부산대학교 도서관

This paper presents the development of an energy management system (EMS) dedicated to an aggregator for energy dispatch among a fleet of electric vehicles (EVs). This EMS main objective is to manage the total energy demand from plugged-in EVs to reduce the overall load on the grid, considering the effects of cold temperatures and human habits on the total demand of energy. The approach used to achieve this objective is training an artificial neural network (ANN) using Levenberg-Marquardt algorithm (LMA). The training dataset is based on climate data, on actual energy consumption in Quebec (Canada), and on real energy and power consumption from an EV charging station that can accept up to 30 vehicles. The results show that the LMA was effective to train three specialized EMS to predict accurate reference power for charging and discharging EVs in summer and winter with a minimal mean square error and maximal R-squared.