 Welcome to the SmartCris seminar. This seminar is sponsored by the Bits and Warts Initiative. Our speaker today is Professor Ariel Leipmann from Monash University in Australia. Good morning Professor, it's 8 a.m. over there. Yeah, so I want to remind everyone that our next and final presentation for this quarter is in two weeks, and the start time again is 3 p.m. The speaker is Gregor Verbic from University of Sydney. Ariel Leipmann is an Associate Professor in the Department of Data Science and AI at Monash University. Since 2020, he has been director of the Monash Energy Institute and needs a reliable, affordable, clean energy for 2030 program. This program optimizes Australia's electricity grid through customer engagement distributed energy resources and network integration. Professor Leipmann also led the Monash Sustainable Microgrid Disability Study, leading to a collaboration between Monash University and state government of Victoria on a 30 million microgrid program. His current research focuses on optimization and machine learning based decision support tools for operational planning of smart grids. Internationally, Professor Leipmann led the Australia Indonesia Center Energy cluster to tackle the challenge of integrating distributed storage and renewables. Dr. Leipmann received his PhD degree from University of Queensland, Australia. Today, Professor Leipmann will talk about Monash University National Initiative and Transactive Energy Management. Let's welcome our speaker. Thank you, Jinwoo, and I'm very happy to be here and really honored to be speaking to your to your seminar community. I'm actually excited that my lovely colleague, Greg Overvich from University of Sydney is going to be presenting on actually Transactive Energy Management as well. In fact, it's a nice leader and I will be doing a more high-level discussion and some of the work that Greg is doing. We're also incorporating into some of my students' research, so it's really nice the way you've staged this. I've titled my talk-talk, NEDZERA Transitions and Campus Precincts, and that's specifically case study of the Monash University NEDZERA Initiative and Transactive Energy Management Living Lab. I didn't really choose to go into the details of the technical options of how you design a Transactive Energy Management Scheme. I think setting the context is actually a nice lead-in, particularly as Greg will be giving you some more technical aspects. So yeah, we'll set the scene here. We have a major challenge. As you all know, I'm sure we have a 1.5 degree target to meet, and that really means in practice. This is from a recent performance report from the UN on how we're going and we're not going that great. We need to be following more or less the green trajectory or at very worst the amber trajectory, the 1.8 degree maybe. We don't want to get anywhere near beyond 2 degrees C, and so things are urgent, and that means in practice global emissions should be heading towards 50% lower by 2030, only nine years away now, 80% by 2040 and 100% by 2050. That to me means that every good researcher, any ability to get involved in this space should be working on some aspect of this problem, and I mean broadly speaking, not just engineering or science researchers. This is one of the biggest challenges facing humanity today. It's not the biggest. So yeah, call to arms, as I say. So a little bit of background on Monash where we started to get, we've always had a sustainability vision, but we started to get really focused on climate change and energy related sustainability in the last five years. We led the Australian community and perhaps the world's university community in establishing a mid-zero mission targeted by 2030, and for which we received the End Momentum for Change Award and the Climate Conference in 24. This is not to boast, although we do like to boast as academics in universities, but this is really just hopefully to get other universities thinking about the same thing. And the reason we do that is not just to be able to say it, but actually there's a lot of learning that can happen from these programs. And because we have much more control of our facilities as universities and our campuses, and they're very diverse facilities, almost mini cities in some cases, like Monash and Clayton campus, where we in normal times have about 50,000 people working and living in one place, one of the biggest universities, biggest university in Australia, and relatively possibly one of the biggest in the world, because we are usually large in numbers. And so, yeah, it's really like a mini city. So we do this under the auspices of the Monash Energy Institute, which is established five or so years ago as a partnership between three faculties and the Facilities Division of the University, which we were presented by our NEDSERI initiative, which is the initiative Help Found. The Faculty of Information Technology, where I am, that sort of information technology is a sort of a synonym for computer science here in Australia, although sometimes people do get confused and cause for IT support. The Faculty of Engineering and the Monash Business School. And so with their generous support, we are able to coordinate researchers across Monash and coordinate their engagement with the industry and government communities and other communities outside the university. So we focus on three impact areas at developing new energy solutions, accelerating the energy transformation to meet climate objectives, and ensuring the consumers are central in this. And we apply our a range of capabilities under five themes, although we are at the