Enhanced intrusion detection in smart grids through integrated pre-processing and classification techniques

The term “smart grid” refers to an updated electrical grid infrastructure that combines conventional power supply methods with modern sensing, communication, and control technology. Ensuring the security and dependability of the contemporary electrical infrastructure in smart grids requires effective intrusion detection. Making sure the grid infrastructure is secure becomes increasingly important as SG technologies become popular. The identification and avoidance of potential threats and attacks in SG environments is a critical function of IDS. In SG, intrusion detection is essential to maintaining system security. Using a deep learning (DL) classification model in conjunction with sophisticated pre-processing and feature extraction techniques, this study investigates a novel method of intrusion detection in SG datasets. The recommended method uses CNN for classification, AE, ICA, and PCA for feature extraction. This study investigates how CNNs, t-SNE, and feature extraction increase intrusion detection dataset accuracy, precision, and recall. CNN with t-SNE and autoencoder has the highest accuracy (92 %), precision (0.89), and recall (0.87). This hybrid technique protects SG infrastructures from cyberattacks by increasing detection. We utilise Python and Jupyter Notebook.