Journal article Open Access
A Methodology to Design an Efficient EM Controller with High Practicability in HEVs Learning
Ehsan Ghasemimoghadam*1 , Kazuhide Togai2 , Hisashi Tamaki1
This paper presented on the methodology of designing intelligent energy management for hybrid electric vehicles (HEVs). This work outlines the use of the deep neural network (DNN) to design learning-based EM controller, which provides a powerful new framework to control the HEV system while improving the HEV performance. The framework utilizes the DNN technology to inference the new knowledge from the non-causal energy optimization results. Herein, we present intelligent energy controller, which was trained to control the HEV system by represented the current state of the system on preset several driving cycles. In order to prepare the data points, the near-optimal solutions are saved in the data store which obtained by the offline optimization processes at various starting SoC. Through computational examples on the series-parallel structure, the designed controller would be evaluated with a canonical rule-based and the non-causal optimized control strategy that fuel consumption improvement on the proposed EM controller observed and the effectiveness of the proposed approach in the designing high practicable EM controller was confirmed.
| All versions | This version | |
|---|---|---|
| Views | 52 | 52 |
| Downloads | 13 | 13 |
| Data volume | 19.2 MB | 19.2 MB |
| Unique views | 52 | 52 |
| Unique downloads | 11 | 11 |
- Publication date:
- April 7, 2018
- DOI:
-
- Keyword(s):
- deep neural network (DNN), energy control system, energy management (EM), hybrid electric vehicle (HEV), machine learning (ML)
- Published in:
- INTERNATIONAL JOURNAL OF ENGINEERING SCIENCES & RESEARCH TECHNOLOGY: 7 pp. 780-791 (3).
- License (for files):
- Creative Commons Attribution 4.0 International