Energy Markets and Quantitative Methods Padova, 17.Oct. 19 hugo@energyquantified.com +47 9187 7970
Disclaimer “The opinions expressed in this presentation and on the following slides are solely those of the presenter and not necessarily those of Energy Quantified (EQ). EQ does not guarantee the accuracy or reliability of the information provided herein.”
Content • What to expect from the future RES data and analysis vendors, and how it could provide an opening for closer integration of academia into the business • Some of the development • What is it foreshadowing? • Integration of academia into the business
What are we doing: Weather2Price Systemic Systemic connecti connecti on on 1. Regimes: regulation vs Market based 1. Climate: Heating, Cooling, Windchill distribution of production… 2. Installed capacity: Wind SPV, 2. Capacity constraints: and Implicit, 3. Historical prices / embedded behavior: Explicit, FB, Exogenous… heating system industry structure 3. Market behavior: Strategic behavior / 4. Social pattern: E.g. workhours and Holidays gaming 5. Observed weather: (Kalman filtering)
Data for the full picture Everything you need to be in power
RES development At the outset: “It can’t go on”…. • Wind power • SPV • Small Run-of-River • CHP etc etc Concerns, among many: • Too expensive • Hard to regulate / balance My take: It will most probably continue to grow. • Political support • Commercially viable
RES development “a landslide of changes” • Trading moving from “Cal 2” to next minute • PPA has taken the role trading previously had. • Olga and Oleg is coming. (our analyst persona) • Two PhDs, in math and programming, • “Do not tell me the price, I will find the solution.” • Give me input, the starting point. How confident am I? • Left a “secure, well payed job”. Started all over. • We follow the plan, involving: Provide the best possible high-resolution data.(and price) One year after launch, Sep. 2018, EQ provides hundreds of users their daily data requirement.
Integration to academia “Making the full circle” I stared career in Statistics Norway. Its purpose is to: • Provide data and foundation for research • Public and private companies (Commerce) Then went on to the vendor industry as of 1999. (EQ’s siblings) Today the vendors (at least EQ) • Commands a growing “surplus” of data • We can never use our systems and data to it’s fullest potential It is an opportunity to corporate. The stage is set • Vendors have matured. Academia may be ready too. • Start simple: Input for students thesis. (EQ does this) • For larger projects: Find questions that could/should be answered, and provide data and infrastructure that is otherwise behind commercial walls.
Some examples on what vendors can provide: “Economics of scale makes the difference” • Data, better than actuals. • Enable you to study intra-hour effects. • Infrastructure to manage large data amounts from a laptop
Actual data is not good enough Why consider using synthetic 1/3 Notoriously Available only erroneous 72% – 98% • “No data” is metered. It is calculated by TSO or other bodies • Some make a good job. Others make a poor job. • E.g using identities/definitions: Consumption = Production + Import – Export Consumption, Consumption in same date SE1 SE = SE1 + SE2 + SE3 + SE4
Consider using synthetic 2/3 Wind and Solar may be Solar embedded in NL embedded or just “off” consumption Wind power i n NL is just way too low.
Consider using synthetic 3/3 • Data from “all countries” MORE OR LESS show these and other kinds of idiosyncrasies. • Measurement error/calculation errors • Definitions that varies • Various time steps and step definition: e.g. measured at the start, the end or as an average of the last hour? • Geo-location/areas may be omitted or partly overlapping As a vendor EQ removes these problems. Bringing order into chaos. One common definition for all areas, for each variable, One common time step (15 minutes). • Historical data prone to revisions • Same for EQ’s Backcasts and Synthetic
Moving into the hour Alt: Why is EQ doing intra hour models EQ claims: Knowing the intra-hour profile in detail is important. • Better understanding of intra-hour markets. Intraday on 30 or 15 min, German 15 min spot market. • Improves the forecast on fundamentals on hourly level, too. • Spot price forecasts (on H level) will probably improve too. • Fundamentals are better described. • But also the price formation, EQ believes.
Get ready for 15min. Not an EQ hype. Hourly resolution, increasingly considered inadequate. PPT, taken from ENERGINET DK
Taking a look at the intra-hour variation Germany …Looking at the numbers they do look small… (Week 37_17 in Germany Zoom in on a Zoom in, and, Not so small at all. day Zoom in on hour 4 - 12 Hourly avg. on increasing path: -above 00 and 15 - undecided for 30 -below 45 and next 00 Dif H – 15min may be amplified or cancelled out by other variables
Dif = Hourly – 15Min Intra hour variation. DE. 1/2 Positive: The hourly average is above the “correct” number Residual Load. ( Con – Wind - Solar) Negative : The hourly average is above the “correct” number Residual Load Workdays. Residual load. Workdays. Jan Feb 18 Jul Aug 17 This is for DE. Max and Min, (H - 15min). MWh/h and %. Weekdays • 15 min step Dif Percent High Solar High Wind 15 30 45 00 MWh/h/ Max 847 584 1539 3028 15 min • 5-6 % may seem little. Min -545 -1097 -2779 -1690 • Within hour same hour, Percent Max 2,1 % 1,6 % 4,3 % 5,5 % Min -1,5 % -2,1 % -5,2 % -4,3 % from -5% to +5%, a spread of 10%. • Context: Consumption models run on a 1-2% error
Intra hour variation. DE. 2/2 Residual Load. The True Spread Dif 00 min Dif 30 min Conclusion: • Dif is substantial. Frequent lager than the modelling error on h level • The averages hide the variation. • Dif varies from day to day, by context, weather and season • May even shift from positive to negative. • Dif varies by context, from day to day, with • Dif is not symmetrical. Linear interpolation on h-level is not advisable. weather and time of the year. • Need to simulate Dif for all variables on a consistent Intra hour model • Even the mid. point, 30 min, exhibits large variation system in order spot the positive and negative correlation.
Conclusion. ½ Why higher resolution matters EQ observes that intra hour modeling: 1. Has a significant positive effect on fundamental models on hourly level. 2. Positive impact from splining weather parameters too. 3. EQ sees that intra-hour models improve the spot models for hourly level too. Theory, market intuition and experience suggest this. The whole “value chain”: . Weather >> Fundamentals >> Price models
How to tap into large data sources • No need to upload data • Minimum time from idea to test/model
Pan European studies made easy Geographical coverage Five flow types 1. Implicit 2. Explicit 3. Combo, Impl. Expl 4. Flowbased 5. Quasi Exogenous Frequency of update: Continuously >> Always updated price expectations. EQ evaluate price movers outside of your area too.
May study local influence too You pick and chose: • Your Stations, points (longitude-latitude) • Your Areas (many stations weighted together) EQ develops, operates and maintains : S i • S i Web pages, consistent with what you see for S i your country • Data feed in the same API and Excel integrator. S i Variables delivered: • All weather driven variables. Coming: • Seasonal Normals Climate info for your locations • All EQs alternative forecasts Lead time: • Backcasted history Historical benchmarks • In 15 minutes resolution Data search
Integration to academia “About time we get started” Making the full circle. From University of Oslo via Statistics Norway to EQ, providing data to academia. There is an opportunity to corporate and harvest mutual benefits. • Interested in elaborating on ideas? • Visit monteleq.com or write to hugo@energyquantified.com
Liked the presentation? Read more on our homepage: energyquantified.com Write me on hugo@energyquantified.com Mob: +47 91 87 79 70
Recommend
More recommend
