In our blog of Nov 3rd, we focused on the traded prices for Week 45 (Week Ahead), which had reached an all-time high of 253 €/MWh (Base price). The weather outlook was at that time quite bullish, with Peak load consumption in the first of the week increasing nearly 10 GW compared to the end of week 44. Spot-prices Thursday/Friday week 44 came out at about 70.00/85.00 €/MWh (Daily Base/Peak) and meant an additional bullish signal to the market.
Easter is not only an important religious holiday and a traditional vacation season in Europe. It is also one of the periods of the year during which the power consumption is strongly affected by the Holiday Effect. The social patterns are different from country to country and reflect a variety of habits and traditions. For us the upcoming Easter is also a moment to reconsider assumptions made within our consumption system and represented by the reduction of power demand attributable to different behavioural patterns. You find some statistics below. Enjoy!
The highest total and relative impact
The highest absolute reduction of consumption due to the Holiday Effect can be observed in Germany, where from 24 March (Maundy Thursday) until 28 March (Easter Monday) 1048 GWh less electricity will be consumed in total. This number contributes to an average of 43,7 GWh/h and 15,1% of normal German consumption during this period. In the “competition” of total consumption reduction Germany is followed by Spain (total reduction of 461 GWh) and Italy (303 GWh). In terms of relative reduction, the second biggest effect is observable in Spain (13,1% of normal) and the third biggest in Denmark (11,9% of normal consumption).
This week we received information that led us to revise the German consumption for 2014 up 6 TWh. This will shift the synthetic data (current MKonline actuals) and simulated normals.
Note that it will also shift the current forecast levels.
We hope to have completed the shifts by tomorrow morning.
We will get back to you on this later.
Cause and effect from adjusted 14 numbers
The monthly consumption data from ENTSO-E for the period Jan-14 until April-15 has been updated.
The data shows an increase of about 6 TWh for 2014 compared to the previous MKonline-data. The graph depicts old and new raw data from ENTSO. Further, it shows the previous MKonline data, which correspond to our synthetic numbers, as they were prior to the change.
The previous MKonline data have been scaled up for H1-2014 as the previous ENTSO-data seemed to be too low, which has reduced the gap to the new ENTSO-data for 2014.
We saw also for 2013 a huge shift in the revised ENTSO-data, and we will now check the new ENTSO-numbers .
MKonline is furthermore working with updating the future consumption growth until 2020 based on the latest ENTSO-E data.
We will keep you informed as soon as the new updates have been finalised.
Head of Continental Analysis
Please be advised that we have re-estimated the consumption models for France, Belgium and the Netherlands.
What changed with the new model?
The model work indicates that the consumption dependency on weather has increased, partly considerably. Most important for power demand derived from the need for heating and cooling. See illustration for Belgium beneath. However, the combined effect from low temperatures and wind – we have coined it WindChill effect -has also increased.
Increase cooling capacity has a distinct tendency to develop stepwise. The high temperatures experienced during 2014, together with improved economy, might have added to both installed capacity, and willingness to run the already installed cooling units.
Power demand derived from the need for artificial light has declined some, compared to the old models. Although not scientifically tested, the latter is in line with what we would expect from the introduction of power saving light bulbs seen over the last few years in Europe. Introduction of power saving light bulbs might also help explain higher sensitivity to lower temperatures.
With the new models we also expect better fit for social pattern in consumption, e.g. vacation and holidays, and bridging. Note that the social pattern tends to change gradually, or occasionally abrupt, following legislative shifts.
As expected, all models and its parameters improved significantly, although we enlarged the training period. Beside better methodology and model formulation, an important reason , we believe, is that we upgraded our data monitoring system for consumption in January 2015. Better consumption numbers, with less statistical noise, now spills over into reduced error margins in the estimated models.
The analyst team look forward to upgrading the consumption models for other price areas too.
Feedback is appreciated
The June edition of MKonline’s Long Term Price Forecast for Nordic Power is set for release on Tuesday 16 June.
