Deep learning technique for identifying false Data Injection threats in smart grid systems

The integration of computers and artificial intelligence (AI) improves SG management and monitors were considered significant. More reliance on technological devices renders one more susceptible to dangerous assaults. The confidentiality of data has been put at risk due to the growing vulnerability of the supervisory control and data acquisition (SCADA) framework towards False Data Injection (FDI) assaults. The behavioral traits of FDI assaults are identified in this work by using historical measurement data and deep learning (DL) techniques. The z-score normalization approach was used for cleansing the historical data and converting continuous-time signals towards discrete form and the discrete wavelet transform (DWT) technique was utilized for data extraction. To find the best FDI assaults in the SG, cuttlefish optimization is combined with enriched recurrent neural network (CO-ERNN) technology. By efficiently reducing assumptions about possible attack scenarios, the suggested cyber-attack detection system exhibits outstanding dependability. Furthermore, an optimization model is put out to describe the behavior of a certain kind of FDI functioning, specifically in situations when accessible restricted positioning measures for power theft in energy systems are at risk. Based on the IEEE 118- bus test system, the versatility of the CO-ERNN proposed attack recognition mechanism is assessed, and the results of the simulation demonstrate the effectiveness of the proposed strategy.