welcome mean i'm really pleased to be here i also sort of a a an economist a member group of engineers and i understand physicist um that i've been working closely with a a a group of cornell that sort of dominated by uh and is uh uh and lang tong is here bob thomas is really are sort of leader who's not here the moment um i one the paper i wrote was really about to why is it that electricity markets a soap or Q rear and and just on like any of the market that i of a looked at that uh and uh i i've got a couple of P here that i'll mention but i really want to spend more time on what i think is going to change in the future um because i think that the way that we've model D my and up until now with completely inadequate for the sorts of things that we going to need to do in the future and in particular that we're going to rely more on price feedback to get a response from the on that really is is pretty a minimal at the moment in the way that systems have evolved over time uh so one of the one of the interesting things that that happened off to markets would deregulated in the U is the prices were so that had never been seen before that all of a sudden prices is were sometimes almost two orders of magnitude bigger than any price is that it that been seen before so you can see you know a you little little little little prices and then talk deregulation regulation yeah and and and we have price is over a thousand dollars where typically there less than a hundred um the strategy in in the east part of the U S is that um but the but this is a bad behaviour and we need to suppress it so they have employ a vast quantities of market monitor to slap people's hands when they misbehave and so you can see that the century no that initial exuberant of of the part of supply as and their ability to speculated and they're get high prices i has really been suppressed if you go to the australian market on the other hand you C Ds price spikes have continued and the really to types of markets one is highly highly monitored markets and the other one energy only markets where you're let these price spikes six this and so an interesting question is why is it that this marketing encourages speculation and this is a a a a a an analysis that we did with computer agents and basically these are identical agents and on the left the agents do not speculate i'm on the right they do speculate and the real issue it is that in an electricity market there's a lot of uncertainty uh a about what exactly is going to happen in the next ten minutes lead a known a day ahead so this uncertainty about how much the system operator is going to buy a essentially makes it possible for some absolutely outrageous speculative off as to be accepted sometimes and to set the market price um so this is a a you know a an interesting feature of of uh market so we model this type of behaviour with the regime switching essentially sensually low price high price for Z and and make the probability of switching uh dependent on on system conditions and i in in particular the expected reserve margin so we were interested in can you predict price bites the day ahead and you know the bottom line is if the or information is good enough yes you can um but when price spikes when away way you know we had to move to other things so the next topic is really a spatial price differentiation so this is a map of new york am for those of you are for not familiar with new york most of the people live down here in new york city and this is niagara falls and basically that's the main feature of the electric system the people live down in one corner and the chief power up the other corner uh on this C here stands for cornell um and uh we are about you know in the middle of the state but the logic of the system is that you you we'll power from the inexpensive expensive west down to the south the east corner there are those a lot of congestion that peak periods so basically the market friend man and you get substantially higher prices in new york city then the price at the same time uh in upstate new york so that that that uh mechanisms what put in to a our generative as to hedge the price variability between two locations and as sensually uh that's what that's what we're talking about so it it's very easy to model uh energy prices as a recursive system and basically um you model uh load is of a a a a a is a sense you model temperature are and and time put your proposed i the sort of source of uncertainty about what's gonna happen next some so i want to hedge prices for the are coming some or all of the upcoming winter um between not i fools and new york city so we model temperature your those locations we model the load as a function of time but your and that we and model the price a a different locations as a function of the load and the the price of natural gas to sort of a price of an input but basically this work or stiff structure um is is a while i should say appropriate for the in this story most customers don't actually see the true market price that just paying regulating prices so essentially those there's no price feedback there's is a straightforward forward um V i ar model vector autoregressive regressive model and we just simulated different uh us mos and then looked at the price differences and you could get a uh but a distribution of the payouts from owning one of these uh for contract so you're contract thing for the price difference as you put your money on the table and you're buying an on so it an income re so there's essentially century it's the simulated density for the house hey out and uh the system operators were very can so and that they kept on paying out more money in real than than money they talk in from running this distortion and they eight they were suspicious that a might be a collective behavior in this market but in the analysis that we did but this is a typical result that the sure pay a a a a the actual price with where the above of the mean of the payouts all the was no the dense that we could find of a sort of big risk premium in this market and therefore the