Major topics in the June update, having implications for the Nord Pool area:
- Swedish nuclear reactors expected to be decommissioned earlier than previously planned
- Further analysis of the UK long term power balance and new assessments of power flows from the Nordic countries to UK
- Updated perspectives on coal prices
The Track Record pages for the fundamental variables (wind production, photovoltaic production and consumption) can be found in the menu-tab of the respective variable. In this blog post we explain the different graphs and tables on the page. As an example we use wind power production for Denmark.
The Track Record Pages give information of the performance of today’s model, not the historical performance. At times we update the fundamental models used on MKonline and do a full recalculation of the historical prognosis. Hence, the prognosis used as input for the statistics on this page might deviate from the prognosis published in the past.
On Friday 23 January, we aim to update our historic data and implement a new consumption model in Germany.
Power load in Germany is an elusive variable . Not because one lacks sources but because of poor updates and internally inconsistent sources. This could be e.g. hourly numbers not summing up to aggregates over months or years, or summing up individual TSOs to national numbers. Another known issue is that they do not meet simple econometrical criteria one would expect from such data, e.g. sign of derivatives etc.
The consumption model upgrade is completed.
Please be advised that we have started to upgrade the consumption numbers and rerun our forecast models and re-simulate the normals. It will take some hours before all forecasts are reissued. For background see this blog.
The table beneath shows the status and progress. Click on the links to see the discrepancy between old and new numbers or be directed to the forecast for the area of interest.
|Diff. old and new data||Progress||Comment|
|NordPool||OK after EC12 21 January||We need some days to bring these pages in sync:
|CWE||Expect to issue new data and upgrade the consumption model Friday 23 January after spot nominations|
|CEE||OK after EC12 22 January|
|IT||OK after EC12 22 January|
|SEE||OK after EC12 22 January|
|IB||OK after EC12 22 January|
|BLT||OK after EC12 21 January|
|UK||OK after EC12 22 January|
Further, we need some days to bring these pages in sync:
- Normal assumptions
Feedback to firstname.lastname@example.org would be appreciated.
22 January 2015: Started to update CEE, SEE, IT, IB, and UK. If this works flawless, we continue with CWE tomorrow. We will keep you up-to-date on our blog and via tickers on MKonline.
21 January 2015: BLT and NP are updated.
20 January 2015: We made some revisions to the new monthly levels in LT and DE.
20-24 January 2015: Only hourly gaps in historic feed remain to be closed.
19 January 2015: All areas except some zones in IT and BU are mended until 14 September. Note, after we implement the new data, we rerun the model and bridge remaining gaps as the one of 14 September until “yesterday with our model generated synthetic consumption numbers, AKA backcasted consumption.
18 January 15: Unfortunately we have gaps in some of the test data files, MK02. We expect to have fixed this by 20 January.
Following customers requests, we have started to publish all 51 ensemble runs in the Data Center.
At this round we added 4 variables: Temperature, Consumption, Photovoltaic, and Residual Load (Con-Wnd-SPV). Later we will add many more ensemble forecasts, such as Precipitation Energy, Price, Exchange, Hydrology and many more.
How to find data?
MKonline covers all of Europe (except Iceland and Ireland) in all price areas/areas 57. How can you find the data?
Simply go to the Data Center, place your cursor in the search field (top left corner) and type in the name of the price area/country and variable you are looking for. The system filters it out as you write.
Let us assume you search for Ensemble Switzerland. 2-3 letters in each word would filter out the files for you. We have also included some predefined filters, see on the left side of the Data Center.
Update and Structure of the files
The files follow the updated schedule for the ECMWF forecast and contain the two latest ensemble forecasts. The file will rotate in order that the first 51 columns always contain the most recent ensemble forecast, whether this is EC00 or EC12.
Each file contains only one variable for one price area, as indicated in the Ensemble Switzerland example.
In order to avoid any confusion regarding the data’s freshness, there is a row above the column names containing the issue date of the data.