 Okay, let's get started. Welcome to the final Smart Grid Seminar of this quarter. Our speaker today is Professor Grega Verbich from University of Sydney in Australia. He's going to talk about a transactive energy system for integrating distributed energy resources. I want to remind everyone that all the presentations have been recorded so if you miss any of them, you can contact us or you can search on YouTube. Professor Grega Verbich received his degrees in electrical engineering from the University of Rubiana in Slovenia. He was a postdoc with the University of Waterloo in Canada. Since 2010, he has been with the School of Electrical and Information Engineering at the University of Sydney in Australia. His expertise is in power system operations, stability and control, and your system markets. His current research interests in grid and market integration of renewable energies and distributed energy resources, demand response, and grid modeling. He was a recipient of the IEEE Power and Energy Society Prize Paper Award in 2006. And without further delay, let us welcome our speaker. Good afternoon everyone. It's early morning here in Sydney. So my presentation today will be, we'll talk about transactive energy systems in the context of integration of distributed energy resources. And maybe I should put this slide first. The presentation is mostly based in this paper, which was recently published in Renewable and Sustainable Energy Reviews. It's a review paper that reviews our work we've done over the last several years in this area. So if you're interested, you may refer to this paper or the references contained in the paper. So the agenda for today, I will first talk about the background and motivation. Given overview of management and coordination of distributed energy resources, then I will focus on specific approaches discussing home energy management first. So when you manage distributed energy resources in your home or maybe a small commercial building. Then I'll move on to coordination where you coordinate several distributed, so actually several prosumers into virtual power plants. You can do so, you can do so with or without taking into account the network. So the next one will be network or work coordination where you explicitly take into account the impact on the network. And finally, I will discuss a conceptually quite different approach which is peer-to-peer energy trading using auctions. And finally, I will conclude with a comparative analysis comparing all the approaches using technical and economic criteria. And finally, some conclusions. Okay, so I usually start my presentation going a little bit back into the past. This is from the CSIRO CSIRO, you can think of that as a national labs, equivalent to national labs in the US. So back in 2013 CSIRO convened a so-called future grid forum where they asked several industry participants to come up with possible evolution scenarios for the electricity system in Australia and also globally. And they came up with four scenarios and the first one is quite interesting because I find it quite visionary remember this was back in 2013 almost eight years ago. It was called Rise of the Presumers. So they predicted by 2050, the consumers will take up on so 46% of consumers will have on-site generation and electric vehicles, and the role of centralized power will decline considerably. So there's much more emphasis on decentralized power supply and importantly consumers will retain certain level of comfort. So the autonomy is quite explicit in that scenario. And the network then becomes a platform for transactions. So those two issues of consumer autonomy and the networks will pop up several more times in my presentation. And another scenario was renewables thrive, where they predicted that centralized power supply will almost exclusively consist of renewables, large scale wind and solar. They also predicted high electric vehicle uptake and strong demand control and batteries will feature prominently in future supply system. So at the moment we are probably witnessing a combination of the two. So large penetration of utility scale renewables at the transmission level and distributed energy resources at low voltage level. So now if we move a little bit further into the future, so this is from 2015, where the Australian energy market operator came up with the prediction of the cost decline of PV battery systems. In 2020, and this is from New South Wales, which is one of the states in Australia, where I'm based in Sydney. The payback period of residential PV battery systems will be around 15 years. Now, today it's probably closer to 10. And this is predicted to fall further, further moving into the future, which means that PV battery system residential small scale PV battery systems will become economically viable or already are economically viable without any subsidies which obviously drives the the uptake of these technologies. Now if we want to look a little bit deeper into the drivers of what drives the uptake of these technologies. I'm showing here a slide from Morgan Stanley research from 2016. Where they so on the left hand slide, the left, sorry, on the left hand side we have installation cost of a generic seven kilowatt hour battery system. So they predicted that the cost will drop quite dramatically from 2016 to 2018 to be perfect on as the cost decline for battery storage is not as fast as they would be but it's still dropping. More importantly, here on the right, I'm showing battery cost per kilowatt hour use versus value opportunities. So this chart on the left tells you that the battery cost of what if you use one kilowatt hour of energy if you move one kilowatt hour out of the battery that costs you around 50, between 50 and 100 cents per kilowatt hour. And on the right, we have we have the revenue opportunities. Now at the moment, customers can only use the bottom two which is solar self consumption. So basically charge the battery when the sun is shining in the middle of the day and nobody is using that energy and tariff arbitrage. So if you are on a time of use very few may charge the battery when electricity and