Send Seer Company with airSlate SignNow

Send seer company

hello everyone thanks for joining us today we're going to take a couple of minutes just for any lead stragglers to join the the presentation will be back with you shortly bye everyone thanks for joining us today for the market intelligence webinar we're going to be talking about some of our new modeling capabilities internally here with the CEO model so before we begin if during the course of this presentation you have any questions there's a little questions box on the slide there that you can ask any questions to the team here there'll be a little bit of time at the end to go through some questions any questions we we aren't able to answer in this session will come back to you these slides will be available will be sending them out tomorrow morning after the presentation and my contact details and links to the products are available with this presentation and with these slides so to introduce myself my name is rod williams i'm the okok analyst team lead so I'm responsible for the team that's creating some of the daily analysis running some of the models and talking to our clients some of some of our clients are joining us on this call today other view other members of the Corps may not be as familiar with our the ER kottpower IQ service so what I'm going to do is I'm going to walk through some of the main points of the seal model I'm then going to talk briefly about some of the specific features we built in for a cot and then we're just going to do a general refresher for some of you who may not be familiar with the power IQ service we're also really really excited to share with you some of our key metrics since since last summer we're really looking forward to sharing those with you and then at the end we'll will just summarize some of the key points and and go through some questions so getting started what is here so seriously the in-house tool that powers a lot of our forecasts with the window market intelligence group specifically for okok its fundamentals based a price and power flow model it really represents this really significant upgrade from the existing model we have essentially it's a security constrained economic dispatch model the output of which is being used in both our power IQ our power buyer models we also provide some of the big data as well that we can go into so it's a it's a network model that calculates all the prices within the earth footprint the actual outputs of those are armed on a hub and load zone level so and as you know we're utilizing a lot of proprietary data within this not only our Genscape real-time monitoring but also our in-house demand and wind forecasting and we can talk a little bit about that as we get into it so firstly let's just really kind of hammer home some of the benefits of why why we're we're presenting this to you today and why wifey is so important versus our existing model capabilities so firstly it's a really significant increase in speed generally model runs take about 120 seconds so an analyst that's fantastic for us we can get our work done quicker more time for analysis as clients the the opportunity is really there to run a lot of custom scenarios that's one of the kind of the key the key value adds of having an analyst team behind this model is that we're able to to look at specific scenarios based on client amongst oh really double examples for example what happens if the wind forecast mrs. what happens if there's a ramp of wind can you plug that into your model rerun give us some scenarios around that what happens if we take this line or this generator out of service how does that affect flows in the model what kind of constraints that that might create in the model the other advantage of seer and loading in even more data than we've initially had is the acura Civic accuracy of it so I'm going to share with you some of the metrics that we we publish externally with our with our reports as well as some of the internal metrics we use to to judge our performance so firstly I mean it can't be understated how much of a advantage we have with some of the Genscape power RT data that's going into the model we recalibrate every single monitored units and about sixty-five percent of the airport stack is monitored including closer around all the baseload generators in our we're also pulling in all available public data for example gas nominations line the gin outages from our clock we're also looking at how how units are offering in on say dad market energy sold we're also looking at the 60-day delayed ERCOT generation data that actually shows how every specific unit in the in the footprint ran on any given day we're backing that with our neural network models so our model consistently outperforms the ISO anyway added to that we have a team of in-house meteorologists who are actually tweaking those forecasts again putting their their own analysis on the raw data of the forecast and again that plant economics the big data we unmask it for other regions oh god publishes it on a delay we do do some work around the big data to make it more manageable to give us more flexibility in terms of how we bid on a supply stack and really flexibility i think is the key it's a full network model so we've we've accession replicated the entire network so we can study flows on any line within the region with with forecasting zone and hub forecasts day ahead and then again we're pulling in all that publicly available data order cod available data as much really as we can get our hands on so a lot of recalibrations based on the day ahead energy sold the capacity available the current operating plan so this is just some of the few benefits I'm going to talk a little bit about some of the key points within CSO firstly starting with the input and output so really some of the fixed and constant inputs are going to be the generating units of each each unit with another the actual network models for the busses the branches