over the neighbors of SoCo and the corrective data It's now going to match the TI line And if you look at generation by source and you sum that up in the corrective data, it's going to match Total generation, et cetera And so I wanted to give you maybe another example of how The algorithm is is doing this to give you a sense of what's happening And this is also sort of a sensitivity check for me. So here I took data from the california iso So that's what SISO stands for here And so I took real data and what I took out some data, you know So I took here 10 ways 10 days work of data on November 10th 1st through 10th And then the plot on the top I just deleted five days to try to see how the algorithm recovers to get a sense for that And so the whole part here was missing you can see the dashed line there is the first guess The orange there is my reconcile data So once I've solved my acquisition program and everything that would be my new data And the blue here is the source source is true. So that's what the actual data was So what this tells you is it's not perfect, but you know, I basically recover after You know, if I if I just take out all that data, I'm the orange is pretty close to the blue, which tells me You know, it's not doing such such a bad job And then the bottom here is I just said again, I'm just going to move the whole thing, you know Just remove 10 days worth of data for demand and what this tells you is well By using data on generation exchanges everything else you can reconstruct the missing one and one of the reasons this is super valuable is Um Well, when you get when you look at the raw data that's coming from the EIA It's never the same balancing area that has missing data You know, it's like one a couple hours here a couple hours there for some channel some other channels in the original work I talked about I did all the data things by hand and it took me three months Um, this one, uh, what was nice is, you know, well now that I've done this it just runs Online and it presses for me. Yeah, what causes all the missing data gaps? They're just like the world or I mean the world isn't like people just being unorganized or it's just like I think sensor's going offline like what? I mean, I guess my My experience is as soon as you start working with real data, you're going to have bad data You know, I just I I don't think I've ever talked to someone who works with real data You know, and they're sort of daily life where they don't have issues like this and why uh, you know, it's like The data glitches, uh, it can be um, so sometimes they're systematic issues like Um, some things you can see when you look at the ISO data is Uh, part of the solar generation Looks like it's counting. You know, it's like depending on which source you're looking at is counted in California versus Uh, in Nevada, you know, so that that's something we think okay like that's something we can go fix and it'll be better But sometimes, you know, when it's like two hours missing data Is it the system in in the ISO that was where there was a bug? Is it somewhere in the database? Is it in my system? You know, it's like they're they're I think one of the difficulties is there are multiple layers of software systems That are talking to each other here. I think as soon as you have that Um, it just becomes, uh Yeah, I just don't think whenever you're working with real data, you should just assume they're going to be issues Yes, um I was just curious how you're doing, um Two factors. So are you doing with losses related to sort of, you know, from generation points to two months specific points um, and then also Partailments of using renewables might generate 100 megawatts, which are here only going to give in 60 megawatts, of course That's what you can inject with that point. So how many are those two? Well, so for the Tail months, maybe let me so for the commitments if you're if you were going to generate 100 megawatts But that's how that's sent to the ground. I think, you know, that basically number leaves the power plant So so in the data that I'm getting here, I'm just looking at Well, well I guess my easy answer is I'm just looking at the data that the the EI is giving me. So in a way, that's kind of their problem But Look at the data what you consume. Exactly. And that just gives you a production base So it's after you know, it's like it's after the losses and after So in other words, I don't separate that And I don't have an easy way to do that. But yeah, I just look at what's going on at the end and where I guess For the first one for the I guess one thing to do one thing more on that for the losses I think that in this data set is basically accounted for in the In the in the exchanges the transfers Yeah, so Okay, I guess back to some results to give you a sense of you know, what kind of Stuff comes out of this and all of these are plots that are Are generated on that web app. I was I was showing and we released all the stuff to do this and so this is showing so the top is LA and so LA has its own little balancing area And kyso is the bottom and you can kind of see it's symmetric here. So blue is demand That orange color is generation The green is imports and this is showing carbon intensity for these different for these two different places and The the full line that's the weekly average of the the weekly median of the data and then the Shaded regions you can see that's the 10th to 90th percentile to give you a sense of the spread And so one thing you can see is kind of in LA the power that they Produce is much dirtier than the power that they consume because they the power that they import is much cleaner And it's it's the opposite in california