discharge is expensive. So you can see that currently those two value streams are not quite sufficient. I'm not, I'm not sufficient. This was in 2016 when this prediction was made was not sufficient to cover the battery, the battery cost, but in the future, they predicted that there will be several other opportunities will emerge for customers to play in different markets or they call this collectively network services which might be ancillary services for low voltage networks as well as, for example frequency control services for the wholesale market. So once those new value streams become available, you can see now that the total value of those revenue opportunities exceeds the cost of the battery and this does goes to show that when those new value streams become available. And this will drive the uptake of residential batteries even further now here in the middle we have VPPs as I will discuss in my presentation, the boundary between network services and VPPs is is is blurred it's not it's not crisp so you can argue that basically coming in one bucket. So, and then this is from 2018 again that prediction from the Australian energy market operator, where they predicted the uptake of batteries and rooftop solar. Now just from here on the lab we have capacity. So this is roughly 20 gigawatts and peak capacity in the national electricity market which is interconnected some on the eastern seaboard of Australia is probably double that number. So you can see that a emo agrees the Australian energy market operator agrees that the uptake of TV will indeed be quite significant, which then leads us to this prediction which is quite remarkable. This is again from the Australian energy market operator and energy networks. We're in 2018 so they copied, they took this from blue blue Bloomberg. Bloomberg predicted by me 30 30s in Australia so that's the stop graph 45% of power will be generated in behind the meter behind the meter means in either residential or small commercial buildings. And that's quite remarkable because if you think if you think about the implication of that we are effectively predicting that will flip the operation of the power system from generation following load to load following generation. And so, for the sake of the, sorry, excuse me. For the sake of this presentation I came up with this diagram or schematic diagram on the left, which shows a typical composition of a power system where we have on the top so this is this reddish red color where we have by generation. And then we just need power, typically over long distances over the transmission system down to distribution and finally, finally to residential low voltage levels. So this is how a power system was, or still is operated traditionally. Now, if we, if we look at the slide, sorry, the picture on the right we have this like a simplified diagram that shows that this transmission system in the middle. And then the power flow, not the direction of those arrows the power flow is downstream so from generation down to medium voltage levels and finally to low voltage distribution where the customers are located. Now, when customers become prosumers so when they take up distributed energy technologies, they can generate power themselves, and with the aggressive uptake of DRs. We are already witnessing in some parts of the network here in Australia South Australia and Queensland are two examples where the power already flows upstream. And so you note here that the arrows now are two way, which means that the power can flow from low voltage of course up to medium voltage and even high voltage networks. And if that prediction eventuates where we get almost half of the power from behind the meter. The power system will become really a two way. So the network that the power flows will become predominantly predominantly two way. But there's an issue with that because know that prosumers are located at the low voltage level, so they are located at the fringes of the grid. As you increase the penetration of prosumers or you can think of that penetration of rooftop solar which is the main distributed generation technologies. As the penetration goes up, you start witnessing networks problems so note here that those bubbles become darker and darker red and that indicates voltage problems. Network problems either voltage problems or congestion problems where you exceed the capacity of the transformer and the cables. So as the penetration increases. Low voltage networks become congested and voltage limits are violated. And this is already happening in many parts of Australia to the extent that Australian energy market commission came up with a proposal to charge customer for solar exports. When you export solar into the grid in the middle of the day. They're suggesting that the customers customers should be charged for that, because the network. To get revenue required to augment their network. Now, note here that customers in Australia all also get a so called feeding tariff so when they from the retailer typically so when they fit power into the grid they get paid the feeding tariff, but on top of that the networks are now would like to charge customer for exporting the grid which would effectively reduce reduce the revenue, the revenue for the customers. Now what that means is that this will probably drive the uptake of residential batteries even further because if you're only alternative. When you generate power and you're not consuming it is to dump into the grid and get paid nothing customers will likely invest and invest in the batteries. So that's the issues we have we have today. Now at the same time, we are witnessing a rapid development in several important technologies. And so the first one is you pick it. The concept of internet internet of things were effectively every device in the system and talk to any other device and what that means in our context is that distributed energy resources can now become system system players on top of that smart devices have now a lot of computing power which means that we can do a lot of computation on those small devices themselves for example in your smart meter you may you