Salp Swarm-Artificial Neural Network Based Cyber-Attack Detection in Smart Grid

Smart Grid (SG) can be easily attacked by smart hackers who try to corrupt the data aggregated by the acquisition system and supervisory control. The hacker has the capability to cheat the Bad-Data Detector (BDD) and compromise the system by injecting malicious data into the meter measurement data. This can lead to wrong decision-making, economical loss, power outages, and so on. To address these issues, a bio inspired Salp Swarm Optimization (SSO) based cyber-attack detection technique is proposed. In the proposed salp neural model for cyber-attack detection, the State Estimation is initially done and the bad data is identified by a BDD. Then, the features are extracted by Discrete Wavelet Transform and the dimensionality reduction process takes place. Here, we use Kernel Principle Component Analysis for reducing the dimensionality. Once the data is decomposed to lower dimensions, the presence of attack in SG is detected by the Artificial Neural Network (ANN) classifier. The SSO determines the most optimal weight values of ANN. This improves the classification accuracy. The performance of the proposed salp neural model is tested in standard IEEE 118- bus and 57-bus test systems. Based on the evaluation and comparison with the existing schemes, the proposed salp neural technique has observed to have better performance in terms of metrics accuracy, Receiver Operating Characteristic curve, and F1-score.