Enhance EV station charge and discharge based on deep learning forecasting method incorporating renewable energy sources and multi-objective optimization algorithm

Abstract

As electric vehicle charging station operations become more common, proper coordination of energy management is necessary to avoid high operational costs and carbon emissions caused by electric vehicle charging requirements, intermittent renewable sources, and fluctuating electricity rates. In this research, a forecast-based optimization methodology is created with the objective of optimizing the charge-discharge management of a renewable-powered electric vehicle charging station system involving photovoltaic generation, wind generation, fuel cells, electric vehicles, and stationary batteries. First, a recurrent neural network is used to predict relevant time series data, such as photovoltaic energy, wind energy, electric vehicle load demand, fuel cell energy production, and electric vehicle energy consumption. These forecasted data are then applied to a bi-objective energy management optimization problem, which is solved using NSGA-II. The objective is to minimize operating cost and  emissions while considering power balance, grid energy exchange, and other operational constraints associated with the fuel cell, battery storage system, and electric vehicles. In this study, two scenarios are examined using one month of hourly data; Scenario 1 does not use stationary batteries, whereas Scenario 2 uses them. The forecasting results show robust predictions, with one-hour-ahead  values ranging from 0.9162 to 0.9708. The optimization results show that Scenario 2 provides better energy shifting and peak-period support through battery charging and discharging management. Compared with Scenario 1, the operating cost is reduced from $171,833.86 to $160,388.44, and  emissions are decreased from 13,485.48 kg to 12,892.23 kg, leading to 6.7% and 4.4% reductions, respectively.