moment splitting out out of energy resources, hydrogen, green hydrogen is a separate theme to ensure we give this proper attention on several major initiatives and various stages of delivery that we focus our work through. One of them is the NEDSERI initiative, which I'm going to talk about. We have our industry partnerships little center called the Grid Innovation Hub, where subscription based collaboration with various industry partners, including our grid utility here in Victorials Net Services, large precincts operator, vicinity centers, and a consultant, global consultant, Wally, which is engineering and business consultants. We have a partnership with Monash with Woodside Energy. It's an Australian, Australia's only oil and gas company. And then also the race for 2030 CRC, a reliable clean energy for 2030 CRC, which I might slightly correct. I'm not directing the whole center, that's got a CEO. It's established as a separate company, but I direct the electricity networks program, one of four programs within it. So these are some of our partners we work with. Zadvisian is out of date. That's now become Wally and Australian energy market operator and others. So what is our NEDSERI program all about? Creating grid interactive precincts is another way of looking at it. So I'm going to focus mostly on the microgrid, which is the research oriented aspect of it. But I will just talk mainly about initially about the other aspects, which is the NEDSERI aspect. So Scott Ferrero is credited with these slides. He's the director of the NEDSERI initiative. So we're investing about $100 million of Monash money to decarbonize our grid through various ways. We're expecting this is done before COVID. A student growth forecast will probably be revised. We're at about 50,000 on all Australian campuses. We're expecting about 80,000 2030. This led to a serious look at what our energy costs were, which we're going to explode. This led to a clear business case for reducing emissions through basically energy efficiency retrofits plus local generation rooftop solar, which forms part of our microgrid living lab. And then also of course, this is the lesson, a big intense energy intensive precinct. You cannot generate all your energy locally if you want it to be renewable. For example, UCSD can do that at times, but that means they've got a great big natural gas, open cycle gas turbine, okay, generation turbine on campus, which they can do that. But we chose not to do that, particularly in Australia, that it's never going to be economic with our gas prices. So now we source energy from a renewable, from a wind farm in Victoria. We will be basically net zero emission electricity by 2022. And then we need to electrify all our gas burning hot water. And in that switch, we'll be 100% renewable on that zero by 2030 or earlier. So this is a nice little wind farm here. So this is our campus and it's got three sub grids. And one of these grids is called, we call ring three. This is where this purple, sorry, this purple area is highlighted. This is our microgrid and we call it a microgrid. But I know it's debatable whether it's really a microgrid because we don't plan to disconnect it sometimes. And that's the technical definition of microgrid. The objectives, the main features of it are flexible buildings. This is our research area using where we use the transactive energy management concept to coordinate building flexibility, storage, there's some energy storage, one megawatt hour, vanadium redox hybridized with lithium ion, and solar about one megawatts of solar, probably more, and some EV charges. So what do we actually want this microgrid to do? We wanted to be able, in it, we optimize investment in local generation assets, because effectively want to integrate local electricity supply with grid supply, provide dynamic control over quantity and quality of multidirectional supply. It's exporting and importing in a way that also takes into account market signals, whether dynamic pricing based or some direct curtailment or grid constraints, critical peak pricing, even and things like that, and do it through providing access to local resources. So to do that, sorry, we are partnering with a Spanish IT infrastructure company called Indra or Minsate, who have what they call an active grid management platform, which is essentially a high reliability, high throughput, publish and subscribe platform with edge computing devices, which we are putting on each of our 22 buildings, plus the energy storage, and our inverters on our solar set, so we can then coordinate all of this in an efficient way. And so this is a nice visualization of the campus by a student of one of my colleagues, Sarah Goodwin. Sarah is a visualization expert working in the energy space, and this is actually a very early stage diagram, which we keep using, but it still looks very nice that they do a lot more interesting work than that. Look her up, and there's ACME Energy Conference being held soon, where she's hosting a workshop on energy visualization, that's ACME Energy. And so you see here, the red lines are their ring, we call it a ring, but of course it's not a ring, it's, but it can be switched. It's designed in a ring format. So we've got our resources there, and then we've got our facilities, sorry, our customers, these are all our faculties, our sports fields, our performing arts centers, residential services, we have a few thousand students living on site in our dormitories. We've got, of course, our stakeholders, the, the, the governance folks for the, for our university. And we've got a lot of on campus, mostly eating