and in california So one thing that's interesting to me in california is you can see the so that that was one of the numbers I was initially when I started all this I was trying to get is the carbon intensity of imports for california Because at the time there was basically the standard number, you know, like one face number that When you looked at card the california air resources board one number for imports that they were using There was always the same and you know, it's an issue my question was a lot, you know Is that is that a good number how much does it change you can actually see the carbon intensity of imports? So this is about Three years worth of data and it changes quite a bit. You know it changes from Here maybe peaks around 400 and all the way down to 200, you know, so from like one to two Is the difference in the carbon intensity of imports and that just depends on what the other things that were on that grid So, you know come back to this style of heat map I was running for Stanford Stanford's electricity consumption. So here again every column Is a day of the year either offer 2020 and every row is an hour of the day And and the color here is the carbon intensity consumption based carbon intensity of the grid in these different areas And you can kind of see so california, you know, the top right You can really see the impact of of solar there, you know, and and actually when you look at these heat maps here by year Kind of see it changing, which I think is nice if you look at the midwest independent system operator So there you see their power is much sturdier in the In the summer when there's a lot of AC then it is in the winter and actually gets, you know in the middle of the night in the In the in like march 2021 It actually gets pretty clean. Sorry. I said this was for 2020. This is 2021 So this is basically last year um and one thing that's different from a midwest independent system operator to california independent system operator is that In their case, the night tends to be cleaner than the A day whereas for us, it's the opposite That's also true. I can't quite see it here. You can see better in the raw data for her pot And actually maybe it was within another year. Yeah, so for it for her pot that's texas on the bottom left And then southwest Powerful on the bottom right, which I think is now. I think is interesting because you see a big variation Um seasonally, uh, you know, if you look at march there versus august You see a really big difference in the carbon intensity of their grid, which tells you, you know So if we start moving towards a grid there where things are changing like that Well, starting to do Some what we're calling carbon work scheduling is really going to start to matter to take advantage of this So what's behind this of course is What what we're putting on to the grid and so what's also What's also available on on this web app is showing is the generation mix by source For generation by source the generation makes for each of these areas And so you can so basically these maps that are these time series plots Are the explanation for the generation side of things on the the box I was showing before and then if you add in the imports Kind of completes the picture And let me finish with this perfect. Um, so You know a couple of quick time takeaways, you know for some of the questions I was asking going into this And going on the sort of does it matter a question? Which I think is One of the first questions should always start before when you're working on a research project Well here, I think yes, it does matter because the carbon intensity of consumer electricity varies in time and space For imports I think the answer is it depends if you're looking at and you can kind of see it from the map here If you're in a place in the wek like in california your imports matter a great deal the number for 2016 that first paper I was talking about is I think california imported in 2016 roughly third of its electricity consumption or price of it and About 40 percent of the co2 emissions associated with electricity consumption in california were all swear So if you care about the carbon intensity of your brood in california, you really care about The imports and and measuring that if you're in a place like the midwest independent system operator Much less of a difference because it's sort of one big boxing area where everything's internalized So for the imports, I think the question is the answer is it depends And then and so here I'm talking more about current and ongoing and future work The fossil fuel plants play a key role in integrating The renewables and you start to see that when you're looking at day like these where you're seeing The the renewables going up and down and the fossil fuel plants going up and down and one of the questions I'm very interested in today is understanding how the renewables impact the fossil fuel plants And you know and how will that continue to impact them in the future in places where we started to have more and more Solar and wind based generation So a lot of this work is trying to provide tools to track, you know, these are these are measurement tools to try to track what's happening in the in the power system and Okay, and I'll stop there. I guess yeah, and and if if you want to learn more about this We've written some stuff and Yes, okay Um, the different utility electric carbon factors to look at the policy Uh, do you look at the plus margin 60 percent or ps policy means gas cannot exceed 40 percent of new supply as electric loads grow or do you use backward looking at what supply the old loads? I I'm not looking at At decisions here. I'm just looking at historical