A Hybrid ResNet-iTransformer Model for False Data Injection Attack Detection in Electric Vehicle Fast-Charging Stations

As a crucial flexible regulation resource in distribution networks, electric vehicle fast-charging stations exhibit high-power, stochastic fluctuation characteristics. This randomness makes false data injection attack (FDIA) particularly challenging to detect with conventional methods, thereby posing a significant threat to the secure operation of vehicle-to-grid cyber–physical systems. To address this issue, this paper proposes an FDIA detection method based on ResNet-iTransformer. Firstly, a residual network (ResNet) is employed to capture abnormal patterns in the measurement data layer by layer. Secondly, the dimension reversal technique of iTransformer is utilized to model the interrelationships among channels, enabling information exchange among feature variables through multi-head self-attention. Then, attention pooling is introduced to filter multi-channel features and focus on key channels, thereby achieving accurate attack detection. Finally, simulation tests are conducted on the IEEE 33-bus system. The results show that, compared with existing common detection methods, the proposed method achieves significant improvements in detection precision, recall, and F1 score, enabling more accurate detection of FDIA in the vehicle-to-grid cyber–physical system.