facilities. There's a lot of people, people need to be fed. So we've got a lot of independent businesses who are required to, to, you know, require electricity. And so yeah, we have a city here basically, and we can show, and you can show too, and I'm sure I stand for doing things towards that. Many US universities are, and others around the world, that this is, there is a way to manage these, this infrastructure in a way that is innovative in order to provide true sustainability outcomes, true decarbonisation. So here's our happy students graduating as well as working in the lab. So one of the things that is, I'm not going to focus on from a research perspective, but it's very important is actually moving our buildings up to high energy efficiency standards. We are relatively unlucky here for various reasons that in Australia, our building stock is, is really quite poor in energy efficiency. We're the least energy productive country in the world from a building and industry perspective, which is quite challenging. There we go. Thank you. Here's my coffee. And it's because Australia recently had very cheap electricity and I think we just didn't worry about, you know, wastage, which is not good, but we are here. So we need to really lift our game on reduction of our energy use from buildings. So a lot of our mission reduction on campus will come from this as well as the electrification of our thermal precincts, some of which is being done through some more novel technologies, district solar heating, replacing natural gas, heating up hot water. And we have an old, you know, gas fired boiler system that we will replace with electric heating that is, of course, powered by renewable energy that we're sorting, sourcing locally and off grid. Now, that's all great, but it's not so cool from an electrical engineering smart grid perspective. And so when you get to the next stage, this is the actual project that we're doing with interest funded by the funded by the Australian renewable energy agency, as well as interest themselves and Monash together. So this is sort of the plan. This is how we're doing this. We are now basically somewhere around here. We've installed all these edge devices. We've got simulators. We've got the battery installed. We've got our EV charging stations. And we are now integrating all these things with our peer to peer or sorry, not peer to peer, but transactive energy management algorithms. And it turns out that, you know, if you want to do it in practice, you start having to make some serious design choices and practical design choices about what algorithms you use. You know, it's nice and dairy to play with all these peer to peer concepts, block with blockchain or without blockchain or complex distributed energy market type mechanisms. But when it comes down to implementing it in the hardware and software that has never been written before, you have to make some interesting choices. So these are the things we are working on this year. This year is sort of the crunchier for integrating everything and getting it to work. So what is our vision? Well, in the end, we want collectively of buildings on that campus will be one commercial sized power, virtual power plant. And which both, which, you know, where more and more demand can follow supply. The dream is that ultimately the whole campuses demand will follow the true gross production of our share of the wind farm plus our solar that that we're procuring so that we don't have to buy the balance from the wholesale market, which of course adds price volatility and therefore puts a premium on our costs. So whether we'll get there ever is a question, but we will strive to get as close to that as possible. This is just another slide and again reminding us that we have this. This is the technology stack where we effectively building underlying it is the edge devices connecting to all our resources. Then we have the Indra active grid management system, which is remember the acronym now. I think it's DDS. Is the public subscribe high reliability platform, which is actually based on some military technology developed either by Indra or sourced by them with the platform. And then say in this is to make sure that all your information packets get to where they need to go through a flexible approach. And then top of that, this is where we put our transactive energy management or market, if you like, layer, which is where we build our intelligence agents in a way that basically can just be deployed into these edge devices with some central coordination agents, which integrate with some of Indra's other software. But some of it is actually a new layer of effectively cloud based hybrid cloud and edge computing distributed management software. This is another view of it. This is more like the traditional electrical engineering computer science view. We've got the flexible distributed energy resources here. We have the kind of actors that will ultimately play in the space, but we might have an aggregator or maybe different or an orchestrator and then the planner, which would be actually the owner of the assets that needs to understand the future outcomes of the pricing signals. If you look at that on a translated to a wholesale market paradigm, we are thinking these days not be admonished, but the world is thinking that somehow we will take these wholesale spot markets with the derivatives markets down to the distribution, electricity distribution level. Now, I think there's a lot of work to be done here and I'm pushing very hard here at Monash and in the CRC that we really need to ask ourselves what this means because