| Title |
Few-shot learning for triplet-based EV energy consumption estimation / |
| Authors |
Čivilis, Alminas ; Petkevičius, Linas ; Šaltenis, Simonas ; Torp, Kristian ; Markucevičiūtė-Vinckė, Ieva |
| DOI |
10.1080/08839514.2025.2474785 |
| Full Text |
|
| Is Part of |
Applied artificial intelligence.. Philadelphia : Taylor & Francis. 2025, vol. 39, iss. 1, art. no. e2474785, p. [1-20].. ISSN 0883-9514. eISSN 1087-6545 |
| Abstract [eng] |
Predicting the energy consumption of an electric vehicle (EV) is often relevant when planning and managing electric mobility. The prediction is challenging as EV energy consumption is highly variable and dependent on context. First, this paper proposes an integrated framework for the collection of online telematic data, processing of this data, online maintenance of statistics, and machine-learning-based prediction of travel time and energy consumption. A key feature of the proposed framework is the preprocessing of the trajectory data into triplets, a convenient data unit that captures the relevant context necessary for effective energ y prediction. The second contribution of the paper addresses the effective management of drastic change in context through robust energy prediction models. In particular, using few-shot learning techniques, we tackle the problem of the need to create different energy prediction models for different EV types, from small EVs to electric buses. Experimental results on three different data sets demonstrate how energy prediction models adapt to different EV types. |
| Published |
Philadelphia : Taylor & Francis |
| Type |
Journal article |
| Language |
English |
| Publication date |
2025 |
| CC license |
|