the flow gates beyond the aggregations the price aggregations whether it's our hub loads own and then the unit big data that's published by okok on a delay other regions we actually unmask that big data as well and and publish that with it we do use a number of other parameters to actually get over that 60-day delay Donna the time series the manually adjusted inputs we're looking at so firstly it's our load forecasts our meteorologists who are creating hourly demand and wind curves its fuel prices it's generation forecasts of how the analysts are actually tweaking the tweaking that the each individual unit within the model Jen and transmission allergies those actually get pulled directly into the model and we are studying line flows on specific constraints based on transmission and generation outages around there I'm some of the outputs so I mentioned Hubbard zonal prices to see in north houston south west hubbard load zone additional to see r is actually some of the non opt-in entities than noe so aen cps LCRA and Raburn load zones so that actually means that we're now forecasting every major price point innerclock in terms of the load zones and the Hobbs we're also looking at the the hourly IPC price adder the marginal unit where we're identifying a marginal unit based on the big blocks and then we're actually publishing flow and shadow prices for specifically modeled constraints within a pot so let's talk a little bit about the recalibration process so we kind of really one of the cornerstones of the CM modeling process is the recalibration with power RT so we have a large percentage of the aircon stack monitored we have sixty-five percent of all total generation the generation we tend to monitor our the the more significant unit so we have pretty much the baseline monitored a lot of the steam and combined-cycle stack really the units that we don't tend to monitor tend to be the smaller units that may not be having a huge impact on the grid or consistently offering in we also have a large amount of wind generation monitored as well in our cotton that's that's part of the recalibration process on at least the wind forecast there so it's actually an hourly recalibration this is really the first thing the analyst is doing as they when they come in in the morning we we create a recap report which is basically replicating the the previous day in terms of price constraints generation behavior any transmission outages that that's being accounted for outside that visibility that we don't have directly in the recalibration and this recalibration happens on each individual monitored unit we're also using the day ahead energy sold data as well so say for example of course Corpus Christi generators that don't have don't have monitors on them currently we actually use a day ahead energy sold data to see how typically they would have run the process here is really recalibrating the the actual output of the unit or the actual commitment of the unit for the previous day to the to match the best bid strategy the the bit behavior whether it's off a day ahead real-time or self scheduled to the actual model and then re running the model to recalibrate as closely as we can to the actual price on num on the hubs and so that's a daily process and actually when you we do our midday update we again recalibrate our models integrate new data for rum for the spice back and dumb and rerun our models on that another big part of this that our development team has worked on is the bid analysis so we have bed blocks and he rates based on historical analysis reached unit it oh god we do quite a lot of work around building in predictive parameters based on similar days so that really helps us overcome they kind of the 60 day delay that you're going to see with our con releasing the data so we can really apply the apply the most appropriate bids and he rates to our units the multiple bid strategy is actually available to the analyst in order to dispatch the unit so we can really have that a level of flexibility and 22 really adequately forecast what we think the what do you think the the unit is going to run out we also we fully decode that big data for other regions and that's something we actually make available as a separate offering so clients can use that that data in parallel in their models for example another big part of its flow gait analysis so flow gates are essentially plugin is it is really an internal term that Genscape uses here but essentially what what a flow gate is is it's every single pair of lines within the network so we've actually replicated the entire network model with a view to being able to control flows on that and see and study flows so the first instance is is we recalibrate our generate generate the stack whether it's based upon day-ahead energy sold whether it's based on previous previous day unit behavior we're going to run that based on with our demand and wind forecasts the first thing we do we study the decision of the line flows on each line so C is actually going to highlight to us multiple line constraints the previous model we had a reedus patch model that really only reduce pass on some of the major Parr say like not the Houston Valley important to create some of those congestion patterns when this it gives us each single available line it's pulling in each each line outage on a per day basis so we're able to capture new line outages branches that may not be available and visit we actually are able to dynamically change that on an hourly basis so we can to a certain extent override it and create congestion with the analyst has a has a view on whether we think let's say for example Silas rate congestion temperatures in Brownsville hitting the mid-80s we want to see that Silas rate congestion we want to create that South load zone basis the model may not be giving us as much as we think within Silas for a unit may deviate around it it's day ahead of schedule as it commonly does that's something that we're