computationally and engineering wise, it's a diabolical problem and computational challenges are everywhere there in terms of speed and resilience of the algorithms and cybersecurity of the algorithms. Here's an example of how we actually put this kit together. This is our red T flow battery founded by an Australian alum and manufactured based in the UK. Now, part of, just go and remember the name of the company that they merged with recently off the top of my head. Now, solar locally and these are our edge devices here built in by Indra and so the objective is we have these sensors and systems of systems in the sensors that connect to them. Then we have information coming in and then we have our local resources and then we are building an IoT enabled set of IT solutions or hardware and software solutions then in the end can help us flex the demand of our whole campus with a market like request coming in between what we want to do is collectively between time one and time two reduce demand by X kilowatts or you could frame it as X percent. Now, this all sounds very nice in practice especially if you have, you know, most of your demand can be offset by batteries if you're talking about sort of traditional residential, virtual power plant kind of concepts, but we have some large commercial type buildings which are really not designed to flex at all. They're managed by very black box building management systems which are actually very hard to modify retrofit any sort of intelligent control over. So we're actually doing some work around using reinforcement learning to learn the behavior of these buildings together with some forecasting methodologies and optimization so that's we published a couple of papers on that which I can share with Chinu later for putting somewhere on your website, but I won't go into detail but basically we have very complex physics problem underneath all this to integrate with. So and this is another view of how we're thinking. We did the original pretty much the current paradigm is passive direct management. There's a direct control mechanism. We are moving through active management which is a more distributed control but it can still be basically without any real choices for the consumers or any sort of market concepts and then we're moving through that through to the transactive energy paradigm where there's actually consumers preferences can be encoded in their local energy management systems and in a way that flexes based on pricing signals through some sort of coordinating mechanism. And so one of the paradigms we're exploring is the concept of micro grid electricity market operator which we're about to complete research on funded by the state of Victoria's government and there are some preliminary reports which I will show you later in the talk. So a little bit more about the architecture of the active grid management system. You can find more information on Intraminsates website. They've got some pretty interesting and impressive technologies they've partnered with Intel to develop some specialized chips to run these edge computing devices that are called OneSate node number one and they sit at all levels of the active grid management network. And on top of that integrating a distribution management system called Prism advanced distribution management system which distribution companies use to manage grids but we will be using it mostly as a power quality management system to ensure that whatever decisions are made from our energy management point of view do not lead to power quality issues like voltage excursions or or potentially harmonic created by somehow kind of strange control of the inverters we might want to do. So this is the the prism system was in fact acquired from another company called if I remember correctly ACS a US independent quite innovative company. So that's pretty much it about our transactive energy management project. There are several reports that you can look at on our website and there's a link at the end of my talk to the page where you can download these from. So now what I wanted to do at this point is just so okay this is very nice this is in a sense the if we have a parallel this is our living lab this is a hardware of research computer we need to write the software the software is the research itself so I'll talk about a few examples of the research we're doing related to this. The research itself that underpins the deployment is still sort of being developed and I elected to leave that for another talk but the components of that how they're coming together I'm going to talk about the sort of research we do at Monash research that that we get published and it's relevant to the research that you do it in Stanford and in your community. So I divide the research areas broadly speaking from a computer science perspective into areas which are more machine learning based more optimization based and then there's a new concept we are developing here at Monash in there and our leading discrete optimization group here which developed software such as Mini Zinc and now from which pioneered the field of constraint programming and constraint processing and so within the machine learning area of course you're familiar with a lot of these things that the forecasting problem is still very much alive but another area that you're also well aware of is non-intrusive load monitoring and then slightly newer area where we look at how do you use machine learning to detect false data injection attacks the sort of cyber attacks that we think would be quite potentially prevalent in virtual power plant configurations where people