data and what happened on the way So, you know, this is really just measuring what happened and then for the second one. So everything okay So I didn't talk about the thanks for the plug I didn't talk about the difference between marginal and average carbon density here at all the plots They showed today are about what's called average emissions factors. So and uh, so here the question They're asking is what's the footprint Of g consumption and to answer that question. What I care about is the average carbon density The margin was yeah, I want to add. We will have another talk. I don't know as uh, apple 28 all Which they uh from professor in ace as a duo He's going to talk about, uh, you know marginal Carbon emission factor the research that she has been doing for many years Yeah So I've never seen one to deliver to students if you're interested in this topic No summation factor whatever average or marginal to what we're going to do is kind of consolidate These two things together now kind of comparison. What's the difference? Is the same source of data you can show the difference between average versus marginal carbon emission factors And we're thinking about do more than just the united states people can have this kind of global emission Tracking system. It'll be very interesting to a lot of users I don't know how talk is in three weeks in three weeks. I'll have to make any questions from the student Other plans to track other uh, greenhouse gases or methane emissions is that even possible? um So not methane so so so so there is so criteria air pollutants are also so and the first thing I was also looking a lot at that so and this is not the other greenhouse gas emissions. This is more Air quality questions and and uh, and so these are uh, the criteria or pollutants are typically associated with coal and gas. So these are sulfur and nitrous Oxides, so those we were looking at as well. I just didn't talk about it here For the methane. I have no idea For the methane. I think it's uh Well, actually sorry, sorry, I do have an idea so so in the in the data So methane is uh, an emission. Yeah, so so it's one thing that we are thinking of doing here is to generalize this framework for getting the you know the the the content of electricity for x, you know where x could be methane could be um, could be a criteria for pollutants But also things like uh, so I recently did some work for someone who was interested in water What's the water footprint of electricity consumption, you know, so I basically I'm modified the Who is in here to get the that out and so basically if if you have if you can give me um Something in units of you know gallons of water per megawatt then I can plug it through this and get something at the end For the methane specifically Actually, I said I don't know but I do know that's in the in the EPA, you know, they also measure as CH1 Put this question to explain wix ask in this again, and she will make also give you some ideas as well Are you looking at all into factoring in things like t-rex time-based your mission credits not that will affect these curves So I'm not so much thinking about Factoring that in as much as I think this kind of work informs the design of tools like that You know, so so this you know things like t-rex that's about accounting and responsibility and you know and You know basically if you had a if you want to put carbon on your balance sheet Is the way I see it and see how it impacts your bottom line if you put a price on it Um, this is really just about measurement, you know, all of this work is about trying to say Uh, how much CO2 is there? In the electricity that you're getting out of the socket. Um, you know, so then you know, then you want to use that for something For for t-rex then yeah So I think it's very related in that in that way the one thing that I haven't worked on so much But I think it's a difficult topic honestly is how do you assign responsibility? Which which I think you know is what people are trying to do with the t-rex or your I guess the whole discussion around marginal versus average of factors is also about that Um, and that I think Um, I think the the the answer in personally is not so easy And not as easy as some people say Yeah, um, when you show the map of the like carbon intensity across like the country I believe it looked as if um, it used coasts have like much higher carbon intensity than west coast Do you know why that is so it depends where well, I guess that goes back to the last picture Thanks for my screen again The last picture I was showing about uh gerations versus So the north west of the country has a lot of hydro power And so that's actually a lot of uh What's clean up here? In over here and so over here and over here you have Some nuclear you have some gas one of the things one of the things that's making uh, you know these Uh colors go more towards the red is coal and coal. There's a lot of around here Wachma has a lot of uh coal and then you have some around here as well So, you know It's basically to do with generation sources The northeast also has quite a lot of hydro and the imports in the northeast are super clean because they're coming from Canada Uh, and that's uh, and what's coming over that line A lot of it is is produced by hydro on the other side. Okay, and is there any particular reason why there's like tends to be more coal? Like in this region because of our legislation So why there's more more coal in the northeast? Yeah, or like and then also I see you in exploited too I think a lot of that has to do with historical reasons. Yeah Like a bit available coal mines Yeah Yeah, why would yeah, I don't think that has to do with the regulation