Detection of false data injection attack in power grid based on spatial-temporal transformer network

Abstract

False data injection attacks (FDIA) severely impact the secure operation of power grid, making accurate FDIA detection crucial for stability of power grid. Considering the complex spatiotemporal dependency of the power grid, this paper proposes a new FDIA detection method based on the Spatial-Temporal Transformer network to effectively integrate the topology and data of the power grid. In the Spatial Transformer network, the self-attention mechanism is utilized to capture the global dependency of the power grid, while the graph convolutional layer is employed to establish the local dependency of the power grid. In the Temporal Transformer network, the self-attention mechanism is employed to capture the long-term sequence dependency of power grid data, overcoming the limitation of long short-term memory network (LSTM) which can only model limited temporal dependency. Therefore, the Spatial-Temporal Transformer network can effectively mine spatiotemporal features in power grid, improving detection accuracy. The proposed detection method is evaluated using real load data from NYISO in IEEE 14-bus and IEEE 118-bus systems. Simulation results demonstrate the effectiveness and robustness of the proposed method.