A distributed dual-layer collaborative energy management strategy for CAV platooning on superhighways

This paper proposes a distributed dual-layer cooperative energy management strategy for electric connected and autonomous vehicle (CAV) platoons operating under superhighway conditions (120–160 km/h), addressing the challenge of simultaneously optimizing energy efficiency and control stability. The upper layer employs an adaptive Delayed Leader-Follower Control (DLFC) strategy operating at 10 Hz. It dynamically adjusts control gains and desired headway based on lead vehicle dynamics, battery state of charge (SOC), and V2V communication quality to ensure string stability. The lower layer formulates the HESS power allocation problem as a Markov decision process and solves it using Proximal Policy Optimization (PPO). Upper-layer power predictions are embedded into the seven-dimensional state space to enable proactive rather than reactive energy dispatch, and the reward function jointly penalizes energy consumption, power fluctuation, and excessive battery loading. The policy network is trained over 6 million interaction steps across four representative driving cycles. Validation across 180 test trajectories demonstrates that the proposed method reduces peak battery C-rate by 80.6% and power fluctuation variance by 49.3% relative to a battery-only baseline, while supercapacitors contribute 48.9% to peak shaving under acceleration. Under extreme step-acceleration conditions, battery current root mean square (RMS) decreases by 75.7% with peak C-rate bounded at 1.4C, confirming effective protection against high-rate degradation. A V2V communication impairment handling mechanism is integrated into both layers, including first-order prediction compensation for transmission delays and adaptive parameter adjustment for packet loss.