able to actually create in the model so it's not just simply a throw throw inputs into the model and get the output it's again squeaking the output with a team of analysts behind it one of the keys as well we're able to run customized scenarios around these flow gates so we say hey what happens if we take this line out what happens is this unit brings on a brings on a second part of say a a combined cycle what happens if we lose a unit trips in the real time later if you're an asset owner what are your plan to take your unit down you can ask us to rerun scenarios around hey what happens tomorrow if I decide to bring my unit offline so the high level of flexibility within the model that we can actually play with so a couple of ergot specific features that we actually built into the co model so firstly that surplus capacity calculation so big big part of our God for identifying volatile hours so using our model supply stack our wind forecast and our demand forecast we can actually take a alternate view to the ISO to that the current operating plan identify low surplus capacity hours that might be at risk to see any DC price added this is something that becomes key in the summer when you especially last August for example will show you some of our performance during last August this really becomes key in terms like 10 * identifying hours for rum potential volatility that OID seen calculation is also part of of the model so using that value of loss load the available capacity aunt Jen skates model demand wind and supply and then in addition to that we we recalibrate based on day ahead energy salts we pull that data really in a parallel to a certain extent to some of the Genscape data so not only we can see unit behavior from previous days we're also able to predict when a unit is likely to run that includes when unit potentially don't participate in the day ahead good example is some of the steam units in North that may not offering to the day ahead over the midday hours and then choose to ramp in the real time take advantage of any real time price trend so again that's that's something we can capture in our model so let's talk a little bit about how we use here so this meant some of you who already scribes subscribe to some gin escapes products and use our service may be fairly familiar to this this might be something to to really just go over for some of our listeners who might not be as familiar with Jen's King so the output of sia that the clients will see it is really a package report it's we're forecasting that by cell stratum trading strategy based on our model base cases and we'll talk about the base case and the importance of that but essentially our forecast price for the on-peak whether it's a weekday or or the seven through 22 we weekend so we're issuing those demand and wind forecast that's part of the service we're actually delivering those those those price forecasts those base case forecasts with sensitivities around them prior to the day ahead close so part of that is we're patching a report that you we're putting out a about a real time and next day real time in the next day day ahead using our model our load forecasts our wind forecast in addition to that we're within congestion analysis to identify any likely constraints and again those custom scenarios for clients the aim is really that that is available to clients prior to the dead so they can take that information and make a more informed trade around those decisions so let's just drill down some of those deliverables so on pecan weekend down in price forecasts this is really what you're going to see from the model so we sear yeah it's actually going to be on every hub every load zone every noe if that's what bout a real time the next a day ahead in the next day real time so the next day day ahead we're looking to forecasts are using our God slow forecaster thought spoon forecast before gusting the day ahead clear with the next day real time when we're using Jen escapes inputs and we actually derive a product from that that's going to look at any day ahead and real-time opportunity to realize a lower cost savings on power bi product we're going to move into that in a moment so again it's hourly price players and this is inclusive of congestion impacts we're actually going to call out the likely constraints that show up for harm for any given day and their impact on the hubs again it's using Jen escapes approaching Windham semarang forecasting so we're actually you have a team of meteorologists who are dedicated and again those those reports and analyst discussions delivered prior to the day ahead market close to enable you to make trades most information so let's just talk about that demand and wind Jen forecasting that's really a big pieces of it we we could just use the iso's data we could just use a third-party we choose to do it in-house really because having a meteorologist on staff and able to talk to the analysts and talk to the clients about some of the risks is really is really what sets us apart and allows us to to outperform say oh gosh broadcast for example so we're providing hourly adjusted demand and wind forecasts on a zonal or wind farm level part of it as I mentioned is the briefing on that forecast we're talking about the risks around the forecast that are matter able to give directional risk they're able to update their broadcast as as new and new and models come in and again that custom scenario again moves in so you can say to us hey get your math to put another gig in in the evening peak tonight based on cooling load let's see what happens let's rerun the models to that so it's really the complete come to support their from both analysts and meteorologists um what I want to show you as well as some congestion analysis so this is a page directly from our reports this is actually um there's that I think this actually might be today's report so we actually show can we do congestion analysis based on every line outage the