Loading...
Description
Information Systems play a central role in the energy sector for achieving climate targets. With increasing digitization and data availability in the energy sector, data-driven machine learning (ML) approaches emerged, showing high potential. So far, research has focused on optimizing ML approaches’ prediction performance. However, this is a one-sided perspective. ML approaches require large computation times and capacities leading to high energy consumption. With the goal of sustainable energy systems, research on ML approaches should be extended to include the application’s energy consumption. ML solutions must be designed in such a way that the resulting savings in energy (and emissions) are greater than the energy consumption caused using the ML solution. To address this need, we develop the Sustainable Machine Learning Balance Sheet as a framework allowing to holistically evaluate and develop sustainable ML solutions which we validated in a case study and through expert interviews.
Recommended Citation
Wenninger, Simon; Kaymakci, Can; Wiethe, Christian; Römmelt, Jörg; Baur, Lukas; Häckel, Björn; and Sauer, Alexander, "How Sustainable is Machine Learning in Energy Applications? – The Sustainable Machine Learning Balance Sheet" (2022). Wirtschaftsinformatik 2022 Proceedings. 1.
https://aisel.aisnet.org/wi2022/sustainable_it/sustainable_it/1
How Sustainable is Machine Learning in Energy Applications? – The Sustainable Machine Learning Balance Sheet
Information Systems play a central role in the energy sector for achieving climate targets. With increasing digitization and data availability in the energy sector, data-driven machine learning (ML) approaches emerged, showing high potential. So far, research has focused on optimizing ML approaches’ prediction performance. However, this is a one-sided perspective. ML approaches require large computation times and capacities leading to high energy consumption. With the goal of sustainable energy systems, research on ML approaches should be extended to include the application’s energy consumption. ML solutions must be designed in such a way that the resulting savings in energy (and emissions) are greater than the energy consumption caused using the ML solution. To address this need, we develop the Sustainable Machine Learning Balance Sheet as a framework allowing to holistically evaluate and develop sustainable ML solutions which we validated in a case study and through expert interviews.