try and disrupt it either for for hostile purposes or just to extract value to themselves and optimization this is where I why I don't use terms like AI very much because really AI is a misused term people often think it's a good one in the greater community that to machine learning or forecasting even but actually it's much more than that so optimization is where we really where we talk about transactive management and really we're talking about a form of distributed optimization so we have some work in that and then some of the other aspects is when we're working on this coordinate inverter PQ control for enhancing grid export reducing curtailment and minimizing voltage issues which are sort of the flip side of the same optimization problem normally you will just curtail PV export or even turn the inverter off completely to manage voltage and so this this is an unsatisfactory problem so I'll talk solution to a problem that can be solved in in other ways and then we work in a sort of predict and optimize space so we'll talk about that as a general concept so some of the work we do is of course maybe more traditional forecasting but pushing that to the next level world expert Rob Heinemann is here one of the leading demand forecasters in the world and some of the work that's emerging out of that is now towards multi-agent sorry residential type forecasting where okay in the past we're just happy to forecast regional demand for the independent system operators but now we need to understand much more what's happening with deeper in the distribution level especially as we put in smart resources so hierarchical forecasting is an area you might want to look at if you don't work in it already this is some work by my colleague Shuru Pan who's a machine learning expert and this is looking at how do we actually detect various devices that are you being used particularly with very low resolution data well a lot of people really do a lot of work with high resolution demand records down to a few seconds a lot of utilities only really have very coarse demand out of half-hourly or hourly you know if you're lucky they have five minutes resolution so what can you do with that and we've shown that with a clever twist of logic in fact inspired by one of my colleagues from Albany and Lachlan Andrew we can use machine image recognition techniques to detect various behavior patterns if you simply turn your demand into a into a heat map and then it looks like an image and we can do edge detection to detect particularly things like pool pumps you can see here if it comes up got these blocks that come through and these are actually the regular on-off loads like pool pumps or hot water systems so this is our cybersecurity work my student together with associate professor cast and Rudolph the head of our software systems and cybersecurity department here who's a cybersecurity expert and we are looking at false data injection tax on distributed demand response systems which I'll describe the actual foundation of this in the next slide and so he's looking and so he's looking to using anomaly detection which is a form of time series forecasting or clustering to detect excursions away from a typical load pattern and and yeah he's which is excuse the order of these gas popping up which is generated by a transactive energy market like mechanism and we find that we are able to detect and then prevent these cyber attacks through appropriate intervention in the day head optimization mechanism so the overall work we've done on designing a market mechanism is basically a highly scalable frank wolf algorithm based distributed optimization approach which enables us to optimize millions of actually of distributed energy resources very fast through a somewhat clever application of the frank wolf algorithm that means we can sort of clear these transactive energy markets in seconds so this is work by our student Dora here who's just submitting her PhD shortly and there's a few papers on that you can find in the in the literature and so another area that I'm very excited about is the work of Peter Lucis who's a PhD student also just finishing up and it's saying well what can we do if we could directly control the some of the inverters in an energy in a distribution network and to manage voltage and you can see some of these simulations here would show well what happens if you don't do anything when you get these amazing for a start you get curtailment the solar output would have ideally provided this envelope which fits this little black line on the right which is on a different scale unfortunately so it should be probably double the size and this is where the voltage starts to max out at the limits and traditional inverters just turn off and then they'll turn on again as voltage drops again and you get all sorts of horrible behavior and what we have shown you can do through coordinated control is smooth this out and reduce and reduce the energy unused energy and and reduce losses through absorption of reactive power at the inverter so I'm getting close to time but I've only got two more slides to go so I think we're doing well and so the latest work out of Peter's PhD which we're about to submit is actually a fully integrated model using we had to build some our own optimization algorithms for AC power flow coordinated with inverters control which have no linear characteristics and then the non-linear losses and now we've got the batteries in there and we've controlled independently both the inverter batteries and the sorry battery inverters and the solar inverters and as well as optimize the charging and