PoPEV: POWER LOAD PREDICTION FOR ELECTRIC VEHICLES CHARGING IN VEHICLES-TO-GRID USING DEEP-LEARNING ENSEMBLE
| Main Authors: | Shivam Sharma, Dr. Aman Kumar Sharma |
|---|---|
| Format: | Article Journal |
| Bahasa: | eng |
| Terbitan: |
, 2021
|
| Subjects: | |
| Online Access: |
https://zenodo.org/record/5887034 |
Daftar Isi:
- The rapid growth of electric vehicles (EVs) can potentially cause power grids to confront new challenges due to changes in load profiles. In this context, this work investigates the charging mechanisms of EVs in V2G, and proposes a new ensemble method called PoPEV, which is based on a deep-learning approach, to predict the power load. Three types of charging mechanisms (coordinated, uncoordinated, and smart) are currently being used for the charging of EVs. The deep-learning-based approach is an ensemble of two current state-of-the-art long short-term memory (LSTM) along with gated recurrent units (GRU). Both LSTM, as well as GRU, are based on recurrent neural networks (RNN). We investigated PoPEV by comparing the results with LSTM and GRU for predictions. Results show that power load can be accurately predicted by using the three deep learning techniques, with the proposed method having an edge in terms of accuracy and high speed.