Plug-and-Play Power Management Control of All-Electric Vehicles Using Multi-Agent System and On-line Gaussian Learning

This paper presents a new architecture for the next-generation reconfigurable power management control (PMC) system for all-electric vehicles, which can handle the personalization and evolution of vehicle architectures in a plug-and-play (PnP) fashion. The proposed control system utilizes multi-agent system (MAS) for reconfigurability and integrates on-line model learning with Gaussian process. In the two case studies of PnP scenarios, one with limited power supply and another with excessive power supply, the simulation results suggest that the proposed PMC system is capable of stabilizing the system and performs better than the baseline controller. The improvement in performance over the baseline voltage-power droop controller can be up to 59% when parameters are optimized using on-line learning.