discharging for a day ahead voltage secure VPP and I am quite excited that looking forward to the community's feedback on that when I first submitted through reviewers obviously and get the typical kind of we don't understand why you have to do this but I think we will be able to explain it to them and I think this is kind of the solution that really then brings us to the true transactive energy management paradigm where you have to optimize for both real and reactive power ensure voltage is is is properly managed and of course Peter's approach is sort of command and control. We didn't want to bite off all the different problems at the same time but then you turn this algorithm into a transactive energy management algorithm with a distributed optimization base so yeah you can see again a little bit closer what it looks like and what I neglected to put here was basically another line which shows am I smooth the version of these lines when when we do enact the coordinate and inverter control okay so you can probably see it here the ideal would have been the purple area this is if you could export all the energy from the inverters to the grid the the blue line here area here is the internal consumption of the houses on that distribution feeder and what you find is that basically all the purple energies strain away at 90% penetration level that's 90% of the houses have 5.4 kilowatts of solar and as you can see most of the energies are needed by the houses because of this typical load here and then this purple just thrown away completely some of it ends up being actually exported from the house but it's consumed in quite high losses because of the grids just not optimized for that direction and then the rest is is exported from the houses and in fact this is exported through the distribution transformer and to the medium voltage level now if we this is with volt va volt what inverter only so this is what they would do on their own it looks like corner inverter control buys us a little bit of improvement by reducing losses but the lesson is also well the energy's got to go somewhere and if you know there's no load you just have to throw it away it doesn't matter how clever you get so there is a layer of analysis that we need to do then to say well what if happens if we can change things on the medium voltage network to keep that energy flowing then maybe you can actually export more of this but if you have standard configuration of distribution transformers built and their protection built for one mostly one directional energy flow you get this but we're able to reduce the energy loss the line losses by about 50 percent right by by i inverted control but that's all you can do but what we also show is like it gets much smoother so these fluctuations here which are already not as bad under pure autonomous inverter control volt va volt what inverters are smoothed out but in reality some of the systems we all have today will be a mixture of really old inverters that just trip on and off autonomous inverters and so the voltage fluctuations will get much worse and then this is an example of yeah even with autonomous inverters they still have to turn off because at some point you just cannot curtail back them off anymore and this shows throughout one of these simulations even in on the autonomous side they're starting to turn off throughout so in fact half the inverters in the middle of the day are just off completely and then finally an example of a predict and optimize methodology using a quite a different method our my colleague john betz is an inventory management specialist in logistics and using optimization techniques developed for that which are really quite different to standard you know next to the linear programming type optimization and he is able to do a stochastic optimization and prediction and optimization of battery storage to maximize utilization of solar and he gives you here the optimal sizing and for securing maximum utilization of solar there's a paper published on that so in summary i guess i've discussed how we work at bonnet on all these different um uh transactive energy management uh systems inspired and led by our net zero initiative in our microgrid lab and um our microgrid living laboratory i should say and so here are some links including um uh the microgrid uh so what am i wondering with the microgrid reports and then you can see our dashboards showing what our microgrid uh resources are actually doing and then a bit of information about the energy institute so with that i'm i'm finished sorry i've got a little bit over time i think the timing is fine uh so there are some questions in the q and a uh i believe the first three are all about uh using eevee uh so so so what was what are your thoughts on in uh on having incentives for incorporating uh eevee charging discharging i think oh look so that's something i didn't talk about but obviously we're going to be working on that uh i think this person is particularly interested in the uh v2g yeah so um okay i think it's a chance to read all the details here so maybe if i don't answer all the question um uh look like laurence is pretty enthusiastic about eevees which which i totally appreciate we are a big fan of electric vehicles as a as a decovernization method and also a potential grid management resource and so my thoughts are yes absolutely this is a major uh opportunity and you know if i if you if you just cast your mind so what's the world going to look like in 2035 say um and i do like to go that far ahead in this case because i think there's some technical and policy challenges that will take a while