A novel data-driven time-delay attack evaluation method for wide-area cyber–physical smart grid systems

Abstract

One of the main challenges in information and communication technology, as well as wide-area power system researches, is the inability to stabilize the voltage and frequency for cyber–physical power systems under the communication time-delay. Thus, it is imperative to investigate the stochastic characteristics and quantitative method of time-delay. In this paper, a novel data-driven time-delay evaluation method for cyber–physical smart grid systems is proposed. The cyber layer and physical components in wide-area power systems are modeled. The transmitted process of the data packet in a cyber layer is considered, and the probability distribution function of multiple latencies in the spatial sequence is depicted by using M/M/1 queuing theory and signal convolution methods. The Sklar’s theorem and Copula’s theory are utilized to integrate distribution functions of multiple delays into a novel delay quantitative model. Furthermore, the time delay is divided according to the reason of its generation as two parts: (1) inherent delay by nature, and (2) delay caused by network attacks. Various sizes of delays caused by cyber-attacks are detected by using communication principle and likelihood ratio. Finally, the case studies under various scenarios are carried out on the New England ten-machine power system. The simulation results show that the stochastic latencies obey the proposed delay distribution model with probability 99.9%. The comparative results illustrate the mean-square error between the proposed and actual delay distribution probability function as 10 −4 . The delay caused by cyber-attacks can be separated from nature latency with the threshold η = 0 . 0101 .