idea is really to we do independent congestion analysis each each analysts will look at transmission outages how generation is running we'll also look at that in the context of how the seer flow game module is showing us likely transmission constraints so what we're showing you here and I apologize if that the slide is a little little small to see we're listing the constraint we're listing the arm the bullish and bearish impact on hub or load zones and this is useful for those of you who are maybe more concerned with basis risks as opposed to as opposed to actually just kind of hub impacts so any kind of price difference between hub and loads over we're also identifying the historical historic top source and sink nodes so looking at what what are the what are the likely highest and lowest nodes when this constraint appears we also look at whether this because constraint shot up in the dead the drivers and then we also call out the risk hours so we say so for example a typical typical outage for the situs rate congestion loss presence to Omaha we'll talk about the drivers of the line outages so that Laura little apartment for eid assembly central line outages we'll talk about likely how is that we expect it to appear and so this is really based on on where we see the strongest flows in the seal model as well as as the analysts call on that we're also talking about the level of risk so for example today with about 86 degrees down in the valley that could be a higher risk tomorrow when we drop about 10 to 10 degrees we think with Silas rain continuing to be committed in the day ahead we can see that that generation on power Artie we think potential that may be slightly less of a risk and maybe may have a lesser presence in the day ahead with potentially lesson in the real time as well so again all this data is packaged and sent our clients prior to day ahead clothes big part of this as well is that you have the ability to talk to an analyst about the assumptions another way that we're using the seer data is arm is in a product called power buyer now probably a lot of you are familiar with this this is this is really optimizing between a day ahead and real time buying strategy so this is the combination of our next day models on next day day ahead model which is using earth cots inputs so I God's load forecaster that got wind forecast and then our next a real time model which uses our load our wind and we're identifying the the difference is so the really the the key is identifying the lowest-cost hours for buying power or conversely the highest cost management costs hours for selling power so what we do is we actually had a risk risk adjusted biol cell signal based on whether you should buy in their head or buying certain hours in the real time as well it also we're also able to identify with that any of the basis risk so for example if you're transacting power of the hub level serving serving load on the on the load zone level where I was able to identify by looking at that congestion any kind of price separation risk that you might see so wanted to get into some of the performance metrics we use quite a few different performance metrics some of the key key to this is we do actually publish our daily performance metrics in terms of our mock P&L as well as our mean absolute percentage error for for demand forecasts there's also a couple of internal metrics we use to to just measure how we're doing I'm going to show you some metrics way back to July that shows how we've been performing so key metrics we uses is the market opportunity captured and then the savings arm in based on our power buyer compared to purely a day ahead strategy so just to go through that first even market opportunity captured so what we doing this this is really based on how much market opportunity there was on each month so for the bowel day and the next day we're looking at okay where was when the report was issued was there an opportunity to send a signal based on whether our base-case forecast was above above the offer or below the bid so if the bid offer was 20 or 25 in our base case was was $18 then we're sending a bearish signal to the market if the market actually cleared below that they'd ask that that counts as a market opportunity and whether we send that signal in the correct direction so you can see that you can see market opportunity we fluctuate a little bit during during the bowel a consistently strong market opportunity captured the percentage captured versus the overall that's on our P&L you see there so you can see then maybe the argument okay lower lower demand level the winter means that there hasn't been has much market opportunity but we're actually capturing a pretty significant amount Fergus really stands out in the it was a very volatile month we were close to the top of our God supply stack record demand levels and again there was a huge amount of market opportunity in both the the validate and the next day and really using our models new using a lot of our proprietary software we were able to to capture that capture that opportunity the next metric we also look at in the correct signal so again whether we are sending a signal above the offer below the bid and the market clears on either side of the bed offer at the time of report issue where we gave ourselves based on did we send us a signal did we send a signal in the correct direction so you can see the percentage of signals then invariably we do try always send a signal so our signals are on north of about fifty sixty percent on both the bowel day and the next day we also look at the number of times that we actually sent a correct signal so not only did we send the correct signal the day settled above above or below the the bid-ask and we were said we were sending a forecast signal in the right direction again those mid market base case forecasts aren't included if we didn't send a signal if we were say market was 20 a bit bit at 20 often 25 we sent a $23 signal