Vehicle-to-grid communication system for electric vehicle charging

被引：16

作者：

Lim, Yujin ^{[2
]}

Kim, Hak-Man ^{[1
]}

Kang, Sanggil ^{[3
]}

Kim, Tai-Hoon ^{[4
]}

机构：

[1] Univ Incheon, Dept Elect Engn, Inchon 406840, South Korea

[2] Univ Suwon, Dept Informat Media, Gyeonggi Do, South Korea

[3] Inha Univ, Dept Comp Sci & Informat Engn, Inchon 406840, South Korea

[4] Hannan Univ, Dept Multimedia Engn, Taejon, South Korea

来源：

INTEGRATED COMPUTER-AIDED ENGINEERING | 2012年 / 19卷 / 01期

基金：

新加坡国家研究基金会;

关键词：

Electric vehicle; geocasting; neural network; predictor; time series; vehicle-to-grid communication; LOCALIZATION; NETWORKS;

D O I：

10.3233/ICA-2012-0391

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Recently, the attention on electric vehicle (EV)/plug-in hybrid electric vehicle (PHEV) has been growing. The EV/PHEV will be one of important electric loads from the viewpoint of smart grid in near future. It is anticipated that the EV/PHEV will affect the load pattern of power grids. For this reason, the effective management of the EV/PHEV based on the information and communications technologies will be a major function of smart grid. For EV/PHEV applications, a user interface device equipped on EVs/PHEVs allows the driver to receive instructions or seek advice to manage EV's/PHEV's battery charging/discharging process. In this paper, we present a design of vehicle-grid communications system. To improve the performance of the system, we customize our communication protocol for distributing EV/PHEV's charging information reliably. Also, we model a one-step ahead nonlinear predictor of the charge or discharge price using a neural network ensemble technique. In the experiments, we verify the performance of our protocol with respect to the data delivery ratio and the number of message forwarding. We also compare the price prediction accuracy using the real energy price data, compared with the conventional methods.

引用

页码：57 / 65

页数：9

共 25 条

[1] Day-ahead electricity price forecasting by modified relief algorithm and hybrid neural network [J].