Optimized Scheduling for Urban-Scale Mobile Charging Vehicle

Abstract

With the development of electric vehicles, more and more owners choose to buy electric vehicle. Electric vehicles have many advantages, such as environmental protection and efficiency. Nowadays, a lot of charging stations have been established, but the situation of drivers queuing for charging still happens. In this paper, we first analyzed the charging behavior in Beijing and learned the behavioral characteristics. then, we propose a new charging framework based on mobile charging vehicles (MCV). In this framework, we propose a Deep Neural Network based on Temporal Characteristics (TCDNN) to predict the charging demand of all regions of the city in the future. This helps us find areas where charging demand are high. Next, we dispatch MCVs to those areas to relieve the pressure on busy stations and also consider the scheduling cost of MCV. we find the best parking strategy of MCV by Teaching-Learning-Based optimization (TLBO). Finally, we evaluate the framework and the result demonstrate that it accurately predict all the moments with high charging demand and RMSE is 31.1% smaller than baseline model. TLBO-based MCV scheduling shortens the user waiting time by up to 73%.