to resolve but certainly by 2035 we will see eevees doing a lot of uh of the heavy lifting as we say in australia around um grid management now there's a lot of um through v2g or v2x is now the whole v2 everything um concept v2 home v2 grid v2 building i don't remember all the different x's here so um does that sort of answer it um i mean you can get into detail of fast versus slow charges and all that and so um yeah let me just say that um you have to go through stages first of all we just need to make sure um we understand the rate of uptake of eevees and different countries and different stages and the grid operators or the grid utilities have to respond through the right kind of investment and the question is how much investment do you need and that's a function of how much first of all managed charging you have so clearly just unmanaged charging in some cases just can be very expensive to accommodate so you need either um uh pseudo price based or price based managed charging but then obviously if you do managed charging you should easily be looking at doing v2g and um that's certainly going to be huge okay uh i think the next question is did you have any retrofit to the building that i think this is what the building energy control um so so in terms of the envelope and the materials um we are certainly looking at that we are doing some of that um it's not my area of at least certainly not my area of expertise but i haven't got the latest information on what portion of our um energy efficiency will be done through material retrofits um and so Jorge sounds like he probably knows more about it than i do and you can imagine the decisions you have to make and it's a cost benefit thing you know the more retrofitting you have to do the more disruptive it is to the building uh operation and time and cost of doing that so we're looking at that um and hopefully we'll get the up close to the optimal decision okay uh the next question is how do you assure zero i think it's zero injection to the network and stability of the system uh yes yeah well so so um you may have missed a slightly um brief uh illusion to the fact that we never operate in standalone mode um we we elected to not try for that that's when that's why i said microagreed a bit is a bit of a misnomer in this case uh uh really it's unnecessary and uh and um uneconomic uh basically we're deep inside a large city's distribution network um uh i know ucsd was able to um even claim supporting the the grid during some i think it's 2017 or 16 grid events that california had um but really what can we do we're a 18 megawatt peak demand campus inside a city of four million people i don't think we can help the grid for stability purposes that much so we just elect to think of ourselves as a giant virtual power plant where we can maybe support voltage issues at the local distribution network by not exporting or importing as as much as possible at the right time okay uh the next question have you compared nox with the three campus aphids this is uh pacific northwest and national lab you of washington and washington state um yeah so i look good idea uh we we do talk to pnl sometimes we work closely with um a company here called strategy run by or the office here in australia is run by um mark paterson who is close to i have to think of his name now pnnl uh who in fact uh came to talk who's a transactive energy management guy and uh he came and i met him here and he inspired a lot of our research and and ultimately in fact the this transactive energy management work so in that sense we we are comparing notes with him in one directionally um but we do seek to then now once we are um a little bit further down the track with the actual practice we will certainly um keen to uh to meet uh and compare notes so uh yeah next question i think it's on uh pricing optimization have you considered that in the in your design well so yeah the design is very much um considering price based now you can do um optimization without price you don't need to have a price um i mean you have um shadow prices to the constraints obviously that mean that assuming you've got the actual real um uh costs that are being um calculated in your objective function and and uh the decision variables uh cost related um but our design is definitely um uh the derived from some external market feed through but it's really one directional right we're not going to optimize with the wholesale energy market in a way that actually means that our decisions affect the um the wholesale market outcome so we will probably we'll take a wholesale market price and then have a local um additional price um uplift shadow price if you like uh on our constraints now we don't plan for our actual uh energy users to be see the prices and actively respond to themselves what we are doing is basically developing a utility function or price or demand function for each of our buildings that will be programmed into our intelligent agent which will be its own basically um uh uh preferences or utility function or its demand curve and then a um a central price will emerge so they all are paying the same price and therefore curtailing or not um flexing or not as their local preferences require uh how are you engaging the university community and no one citizens uh very good question um so we are uh working through various channels on that we had a um participatory um process where we first of all uh informed the I went through talking to the um to the community the university facilities management community and the departmental um management for all the departments or faculties in the university