Forecast Competition

Announcement:

  • EEM 2017 Forecast Competition has ended. Thank you for your participation! Final ranking and first short analysis are available here.

Forecast Competition Announcement

In the run up to EEM 2017, the EEM 17 Organizing Committee is launching a wind power forecast competition and is excited to have you participate.

Why?

Forecasting weather-dependent electricity feed-in is of crucial importance for electricity trading as well as a stabile system operation. So far, this topic has not been addressed enough in research and industry circles. Additionally, with growing shares of intermittent renewables in Europe, this topic is growing in importance. Hence, we want to attract attention to this crucial topic by offering a competition that aims at stimulating interest in devising creative methodological tools to tackle this challenge.

What?

In cooperation with our technical sponsor, we will provide you with a set of different weather input factors, e.g. wind direction, with which you are to forecast the power generation of a wind power plant portfolio. You may participate individually or as a team. The data input is organised in a realistic setting. 

What’s in it for me?

The winner of the competition:

  • gains notoriety
  • receives (either team or individual participant) a monetary prize of 500 €
  • is invited to dine with keynote speakers at an exclusive dinner on the evening of the 7th June (Tue.)

The top 3 finishers:

  • are invited to present their forecast models at the EEM17
  • take part in a special guided city tour of Dresden
  • will receive a regional quality product

How does the challenge work?

You are asked to forecast the quarter-hourly wind power generation of a wind power plant portfolio for the following 38 hours (152 quarter-hour values) on a daily rolling basis for the competition period of 14 days.

The competition period is to take place from 3rd to 16th of April 2017. During this period you must submit your 38 hour forecast every day by 10.00 am (UTC+2). That means you are asked to forecast the current day (D) from 10.00 – 24.00 and the entire following day (D+1).

To develop, test and calibrate your model, historical data will be provided to you. During the competition period, you will receive a data update on a daily rolling basis.

Timeline

timeline_highres

Start Example

To facilitate a better understanding of the concept of the competition, the following example presented for the first two days illustrates the detailed daily data provision and forecast submission procedure. In total, you are asked to submit 14 forecasts, the last one on the 16th of April 2017. Each submission contains 152 data values which represent the quarter-hourly generation of the wind power plant portfolio in MW for the rest of the current day (D) and the entire following day (D+1). Hence, the first value of every submission is an average value for the quarter-hour between 10.00 – 10.15. You will receive real world historic data on a daily basis, but not exactly period specific to the competition (i.e. not real data from 02.04.2017, 03.04.2017, etc.). However, seasonal effects are considered. These daily realworld data are provided to you at 08.00 am on day D, covering the last 24 hours between 08.00 (D-1) and 08.00 (D). The data set consists of meteorological as well as power generation time series as described in the next section and is subsequent to the training data set.

 

start-example_highres

Data Overview

You will be provided with two types of data. The first one contains 4 meteorological input factors in a three-hours-resolution. These include the following:

  • wind speed u
  • wind speed v (rectangular to wind speed u)
  • temperature
  • global radiation

These input factors are depending on the „height above the ground“ and the location. In total you need to consider 10 locations. The following table describes an example for the meteorological data that will be provided to you. Once again, please keep in mind that these are real world historic data on a daily basis, but not exactly period specific to the competition (i.e. not real data from 02.04.2017, 03.04.2017, etc.).

 

generation-data-example_highres

 

You are free to consider as many of these input factors as you wish, i.e. you do not have to include all of them.

The second table represents the related quarter-hourly generation of the wind power plant portfolio in MW as the following table illustrates.

 

capacity-timeseries_highres

 

You can find the above mentioned historical training data set for a consistent period of time here in the subfolder “01 Training Data”.  This data set should be used to create or calibrate your forecast model. During the competition you will additionally get the real world historic data for the last 24 hours every day at 08.00 am as described in the start example above, subsequent to the training data set.

Awarding Procedure

The evaluation will be based on two simple mean absolute errors (MAE). The first MAE measures how well you have forecasted the generation (G) in the near future for the current day (D) from 10.00 to 24.00. The second measures your prediction for the entire following day (D+1) from 00.00 to 24.00. Both are weighted equally and will be aggregated for the competition period. The participant with the lowested cumulated MAE (weighted) wins the competition. The final score will be based on the “cumulated weighted MAE” removed by your two submissions with the highest (worst) „weighted MAEs“. The following table and formulas illustrate the evaluation procedure. The red numbers are the results of the final ranking.

awarding-example-with-formulars-and-arrows_highres

 

The ranking of the participants and their MAE’s will be updated by 12.00 every day.

Registration is closed

Registration period for the EEM 2017 Forecast Competition has ended! Further registration requests cannot be considered. The registered participants will receive more details regarding the competition procedure soon.

Please send an email to forecast.eem2017[-at-]ee2.biz to register yourself or your team for the competition by the 15th of February 2017. We will send you a confirmation of your registration.

You do not need to be registered for the EEM 2017 conference to register and participate in the forecast competition. If you finish under the top three participants and would like to receive your prize and hold a short presentation about your model/approach at the conference, you will, however, at this point need to register for the EEM 2017. In this case, even at this late date (April), you will be eligible to register for the conference under “early bird” conditions.

Data Provision / Download

Historical training and daily data as well as the updated standings during the competition period will be provided in the EEM2017 Forecast Competition Folder .

Submission

Please use your personal submission template that will be provided to you after the registration period for your forecast submissions. Upload your daily submission every day within the competition period by 10.00 am at the latest into your personal Forecast Competition folder. Times are local time of Dresden equally to CEST or UTC+2 for the competition period. Further details of the submission procedure are provided to the registered participants.

Help & Questions

If you have any questions or need help regarding the competition, please feel free to contact us at forecast.eem2017@ee2.biz.

Kindly supported by:

EnBW.svg

 

 

ARE YOU READY FOR THE CHALLENGE?

We are looking forward to your participation! 🙂 

< Please note that no legal claim can be made. >