Home // International Journal On Advances in Intelligent Systems, volume 5, numbers 3 and 4, 2012 // View article

Rapid Energy Consumption Pattern Detection with In-Memory Technology

Authors:
Christian Schwarz
Felix Leupold
Tobias Schubotz
Tim Januschowski
Hasso Plattner

Keywords: energy pattern recognition, smart meters, machine learning; in-memory database; in-memory technology; inter-quartile range coverage

Abstract:
The transformation of today’s energy market poses new challenges for both, energy providers and customers alike as the usage of renewable energy sources and energy-awareness increases. Additionally, the energy infrastructure is changing fundamentally. On the one hand, the installation of so called smart meters offers the possibility of more detailed monitoring and fine grained electricity billing. On the other hand, the amount of data produced within the power grid increases dramatically. Utility companies will use such data to increase prediction accuracy and to improve energy production, while consumers will more and more transform to prosumers. Within that environment the necessity of short-term predictions increases to improve the power grids stability. In this article, we respond to some of the challenges that energy consumers and providers face by an implementation of a prototypical recording, monitoring and analysis landscape that uses smart meter data. The challenges that this article tackles include: real-time energy consumption classification; mass energy consumption data classification; and early short-term energy consumption prediction. In extensive experiments on real-world data, we show that such challenges can be handled effectively. We leverage smart meter data via a novel combination of machine-learning algorithms and latest in-memory technology.

Pages: 415 to 426

Copyright: Copyright (c) to authors, 2012. Used with permission.

Publication date: December 31, 2012

Published in: journal

ISSN: 1942-2679