may run your home energy management system on a small a single board computer. Another one is blockchain or distributed ledger technologies that will enable a distributed energy marketplace. Now to drive this message home here at the bottom may have a slide from international energy agency. So how much how much money globally utilities are investing in digital electricity infrastructure digital technologies in general. This is from 2016. So you can you can see that even five years ago, the total global investment by utilities utilities in digital technologies exceeded the investment in global gas powered generation. So this just goes to show that those underpinning technology will enable this vision or fully decentralized power supply. So now I'm at the like a main part of my presentation, what I will review management and coordination approaches of distributed energy resources. Now for the sake of this presentation or in this paper, this paper here at the bottom which I mentioned initially. So let's start with with this diagram where we classify coordination approaches based on two criteria. The first one is the extent to which they take into account networks, and the second one is the extent to which they explicitly focus on customers or system and the approaches can be either network aware or completely network oblivious. So the first one is home energy management system. So you can think of that you are an end customer you buy a rooftop solar system a battery system and you want to manage it to minimize your cost. And that obviously sits in this bottom right corner, where you are fully customer oriented and you don't care about the network at all. And we saw that this can lead to network problems so then the next approach is home energy management with with operating envelopes and this operating operating envelopes you can think of that as distribution system operator telling you what you cannot you can or cannot do in order to prevent event network problems. So now see a note here that this approach is now much more network aware. Then we have peer to peer energy trading sit somewhere in the middle. And finally virtual power plants. So I put the VP P. So I can call that network oblivious VP P which is state of the art currently in Australia and also globally. I put them in the top right corner because they are completely system oriented oriented they don't. And they implicitly care about the customer of course of course, but customers are not explicitly taken into account in the objective, as we will see later. And finally we have network aware virtual power plants. And they are based on the solution of the optimal power flow problem so they are fully network aware, but then the extent to which they take into account customers in the objective can can vary. So first of all home energy management. So as I mentioned, when you invest in distributed energy technologies PV battery systems are the obvious ones, then of course you would like to come up with an efficient way of managing these technologies to minimize to minimize the electricity cost. You can also focus on improving comfort for example if you are also scheduling your air conditioning and maybe some some other objectives as well. Now if we look at this more holistically electric vehicles are probably becoming an important part of that but you might also have fuel cells or micro turbines for combined heat and electricity generation, and possibly fuel pumps are now quite quite popular in cold climates for for heating. So, when you have a range of devices then obviously you would like to optimize their performance or their operation, and this can be cast as a sequential decision making process under uncertainty, where you minimize some cost reduction as I said you can think of that as minimizing electricity electricity cost. And you do so, you want to find a policy so a policy you can think of that as a rule that tells you what you do in a particular situation so for example do I charge the battery do I discharge the battery and so on. For example, you have some randomness so this is not fully deterministic. So that means that you have you minimize active cost horizon. Now there are several solution techniques mixed integer linear programming is probably the most popular one, the problem can also be solved dynamic programming, but it has to be emphasized that this problem is quite computationally challenging in itself. For the sake. So in this paper, we use the fairly simply simplified home energy management formulation where we only use this is the most typical situation currently in Australia. So we assume that we have a hybrid inverter so we have one inverter to interface the PV battery system with agreed, and we have this user agent that uses electricity. Not not flexible demand. So now this problem. If you have a more, more devices than than you have you probably have to formulate it using mixed integer linear program, but in this setting, assuming that the user base, a time of use tariff when when import electricity from the grid and point when the power flow is from the grid to the customer, the user plays a time of use tariff, but when it exports power to the grid so for example when TV is generating more than the user can consume the user gets feeding tariff now because the feeding tariff is always less than the time of use tariff, we can, we can linearize this problem or we can substitute constraints completely and using a two auxiliary variables to model that net power flow exchange. So, so with this linear formulation, the problem becomes minimize the amount of money, the user base when imports and minus the money, the user makes when exports power into the grid. And again, the retail tariff can either be time of use or FT means flat state so time of use. And the feeding tariff is what you get paid when you export power to the grid and this is always less less than the time of use. So I won't dwell on that. The details are in the paper. It's just a simple linear linear problem. And again, this is with complete disregard of, of the impact on the network. Now we can rectify that by introducing some operating