I think the Yeah, I don't know enough about coal, but I but I think there's I mean there's a lot of mining for coal in In those regions and it's just pretty close by So this uh, this consumption based carbon intensity It's getting measured as like a consumer or is this is this like a life a life cycle? part of it So yeah, so I talked a tiny bit about that earlier So this map the one that's live is life cycle analysis base, but it's looking at uh, sort of the life cycle analysis estimate for co2 associated with the electricity you're taking out of Out of the club. So it's a consumer perspective looking at life cycle analysis You could so I was actually talking about this with the mesh earlier this week Um, you could so the the first one was looking at direct emissions So so the sense the continuous emissions of mining systems. Those are on the sacks in the power plants That's actually measuring what's coming out. And so the first maps I was showing You know with the same color scheme was just looking at how much is coming from just burning stuff There's not full life cycle. Um, so I don't have a good sense for how different the two are That's what we're talking about with Inesh and what she told me to make ask her again the question of three weeks But what she was talking about Monday is for her most of the emissions are associated with the fuel burning And the process, uh, you know, actually for like things like coal and gas is actually, you know, maybe 10-15 percent so, uh So the direct emissions estimates should be all you know Are you like almost everything? Yeah for for coal and you know, those are interesting to see like, you know Transporting the truck versus pipeline or something like that. Well, so so that's that's supposed to be all the kind of foreign life cycle Um, but then it's also for like, you know solar panels and right farms. So also So well, so yes, so so one and these numbers I'm using here. This is sort of all in number So here, for instance, um, so the numbers that are behind this solar and wind are not zero They're they're They're they're more like, uh, uh, 40 Yes, so you kind of lump all that in with, uh, like price per kilogram kilogram to be or two for Oh, yes, yes, so so this is all I'm doing. Um, you're talking about emissions factors. Yeah Yeah, this picture. So Here, so this is all in number you know, sort of everything into Co2 for a megawatt hour and so um, but so what what initiatives on me on monday is the For the coal for instance that thousand 90 percent of that is just the direct emissions estimate for Or you're right for the solar and wind, you know, it's probably, you know, everything is basically the life cycle because there are no emissions when you're Unless you can't do like operating emissions, you know, if you have to clean this water panel and camp those emissions But really there are very little operating issues. It's mostly about Uh, the manufacturing and the supply chains on the coal and the gas Almost everything is on when you burn the stuff It's kind of fascinating that they're all pretty similar At the bottom Well, well, actually, you know, some of them are like double the others. Okay, the reason I say they're similar Exactly. I think that the number of it. So these are well, so here I was looking at IPC 21 but actually They're IPC 2018 numbers that are more recent and it's You know, it's like 12 50 or something, you know, so 12 to 50 when you're looking at one two way, that's a big difference But when you're looking compared back to 450 the gas Um, it's you know, it's like the way I see it. It's another zero But yeah, there's none that's five and 150 there's not an order of magnitude difference between Uh, there's not like a five and a 15. It's like 15 to 15 Yeah, I think so Yes, so as you kind of been talking about this really gets at like the measure the issue of tracking emissions And to your knowledge, do you know if any of these results or data are actually being used by independent system operators at the moment for any players, um To actually optimize imports versus exports to minimize carbon emissions or are there any Plans or steps for that to happen in the future Uh, that's a lot of questions. I like the one So are they looking at this? Yes and more and more like if you go to the So like uh, yeah, like if you go to like the today's outlook for piso Uh, I guess you can show you that okay You'll see that Yeah, and they didn't used to do this. So, you know, so so they care piso at today's outlook You'll see that now they Do you have carbon emissions? Yeah, now they have it. Yeah, exactly. If you go here to california So this is the california. So that was what I was calling c iso and my plots So they have a their estimate, but you know, so here, you know, now they give you the same way they tell you what's demand supply They also tell you the co2 emissions and so here So I don't know if they've updated, but so it's gonna like initially like what was you know Dating into that imports was you know, what kind of got me started into all of this But you know, they're they're so I guess they care We'll be here enough to you know, set this up Whether they're optimized So I don't know whether they're like actually changing the scheduling to take this. No I don't think piso does but uh, sometimes we're like, you know, I saw with the piso carbon And they have that to impact their dispatch So some going on kind of research early demonstration work at you know, I saw I You know, we're not at I think we're at the we're starting to be at the tracking stage, but I think we're not here at the pricing stage Okay Thank you. Thank you. Our next seminar is in two weeks And Central image Thanks