The Charging Scheduling Study Of Electric Vehicles Based On Markov Decision Process

The role of electric vehicles in the smart grid is very important in future, it can reduce dependence on petroleum resources, and it can reduce exhaust emissions, so it plays an important role in our daily lives. But electric vehicles need to be charged in our daily life, coordinated charging schedule is an emerging problem, the uncoordinated charging will cause serious impact on the grid system, in order to reduce the electric cars’ impact on the grid, and ensuring the experience of the user, the study on coordinated charging is significant.This paper describes the development status of electric vehicles in our nation, our country is developing electric vehicles, car companies are developing electric vehicles, too. Electric vehicle charging equipment including: battery, charger, charging pile. Electric car’s battery inludes:storage battery and fuel battery. EV charger includes:DC charger and AC charger, which can charge the battey of electric car. Electric car charging pile can charge all kinds of electric vehicle.In this paper, we observe that the coordinated charging problem could be described as a semi-Markov Decision Process (SMDP). We consider: charging efficiency, charging fairness, less power losses, we employ a neuro-dynamic programming(NDP), which has in-built capability of studying from outside environment and adaping system parameters based on the performance requirements.We use the linear feature-based approximations architecture to approximate optimal differential cost function. After establishing the approximating architecture, we employ the technique of Temporal-Difference learning to perform online parameter tuning. Simulation results indicate that NDP outperforms two scheduling algorithms and simultaneously considers: charging efficiency, charging fairness, and less power losses.