Unsupervised detection and open-set classification of fast-ramped flexibility activation events
share
The continuous electrification of the mobility and heating sector will introduce new challenges to distribution grid operation. Uncoordinated activation of flexible units, e.g. simultaneous charging of electric vehicles as a reaction to price signals, could systematically trigger transformer or line protections. Real-time identification of such fast-ramped flexibility activations would allow taking counteractions to avoid potential social and financial cost. In this work, a novel data processing pipeline for identification of fast-ramped flexibility activation events is proposed. The pipeline combines techniques for unsupervised event detection and open-set classification. The systematic evaluation on real load data demonstrates that main building blocks of the proposed pipeline can be realized with methods that fulfill important requirements for an application in a distributed event detection architecture. For the detection of flexibility activation events an upper performance limit is identified. Moreover, it is demonstrated that application of an open-set classifier for classification of flexibility activation events can improve the performance compared to widely-applied closed-set classifiers.
READ FULL TEXT