Renewable Energy Forecasting (ReEF)

About

Renewable sources of energy display distinct characteristics, like the fluctuation of energy production and a decentralized regional output. Thus, they differ decisively from conventional sources of energy. Integrating the increasing share of solar energy and wind power into the overall energy production is a challenging task for all partners involved. The forecast of the amount of produced energy is a crucial aspect, as this is the basic information needed for regulating the balance between energy supply and demand.

Having this in mind, the aim of the project ReEF is to evaluate the existing multitude of forecasting methods. Furthermore, ReEF is striving for the development of optimization methods. The project focusses on identifying an optimum aggregation point for forecasts by using hierarchical clustering algorithms including the use of appropriate similarity measures for energy time series and making maximum use of additional context information. Therefore, a suitable benchmarking method is designed and developed for these forecasts. This framework will give technical support when deciding for an ideal model or a combination of models and will simplify decision-making. By the integration of necessary data pre- and post-processing measures this process will be mainly automatized and can be executed by a wide range of users.

Online Access

The demom version of the ECAST benchmark framework can be accessed here. You can either register for your free personal account or use the public demo user (login/pwd demo/demo) to explore the world of online forecasting.

Demonstrator

History & Milestones

2013: Start of development. First prototype version of ECAST was implemented by the student team Lucas Brünings and Willy Zobel. The focus lies on the integration of publicly available forecasting libraries OpenForcast and Mirabel together with an commercial product for benchmarking purposes. An open R* interface was added by Lars Kegel.

2014: The ECAST idea is presented by Robert Ulbricht at an EDBT conference workshop. Furthermore, an web-Interface is added by Lucas Brünings. The benchmark framework prototype is now available as a stand-alone online service.

2015: ECAST prototype was transformed to a stable demo version by the student team Johannes Wilke, Dennis Veit, Sebastian Freund, and Frederick Huschebeck. Some core modules are rebuilt from scratch, web techology is changed to offer a more user friendly and flexible interface.

2016: ECAST demonstration is successfully presented by Wolfgang Lehner at SIGMOD conference in San Francisco.

2017: An ECAST-branch designed by Lars Kegel is used as the central benchmark service in the GOFLEX project. Additional external data sources are integrated exlusively for that purpose. Data exchange between the participants is highly automated.

2018: A new core module for forecast combination tasks is added. The conceptual analysis and development were conducted by the student team Dustin Laudahn, Thomas Olszewski, Amir Rahimi, Nick Scheider, and Alexander Volkmann.

2019: The evaluation module is extended to cover a wider branch of state-of-the-art performance evaluation measures and an additional interface for tailor-made measures. Additionally, ECAST now uses intelligent parallelization techniques in order to enhance servic quality and reduce respone times for users. (Work in progress)

Acknowledgment

This research project is funded by the European Regional Development Fund (EFRE) and the Free State of Saxony under the grant agreement number 100269304 and co-financed by Robotron Datenbank-Software GmbH.