Secure and trust-aware pricing mechanisms for flexible energy markets in electric vehicle charging station integrated microgrids: A multi-agent reinforcement learning approach

This paper introduces a novel cyber-resilient and trust-aware multi-agent learning framework designed for secure pricing and flexible energy market operation, particularly in microgrid environments with high penetration of electric vehicles (EVs) and charging stations (EVCSs). The proposed approach seamlessly integrates multiple essential components, cyberattack detection, trust evaluation, and market control, into a unified learning architecture, enabling real-time secure and efficient energy exchange despite the presence of adversarial conditions. The framework is built upon a hierarchical multi-agent reinforcement learning (MARL) structure, where decentralized EV and EVCS agents make local decisions based on partial observations, while a centralized critic coordinates network-wide interactions through graph-based aggregation. This structure ensures that agents make informed decisions about energy pricing, charging, and vehicle-to-grid interactions, while remaining resilient to cyber disturbances. The framework's performance is validated on three increasingly complex power systems, demonstrating robust resilience against common cyber threats such as false data injection, denial-of-service, and latency-based attacks. The simulation results reveal significant improvements in market stability, trust retention, and pricing efficiency, including price-volatility reductions of up to approximately 78% relative to the rule-based pricing baseline under challenging attack scenarios. This work provides a scalable and adaptable solution for the secure operation of future flexible energy markets.