Load Forecasting for EV Charging Stations Based on Artificial Neural Network and Long Short Term Memory

Due to the increasing prices of nonrenewable resources and their fast depletion, Electric Vehicles (EVs) are proposed to be a viable and eco-friendly option for transportation. However, the increasing number of EVs poses a challenge for EV charging stations to fulfil their charging demands. EVs can manage and derive their charging strategies if the future charging load is known. This work addresses the EV load forecasting problem at a charging station and tries to predict charging requirements for the next hour, day and weekly basis through Artificial Neural Network (ANN) and Long Short-Term Memory Network (LSTM) approaches. In this work, the real dataset of the Adaptive Charging Network stationed at California Institute of Technology, USA is used. Experiments results show that LSTM proves to have more accuracy in terms of RMSE compared to ANN for all types of forecasting.