Incremental Ensemble Learning for Electricity Load Forecasting

• DOI: 10.12700/aph.13.2.2016.2.6
• Publication Date: 2016-03-03T02:17:20Z
• Authors: Gabriela Grmanová; Peter Laurinec; Viera Rozinajová; Anna Bou Ezzeddine; Mária Lucká; Peter Lacko; Petra Vrablecová; Pavol Návrat
• Keywords: Ensemble Learning; Incremental Learning

The efforts of the European Union (EU) in energy supply domain aim to introduce intelligent grid management across whole EU.The target is planned contain 80% all meters be smart generating data every 15 minutes.Thus, EU will grow rapidly very near future.Smart are successively installed a phased roll-out, and first meter samples ready for different types analysis order understand data, make precise predictions support control.In this paper, we propose an incremental heterogeneous ensemble model time series prediction.The was designed electricity load taking into account their inherent characteristics, such as seasonal dependency concept drift.The proposed characteristicsrobustness, natural ability parallelize incrementally train modelmake presented suitable processing streams "big data" environment.