envelopes. And again, you can think of that as the distribution system operator telling customers what they can or cannot do and there are several ways you can do this. The approach we use is to use the power flow Jacobian, which effectively gives you sensitivities of active and reactive power injections on voltages and phase angles in the network. And this obviously assumes that the distribution system operator runs state estimation so and also has a full observability of the network. When you have access to the power flow Jacobian, you can quickly assess the impact on the voltages by using the sensitivities from the load flow Jacobian and looking at the power the customer injects into the network. Now, I haven't mentioned, I did mention this explicitly, but voltages are probably the biggest issue when you increase the penetration of rooftop solar, because in the middle of the day, when the demand is low and the sun is shining in the systems generate power. So what I said before that home energy management is completely network oblivious. That's not quite technically not quite true because the inverters now have to be what they do have protection setting so the voltages in the network. Go go too high than the inverters will trip so you can argue that home energy management of consumer battery inverters TV battery inverters are are to some extent network. Okay, but a principal way of doing that is having system distributed system operator in place that monitors the system by doing state estimation, computing in real time. The impact of of customers exporting power to the grid and sending back power limits to the customers which they have to obey in order to prevent network problems. Okay, so those two approaches were full course focused explicitly on on customers. So virtual power plants, that's conceptually a different approach and the idea is to aggregate a large number of consumers so we can think of that again behind the meter distributed energy resources located in residential buildings or maybe commercial small commercial buildings equipped with a home energy management system and communicating with the rest of the systems to a smart meter. Now, here I am assuming that we have one physical network so here I'm assuming that all the customers are connected one low voltage network managed by the distribution system operator. In principle, every customer can be under a different retailer so you can buy electricity in the market through a different retailer we have several options. And then on top of that, you might also have an aggregator and in principle that can be many of them. So services that your devices can provide to the grid sell those services into the whole cell market. So that's the idea. Now, if we now look at the implication of that that model onto the system. Here on the left I have a deregulated power system, but with passive demand so that would be probably what we have today or what we had maybe 10 years ago where customers are fully passive. So the power flow is from generators to transmission down to distribution and to the customers and the financial flows are then in the other direction so through the retailer, the market operator and finally to the generators. When you when customers become active participants in the system so they become consumers. So now this changes so you still have, I mean customers can still buy electricity from from the generators so that would imply that the power flow is downstream, but they can also sell electricity into the grid so this power flow here becomes to way. Now, the retailer is still there as before but on top of that you also have an aggregator that can sell services into either, either to the distribution system operator or to the whole cell market and this is facilitated by by an aggregator. And in principle that can be the retailer and the aggregator can be can be one company so this is still still early days. There is no this sector is not not far from mature yet and the financial flows now note here that again you still have a contractual relationship with the retailer, but you can also make money by selling services. That will run aggregated. So now in principle this can be formulated as an optimization problem where you minimize some cost function. And so and that the X are the decision variable so that's the feasible feasible set and includes the aggregator, which is this entity here in the middle user agents sitting at the bottom, and the network variables system operator. So that's a canonical aggregation problem. Now before I go into specific details, I will first mention the state of the art, which is the business model that we already have in Australia, I call that retailer VPP or VPP 0.0 which is probably the simplest VPP model you can you can think of. Now note here that is, there is no optimization. So the VPP controller, which is owned by a retailer controls, or has a contractual relationship with customers in different parts of the network so that these consumers are connected to different low voltage networks so they might be electrically far apart. And the business model of the retailer is to mitigate price exposure. So the retailer effectively uses customer owned batteries to minimize the exposure in the wholesale market. So when the electricity price goes up, they discharge batteries to avoid buying electricity in the wholesale market because it's, it's too expensive. And this is this uses direct load control so no, no, no optimization. And that battery controllers still still maximize self consumption for the customer that can do price arbitrage for battery controllers still optimize the value of that before the customer. And the retailer when it needs the capacity, it can override, override those decisions. Now, battery controllers, the state of the art currently use some heuristics so there is no, no optimization involved in that, even though there are companies who already do that. So which means that customer demand profiles are not considered. There is no forward looking optimization. So for example, if you are not at home, then tomorrow, this is not taken into account in optimal of