The vehicle-to-grid (V2G) concept is a key element in the successful transition to smart grids and smart cities. The optimal integration and operation of V2G is challenging owing to the fluctuating nature of energy demand and grid constraints. This study addresses these challenges by optimizing residential V2G powered by a hybrid energy system coupled with novel gravity energy storage and batteries. First, the optimal sizing of the hybrid photovoltaic (PV) /wind system was determined to meet the residential load demands with the primary objective of minimizing the cost of energy (COE). To achieve this goal, a dynamic smart home energy management system was developed. The proposed model compares the performance of a battery electric vehicle (BEV) and fuel-cell electric vehicle (FCEV) for V2G applications. Forecasting of PV and wind production, as well as load demand, is performed using recurrent neural network (RNN) algorithms, including gated recurrent unit (GRU) and simple RNN models. Six years of historical data were collected to train the models and their effectiveness was validated using a measured PV production dataset. A sensitivity analysis was conducted to evaluate the parameters that significantly affected COE. These findings suggest that a hybrid renewable energy production system with battery energy storage is highly reliable. The BEV is proven to be more efficient than the FCEV in the studied V2G scenarios, with an energy of about 0. 29 ·kWh −1. The forecasting model based on the GRU architecture successfully predicted both energy production and load variability. The obtained results demonstrate the usefulness of deep learning models in leveraging the transition towards smart grids.
Achour et al. (Sun,) studied this question.