Hybrid Energy Forecasting and Trading Competition

Submission Dates:
02/19/2024 to 05/18/2024
Citation Author(s):
Jethro
Browell
University of Glasgow
Sebastian
Haglund
rebase.energy
Henrik
Kälvegren
rebase.energy
Edoardo
Simioni
Ørsted
Ricardo
Bessa
INESC TEC
Yi
Wang
University of Hong Kong
Dennis
van der Meer
Ørsted
Submitted by:
Jethro Browell
Last updated:
Mon, 07/08/2024 - 06:27
DOI:
10.21227/5hn0-8091
Links:
License:
Creative Commons Attribution

Abstract 

Forecasting production from wind and solar power plants, and making effective decisions under forecast uncertainty, are essential capabilities in low-carbon energy systems. This competition invites participants to develop state-of-the-art forecasting and energy trading techniques to accelerate the global transition to net-zero and to win a share of $21,000 in prize money. It aims to bridge the gap between academic and industry practice, introduce energy forecasting challenges to new communities, and promote energy analytics and data science education.

The competition is organised by the IEEE Power and Energy Society Working Group on Energy Forecasting and Analytics and is sponsored by Ørsted, one of the world’s largest renewable energy companies, and rebase.energy, and tech start-up based in Sweden. Forecasting power production and participating in energy markets is a core part of Ørsted’s business, and an area where data and decisions science can bring great value. The competition platform developed by rebase.energy gives participants access to a wide variety of data, including weather forecasts and real electricity market data, to generate their forecasts and trading decisions.

The competition will run from February to April 2024, but the platform will open in advance for participants to begin developing and testing their solutions. This is a genuine forecasting task! It requires daily submissions of forecasts and market bids. Automation is strongly encouraged.

This competition is organised by the IEEE Power & Energy Society Working Group on Energy Forecasting and Analytics and sponsored by Ørsted and rebase.energy. The organising committee are: Jethro Browell (Chair, University of Glasgow), Sebastian Haglund (rebase.energy), Henrik Kälvegren (rebase.energy), Edoardo Simioni (Ørsted), Ricardo Bessa (INESC TEC), Yi Wang (University of Hong Kong)

 

Instructions: 

There are two tracks: forecasting and trading. Participants are encouraged to compete in both tracks, and an additional prize is on offer for combined performance across forecasting and trading. Submissions to both traks must be made by 9:20AM UTC on each day of the competition period.

The forecasting track requires participants to produce probabilistic forecasts of the power production from a hybrid power plant comprising the Hornsea 1 wind farm and the combined solar capacity of East England, totalling approximately 3.6GW. Forecasts in the form of quantiles from 10% to 90% in increments of 10% are required for each half-hour period of the day-ahead. Forecasts will be scored using the Pinball Loss.

The trading track requires participants to trade the energy produced from the hybrid power plant in the day-ahead electricity market. This track is based on Great Britain’s wholesale electricity market, which features a day-ahead auction and single-price imbalance settlement.

Key Dates:

Following an unexpected reduction in the export capacity from Hornsea 1 wind farm, the competition start and proceeding dates have been revised to give participants time to adapt to the new situation. The export limitation at Hornsea 1 is expected to end on 17 February, however, the competition start will not be pushed back any further. Participant should consider any potential reduction capacity in their submissions. Data on capacity reductions is available via REMIT (see FAQs in the documentation).

Registration will be open from 1 November 2023 to 26 January 2024.

1 November 2023: Competition platform opens for development and testing

19 February 2024: First submission of the competition period (forecasts and bids for 20 February 2024)

18 May 2024: Last submission of the competition period (forecasts and bids for 19 May 2024)

24 May 2024: Deadline to respond to participant survey

31 May 2024: Announcement of final leaderboard and prizes

Scoring and Prizes:

A live scoreboard will be maintained on the competition website. This will display team name and their score in each track. The scoreboards will be updated as data become available, which is typically with a lag of seven days.

During the test and competition period, a scoreboard displaying scores for the most recent full week of evaluation data will be displayed. During the competition period, scores will be displayed covering the period from the start of the competition to the most recent evaluation data.

The following cash prizes will be awarded:

Rank

Energy Trading Track

Forecasting Track

Combined Ranking

1

$3,000

$3,000

$3,000

2

$2,000

$2,000

$2,000

3

$1,000

$1,000

$1,000

Best placed student team

$1,000

$1,000

$1,000

A team’s score in the “Combined Ranking” category will be the sum or ranks from the Energy Trading and Forecasting tracks. Ties will be broken based on ranking in the Forecasting track. A student team finishing in the top three may claim the main prize and the prize for the placed student team.

