Cyber-Resilient Smart Grids: A Randomized Coefficient-Based Defense Mechanism Against False Data Injection Attacks

The evolution of intelligent power grids into integrated cyberphysical systems has unveiled significant vulnerabilities within their digital measurement frameworks, with false data injection (FDI) attacks presenting substantial threats to grid stability by potentially inciting cascading failures via manipulated control interventions. Although the existing technique of the Variable Dummy Value Model with a nonlinear functional form (NF-VDVM) enhances the detection of FDI, it continues to be susceptible to adaptive adversaries who capitalize on deterministic patterns. This study introduces a Dummy-Variable Model with randomized coefficients (RP-VDVM) that fundamentally fortifies security measures through the implementation of dynamic coefficient randomization, effectively eliminating predictable attack vectors. Rigorously assessed on the IEEE 14-bus and other bus systems using authentic PJM load profiles, RP-VDVM exhibits remarkable 95% detection accuracy under a 5% attack intensity (representing a 2.1 times enhancement compared to NF-VDVM 45%) while maintaining a false negative rate 2% according to North American Electric Reliability Corporation (NERC) standards. An analysis of computational complexity validates the model’s theoretical scalability to more extensive systems, with anticipated execution durations of 480 milliseconds (ms) for 300-bus networks. The proposed model is also compared to other machine learning detectors in order to position it as a viable, high-performance resolution for AC power networks that surpasses both conventional state estimation methods and machine learning-based detection systems in security-critical scenarios.