the charging and discharging profile decided by that. That's suboptimal. Now I have to mention that in October this year, the, it will come possible for aggregators to beat also in the wholesale market so this is called so called demand response mechanism rule change, which Australian energy market commission put in place. Now, interestingly, these aggregators will only be allowed to aggregate large scale loads so not residential customers, even though if you if you read the report from the Australian energy market commission. I want to acknowledge the fact that that not in very not so distant future. The main technology for aggregation will be distributed energy resources, but because they consider is considered these two complicated at the moment, simply because you have so many of these small devices at the moment this rule change only allows aggregation of large scale customers. Okay, so now, if you want to do this retailer VP model in a principled way you can formulate that as an optimization problem where you maximize social welfare. So if you can think of that, you can think of the cost function as a quadratic cost function that that models, the price of electricity in the wholesale market, but not here. The decision variables now also include the user agents, which effectively means that the retailer now uses customer batteries to minimize their own cost. So social welfare so you can argue that customers also benefit, but customers are not explicitly taken into account in the objective. Now, you can rectify that by putting in the objective also the objective function of the customer, which might be the same objective function as we have seen for the home energy management problem. So you can then so this now becomes a biker by criterion optimization where you have this gamma here that tells you how much weight you put on each on each component. You can argue that when you have a large number of customers possibly in the thousands that solving that centrally so this is a centralized solution, it's not feasible. You can, you can solve this problem in a distributed fashion, where you use dual decomposition, and this effectively becomes a price based coordination, where in each iteration. You start with an initial price update, which is the dual variable associated with the power balance constraint. The user agents optimize based on on the existing price update. The aggregators do that for their own sub problem, and they update those like range multipliers you can think of that as a new price, they send it back to the users, and that after a couple of iterations, you reach a solution. It's scalable scalable, and that computation is done locally on smart meters on the devices owned by the customers, which again allows you to scale that to several thousands tens of thousands. Okay, if you have also integer variables in the problem that becomes a bit more complicated because you have a non zero duality gap, but we came up with a solution. Now, so far we haven't considered the impact on on the network. And this can be taken into account by reformulating the VPP 1.0 and 2.0 optimization problem into an optimal power flow problem. You can think of that as economic dispatch subject to network constraints. So again, you are minimizing some global cost function you can put in the objective also the cost function of the consumers, so that part remains the same. Now, note here that network constraints are now also also explicitly taken into account in the objective. So here we have the power power flow constraints. So that then becomes conceptually similar as before except that now that network constraints are explicitly taken into account in the optimization as well. Now, as a result of having the network constraints in the objective the sub problems become coupled so that the composition I mentioned before doesn't work anymore. And so you have to use another approach and we use the ADMM the so-called alternating direction of multipliers method, which again you can think of it as a price based coordination so you also exchange prices in each iteration. There are some other technicalities you have to make, you have to satisfy in order for this to work so I won't really go into details but you have the references available in the paper if you want to learn more details. Okay, so now the way I see this is, so we start with this retailer VPP 1.0 and then sorry 0.0 and then we move to an optimization base which is VPP 1.0 and you can take into account also customer objective. You take into account also network constraints and then you end up with this VPP 3.0 which is fully network aware so network constraints that are taken into account explicitly which is important in this possible future where half of the power comes from behind the meter resources and customers are also included in the objective. So we actually trial that technology in a recently completed trial on Bruni island. It was funded by Australian renewable energy agency. There were three universities, Australian National University, University of Tasmania and us, network service provider does networks and technology provider to deposit power. So basically what we did, we coordinated 32 batteries while taking into account both the demands of the network as well as objective customers to manage network constraint on a constrained part of the network from Bruni island in Tasmania. Okay, now so finally peer to peer energy trading. Now this is conceptually quite, quite different. So the idea here is and this is similar to on this borrows concepts from shared economy like Uber or RB and B, Airbnb, where in principle, every consumer can trade power with any other consumer in the network. And this is different from from the existing model where you have to trade through a central central pool. Now in the context of low voltage electricity markets or in the context of distribution of distributed energy resources, you can only do so on a single low voltage network. So peer to peer is not possible if you have consumers located in different parts of the network, because if you were to send power to the other part of the network the losses