All prize-winning teams will be invited to submit a paper describing their approach to the International Journal of Forecasting (which must be fully reproducible). Furthermore, the top three teams* and best placed student team** will be invited to:

  1. present their approaches at the International Symposium on Forecasting, which will take place 30 June to 3 July 2024 in Dijon, France. ISF is a fantastic conference attended by forecasting researchers and practitioners from around the world. HEFTcom has been given a special session at this conference, and financial support of $1000 per team is available to assist with travel and registration fees.
  2. present at Ørsted’s internal Science Talks seminar series in September 2024 (online), which will include the opportunity to interact with forecasters and traders at Ørsted.

*1st place in each track, or next best placed team if a team places 1st in multiple tracks, with preference going to the combined ranking track, followed by trading track, followed by forecasting track.

**Best placed student team in the combined ranking track

Getting Started & Open Benchmark:

The organisers have provided an open-source benchmark which will serve as a reference for performance in the competition and as a quick start tool to aid teams in their participation. See this GitHub repository and attached documentation for details.

For competition networking, team formation and support using the competition API, you are welcome to join the Rebase Community Slack channel through this link.

Competition Rules:

Teams: You may participate as an individual or team. There is no limit on the size of teams. Only one registration is required per team.

Student Teams: All members of student teams must be registered full-time students (or part-time but not working in the energy industry) for the duration of the competition period to claim a student prize. The competition organises will verify student status as part of the registration process.

Data: rebase.energy have provided convenient APIs for most of the data you will need. However, there is no restriction on the use of additional data. Additional open data sources are suggested in documentation.

Missed submissions: A team may miss up to five submissions during the competition period. Missing entries will be filled by the Open Benchmark and scored accordingly.

Public dissemination of entries: by participating, teams grant the competition organisers the right to publicly disseminate any submitted data (bids and forecasts) after the end of the competition period. This is to enable reproducibility of the results table and further analysis.

Reports: To retain a position in the final leaderboard and to qualify for a prize, participants must respond to a survey that will include providing a high-level description of their methodology and any additional data used by 24 May 2024. A high-level summary of participants’ methodologies will be published after the competition.

Competition back-end: Only staff at rebase.energy will have access to the back-end of the competition platform (submission data, scoring algorithm, leaderboard etc) until the end of the competition period. Participants affiliated with rebase.energy or Ørsted are not eligible to claim prizes.

The organisers reserve the right to disqualify teams deemed to have broken rules or undermined the spirit of the competition, and to update the competition rules if necessary.

Notices:

This dataset contains BSC information licensed under the BSC Open Data Licence, the Open Government Licence and data from Sheffield Solar PV Live (Terms). See attached documentation for details.

Final Leaderboard:

The prize winning teams are:

  • SVK: Jon Olauson, Olle Viotti and Jakob Huss (Svenska kraftnät)
  • Rnt: Harisankar Haridas and Tarun Raj (REint.ai)
  • UI BUD: Gergo Barta (Vestas Wind Systems)
  • GEB (student team): Chuanqing Pu, Songyuan Liu, Jinming Yu (Shanghai Jiao Tong University)

Team

Pinball [MWh]

Forecasting Track Rank

Revenue [£m]

Trading Track Rank

Combined Track Score

Combined Track Rank

Student Team?

SVK

22.18

1

88.88

1

2

1

 

Rnt

24.64

3

88.29

2

5

2

 

UI BUD

23.18

2

88.07

4

6

3

 

GEB

25.16

4

88.18

3

7

4

Student Team

BridgeForCast

25.34

5

87.67

6

11

5

 

Sukantabasu

27.04

7

87.83

5

12

6

 

Stochastic Parrots

27.50

8

87.53

8

16

7

 

EnergiWise

27.65

9

87.43

9

18

8

 

Ihubex

29.22

12

87.64

7

19

9

Student Team

LSEG Power Team

25.74

6

85.71

13

19

10

 

NICE_Forecast

27.98

10

87.21

11

21

11

Student Team

Oracle

28.34

11

87.20

12

23

12

 

RE-Cast

30.04

13

87.31

10

23

13

 

GM Team Mannheim

33.92

15

85.55

15

30

14

 

༼ つ ◕_◕ ༽つ

31.01

14

84.29

18

32

15

 

Zzblu

35.04

16

84.41

17

33

16

 

OLPZR

44.84

20

85.58

14

34

17

 

tradRES

37.19

18

84.45

16

34

18

Student Team

Eguzkinet

36.68

17

83.83

21

38

19

 

6340

47.37

21

84.09

19

40

20

 

KittenKilowatt

41.67

19

83.71

22

41

21

 

power_rabbit

53.96

23

84.03

20

43

22

Student Team

FCOR_BL

51.66

22

82.26

24

46

23

 

Enerweb

55.13

24

83.22

23

47

24

 

Benchmark

53.58

 

82.23

(Teams Associated with the Organising Committee)

quantopia

25.38

87.96

 

Rolling Leaderboard - Competition Period (Preliminary Results)