Deep learning-driven False Data Injection attack in renewable integrated smart grids

The modernisation of power grids accompanies growing reliance on information and communication infrastructure, raising the vulnerability to cyber threats. Even a trivial, undetectable manipulation of big data can seriously compromise the data integrity of the targeted power system, resulting in adverse consequences, e.g., power theft, misleading control actions, false tripping, etc. The integration of distributed energy resources further amplifies data complexity, introducing new challenges in ensuring data security. This paper proposes a novel deep learning-based False Data Injection attack scheme utilising Conditional Generative Adversarial Networks to assess potential breaches in data integrity within renewable-integrated power networks. The study evaluates the effectiveness of the proposed attack against advanced False Data Detection mechanisms in complex smart grid environments. The IEEE 5-bus, 30-bus, and 118-bus systems incorporating photovoltaic and wind-based distributed generation with realistic load and generation dynamics are modelled as target networks. Simulation results critically analyse the limitations and effectiveness of the proposed attack, demonstrating its capability to bypass state-of-the-art False Data Detection safeguards.