will simply be prohibitive. So you are restricted to doing that in a single low, low voltage network. So that's conceptually how this peer to peer trading looks like there are several approaches. I just mentioned one for the sake of completeness completeness which is multi bilateral economic dispatch, which is similar to economic dispatch in a wholesale market except that now you trade power between consumers, and you represent the welfare of sellers with a utility function welfare of buyers with another utility function which appeared in the objective. So I won't go into into details. Instead, I will focus on, I will focus on auctions, which we did quite a bit of researching. So, some preliminary concept so if you you can think of you can think of a peer to peer market that consisting of two sets the first is the set of sellers and then you have then you have the buyers. This is a four tuple where you have a buyer, seller, transaction price and section quantity or the power you sell, where buyers utility is modeled well with this utility function and the seller has another utility function. So this borrows concepts from from micro from micro economics. So, and the way to realize this market is to use an auction of some some some some type. So the first option we used is a continuous double auction, which consists of multiple buyers and sellers so this is typical similar to eBay. And this auction requires an auctioneer so it's like a third party that that that clear transactions that clear transactions between buyers and seller. This can be supported by a distributed ledger like blockchain so that you can get a completely decentralized decentralized market place. Now of course you can ask you can ask yourself a question so how do you come up with an optimal beating strategy for for an end customer. Now you have to understand that this is a thin market so the number of players in the market is not huge. So which means that finding an optimal beating strategies rather difficult. So that's why we use this concept of zero intelligent class traders. So where you where you have some. I'm conscious of the time so I'm read this up over in five minutes. Where you have some automated mechanism that adaptively adjust the beating prices in from for the buyers and seller. Okay, so I will skip the details I mean this is quite, quite dense so if you're interested, there are papers. So this overview paper plus the references containing that paper you can look for more details, details there. I will skip all of that so just maybe just one thing I would mention is that peer to peer energy trading is probably easier to implement because it doesn't require the central optimization it can be, it can be bolted on on the existing market model. And the net the impact on network constraints can also be taken into account using this network permission structure which is conceptually similar to the operating envelopes. I mentioned before so we have a distributed distribution system operator that constantly monitors and then prevents buyers or seller doing something that will, that will violate network constraints. Okay, so I will conclude this with a comparative analysis. We use, we use a distribution network model low voltage distribution network model consisting of 100 customers 30 customers didn't have any DR so these are the. The green ones in black, the green ones at TV and battery systems where whereas the yellow ones. TV only, and the prices we use again is the time of use which is this one in red flat there if in blue and dark blue is the feeding. Um, so before that we tested. So we increase the penetration of TV and we found out that as you exceed three kilowatts per customer then you get network problems, network problems. Or voltage voltage problems. So this line is, these are the voltages at each note in the system as you increase the penetration. So this year so what it means is that you pretty much have to limit, you have to put a kilo on the export for the export of the customer in order to prevent voltage violations in the network. Okay, so. So here I'm showing network congestion or transformer capacity. So maybe this is a fairly, fairly busy slide so I won't really have the time to unpack all of that. So maybe just one thing to highlight is when you do this home energy management with complete disregard of the network, you can see that in the middle of the day you get power flows that exceed the transformer capacity, but this is obviously avoided in opf time approaches where you where you explicitly take into account network constraints and peer to peer with network permission structure is also effective in reducing, reducing network problems. So the comparison here is shown for flat areas and time of. Now for voltages, similar story, you can see that in the middle of the day so these here is time. That's customer index, you can see that in the middle of the day. You get voltage problems which is indicated by this dark red color. So obviously for home energy management with the complete disregard of the network, you get voltage problems body opf time approaches. Now finally some economic conclusions. So here I have total energy exported by users and for platen and time of use. So maybe one thing to notice that peer to peer trading. Significantly limits the amount of power users and exporting the network and this comes from the fact that you are only allowed to sell power into the week if you find the buyer, which doesn't happen. Always, which quietly means the amount of power you can take. Here I cash flow comparisons. One thing to notice that this peer to peer with network permission structure, actually, actually ends up ends up making money for the customer so the amount of money you pay to buy electricity is less than what you get when you sell electricity to the market, and that's simply because you can sell power at a price that is higher than the time of use 30. But note here that also the customers who buy power, they do it for a price that is less than the applicable time of use eventually eventually everyone benefit. So here we ranked customers. So this here was cumulative, we