Results

The challenge consisted of two evaluations, one online where teams continuously submitted their solutions to a set of 50 instances, and a final evaluation where each team's code was evaluated in a consistent environment on 50 new unseen instances.

Public Leaderboard

The rankings from the public leaderboard are show in the table below. More detailed information is available.

  Team Affiliation Score
1 SENCOM CODeS, KU Leuven 0.75
2 OptaPlanner Delirium Red Hat and unaffiliated 9.99
3 Kadioglu, Pacino, Tierney Mixed 34.87
icon Deepak Mehta Insight, UCC 35.34
icon KULZ DTAI, KU Leuven 50.37
icon Team Prolog Insight, UCC 76.42

Longest Period on Top

Teams were encourages to submit regularly to the public leaderboard, the prize for the longest period in first position goes to team SENCOM.

  Team Name Time at the top
1 SENCOM 54 days, 23:53:55
2 OptaPlanner Delirium 1 day, 21:51:52
3 Deepak Mehta 1 day, 5:06:16
4 Team Prolog 20:44:01

A visualization of each team's score over time.

Final Evaluation Leaderboard

1st place goes to team SENCOM from CODeS, KU Leuven. Tony Wauters, David Van Den Dooren, Thomas Sys

  Team Affiliation Score
1 SENCOM CODeS, KU Leuven 3.87
2 OptaPlanner Delirium Red Hat and unaffiliated 26.13
icon KULZ DTAI, KU Leuven 45.18
3 Kadioglu, Pacino, Tierney Mixed 60.37

Penalty of Forecasted Energy Price

The ultimate goal of the challenge is to investigate the penalty of scheduling with a forecasted energy price, and to develop improvements to this either by producing a better forecast or by adapting the scheduling task. The plot below shows that two completely different schedules are produced if we schedule with the forecasted energy price (blue) or with the actual price data (red).