to explain where we headed we are still probably a bit far away from actually actioning any real control on campus you know that's still a uh that's to be honest about a year away that's just tells you how complex this is in the real organization we can imagine this we imagined this five years ago we started the work and then you know we're just getting to the point where this can actually you know do something on on camp campus so um uh this not to underestimate the complexity in the real world of of actioning all these things that in the lab in silico on the computer look you know oh yeah we just do that and bang we'll have a super flexible demand side which would just absorb all these variable renewables at grid uh scale so yeah not so easy and the engagement is still ongoing uh Melbourne citizens are not so much our um direct constituency as um you know they just see this campus that that's you know relatively small part of the city's demand anyway but once um you know we we will have a process where we tell people you know how things are going once the campus actually is flexing uh some demand and then we will work more with city councils and other um organizations we are working with them to show how this can be done in their precincts in a in a equivalent way but um and I believe that there's still a long way to go for this globally and we need to do a lot more of this and universities really uh need to lead the way because we don't have to worry as much about regulatory constraints about how we manage our facilities and grids so if we all get together learn from each other we can then show cities how to do this um at least we solve the technical and social and economic problem of it and then they can worry about the the political problem and the regulatory problem that uh together with us and uh so yeah I invite all the universities to join us in this transformation in fact and um please let us know if you want to collaborate on this if you're not already working on this we have a team standing by to to tell you how we did it not that we were the best at this maybe we were one of the first but um um but there's certainly a lot more to learn I have to have a follow-up question to this question does your customer engagement evolve students living in the dormitories yes yeah we did some actually um quite explicit studies of that through um a sister organization Climate Works Australia which is a um a kind of a think tank and policy uh and development advocacy organization it's actually part of Monash but it operates in separate organization and they did work with us on some studies in the dormitories and if you email me I'll put you in touch with with the people who did the study and um we've got some reports on that yeah okay great uh i'm not sure if the cooking how do you manage the relation with regulator regulator and grid power regulator well so luckily the regulator is not the problem that's the whole uh the wonderful thing about our um being a campus is we have a private network here um in some cases we call it an embedded network so we can do whatever we want from a grid and local perspective that that gives us a lot more freedom to experiment and it's a lot of other campuses in the in the world are the same right and probably Stanford is the same some of them are more urban um campuses that um are really deeply embedded into the city's grid and therefore you do have to work with the regulator but we decided not to bite that problem off yet um we'd work with the regulator to get their views on and then part of our microgrid energy management operator MIMO project which there's some interim report there in the links it is around how you would then do this when you do have to work with the regulator and we had some regulatory experts working with regulators and governments to to tease that out um the second part of the the grid operator of course is a much more challenging thing even though it's a virtual it's a private network because we connect to them even connecting one megawatt of solar on a 18 megawatt peak demand campus was a problem for them they're worried about short circuit ratios and and other things and it took me a couple of years probably to negotiate that with them so um you know it it's quite quite diabolical um and so uh yeah don't um don't underestimate that I mean some utilities are better than others some more conservative than others um is it a good business to be a virtual power plant um the answer is how long is a piece of string is is uh uh the way we say this in Australia and that's like one day it will be a good business at the moment I don't know that any virtual power plant operator is making money as a as a standalone business that the costs of enabling this are still um you know emerging as any new technology um it's expensive um when you're learning how to put all the bits together so so I assume that's the meaning of the question if you look in pure um well actually there is no other way to answer it I don't think because enabling the operation of virtual power plants got so many layers um the cost of it is is is is a function of all these layers so thank you for that question are they all good questions we're at the end of the presentation if there are no more questions yeah and thank you very much for the wonderful presentation yeah thank you for the opportunity and the honor of addressing your uh very um uh expert audience okay uh okay so thank you everyone and uh again our next seminar is in two weeks uh June 3rd thank you Gino and Mahila and the rest of the Stanford Smarterhood community okay thank you okay all right