A Synergistic Learning Based Electric Vehicle Charging Demand Prediction Scheme

Transportation electrification has been seen as a potential solution to the depletion of fossil fuels as well as global warming and air pollution. However, promoting electric vehicles (EVs) encounter great challenges due to insufficient EV charging infrastructure and the lack of EV data. To address the challenges, this paper proposes a synergistic learning-based EV charging demand prediction scheme that takes advantage of both data-driven and analytic approaches. First, using the prevalent regular traffic data, a long short-term memory (LSTM) neural network is introduced to predict the on-road traffic flow. Based on the prediction result, the EV charging demand is analyzed by considering a variety of factors (e.g., travelling pattern, EV model, etc.) and simulating the demand using Monte Carlo simulation (MCS) to achieve an accurate prediction. The synergistic learning method ensures an accurate EV charging demand prediction without using detailed EV data. The accuracy and outcome of the proposed learning method are then validated by a case study using real data from the California Department of Transportation.