Abstract
This work proposes a novel distributed set-membership estimation for the locally observable system state with random delays and false data injection (FDI) attacks. By reconstructing the delayed measurements, the random delay system becomes a delay free one. An ellipsoid based dynamic attack detection strategy with adjustable thresholds is established to detect FDI attacks. To reduce the computation load, the observability decomposition approach is used to obtain the observable substate of each node. Then the distributed estimator is designed to estimate the observable substates, and the unobservable ones are estimated by exchanging information with neighbor nodes. A sufficient condition is then obtained to derive the ellipsoid sets of the system state through iteration. Furthermore, an optimization method is used to reduce the estimation errors. Finally, the estimation performance is demonstrated through numerical simulations, which indicate the efficiency of the proposed method.