or was $23 price there's no signal there we also don't count days where the market spelled within the bid-ask lastly what I wanted to show you was some of the power buyer savings so one of the one of the one of the common common things I hear is that well the market is is relatively quiet over specifically over the winter and so there's not really not much opportunity to realize any cost savings of us it's just purely sharing at a just taking a day ahead strategy so what we do is we look at power buyers hourly risk-adjusted next day guidance so we're saying what is the least cost strategy if you can take some of your loan to the real time versus a hundred percent day ahead strategy what is the what is the least cost strategy you should be doing and so based on that we look at these these performance metrics based on if you took a purely one hundred percent day ahead so every single day you just saw all your loads into the dad nothing is in real time we actually look at that versus following our guidance what would have been the percentage savings do you can see we've gone back to July 2015 through through to March twentieth so a few days ago we can definitely break down these numbers for you on a month-by-month basis if anyone needs those numbers please feel free to reach out to me but again we show pretty consistent savings and again it's really down to a lot of that proprietary data we have the the availability of the big data we have and how we're using that big data and the bit strategies as well as being able to identify both how monitor generation is running through some of our proprietary monitoring as well as arm as well as utilizing all available gen data that we have so that pretty much wraps up the presentation that I had so just just really to clarify and just go over a couple of points so this is an upgrade on an existing internal internal model process that we're able to weather we're actually running our models through biggest biggest changes really speed the accuracy and the flexibility be in the ability to use some proprietary data key to our clients use of this is being able to to have an analyst team behind it and a big big advantage of a lot of our clients use really the most value they see is when when you're able to engage me in my team on talking through some of the forecasts it often gets the point where our clients were actually where the report serves as a general summary will follow up with a phone call prior to the day ahead clothes and will actually talk through specific risks to the client another big advantage of the service is being able to walk through custom scenarios with with the analyst and actually being able to rerun those scenarios for for anything and again anything that you share with us if we don't share with the rest of the market or other other clients that's again there's a customized confidential scenarios to our clients so up on the screen is a link to market intelligence where you can find a little bit more information on some of the products my contact details are also Beth encourage you to reach out to me with any questions in my heart love to hear some feedback anything you you had more questions on we're going to take a short break now and we're just going to have a look at some of the questions that came in on but thank you everyone for joining us and we'll be back shortly okay we had a few questions just to go through and so firstly socia it's actually an acronym we don't have a dedicated kind of name for this year we're really just looking for a good model name that shows kind of forecasting and accuracy so we chose CR there was another question about getting the no e LZ price forecasts so we're going to be rolling that up out over the next couple of weeks so for some of our are no e LZ clients who are interested in doing that when you love your feedback on some of that and again any custom scenarios around that and any specific units that we can run scenarios around we're definitely able to do and then there was another question in terms of just monitoring how much how much generation do we have monitored so the actual number inner thought that we have monitored is sixty-five percent that breaks down in terms of I think it's it's basically every single base load unit in a cotton bar one I think it's a coal unit that is monitored in a cop we also monitor a large large amount of the CC steam and wind stack as well um now they are really are our approach for monitoring these units is based on a lot of the time when analysts and looking at units where we try to really identify which are the really important units to have visibility on in terms of in terms of market market variability so one other server case in point one of our most recent additions to the interface was a Silas ray constraint some of you may have seen the the blog post that my colleague Connor Danielson put out recently we just our monitoring Silas ray in the past month or so we're also going to be seeing new additions around combined cycles in the corpus christi area shortly as well as additional wind farms we're identifying really units that are going to drive the most price or or need the most visibility so those units tend to be some of the the largest steam units combined cycles in north and south for example a lot of wind in West some of those key press wind generators that we see there and then again where we're able to really focus on monitoring based on on on what clients need to see we one of the advantages of having analysts desk here is that we're able to actually see what really are the important units to have as well so I think that pretty much wraps it up for questions for today again if you do have any other follow-up questions and comments any feedback there will be a survey at the end of of this for you to fill out we'd really appreciate your feedback and thanks for joining us today I was Rob Williams I'm the team lead and we look forward to hearing from you soon thanks bye