Authors:
Marcus Voß
;
Mathias Wilhelm
and
Sahin Albayrak
Affiliation:
Technische Universität Berlin, Germany
Keyword(s):
EV Flexibility Modeling, EV Flexibility Forecasting.
Related
Ontology
Subjects/Areas/Topics:
Case Studies
;
Energy and Economy
;
Health Engineering and Technology Applications
;
Load Balancing in Smart Grids
;
Neural Rehabilitation
;
Neurotechnology, Electronics and Informatics
;
Simulation and Modeling
;
Simulation Tools and Platforms
;
Smart Grids
Abstract:
Electric vehicles (EVs) have been proposed to provide flexibility to the energy grid in various ways. With EV exhibiting very diverse usage patterns on the one hand, and many demand response (DR) schemes and their respective requirements on the other, aggregators of flexibility, as well as operators of controlled charging infrastructure, need models and methods to assess the suitability of specific EVs for specific schemes. In this paper, we provide an application independent flexibility model that allows quantifying the potential amount of flexibility based on a historical dataset. Further, we provide a process to assess the predictability of flexibility through modeling it as a short-term load forecasting problem suitable also for smaller aggregations. Our key findings using real-world data of over 200 charging points are that up- and downwards flexibility per interval have a similar magnitude, but it is unexpectedly low for the high number of charging points. Further, we find that
forecast errors are quite high, although we can improve upon naive benchmarks by almost 20% in mean absolute errors with learning models.
(More)