bundled all the customers together. Here we ranked customers, all 100 of them based on the performance and we grouped those approaches based on statistical similarity. So for energy exported to consider peer to peer and home energy management is the least beneficial. And the other ones performed fairly fairly similarly so this is for energy exported and the net cash flow. Okay, so some conclusions. And the main conclusion is that the coordination can take many shapes and forms. And also very importantly, active network management will be required moving into the future as the penetration of these devices increases. The finalization based approaches if you do the do this in a distributed fashion that will require frequent communication between between agents, which might impose a significant burden on the communications network and the open question is how to reward consumers for offering services into the network. We need to use dual variables of the power balance constraint which is a locational marginal prices, although that has some problems. So this question is still not not not answered. Approaches are easier to implement, compared to optimization based approaches because as I said they can simply bolt it on onto the existing market framework. And P view curtailment which is implicit in opf based approaches or even network, if you use a network permission structure depends on electrical distance so can be unfair. And one slide from our recently published paper where to show that if you just curtail customers, the customers at the end of the feeder will be curtailed more simply because of the electrical distance and we came up with several approaches with that do that in a more fair and equitable fashion. Maybe finally the final slide is that this aggregation will probably have to be done in layers. So, first level of application on the low voltage network then medium voltage network and finally taking the home network, because simply doing a full end to end market would be would be simply to complex. So, that was all from me so we have few minutes left for questions so please go ahead if you if you have any questions. As I said that presentation was quite dense, but there were pointers to other papers if you're interested in more in more details you can refer to those papers and find more information there. While we're waiting for the audience to type their questions I have one question related to some of your last slide. When you consider the impact from the on the networks from consumers, how big of the networks you consider this is one example I think you have you consider one substation, I guess. Yeah. The low voltage networks in Australia are around 100 a few hundred customers so typically low voltage networks in Europe and Australia are a bit different to what you guys have in North America. So the number of customers in one low voltage networks can be a few hundreds for example between 0.5 megawatt and maybe 1.5 megawatts. Okay, but in a medium medium voltage medium voltage networks can be possibly up to 50 megawatts so maybe up to a sense of low voltage networks in one so that could be a medium voltage would be that so probably several tens of low voltage networks, several thousands of customers. To understand the actual impact, do I really have to consider several thousand you know at that level. I think this problem is like in transportation right if something happens locally, you don't know how the type of impact in the upstream and how far in the upstream so you will consider. This is still this is still being worked out so as I said, currently, network companies in Australia use simply blanket constraints or in congestive network so you can export two kilowatts a bit. But that obviously is not economic it's not optimal because sometimes the network can take more. And there are other opportunities to use that power and using it to peer approaches and so on but this is so but there are other approaches are emerging so the one we demonstrated on Brunia Island is one approach. Network companies here are working on operating envelopes that dynamically adjust the export limit so as to make better use of the available power. Let's go to the question. Yeah, it's currently rejected. Yeah, that's an excellent question so electric vehicles I think are a perfect dancing partner for for distributed energy risk energy generation. Unfortunately in Australia, this hasn't been recognized yet by the federal government so our federal government is not very progressive and they're not not actively helping with the uptake of electric vehicles, but I guess moving into the future to make good use of all that generation that is available in the middle of the day charging electric vehicles basically for free. It's seems to be a no brainer. The big question will be how to come up with business models because suddenly you have integration of several sectors that were currently that currently live in silos, you will have to aggregate them so to probably. So, now you buy electricity, which is one service and transportation is another service. All of that will then in the future probably become part of an integrated energy service or you might have a retailer might also offer you an electric car and rooftop solar and a battery on top of your power connection. Maybe there's a solution but definitely electric vehicles are a perfect dancing partner to solve this problem, while also maximizing the benefits. Now that some studies show that electric is actually used by direction. Yeah, so by directional charging so then again this is, of course, the more you cycle the battery the more it degrades. So, by directional charging or this vehicle to home or vehicle to grid concept, obviously has some downsides, batteries, battery degradation will be affected. My answer to that would be at the end of the day, if the value of all those additional value streams exceeds the cost the additional cost then this will be done. So if you can make I mean if you have to if you have to replace your car battery in seven years as opposed to 10 years but you make more money over those seven years I guess you will