Online Energy Management of Electric Vehicle Parking-Lots

摘要

Electric vehicles (EV) charging scheduling in parking lots has been a hot topic in recent years. Instead of simply starting the charging process with the entrance of the EVs, a parking lot operator can decrease the cost of buying electricity in real-time, when prices are low. However, this decision-making process involves randomness in both price and EVs behavior (arrival and departure times). In this study, we introduce a supervised machine learning framework using a multi-layer perceptron regression that can train an online estimator to help the operator with the aforementioned process. This online estimator uses a small set of historical data and provides values of the amount of energy that should be bought by the operator. We use this method in the online management of EVs within parking-lots and evaluate the performance with a real-world EVs’ charging data.