Safe reinforcement learning-based system for connected and autonomous vehicle charging infrastructure
Abstract
This paper is about creating a system that helps to manage the charging of electric
vehicles that are connected and can drive autonomously taking into consideration
the safe reinforcement learning outcomes in this process. The system is regarded
as an intelligent decision support system (IDSS). In this system, a holding corporation
that works the whole charging infrastructure, installs charging equipment for
both regular electric vehicles driven by humans and autonomous vehicles. The problem
arises when human-driven vehicles ask for more charging time and energy than
they really need and to success charging request competition, which can particularly
lead to cause issues. To address this problem, a proposed solution aims to make
sure the charging equipment is used efficiently minimizing the risk of not having
enough power available as well as considering all the safety of the charging equipment.
Here a system will be introduced where it encourages human-driven vehicles
to make rational charging requests based on data and noting down the parameters
which are the number of DSOs (Distribution System Operator), the nearest finding
of EVSE(Electrical Vehicle Supply Equipment), the association among the EVSEs,
the starting and ending time of the plugin, energy absorption, time duration, request
for charging for the CAV, CV and AVs. Furthermore, the introduction of a learning
system where the charging equipment learns how to schedule charging sessions
based on the procession from the main operator or the distribution system operator.
The conducted experiments will show that this system improves the charging rate,
active charging time, and energy usage compared to existing systems ensuring all
the protection of the electrical and connected autonomous vehicles. Therefore, the
study will contribute to making transportation systems smarter and addressing the
challenges and safeties of connected and autonomous vehicles.