be happy. You will have to do that so again the question is we have to come up with business models that will that will allow that where you might anticipate the next best and most most likely power consumption areas of the application. And this is already happening in Australia, low voltage networks, as I say are saturated with TV so you have too much PV already. So network companies are now are now looking to network companies by network companies I mean operators or flow voltage networks they are the most affected because of course network limits are violated so they are now at the bleeding and they are now trying to come up with solutions as I said blanket constraints are the most the most blunt instruments they are now coming up with more dynamic operating envelopes optimal power flow approach might be another solution. I guess when additional value streams become available. I mentioned this demand response mechanism so when it becomes possible for end customers to offer services for offer services to the wholesale market. So when network companies start using behind the meter distributed energy resources as a non network solution so that's the term they use if you want to deal with network congestion you can use. You can use it in people's homes when all of those services become possible additional value streams will become available then retailers will become part of the story. So currently network service providers don't typically talk to retailers so they are still to separate places, even though their models and the impact they have are highly intertwined so this hasn't quite happened yet but we are moving into that direction. He would be responsible for maintaining the resilience of the power supply. Okay now resilience. I'm not sure if I correctly correctly understand the meaning of the term resilience is typically associated with resilience of power networks to network, sorry to climate disasters. So that would be the responsibility of well the Australian energy market operator they set the target for reliability and also resilience is not currently explicitly taken into account. It is implicitly in the operating procedures. One comment to increase resilience of the grid. If you can operate parts of the network when the network goes down someone that like a bike transmission goes down. If you can operate low voltage networks. In an island it's fashion so you disconnect low voltage networks from the from the main network and you operate them as micro it's in an isolated fashion that would of course increase the resilience of the grid. I think in California it happened that where you have the wildfire danger you disconnected customers from from the grid in order to prevent further wildfires. Obviously not, not a good solution. And also when when, as I said when the network goes down if you can still operate parts of the network supported by distributed energy resources located in that network that obviously would decrease the resilience but we I haven't seen any any business models in that space yet. Have peer to peer energy trading models reached pilot test. Yes, there was a test in in in Western Australia. What they didn't consider they didn't consider the impact on the network. There are currently, there are currently many trials going on in Australia. If you're interested arena Australian renewable energy agency. I had their logo on the slide with this Brunia Island project. If you're interested Google arena and go to the list of projects they currently support many projects in this space. Obviously the optimization by solutions. I mentioned in this slide you can think of that as a gold standard, which requires might be considered to complex. Network companies and retailers and are now coming up with solutions that you can think of that are try to approximate that solution try to come up with something that achieves a similar objective but we let less complexity. So there are many trials currently going on in Australia and also in the paper I mentioned on the second slide if you go at the end we have a list of tries already also included in that paper so that would be maybe a good starting point for you to read. There have been studies for a practical and optimal mix between firm power supply and the eyes. This is another excellent question. Firming, firming is now a big discussion point in Australia. For those of you who are not familiar with the concept of firming basically means that you have variable renewables, and you have to, you have to ensure that you have sufficient supply. Which you can so essentially you are filling the gaps when the wind is not blowing can the sun is not shining curve can fill this gap with batteries gas fired power generation is another option. Distributed energy resources can be can be also a very good solution. And as I said, the main issue at the moment is that it's still quite I mean this is every any solution involving and customers is quite complex, simply because of the sheer number of them, but also the business models currently I mean will spend several sectors of electricity supply chain retailers network operators possibly the system operator and so it can be it can be from the business solely from the business perspective. It can be quite complex so we haven't quite I mean we are not. We are not there yet I guess it will move in that direction. I guess eventually we'll find the sweet spot between the managerial complexity and technical technical and economic benefits but yes to answer your question distributed energy resources. I see them as a as a main ingredient of a of a future power system where we might, especially given the fact that we might get half of the power from from from these devices so we'll have to take them into account and we'll have to make to make this. This happened. Um, Okay, thank you very much. Thanks for answering all the questions. Yeah, thank you for giving us a wonderful presentation. So, we'll take the minutes after the house so if there are no more questions will end the presentation here. And thank you very much again. Thank you very much. Thank you for having me. Okay. Thank you. Bye bye.