Accurate Charging Demand Estimation of Electric Vehicle via Multi-Data Sources

With the penetration of electric vehicles (EVs) increased, charging demand estimation is more urgent and important. It provides constructive ideas and guidance for building an efficient charging pile network. Unfortunately, most of the existing charging demand estimation methods are based on simulation experiments. They lack real data and ignore the complexity of charging behavior. Therefore, the description of the charging demand is not accurate enough. In this paper, we first analyze the proportion of quick/slow public charging piles and their usage situation, compare the distribution of charging specific gravity of different vehicle types. Then we propose a novel algorithm to estimate charging demands. Through the analysis and observations of the data of the public charging facilities and EVs in Kunming, we obtain the spatial and temporal distribution of charging demands of EVs. In the spatial part, electric orders of the pile, parking locations and charging locations of EVs are analyzed with the Gaussian mixture model(GMM).A similarity-based algorithm is provided to fuse the three models. In the temporal part, charging times per hours of EVs are analyzed with the Autoregressive Integrated Moving Average model(ARIMA). Evaluation results demonstrate that our approach outperforms alternative solutions by reducing the spatial error by up to 66.70% and temporal error by up to 63.41%, respectively.