Charging planning model for electric vehicles based on range prediction

Range anxiety continues to be one of the most important factors negatively affecting the transition to electric vehicles (EVs). Factors that trigger range anxiety include EV drivers not relying enough on remaining range indicators. However, it has the potential to reduce range anxiety if the EV driver knows where to stop for charging on the route according to the range information remaining at the beginning of the journey. Although driver information systems of EVs can access real-time data thanks to smart transportation technologies, it cannot be expected that the created charging plan will reduce range anxiety unless range prediction is made considering the conditions of the determined route. The aim of this study is to determine where and how much the EV should be charged for minimum travel time or cost by creating a charging plan based on real-time range prediction for a specified route. The deep neural network (DNN) model was trained by using the inputs of the static features and dynamic features of the journey in the range prediction model. In the charge planning model, the amount of energy consumed between the nodes on the route was obtained by the range prediction model. Within the scope of charging planning, a mixed integer programming model is developed that take into account non-linear charging times, charging prices that vary depending on time slots, vehicle to the grid (V2G) applications, of charging units, multiple charger points with different power levels and their availability (occupied/free). However, since the developed mathematical programming model was insufficient in terms of the solution time, a mat-heuristic approach consisting of hybrid use of genetic algorithm and mathematical programming model was proposed as a solution approach. Test results on 32 various problems indicate that the mat-heuristic approach outperforms the metaheuristic and heuristics for both minimum travel time and minimum travel cost