Enhancing Security in Smart Distribution Networks: Proximal Policy Cooperative Optimization Against False Data Injection Attacks

False Data Injection Attacks (FDIA) present a substantial threat to the stability and reliability of smart distribution networks by tampering with supervisory control and data acquisition measurements, misleading control centers into making erroneous decisions. However, existing research on FDIA for power grid primarily focuses on attack detection and localization, lacking effective countermeasures. To address this gap, this paper formulates the FDIA countermeasure problem for automatic generation control systems in smart distribution networks as a Markov decision process and proposes a real-time countermeasure method based on proximal policy cooperative optimization. The method introduces a cooperative exploration strategy to accelerate algorithm responsiveness by learning multi-style countermeasure policies. Through a centralized information-sharing architecture, it enhances coordination between individual policies, enabling real-time generation of optimized countermeasures. Simulation experiments under multiple attack scenarios demonstrate that the proposed method effectively maintains grid frequency stability and tie-line power balance, outperforming existing reinforcement learning methods in FDIA mitigation.