Electric vehicle charging load forecasting based on user portrait and real-time traffic flow
Haihong Bian, Shengwei Bing, Quance Ren, Can Li, Zhiyuan Zhang, Jincheng Chen
Abstract
Most of the existing research problems focus on: simplifying the users into groups with the same characteristics or not considering the influence of different groups on electric vehicle EV models, travel and charging behavior; EV batteries have the same capacity; charging power is regarded as a constant; travel paths are based on the shortest distance. In order to solve the above problems, this paper proposes a charging load forecasting method for electric vehicles based on user portrait and real-time traffic flow. Users are divided into different groups based on their social and travel characteristics, their probability distributions are extracted separately. A method for evaluating car models is established to differentiate parameters such as energy consumption, battery capacity, and fast charging power. The effect of battery degradation on battery capacity and the effect of power decay on charging power are considered. Road speed is divided into multiple sections according to the different traffic flows in different time periods, accordingly the road time duration is calculated. An improved Dijkstra algorithm is proposed to solve the trip path by taking the real-time minimum time duration as the goal. The path switching strategy is proposed when the remaining power cannot meet the requirements. Finally, Monte Carlo algorithm is employed to simulate the daily profile of EV charging load, thereby enabling the generation of load profiles for different groups and different functional areas. The results show significant differences between user group. Different user groups and functional areas have an effect on the peak value of charging load curves. The power decay model makes the charging time and charging amount more accurate. It is verified that the path selection is affected by real-time and traffic flow, the proposed path optimization algorithm could reduce the trip duration, with estimated maximum savings of 30%.