Analyzing Malicious Data Injection Attacks on Distributed Optimal Power Flow Algorithms

The rapid growth of distributed energy resources motivates the application of distributed optimization algorithms for solving optimal power flow (OPF) problems. Using distributed optimization to solve OPF problems requires repeated data sharing between agents responsible for different regions of the power system. The performance of distributed optimization algorithms depends on the integrity of the shared data and the communications between the agents. This paper investigates the vulnerability of a distributed algorithm called Auxiliary Problem Principle (APP) with respect to cyberattacks on the communications between the agents. We propose cyberattack models that manipulate the shared data between neighboring agents to achieve an objective such as driving the algorithm to converge to a suboptimal solution. We investigate the convergence characteristics of the APP algorithm in the context of the DC optimal power flow (DC OPF) problem with and without manipulating the shared data. Last, we demonstrate how trained neural networks can provide highly accurate attack detection.