Cyber-Resilience Enhancement and Protection for Uneconomic Power Dispatch Under Cyber-Attacks

Abstract

False data injection (FDI), could cause severe uneconomic system operation and even large blackout, which is further compounded by the increasingly integrated fluctuating renewable generation. As a commonly conducted type of FDI, load redistribution (LR) attack is judiciously manipulated by attackers to alter the load measurement on power buses and affect the normal operation of power systems. In particular, LR attacks have been proved to easily bypass the detection of state estimation. This paper presents a novel distributionally robust optimization (DRO) for operating transmission systems against cyber-attacks while considering the uncertainty of renewable generation. The FDI imposed by an adversary aims to maximally alter system parameters and mislead system operations while the proposed optimization method is used to reduce the risks caused by FDI. Unlike the worst-case-oriented robust optimization, DRO neglects the extremely low-probability case and thus weakens the conservatism, resulting in more economical operation schemes. To obtain computational tractability, a semidefinite programming problem is reformulated and a constraint generation algorithm is utilized to efficiently solve the original problem in a hierarchical master and sub-problem framework. The proposed method can produce more secure and economic operation for the system of rich renewable under LR attacks, reducing load shedding and operation cost to benefit end customers, network operators, and renewable generation.

Decentralized attack detection for multi‐area power systems via interconnection‐decoupled sliding mode observer

“…It is known from Reference 27 that the following assumptions are required to accurately estimate the system states via UIO (7):…”

“…It is easy to verify that Assumption 1 is true with C i and Ξ i given in (2) and ( 5), respectively. Under Assumption 2, we can decompose T i as T i = T 1 i + T 2 i , and design these matrices in (7) by the following expressions…”

“…Suppose that the malicious attacks on three areas are: y a 1 (t) = [0.2(t − 2) 0.4(t − 2)] ⊤ for t ≥ 2 s; y a 2 (t) = [0.1 cos(t) 0.3 sin(t)] ⊤ for t ≥ 4 s; y a 3 (t) = [0.1 0.3] ⊤ for t ≥ 7 s, respectively. The comparison detection results between the proposed UI-SMO ( 13) and the traditional UIO (7) are shown in Figures 4-6, in which the detection residual thresholds for the traditional UIO (7) based on (12) are given as (for i = 1, 2, 3):…”

Optimization Models in Cyber-Physical Power Systems: A Review

“…When a smart grid is under false data injection attack, its optimal operation differs from the normal condition. A twostage distributionally robust optimization model for economic dispatch of power systems under the load redistribution attack is proposed in [85], which also considers uncertainties of renewable energy sources. The model schedules on day-ahead and real-time basis in the first and second stages, respectively, and finds a less conservative dispatch solution in comparison with robust optimization.…”

“…Ref. [22] deploys Bender's decomposition technique to solve the MINLP problem, while the column and constraint generation method is used in [2], [23], [85] to solve multi-level models. Both decomposition techniques find the optimal solution in an iterative way.…”