Real-time Identification of False Data Injection Attack in Smart Grid

Modern smart grid operators employ state estimation algorithm to ensure real-time monitoring of the power system. It leads to an accurate identification of the current operating states of the grid. Recently false data injection attacks (FDIAs) have shown their capability to generate a forged set of estimates by circumventing the conventional bad data detector. Real-time successful identification of false data injection attacks is an indispensable requirement to ensure secure and reliable grid operation. State forecasting driven detection models can effectively determine the deviations of the operating states due to FDIA. This work showcases a scalable deep neural network based state forecasting policy which is capable of detecting FDIAs in real-time. With an optimal set of hyper-parameters, an effective state forecasting policy with minimal error has been showcased. A critical comparison between the proposed neural network model with the state of the art forecasting policies showcase its efficacy. An extensive survey on the IEEE 14 bus test bench portrays the effectiveness of the proposed real-time FDIA detection policy.