system operated was not interested in this analysis so didn't sort of support their prize so uh see her we do not when doing so now let's that's got on the importance of subject um the future joe smart rate so where assuming and this you smart grid that we going to switch from fossil fuels to renewable sources and certainly your is leading the way and the us to sort of dragging its feet to get into this new era but this is basically what we have to deal with oops not that more when generation we got have some storage capacity to deal with the fact that that uh we in this not a typical a dispatch able source of supply as a lot of uncertainty and variability even if you've got lots of different wind far so you're displacing fossil fuels wholesale prices go down but this means that the warnings the money above cost that conventional generate get is also going down so that those are growing tension which we the what we've called financial at a course C all the existing generate is and um the the the uh they get in the market so uh in in the new in the uh us markets we we have established uh capacity payments outside the wholesale market to try to supplement the uh learning self generate so essentially century um lower uh a learnings in the wholesale market means higher uh payments in the capacity market a higher price for capacity and more reason for managing the system peak managing to monte in a more uh i it's a russian way we we tend to treat it the demand on as a as a given as something it you and and those days are basically over so customers and not getting the correct economic signals that a lot of them paying regulated prices is this is silly um that that that we need to we need to get a team and participating in this new market in of full way and this doesn't mean just buying when prices a low uh it means to shifting the mine from P periods to one peak period and a little so i think more importantly i lane and celery is that don't actually a is this yet into the mark and ramping service is is the one that will focus on so one of the ways of doing this is with controllable de monte and that doesn't have to B delivered instantaneously like these like so so what electric vehicles as a good example "'cause" charging the batteries in electric vehicles but it's not not enough a an now for a little things around the at to make much difference so we really need a know the source and and the U S with the system that's really dominated by the air conditioning a that's sets the conditions for a systematic C almost or is an obvious alternative to using um air conditioning on the on so basically making a nice when the system finds it convenient and then melting the ice to keep color when you want to keep cool is a smarter way of doing things and just banging on the air condition when you want cool your and and we we what for at how much of the system load is is uh temperature sensitive and in new york city you know this gives you an idea it's about two thousand megawatts out of twelve the potentially is is uh champ should it is temperature sensitive and could be control hot hot water is obviously another example of these sorts of or where you can control them so when when one starts incorporating a control of all and into the system this is really the same as as any type of storage so what we've got here is what customers want to buy the blue line is what customers by now of a way in generation so the blue is really what the conventional generate is have to supply and you can say wind was blowing pretty smoothly and and that that that that that uh we we get some some variability that's in how and with this source of generation so we propose the that that in addition to and a G the or to be a ramp so this where you're was actually hey if you a part of the majority you're moving the same why that the system is moving but if you could move against that you get paid in in in a way if you can mitigate the changes in what conventional generators as a doing you get paid so once you incorporate that over on this side you can see that the the variability of the generation from conventional sources is essentially smoothed out oops so the same thing is true of the energy prices that the energy prices get smooth out by ramping and the corresponding rand being price is a very interesting but the they but the real cost of ramping the the these prices a a very very variable because of winn so you can see you when the wind was well behaved there ramping prices a pretty mode but when the wind with variable you got a lot of variability and the then the marginal cost of ramping and that when you incorporate the ramp cost in the optimisation you were essentially smooth out the ramp the really implication is that and that that this is the sort of a uh and say laurie of is that i think we need to allow but the on side to benefit from that that we have customers who were complaining about paying higher cost so the smart grid i i think the way to make this economically viable is that rather than looking at uh customers as has a sink by energy that we ought to look at cost or where aggregates as potential sources of services to support the great and that's basically what these are concluding remarks say um customers can lower one net payment by having control able uh them on purchasing more energy a night if you like mitigating price spikes and i and and reducing them on your system P periods essentially using the amount of conventional capacity needed for at a course C and finally selling ramping services to mitigate wind variability but if they can get these payments the net payments for providing the sort the so this is that they want to get from electricity will be lower and i think that this is the way for what i clearly there's be back is something that is not really integrated fully into the current grade and this is the challenge for all you people in communication thank you i think we had time are um lots of people work on this project the plotted okay a a i oh i a a a oh a a oh a a a a a a a a i i a yeah a i i a a a i a oh a a a oh oh a a thank you