Detection of Cyber Attacks in Modern Renewable Integrated Power Sector: A Matrix Separation Scheme

The rapid development in the modern power sector has led to the large-scale inclusion of renewables and electric vehicles. State estimation (SE) algorithms in the control centre present an effective way of monitoring and furnishing the operating conditions. This work presents an extended Taylor series-based power system SE where the measurements (active and reactive power) can be made linearly independent, thus enhancing the system redundancy. The presented approach also caters to unobservable systems where conventional SE algorithms can not be effectively executed. False data injection attacks (FDIAs) target the solutions of these estimation algorithms and may lead to critical scenarios by inherently bypassing the conventional bad data detectors (BDDs). For an effective identification of such attacks within the acquired measurements, this work presents a matrix separation scheme-based attack detection strategy. A comparative analysis between three undertaken matrix separation approaches namely the augmented Lagrangian multiplier (ALM), augmented Lagrangian alternating direction method (ALADM) and Go-decomposition technique (Go-Dec) furnish an efficient attack identification framework. Moreover, when tested on large-scale systems (IEEE 14 and 118 bus) under varying attack and noise conditions, the presented approach furnishes a robust and efficient attack identification framework with a computational burden ranging in the order of a few milliseconds, thus resulting in a real-time detection policy.