GECCO Challenge 2012

Optimizing Energy Efficiency by Consumption Data Forecasting

Press Items

Oberberg Aktuell: Internationale „Datendetektive“ auf die Probe gestellt
FH Köln: Stephan Hutterer gewinnt Wettbewerb zur Stromverbrauchs-Prognose
Kölner Stadtanzeiger: Was der Bäcker gerne wüßte
CleanThinking: Verbesserte Verfahren zur Prognose des Energieverbrauchs
smartmeter.de: Vorhersage für den Stromverbrauch auf Basis intelligenter Zähler
koelncampus.com: Doktorand hat neues System zur möglichst genauen Vorhersage für Stromverbrauch entwickelt
currentgame.de: Verbesserte Verfahren zur Prognose des Energieverbrauchs

News

  • We are happy to announce the Winners of the Gecco 2012 Industrial Challenge:
    1. Stephan Hutterer (University of Applied Sciences Upper Austria)
    2. Torsten Hildebrandt (University of Bremen)
    3. Jens Heger (University of Bremen)
  • The slot for Industrial Challenge result presentations at GECCO 2012 has been reserved:
    Monday 9th, 12.40 pm to 13.40 pm.
    We are looking forward to meeting you in Philadelphia!
    Please see this website and the official GECCO program for program changes and updates.
  • Submission Deadline has been extended until June 29, 2012
  • The Resource Package has been extended by two scripts and two example data sets that allow competitors to check their computed prediction against our RMSE calculating procedure.
  • The competitor with the best prediction is rewarded with an iPad, which is kindly sponsored by GreenPocket.

Overview

Offshore TurbineGoal of the GECCO 2012 Industrial Challenge is to develop accurate forecasting methods for electrical energy consumption profiles. Based on real-world energy consumption time series data, as recorded by modern smart-metering equipment, these methods should provide accurate predictions of electrical energy consumption. As smartmetering involves large volumes of streamed time series data, acceptable forecasting methods must be accurate and computationally efficient.
The objective of this competition is to provide accurate time series predictions of electrical energy consumption based on real-world training data. Based on these training data, participants of the challenge must provide quarter-hourly predicted electrical energy consumption time series that extend each of the given training time series into the future.

The challenge is organized in a single round. In order to participate in the competition, appropriate predictions for the four training data sets given have to be supplied. Submissions are accompanied by a two page report describing the algorithm that has been applied to generate the predicition time series.
Participants of the GECCO 2012 Industrial Challenge will maintain all rights to their forecasting methods and systems. The organizers will not share a submission without explicit permission of the participant. Nonetheless, participants are expected to describe their methods to the GECCO 2012 community.

Highlights of the GECCO 2012 Industrial Challenge include:

  • Interesting Problem Domain: Forecasting energy consumption profiles based on smart-metering data offers a challenging and highly relevant test case for modern time series prediction methods.
  • Real-world Data: Multiple real quarter-hourly energy consumption time series data sets are provided for training and assessing forecasting methods.
  • Realistic Quality Measurement: Forecasting methods are assessed by their prediction accuracy for a continuum of prediction horizons, ranging from minutes to weeks into the future.
  • Fair Submission Assessment: Prediction accuracy is determined on test data available to the organizers, which will be made public after the competition ends.
  • Direct Link to Industry: The challenge’s industry partner GreenPocket GmbH, a provider of modern smart-metering software, will evaluate the winning submissions for commercial application
    and will be in direct contact with the winning participants.

For further information please see our Call for Participation which contains more detailed informations.

Downloads and Submission

  • Download the Call for Participation in PDF format.
  • Download the GECCO 2012 IC Resource Package, containing the required datasets, software, and the official rules and regulations in ZIP or TAR.GZ format. Please see the file README.txt for further instructions.
  • Please use the GECCO 2012 IC Submission Form for sending your submission. (Submission closed)

Important Dates and Contacts

  • Software and Data Availability: immediately
  • Challenge Submission Deadline: June 22, 2012
  • GECCO-2011 Conference: July 07-11, 2012
  • Contact: gecco at f10.fh-koeln.de

Organizing Committee and Sponsors

GreenPocket
Gecco 2012

  • Oliver Flasch, Cologne University of Applied Sciences
  • Martina Friese, Cologne University of Applied Sciences
  • Thomas Bartz-Beielstein, Cologne University of Applied Sciences
  • Wolfgang Konen, Cologne University of Applied Sciences
  • Jens Neuhalfen, GreenPocket GmbH
  • Pier Luca Lanzi, Politecnico di Milano
  • Jorn Mehnen, Cranfield University

GECCO is sponsored by the Association for Computing Machinery Special Interest Group for Genetic and Evolutionary Computation (ACM SIGEVO).